A Modular Bioinstructive Platform for Additive-Free, Topography-Driven Stem Cell Differentiation and Patterning

Recreating 3D bone formation in vitro without biochemical inducers remains a longstanding challenge in preclinical testing . We present a scalable, bioinstructive platform based on polylactic acid microparticles with controlled dimpled surface features that direct mesenchymal stem cell differentiation through endogenous topography-mediated mechanotransduction, establishing a mechanistically validated, additive-free platform. These 3D topographical cues drive cytoskeletal reorganisation and induce osteogenesis via canonical Hedgehog signalling. RNA-Seq revealed early significant upregulation of cytoskeletal components and osteochondral transcription factors, including runt -related transcription factor 2 (RUNX2) and SRY -box transcription factor 9 (SOX9), followed by activation of the insulin growth factor -II pathway and osteogenic commitment . To demonstrate translational potential , two - photon polymerisation lithography was employed to engineer precise ly-patterned 3D topographies, inducing graded GLI1 expression without added soluble cues . This establishes a modular, versatile platform for stem cell engineering , offering a topography-driven, non-genetic analogue to mechanogenetics with broad utility for regenerative medicine and human-relevant development of bone models. .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 2 Graphical abstract:

Differentiation; Hedgehog signalling; Mesenchymal stromal cells; Microparticles; Osteogenesis .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 3

Bone formation is fundamental to skeletal development, homeostasis and repair, yet current tissue engineering strategies rely predominantly on costly growth factors that inadequately rec reate the intricate processes of osteogenesis . Although it is well established that cell -matrix interactions can be used to direct cell response [1], a translation-ready approach to unlocking the potential of cell -instructive topographies for modulation of cell signalling has not been realised. This challenge is particularly relevant for tissue engineering, where traditional approaches rely heavily on costly, externally supplemented soluble factors that often fail to recapitulate the complexity of native tissue formation and can mask intrinsic cellular responses. The development of an alternative strategy leveraging cell -instructive biophysical cues to harness the inherent mechanosensory capabilities of cells to guide osteogenesis will transform regenerative medicine by enabling precise control over tissue formation while reducing costs. Engineering functional bone tissue demands biomaterials that can precisely direct stem cell fate and patterning. While biochemical factors, such as dexamethasone, are widely used to induce differentiation of human mesenchymal stromal cells (hMSCs), they introduce confounding effects, which may result in inconsistent cellular responses and unintended lineage outcomes [2]. This includes potential activation of adipogenic pathways and upregulation of oxidative stress -related genes, compromising bone formation [3]. The hierarchical structure of bone’s extracellular matrix (ECM) has inspired scaffold design with customised topographies for bone regeneration [4]. Microtopographies, including pits, pillars, and gratings on titanium implants promote MSCs osteogenic differentiation in the presence of osteoinductive supplements [5, 6]. However, .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 4 titanium’s stiffness and highly adhesive surface chemistry often favours fibrogenesis over the desired osteogenesis [7]. Similarly, tailoring topographically -textured micropatterns on glass slides coated with fibronectin (FN) has also been reported to promote osteogenic differentiation in ra ts [8], but FN suffers from low stability and degradation over seven days when subjected to mechanical forces such as laminar flow with shear stress [9], limiting long-term effectiveness in certain applications. Recent advances in our understanding of mechanotransduction pathways have revealed promising alternatives to biochemical manipulation. The Hedgehog (HH) signalling pathway plays a pivotal role in the development of the skeletal system during embryogenesis [10], and regulates bone remodelling throughout postnatal life [11]. Critically, its mechano -responsiveness allows it to function as a central mediator of biomechanical cues, translating external mechanical forces into cellular response [12]. The core components of the HH signalling pathway consist of HH ligands, with Sonic Hedgehog (SHH) being the primary member, the twelve -pass transmembrane receptor, Patched1 (PTCH1), the seven -pass transmembrane signal transducer Smoothened (SMO), and the effector transcription factor Glioma-associated oncogene homolog 1 (GLI1) [13]. Dysregulation of the HH pathway has been reported in primary bone tumours such as osteosarcoma, and other musculoskeletal disorders such as osteoporosis and osteoarthritis [14], reinforcing its central role in skeletal homeostasis. Topographically-textured 3D polymeric microparticles offer a promising platform for bone tissue engineering . mimicking native bone microarchitecture , while providing cell-instructive capabilities and a high surface area-to-volume ratio for large-scale cell expansion [15]. Previous research has demonstrated that 3D convex curvature enhances hMSC osteogenic differentiation by modulating the cytoskeleton, altering the distribution of focal adhesion proteins such as vinculin (VCL), and reducing stress .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 5 fibre formation [16]. The osteoinductive impact of topographically -textured microparticles on murine mesenchymal progenitors is mediated by the SMO - dependent activation of GLI1, but this has not been assessed in relevant human primary cells [17]. This study aims to leverage these cell -instructive microparticles to investigate mechanically guided osteogenesis in primary hMSCs, mapping the transcriptional landscape of hMSCs cultured on these osteoinductive substrates to identify key bone- specific gen e signatures and downstream molecular pathways governing lineage commitment and driving osteogenesis without the need for confounding biochemical supplements. Additionally, a key innovation of this study is the application of two - photon polymerisation (2PP ) lithography to engineer precisely controlled GLI1 expression gradients, achieving fine -tuned modulation of cellular responses via engineered high -resolution topographies. While 2PP has been used to fabricate surfaces with gradients in roughness or stiffness [18, 19], its application to generate spatially defined gradients of cell -intrinsic signalling , such as GLI1 expression, represents a novel approach, bridging advanced fabrication techniques with the study of spatial regulation of cell fate. This study advances our understanding of how stem cells respond to 3D surface topography by showing that these surface -engineered microparticles with defined microfeatures can spatially modulate GLI1 expression in the absence of exogenous biochemical stimulation that may mask inherent cellular responses. Transcriptomic analysis further supports topography-induced activation of Hedgehog signalling and its role in directing early cell fate decisions. By leveraging high-resolution topographical cues to induce spatially patterned Hedgehog pathway activation, we present a versatile platform for probing the physical regulation of stem cell signalling and directing differentiation with microscale precision. .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 6

Design and fabrication of surface-engineered, cell-instructive microparticles To develop a scalable, supplement-free bone regeneration platform, we optimised our previously developed osteoinductive system [15, 17] for systematic investigation of topography-mediated mechanotransduction. This platform leverages topographical surface patterning to modulate cell responses. We engineered this platform based on three translational design criteria: (1) Microparticles should present precise microtopographical cues to enable spatially resolved, mechanotransduction -driven cell fate decisions; (2) Surface architecture should be tailored to promote optimal cell attachment and additive-free osteogenic differentiation [15, 17], in line with efforts to eliminate the need for exogenous biochemical cues for reduced regulatory complexity; and (3) Injectable particle size, facilitating minimally-invasive delivery and scalability for clinical translation. Smooth and topographically-textured (‘dimpled’) microparticles were fabricated using a solvent evaporation oil-in-water emulsion technique, incorporating phase separation of fusidic acid (FA) as a sacrificial component to achieve defined surface patterning [15]. During hardening, FA is excluded from the bulk of the microparticle, producing distinctive surface patterns (Figure 1A). PLA was selected for its biocompatibility and hydrophobicity, preventing degradation during culture [15, 20]. .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 7 Figure 1: Fabrication of smooth and dimpled PLA microparticles. A) Schematic representation showing the fabrication workflow of smooth (i) and dimpled microparticles (ii) by a modified oil -in-water solvent evaporation emulsion method. B, C) Representative scanning electron microscopy images of smooth B) and dimpled C) microparticles acquired at 10 kV (Scale bars: 20 μm). D, E) Atomic force microscopy topography images of the smooth D) and dimpled E) microparticles (Scale bars: 1 μm). Abbreviations: PLA, Poly(D,L-lactic acid); DCM, Dichloromethane; FA, Fusidic acid. .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 8 Fabrication parameters were optimised to ensure comparable average sizes of the fabricated microparticles, providing smooth microparticles as an appropriate control for 3D surface topography. Distinct surface morphologies were achieved through modulation of processing parameters, including homogenisation speed, polymer concentration and polymer/FA ratio (Table 1). Microparticles of similar size ranges of 46.52 ± 6.81 μm for smooth and 45.79 ± 5.43 μm for dimpled have been demonstrated to be injectable through clinically -relevant 21G needles [21]. The average dimple size falls within the mean size range that we previously reported to induce osteogenesis in hMSCs [15] and C3H10T1/2 cells [17] (Table 1). Surface area measurements (m²/g) indicated minimal porosity in both designs, reflected by low micro-pore (Vmicro) and meso-pore (Vmeso) volumes (Table 1). The successful fabrication of topographically -textured microparticles was confirmed through scanning electron microscopy (SEM) and atomic force microscopy (AFM) analyses, revealing well-defined surface patterns and feature dimensions (Figure 1). Dynamic light scattering measurements demonstrated narrow size distributions (Table 1). To maintain consistency in surface area available for cell attachment across samples, microparticle quantities were calculated to ensur e uniform cell seeding density across 2D and 3D cultures. .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 9 Table 1: Fabrication parameters and properties of the polymeric microparticles used in this study Smooth Dimpled Emulsion settings FA/PLA ratio 0/100 23/77 Homogenisation RPM 3800 1300 Total polymer concentration (w/v%) 20 10 Particle properties Particle size (μm) 46.52 ± 6.81 45.79 ± 5.43 Polydispersity index (PDI) 0.12 ± 0.03 0.13 ± 0.01 Dimple size (μm) - 4.35 ± 1.06 BET surface area (m2/g) 0.22 0.33 Vmicro (cm3/g) 5.10 × 10-6 1.16 × 10-5 Vmeso (cm3/g) 1.16 × 10-4 1.82 × 10-4 Abbreviations: FA, Fusidic acid; PLA, Poly(D,L-lactic acid); BET, Brunauer –Emmett–Teller; RPM, Rotation per minute; Vmicro, micropore volume; Vmeso, mesopore volume (we define mesopore volume as the pore volume originating from the pores smaller than 20 nm). Polymer microparticles are valuable for biomanufacturing workflows that depend on effective cell adhesion [15, 22]. hMSCs cultured on microparticles demonstrated excellent cell attachment and viability in serum-reduced medium (Figure 2A), with no significant differences observed between the two designs at any time point ( Figure 2B). However, cell numbers in 3D-cultured samples were significantly lower than 2D- cultured controls at day 14 (p ≤ 0.0001). .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 10 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 11 Figure 2: Impact of microparticles design on viability, proliferation, morphology and cytoskeletal organisation in hMSCs cultured on smooth and dimpled microparticles. A) Representative fluorescence microscopy images showing high viability of hMSCs on smooth and dimpled microparticles 3 days after seeding in serum -reduced medium. Live cells are stained with calcein -AM (green) and dead cells with ethidium homodimer III (EthD-III) (red) (Scale bars: 200 μm). B) DNA content in hMSCs cultured on dimpled and smo oth microparticles versus 2D-cultured controls quantified at days 1, 7 and 14 days after seeding. Statistical significance is determined by two-way ANOVA with Tukey's multiple comparisons test. Data represents mean ± SD (**** p < 0.0001, N= 3 donors). C) Scanning electron microscopy images showing hMSCs morphologies 3 days after seeding on smooth and dimpled microparticles (Scale bars: 20 μm). D) Representative confocal maximum intensity projection images of hMSCs stained for VCL (red), F-actin (green), and nuclei (DAPI; blue) after 3 days in culture (Scale bars: 2 0 μm; N= 2 donors). White arrowheads indicate VCL localisation. Abbreviations: EthD III, Ethidium homodimer III; VCL, Vinculin; F -actin, Filamentous actin ; DAPI, 4′,6-Diamidino-2-phenylindole Cells demonstrated distinct morphological adaptations in response to topographical design, with a spread -out, flattened morphology on smooth microparticles and elongated, spindle-like appearance on the dimpled design (Figure 2C). This prompted the investigation of cytoskeletal architecture. Dimpled surfaces induced unique cytoskeletal arrangements (Figure 2D) that suggest differential mechanotransduction signalling. Apparent focal adhesions, which reflect cell-extracellular matrix interactions and subsequent cytoskeletal reorganisation [23], were examined by co-staining for VCL with F-actin. Immunostaining revealed well -defined, streak-like focal adhesions at the leading edge and pronounced F-actin stress fibres in hMSCs cultured on planar 2D-cultured controls. In contrast, cells seeded on dimpled microparticles exhibited diffuse VCL localisation and poorly defined focal adhesion structures , indicating impaired focal adhesion assembly [24]. Cells on smooth microparticles displayed mature focal adhesions (Figure 2 D), highlighting the distinct effects of surface topography on focal adhesion dynamics. .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 12 Transcriptome profiling identifies unique temporal transcriptional signatures of hMSCs cultured on dimpled 3D topographical features While our previous work established the osteoinductive capacity of topographically - textured microparticles [15, 17] and the involvement of canonical HH signalling in this response in murine C3H10T1/2 cells, the underlying molecular mechanisms in human cells remained unexplored. Decoding how stem cells interpret osteoinductive physical cues at the transcriptional level will enable the design of bioinstructive systems that leverage mechanically-guided developmental mechanisms for regenerative and tissue engineering applications. To build on our findings, RNA sequencing (RNA-Seq) was therefore performed to determine the genome-wide transcriptional changes induced by different microparticle designs. hMSCs from three independent donors (Table S1) were seeded on smooth microparticles (serving as 3D topographical control) and two sets of dimpled microparticle cultures. On the following day, one set of dimpled samples was treated with KAAD-cyclopamine (a HH signalling inhibitor), generating three sample conditions: smooth microparticles, dimpled microparticles, and KAAD- cyclopamine-treated dimpled microparticles (Figure 3A). .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 13 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 14 Figure 3: Dimpled microparticles show distinct transcriptional profiles compared to smooth micropartice cultures after 3 and 14 days in culture. A) Schematic representation of the experimental strategy for RNA-Seq utilised in this study. HMSCs from three donors were cultured on dimpled and smooth microparticles. On the following day one dimpled group was treated with KAAD-cyclopamine (HH antagonist). RNA-Seq was performed after 3 and 14 days in culture. B) (i) Schematic illustrating canonical Smoothened (SMO) -dependent HH activation, which begins when HH ligand inhibits Patched 1 (PTCH1), allowing SMO to overcome Suppressor of Fused (SUFU)-mediated repression and activates GLI, driving target gene transcription. (ii) KAAD-cyclopamine antagonises SMO, inactivating the canonical HH pathway. C) Principal component analysis plot displaying transcriptomic variance across dimpled, smooth and cyclopamine-treated dimpled samples at days 3 and 14 post -seeding. Each point represents an individual sample. D) Pie charts showing the average proportion of upregulated (log2 fold change > 1, red) and downregulated (log2 fold change < -1, blue) genes, with padj < 0.05, based on RNA-Seq analysis. E, F) Gene ontology enrichment analysis for biological processes in hMSCs cultured on dimpled versus smooth after 3 (E) and 14 (F) days in culture performed with Enrichr. Pathways of interest highlighted by red boxes. Length of the bar represents the degree of gene enrichment. Gene count is indicated on the x-axis. Colour indicates Benjamini-Hochberg adjusted p value. Abbreviations: GLIR, Glioma-Associated oncogene homologs repressor form ; GLIA, GLI activator form ; PCA, Principal Component Analysis; GO, Gene Ontology; BP , Biological Process. Canonical HH signalling is initiated with the binding of SHH ligand to PTCH1, relieving the inhibition of the transmembrane protein SMO. The activation of SMO enables the nuclear translocation of the transcription factor GLI1, driving the transcription of HH target genes, such as RUNX2 [25]. Treatment with KAAD -cyclopamine (SMO antagonist) inactivates this pathway (Figure 3B). RNA-Seq was performed at two time points: day 3, capturing any HH signalling activation and early topographically-induced osteogenic commitment, and at day 14, to assess HH signalling and downstream pathways governing cellular adaptation to the microenvironment (Figure 3A). Principal component analysis (PCA) revealed distinct clustering by topographical condition and time point. At day 3, hMSCs cultured on dimpled microparticles segregated clearly from both smooth and cyclopamine -treated dimpled cultures (Figure 3C). The overlap observed between smooth and cyclopamine-treated dimpled samples suggests strong similarity of their transcriptional profiles at this early time .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 15 point (Figure 3D), with only three genes being differentially expressed at day 3 : Zinc finger protein 117 ( ZNF117), poly(rc )-binding protein 1 ( PCBP1), and thioredoxin domain containing 5 ( TXNDC5). By day 14, cells cultured on dimpled microparticles formed distinct clusters relative to those on dimpled microparticles at day 3, indicating sustained transcriptional divergence driven by the dimpled topography. At day 14, only 0.43% of genes were differentially expressed between cells cultured on smooth microparticles and those on dimpled microparticle cultures treated with KAAD- cyclopamine (Figure 3D). To identify putative drivers of hMSC adaptation to topography, we examined the top 30 up- and down-regulated DEGs at day 3 and day 14 post-seeding (Table S2 and S3, respectively) . Early responses (day 3) were marked by upregulation of genes involved in cytoskeletal organisation, including tubulin alpha 3C ( TUBA3C), and transcription factors linked to early osteoblastogenesis, such as GATA binding protein 4 (GATA4). By day 14, differentially expressed genes were increasingly associated with musculoskeletal tissue development and matrix remodelling, including sclerostin domain-containing 1 (SOSTDC1) and serpin family E member 2 (SERPINE2). Culture of primary hMSCs on dimpled microparticles resulted in 11979 genes (19.11%) at day 3 and 2701 genes (4.31%) at day 14 of all analysed genes (62700 genes) showing significant differential expression (with log2 fold change > 1 and padj < 0.05) relative to culture on unpatterned smooth microparticles. Gene Ontology (GO) analysis was conducted to identify biological processes (BP) that were enriched in cells cultured on dimpled versus smooth microparticles. At day 3, enriched pathways were primarily related to ECM organisation, skeletal system development and angiogenesis (Figure 3 E). By day 14, enrichment of ECM-related processes and skeletal system development remained prominent, with positive .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 16 regulation of angiogenesis and vasculature development emerging (Fig ure 3F). Enriched GO terms with padj-values and overlap metrics are listed in Table S4. Dimpled topographical features promote transcriptional upregulation of cytoskeletal genes and activate canonical Hedgehog signalling At day 3 post -seeding, there was significant differential upregulation of key mechanosensing molecules, including piezo -type mechanosensitive ion channel component 2 ( PIEZO2) and integrin subunit beta 4 (ITGB4), as well as associated structural genes such as LAMA5, which forms a functional complex with ITGB4 [26]. (Figure 4A). .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 17 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 18 Figure 4: Culture of primary hMSCs on dimpled microparticles induces the upregulation of key Hedgehog signalling pathway components. A, B) Volcano plots displaying HH pathway-related differentially expressed genes in dimpled versus smooth microparticlecultures at day 3 (A) and day 14 post -seeding (B). Differential expression was defined as log₂ fold change > 1 (upregulated, red) or < –1 (downregulated, blue) with padj < 0.05 (N = 3 donors)C-E) Quantitative real-time PCR (qPCR) analysis of key HH genes: GLI1 (C), PTCH1 (D), and SMO (E), after 3 days in serum-reduced medium, relative to untreated 2D controls. F) Relative GLI1 expression after 3 days of 300 nM KAAD-cyclopamine treatment or 0.06% (v/v) DMSO in serum-reduced medium, and relative to 2D vehicle-only controls (with 0.06% DMSO). Expression in 2D purmorphamine -treated controls (2 μM) was calculated relative to 2D vehicle-only controls. (N= 5 donors). Statistical significance was calculated using one-way ANOVA with Tukey's multiple comparisons test. Values are shown as mean ± SD (*p < 0.05, **p < 0.01, ***p < 0.001). G) Representative confocal maximum intensity projection images of hMSCs in response to KAAD-cyclopamine treatment, stained for GLI1 (red) and nuclei in blue ( DAPI) after 7 days in culture (N= 2 donors; Scale bars: 20 μm). Samples labelled with "-KAAD" indicate those treated with KAAD-cyclopamine. Abbreviations: GLI1, glioma-associated oncogene homologs 1; SMO, smoothened; PTCH1, patched 1; HHIP, hedgehog-interacting protein; PIEZO2, piezo type mechanosensitive ion channel component 2; VCL, vinculin; TLN1, talin 1; PXN, paxillin; ZYX, zyxin; LAMIN1; LIM domain and actin binding 1 ; ACTG2, Actin Gamma 2, Smooth Muscle; TNNT1, troponin T1; TNNC1, troponin C1; EEF1A2, eukaryotic translation elongation factor 1 alpha 2; L1CAM, L1 cell adhesion molecule; TUBAP2, tubulin alpha pseudogene 2; TUBB4A, tubulin beta 4A; TUBA3C, tubulin alpha 3C; LAMA5, laminin subunit alpha 5; ABI3BP , ABI family member 3 binding protein; FN1, fibronectin 1; ITGA5, integrin subunit alpha 5; ITGB4, integrin subunit beta 4; DMSO, Dimethyl sulphoxide ; DAPI, 4′,6-Diamidino-2-phenylindole; Pur, Purmorphamine; DIMP, Dimpled; SM, Smooth; DIMP_KD, Dimpled culture treated with KAAD-cyclopamine . The marked upregulation of eukaryotic translation elongation factor 1 alpha 2 (EEF1A2) and L1 cell adhesion molecule ( L1CAM) suggests active cytoskeletal remodelling [27, 28]. Several tubulin genes were upregulated, including tubulin alpha pseudogene 2 ( TUBAP2) and TUBA3C, which encode microtubule structural components. The upregulation of c omponents of the actin –myosin contractile apparatus, including Troponin T1 ( TNNT1) and actin gamma 2, smooth muscle (ACTG2), may reflect increased cytoskeletal tension in response to dimpled topography [29-31]. On the other hand, other ECM and cytoskeletal-associated genes, including FN1, ABI family member 3 binding protein ( ABI3BP), and LIM domain and actin binding 1 ( LIMA1) were downregulated . These elements play critical roles in .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 19 regulating cytoskeletal dynamics by modulating focal adhesion and actin filament assembly in response to mechanical cues [32]. Consistent with this, focal adhesion - associated genes such as talin 1 (TLN1), paxillin ( PXN) and VCL were also downregulated (Figure 4A). This aligns with our earlier data, where VCL immunostaining indicated apparent focal adhesion disassembly in hMSCs on dimpled microparticles (Figure 2 D). Notably, FN1 and integrin subunit alpha 5 ( ITGA5) were significantly upregulated at day 14, suggesting temporal reorganisation of adhesion - related gene expression (Figure 4B). Transcriptomic analysis at day 3 post-seeding revealed significant upregulation of key HH signalling genes such as GLI1, SMO and PTCH1 (Figure 4A). The sustained activation of the HH signalling pathway was evident in dimpled microparticles after 14 days of culture (Figure 4 B), although SMO expression was no longer differentially expressed relative to smooth microparticle cultures at day 14 (Fig ure 4 B). Concurrently, the expression of hedgehog-interacting protein (HHIP), a negative regulator of HH signalling, was found to be significantly upregulated in dimpled microparticle- relative to smooth microparticle-cultures at day 14 post-seeding. The differential expression of key HH pathway genes at day 3 was validated by real - time qPCR, confirming significant upregulation of GLI1 (2.30-fold, p < 0.05; Figure 4C), PTCH1 (1.51-fold; p < 0.0001, Figure 4D) and SMO (1.84-fold; p < 0.05, Figure 4E) in hMSCs on dimpled versus smooth microparticles, relative to 2D cultures. To determine the route of HH signalling activation in dimpled microparticle s-based cultures of hMSCs, GLI1 expression was analysed at the transcript and protein levels using real-time qPCR and immunostaining, respectively. At day 3 post -seeding, real- time qPCR revealed that the treatment of dimpled microparticle cultures with 300 nM KAAD-cyclopamine significantly reduced GLI1 expression levels compared to .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 20 untreated dimpled microparticle cultures ( p < 0.01) . This reduction was to a level comparable to that observed in hMSCs cultured on smooth microparticles (Figure 4F). This was conducted using a DMSO-containing vehicle control, which had no appreciable effects on GLI1 levels. Although multiple GO terms appear enriched in the comparison between KAAD -cyclopamine-treated dimpled and smooth surfaces, the overall differences are minimal , as indicated by the low gene counts per term and borderline significance (Figure S1). At day 7, immunostaining revealed that KAAD- cyclopamine effectively inhibited GLI1 upregulation at the protein level, while untreated dimpled cultures exhibited visibly higher GLI1 expression (Figure 4G). These findings highlight the dependence of HH pathway activation on SMO and confirmed the canonical activation of the HH signalling in hMSCs in response to 3D dimpled topographies. Subsequent analyses focused on comparing dimpled against smooth microparticle cultures to isolate intrinsic effects of topographical features on hMSCs, as the cyclopamine-treated cultures closely resembled smooth microparticle cultures, and primarily served to confirm HH pathway involvement. Transcriptomic analysis identifies a transient , developmentally -relevant osteochondral state driving microparticle-induced osteogenesis RNA-seq analysis revealed a distinct temporal progression in lineage commitment and differentiation of hMSCs, driven by the 3D dimpled topographies (Figure 5A). .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 21 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 22 Figure 5: Transcriptomic analysis of hMSCs on dimpled versus smooth microparticles reveals novel insights into mechanically-guided osteogenesis without the confounding influence of biochemical additives. A) Heatmap showing expression of selected genes associated with skeletogenesis across experimental conditions at days 3 and 14 post-seeding. B, C) Volcano plots displaying key differentially expressed genes associated with cytoskeleton, osteogenesis, and chondrogenesis in dimpled versus smooth microparticle cultures at day 3 (B) and day 14 (C) post-seeding (log2 fold change > 1 and padj < 0.05, N= 3 donors). D, E) Relative quantitative real -time PCR (qPCR) analysis of gene expression at specific time -points in serum -reduced medi um, relative to 2D controls: RUNX2 (D; N= 5 donors) after 3 days, and SP7 (E; N= 5 donors) after 10 days. 2D positive control treated with 2 μM purmorphamine is shown as a reference point for RUNX2 expression. (F) Relative qPCR analysis of BGLAP expression after 10 days of either 300 nM KAAD-cyclopamine treatment or 0.06% (v/v) DMSO in serum-reduced medium, and relative to 2D vehicle-only controls (N= 5 donors). Statistical analysis was conducted using one -way ANOVA with Tukey's multiple comparisons was applied. Values are presented as mean ± SD (** p < 0.01, *** p < 0.001, ****p < 0.0001). G, H) Representative maximum intensity projection confocal images of hMSCs stained for COL1A1 (green; G), and OCN (green; H) with nuclei counter-stained in blue (DAPI) after 10 days of culture (Scale bars: 50 μm). A single optical slice is used to represent brightfield for clarity. Abbreviations: SOX9, SRY-box transcription factor 9; COL2A1, collagen type II alpha 1 chain; COL10A1, collagen type x alpha 1 chain; COL1A1, collagen type I alpha 1; COL6A1, collagen type VI alpha 1 chain ; ACAN, aggrecan; RUNX2, runt-related transcription factor 2; SP7, specificity protein 7, also known as osterix; MSX2, Msh homeobox 2; SPP1, secreted phosphoprotein 1, also known as osteopontin; POSTN, periostin; PAX9, paired box 9; BGLAP, bone gamma -carboxyglutamate protein , also known as osteocalcin; ALPL, alkaline phosphatase; TNC, tenascin-c; SCUBE3, s ignal peptide -CUB-EGF domain -containing protein 3 ; SERPINE2, serine protease inhibito r 2; GALNT1, N - acetylgalactosaminyltransferase 1; LEPR, leptin receptor ; SOSTDC1, s clerostin domain - containing protein 1; CNMD, chondromodulin; BMP, bone morphogenetic protein; IGF-II, insulin-like growth factor 2 ; IGFBP, insulin-like growth factor binding protein; DAPI, 4 ′,6- Diamidino-2-phenylindole; OI media, Osteoinductive media ; Pur, P urmorphamine; DIMP, Dimpled; SM, Smooth; DIMP_KD, Dimpled cultures treated with KAAD-cyclopamine . The early mechanosensory response observed at day 3 post -seeding was accompanied by the significant upregulation of genes associated with early osteogenic commitment in hMSCs seeded on dimpled microparticles, including alkaline phosphatase (ALPL) and key transcriptional regulators of skeletogenesis such as Msh homeobox 2 (MSX2), and paired box 9 (PAX9). The expression of RUNX2, a key transcriptional regulator of osteogenesis, was observed in both smooth and dimpled microparticle cultures (Figure 5A). Several bone morphogenetic proteins (BMPs) that are critical to bone formation , including BMP2, BMP4 and BMP7, were significantly .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 23 upregulated in dimpled microparticle cultures (Figure 5B). Interestingly, an upregulation of expression of SRY-box transcription factor 9 (SOX9), the master regulator of chondrogenesis, and its downstream target collagen type II alpha 1 chain (COL2A1) were also observed at day 3 , suggesting a transient osteochondral bipotential state at this early time point [33] (Figure 5B). However, other chondrogenic-specific markers were also significantly downregulated at day 3 , such as aggrecan (ACAN; Figure 5 B) confirming incomplete commitment to chondrogenesis [34]. The substantial downregulation of signal peptide -CUB-EGF domain-containing protein 3 ( SCUBE3) further suggests impaired BMP -mediated chondrogenesis [35] (Figure 5B). By day 14, hMSCs on dimpled microparticles exhibited a clear trajectory toward s a bone-specific transcriptional signature , displaying a profile indicative of osteogenic priming. This was supported by expression of RUNX2 being sustained (Figure 5A) and significant upregulation of osteoblast -associated markers, including secreted phosphoprotein 1 (SPP1; encoding osteopontin), leptin receptor (LEPR), tenascin-C (TNC), N-acetylgalactosaminyltransferase 1 (GALNT1), and periostin (POSTN; osteoblast-specific factor 2). SOSTDC1, a HH-responsive BMP antagonist known to support trabecular bone maintenance [36], was also upregulated (Figure 5C ). Upregulation of FGFR1 but not FGFR2 at this time point further supports osteogenic differentiation [37]. SERPINE2 was among the most significantly upregulated genes (4.47-log2 fold change, padj < 0.01), consistent with its known role in extracellular matrix remodelling and osteogenic differentiation [38]. Moreover, bone matrix -associated collagens such as COL1A1, COL6A1, and COL8A1 were significantly upregulated at day 14, suggesting commitment towards a skeletal or osteochondral lineage [39, 40]. Furthermore, differential upregulation of insulin-like growth factor 2 (IGF-II) along with .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 24 its binding proteins IGFBP2, IGFBP3 and IGFBP6 was observed. While significant upregulation of expression of COL10A1 was observed at day 14 , SOX9 and ACAN were not differentially expressed at this time -point, while COL2A1, a cartilage -specific matrix protein [41] and chondromodulin ( CNMD), a cartilage - associated glycoprotein [42], were significantly downregulated (Figure 5C). Another marker of hypertrophic chondrocytes, matrix metalloproteinase 13 (MMP13) [43], was detected at low levels at day 14; however, statistical significance could not be assessed due to limited read counts at this time point (Figure 5A). Real-time qPCR validation demonstrated significant upregulation of expression of RUNX2 (1.44-fold, p < 0.01; Figure 5D) at day 3 post -seeding and specificity protein 7 (SP7; 10.70 -fold, p < 0.0 01; Figure 5E ) at day 10 post -seeding in dimpled microparticle cultures compared to smooth cultures and relative to 2D controls. Notably, there was no significant difference in SP7 expression between cells seeded on dimpled microparticles and in 2D cultures treated with osteoinductive medi um. Moreover, elevated expression levels SPP1 and integrin-binding sialoprotein (IBSP) were observed at day 10 post -seeding relative to 2D controls (Figure S2), though no statistically significant differences were detected. Immunostaining confirmed enhanced expression of COL1A1 (Figure 5G) and osteocalcin (OCN; Fig ure 5H) at day 10 post-seeding in dimpled versus smooth microparticle cultures. To confirm the role of canonical HH signalling in topography-induced osteogenesis, the expression of bone gamma-carboxyglutamate protein (BGLAP), a late marker of osteogenesis, was examined after treatment with KAAD-cyclopamine. The expression BGLAP, encoding osteocalcin, was significantly reduced by 2.5 -fold ( p < 0.0001 ; Figure 5F) in dimpled cultures treated with KAAD-cyclopamine compared to untreated dimpled samples and relative to the 2D vehicle-only controls, confirming the .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 25 dependence of its expression on canonical HH signalling. Dimpled topographical features induce temporal regulation of IGF-II expression To elucidate downstream mechanisms of HH signalling and cross -talk governing the trajectory towards osteogenic commitment, QIAGEN’s Ingenuity ® Pathway Analysis (IPA) of DEGs in hMSCs cultured on different microparticle designs was performed to provide further insight into the dynamic molecular signatures associated with lineage commitment. IPA was performed on the set of 256 DEGs at day 14 post -seeding to identify the top canonical pathways predicted to be activated, offering insights into the regulatory networks orchestrating this osteogenic progression (Figure 6). .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 26 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 27 Figure 6: Dimpled microparticles promote expression of Insulin -like Growth Factor II (IGF-II) at day 14 post-seeding. A) Network analysis of the upstream regulatory network of insulin-like growth factor II (IGF-II) generated using IPA based on overlaid DEGs in hMSCs on dimpled versus smooth microparticles at day 14 post -seeding, highlighting transcriptional regulators, growth factors, and transmembrane receptors associated with IGF -II signalling. Nodes are colour-coded to represent expression levels: upregulated (red) and downregulated (green). Edges represent predicted relationships: activation (orange), findings inconsistent with expected relationships (yellow), and undefined effects (gray). This visualisation (created on BioRender.com) focuses specifically on u pstream regulators of IGF -II, with full post- trimming analysis provided in supplementary data . B) Representative multiplexed fluorescence Western blot performed at day 14 post-seeding hMSCs on smooth and dimpled microparticles against 2D controls, showing individual channels (i), and the overlaid image (GAPDH and IGF-I in blue, IGF-II in red). (C) Schematic depicting the proposed mechanism of topographically-guided osteoinduction by 3D dimpled topographical features in hMSCs. Abbreviations: IGFBP, insulin-like growth factor binding protein; FN1, fibronectin 1; PTCH1; Patched 1; GLI1, Glioma-associated oncogene homolog; ITGA5 , Integrin subunit alpha 5; GAPDH, Glyceraldehyde-3-phosphate dehydrogenase. Canonical pathways were assessed using the activation z-score- (z-score > 2), which correlates observed gene expression with the expected direction of expression for DEGs [44]. IGF transport and uptake by IGFBPs was identified as the most significantly activated canonical pathway (z -score = 3.21, padj = 1.08 × 10 −11, Table S5). Notably, IGF-II showed no differential expression at day 3 post -seeding (Figure 5B and Figure S3). Interaction network analysis identified key regulatory molecules that may be responsible for the gene expression changes observed [45]. A direct activation link between GLI1 and IGF-II expression was observed, while PTCH1 indirectly inhibits IGF-II expression (Figure 6A). Despite PTCH1's predicted negative influence, IGF-II remained activated, as indicated by the yellow node. Additionally, an indirect bidirectional interaction was identified between IGF-II and ITGA5 (Figure 6A). Experimental validation confirmed this striking dichotomy in IGF expression, where dimpled surfaces specifically induced IGF -II expression, while smooth surfaces promoted IGF-I expression (Figure 6B). The simultaneous expression of IGF -II and .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 28 IGF-I proteins was confirmed using multiplex fluorescence -based Western blot analysis at day 14. This approach allowed concurrent detection of IGF -II and IGF -I, which possess a similar molecular weight, along with GAPDH as loading control, in the same sample [46]. This revealed the mature IGF proteins at their expected molecular weights, approximately 7.5 kDa [47] (Figure 6 B). This topography - dependent regulation of IGF signalling suggests a novel mechanism by which 3D surface features direct cell fate. Dim pled topographies orchestrate the temporal progression of lineage commitment, initially establishing a bipotential osteochondral state driven by the activation of HH signalling, and transitioning toward osteogenesis via the IGF-II pathway (Figure 6C). Precision-engineered GLI1 expression gradients by two-photon lithography for spatially controlled stem cell signalling Building on this mechanistic understanding of topography -induced signalling cascades, we fabricated precisely engineered micron -scale platforms. This proof-of- concept tested whether spatially-arranged microtopographies could generate defined zones of topographically-guided cellular signalling. This approach enables locali sed signalling control without genetic modification, serving as a novel physical analogue to mechanogenetics for controlling cell behaviour through material design. 2PP direct laser writing offers precision in fabricating intricate 3D microstructures while ensuring topographical consistency and uniformity [48]. Harnessing the osteoinductive capability of 3D dimpled topographical features, combined with the exceptional precision of 2PP, we engineered a GLI1 expression gradient by strategically employing cellular responses to topographical cues ( Figure 7 ). GLI1 serves as a downstream transcriptional effector that reflects actual signal transduction and how cells translate topographical signals into differentiation responses [49]. .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 29 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 30 Figure 7: Engineering topography-induced GLI1 expression gradients using high- resolution two-photon polymerisation lithography. A) Schematic of the design strategy for 2PP fabrication. Hemispherical features (27.5 µm height and 55 µm width) served as building blocks that were systematically arranged in x and y directions to form structured arrays. B, C) Grid workflow language used to generate single-topography array (B) and dual -topography arrays combining 2 µ m and 7 µ m dimples (C). D) Computer -aided design of microparticles with different topographical features and corresponding SEM images after fabrication (Scale bar: 50 μm). E, F) Representative fluorescence images displaying GLI1 expression (red) and nuclei (DAPI; blue) in primary hMSCs seeded on single-topography arrays after 7 days in culture (E; scale bars: 20 µm; N= 2 arrays) , and on dual-topography arrays featuring three rows of 7 µm dimpled hemispheres and twelve rows of 2 µm dimpled hemispheres (F, scale bar: 100 µm ; N= 2 arrays ). hMSCs cultured on glass slides served as 2D controls. DAPI channel is presented as maximum intensity projection images, and GLI1 as sum slice projection images. A single slice is used to represent brightfield for clarity. A sliding paraboloid

subtraction was applied to the red channel in dual -topography array s, and standard background subtraction was applied for corresponding 2D controls. G) Schematic of the source-receiver configuration: MSCs seeded on 7 µm dimples act as ‘source cells’, while cells attaching to adjacent 2 µm dimpled rows are designated as ‘responding cells’ [Created with BioRender.com]. H) Representative confocal z-projection image (sum slices projection) displaying graded GLI1 expression, demonstrating the spatial correlation between topographical feature sizes and GLI1 activation (Scale bar: 100 µm). Abbreviations: GLI1, Glioma-associated oncogene homolog 1; DAPI, 4′,6 Diamidino-2- phenylindole Hemispherical microstructures (27.5 µm height) were fabricated with precisely controlled dimple sizes of either 2 µm or 7 µm (Figure 7A) . These structures were imported into the DeScribe software (v2.7, Nanoscribe GmbH, Germany) , and systematically arranged along the x and y axes to create two array configurations: x= 220 µm and y= 220 µm to achieve the single -topography array (Figure 7B), and x= 440 µm and y= 825 µm to engineer dual-topography arrays (Figure 7C). This approach minimised slicing-induced lines during 2PP fabrication, preventing a rtifacts. IP-Visio, a methacrylate-based commercial resin, was selected for its biocompatibility and lower autofluorescence compared to other photoresins [50]. Additionally, its reported stiffness (1.8 ± 0.64 GPa) [51] closely aligns with that of PLA microparticles [15], making it the optimal photoresin for this study. Precision of designs was validated by SEM imaging, demonstrating reproducible topographical features (Figure 7D). Initial .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 31 studies using single -topography arrays revealed that 7 µm dimples resulted in a significant upregulation of GLI1 expression in hMSCs as previously observed with dimpled PLA microparticles, whereas 2 µm dimple sizes showed minimal effect (Figure 7E). Inspired by this differential response, dual -topography arrays were then engineered. These arrays featured hemispherical structures with two dimple sizes: The top three rows featured 7 µm dimples, optimised to stimulate hMSCs as ‘source’ cells expected to induce GLI1 expression. In contrast, subsequent rows consisted of 2 µm dimple d structures, where hMSCs acted as ‘responding’ cells (Figure 7G). The platform was designed around the principle that physical microenvironmental features can orchestrate spatially regulated signalling responses , analogous to how patterned matrix cues guide tissue development [52]. Inspired by principles of localised activation and spatial propagation seen in developmental systems, we hypothesised that precisely engineered surface topographies could generate spatial heterogeneity in mechanotransduction responses. To test this, GLI1 expression in primary hMSCs seeded on this dual-topography array was evaluated at day 7 post-seeding. A visible spatial gradient of GLI1 expression emerged in the absence of any exogenous biochemical a gonists (Figure 7 F, H). Highest GLI1 expression was visibly highest within ~290 µm of the ‘source’, gradually declining on more distant regions. This pattern persisted despite minor lateral displacements within the arrays during pre - seeding preparation, indicating robust spatial control. These findings demonstrate that engineered topographical features alone can induce spatially resolved intracellular signalling in stem cells, generating defined zones of mechanically-induced signalling. This provides a powerful platform for exploring spatial aspects of mechanotransduction and offers a biochemical-free approach for engineering zonally- .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 32 patterned cell ular responses, which is potentially transformative for developmental biology studies and regenerative material design.

Polymeric microparticles with engineered surface topographies offer powerful bottom- up engineering tools for directing stem cell fate through mechanical rather than biochemical cues [15, 20]. This study demonstrates that our topographically- engineered dimpled microparticles induce osteogenic differentiation in hMSCs by mechanical stimul i alone . Canonical HH signalling mediates this response, establishing a direct mechanotransductive mechanism . This materials -based approach offers a scalable, additive-free platform for developing cell-instructive 3D in vitro culture systems, with strong translational potential for regenerative medicine and stem cell biomanufacturing. Topographical cues modulate mechanotransduction by altering cell morphology through cytoskeletal reorganisation [15, 24]. The elongated cell morphology observed on dimpled microparticles suggested increased cytoskeletal tension [53], in contrast to the isotropic spreading seen on smooth microparticles [54]. Elongated cell morphology is associated with altered focal adhesions, enhancing osteogenic potential in hMSCs [55]. Geometrical cues have been demonstrated to promote MSCs differentiation independent of soluble factors, with cytoskeletal -disrupting agents modulating these shape -based trends [56]. The increased expression of the mechanosensitive ITGB4 at the earlier timepoint [57] is known to reduce VCL localisation within focal adhesions [58]. This is consistent with the significant downregulation of VCL, PXN, and TLN1 and diffuse VCL distribution observed at day 3. Furthermore, the downregulation of LIM domain kinases, including LIMA1 and ZYX, has been associated with actin depolymerisation by regulating actin filament turnover .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 33 [59, 60]. The mechanosensitive ion channel PIEZO2, which is critical for proper SMO activation [61], was also significantly upregulated on dimpled topographies. Since substrate stiffness alone does not activate PIEZO2 [62], this suggests a topography- specific mechanical activation pathway. PIEZO2 activity supports calcium influx critical for SMO activation [63], potentially linking external mechanical cues to intracellular HH pathway responses. In our system, dimpled topographies coincided with disrupted actin organisation and diffuse VCL distribution, suggesting a mechanotransductive mechanism involving actin depolymerisation [17]. This is consistent with mechanisms reported during oestrogen withdrawal in murine osteocyte -like cells, where similar cytoskeletal changes were associated with HH activation [24]. Dimpled microparticles activated the canonical HH signalling pathway as early as day 3 post-seeding, with activation sustained through day 14 . SMO-dependence of HH signalling activation in dimpled microparticle cultures was confirmed by the significantly reduced GLI1 expression in dimpled samples treated with KAAD - cyclopamine, and the similar transcriptional profile of cyclopamine -treated dimpled microparticle cultures to smooth microparticles. This aligns with our previous findings on murine embryonic mesenchymal progenitors [17]. While HH pathway components showed distinct regulation at day 3, gene expression profiles converged by day 14, suggesting that topography -driven signalling acts within a narrow temporal window . The significant downregulation of BGLAP in hMSCs cultured on dimpled microparticles and treated with KAAD-cyclopamine relative to untreated dimpled cultures confirmed the central role of HH signalling in topography-induced osteogenesis. Significant upregulation of HHIP in dimpled cultures at day 14 suggests feedback attenuation of HH activity to maintain signalling homeostasis and prevent aberrant pathway activation [64]. As a pro -osteogenic gene, HHIP plays a pivotal role in .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 34 regulating osteogenic mesenchyme in the coronal suture of mice and has been implicated in human embryonic skeletal development [65]. The lack of SMO differential expression at this stage is consistent with its known post-translational regulation, and could be attributed to the activation of IGF -I (Figure 6) , which has been positively correlated with SMO expression in a GLI1 -independent manner [66]. Moreover, sustained SMO expression results in impaired postnatal bone formation in mice [67]. This suggest s that SMO activation is tightly regulated and transient during osteogenesis, in line with our findings. The activation of the canonical HH pathway mirrors processes observed during skeletal development, where GLI1-expressing cells exhibit osteochondrogenic potential by contributing to endochondral and intramembranous ossification through the induction of SOX9 and RUNX2 expression, respectively [68]. It has been reported that the interaction of BMPs with the HH pathway shifts the balance towards RUNX2 expression, driving MSCs commitment to osteoblasts [69]. BMP2 has also been suggested as a direct target of GLI1 [70], with GLI1 serving as a critical mediator between HH and downstream BMP pathway [71]. Early GLI1 expression induced both SOX9 and RUNX2, reflecting a transient bipotential osteochondroprogenitor state. While this early bipotential signature might suggest endochondral ossification , the absence of cartilage -specific matrix proteins , such as ACAN, despite transient COL2A1 at day 3, argues against progression through a full endochondral ossification route. The significant upregulation of COL10A1 at day 14 , a marker of hypertrophic chondrocytes [43], likely reflects residual early chondrogenic activity rather than full hypertrophic transition. Further investigation is needed to clarify the role of COL10A1 in this context. This initial bipotentiality is critical for the development of certain intramembranous bones via secondary cartilage [72]. Significant upregulation of LEPR .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 35 on dimpled microparticles was observed at day 14, which aligns with reported in vivo findings that skeletal stem cells exhibit osteoblast-chondrocyte transitional identities in young bone marrow, progressively being replaced by LEPR-expressing stromal cells at later stages to serve as a source of osteoblasts in adult marrow [73]. RUNX2 remained expressed across both microparticle designs at both time points, likely induced by matrix stiffness of the PLA microparticles [74]. This was accompanied by upregulation of osteogenic co -regulators MSX2, a transcription factor that synergises with RUNX2 and drives SP7 expression [75], and PAX9, a transcription factor required for craniofacial and tooth development and associated with early skeletogenesis, regulates key genes for bone formation such as ALPL and COL1A1 [76]. These transcriptional regulators support osteoprogenitor proliferation, suppress alternative lineages, and promote expression of key drivers of osteoblast maturation. MSCs differentiation to osteoblasts progresse d with the subsequent expression of SP7, a downstream effector of RUNX2 and master regulator of osteoblast differentiation [77], accompanied by increased expression of COL1A1 and OCN in hMSCs cultured on dimpled relative to smooth microparticles. By day 14 post-seeding, hMSCs demonstrated a clear trajectory towards osteogenesis. This was evidenced by the upregulation of POSTN, which is regarded as a marker of intramembranous ossification preceding increase in other osteogenic genes [78]. The expression of TNC, an ECM glycoprotein involved in osteogenesis and mineralisation [79], was also increased and is known to be induced by mechanical stimuli in murine pre-osteoblastic cell [80]. SERPINE2, essential for the early osteogenic commitment of MSCs [38], were also dramatically upregulated. Moreover, GALNT1 is critical for the expression of SPP1 and IBSP in osteoblasts [81]. In addition, the transcriptional profile of hMSCs on dimpled microparticles revealed an ECM niche associated with early osteogenesis. .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 36 Collagen VI is a key pericellular matrix component found in MSCs, pericytes, and osteoprogenitors, supporting adhesion, survival, and mechanosensitivity [40]. COL8A1, more typically associated with vascular endothelium, was also upregulated. Its role in matrix remodelling and angiogenesis suggests a vascularised osteoprogenitor phenotype [82]. Gene ontology analysis confirmed enrichment in skeletal and vascular development pathways, reflecting the coupling of angiogenesis and intramembranous bone formation , where mesenchymal condensation centres facilitate blood vessel formation, enabling nearby mesenchymal cells to differentiate into osteoblasts [83]. The dominance of osteoblast-specific markers, validated on both genetic and protein levels, highlights the intramembranous ossification trajectory of hMSCs on dimpled microparticles. These results underscore the potential to integrate bone formation and vascular growth, paving the way for advanced tissue -engineered constructs. While HH signalling is pivotal in driving lineage commitment, it is insufficient as a solitary driver to fully orchestrate complete intramembranous ossification [84, 85]. Our findings also implicate IGF-II as an effector in later-stage osteogenic differentiation. Aberrant IGF-II signalling in HH -responsive cells severely impairs bone formation in mouse embryo s, highlighting the necessity of sustained HH expression to activate IGF-II to complete the differentiation process [86], aligning with our findings . The activation of IGF-II by GLI1 extends to ECM regulation, with a bidirectional interaction between IGF-II and ITGA5 promoting FN1 expression, forming a positive feedback loop that enhances ECM organisation and induces RUNX2 expression in hMSCs [87]. Furthermore, IGF-II has been reported to promote osteoblast maturation up to bone mineralisation [88]. We previously demonstrated that topographically -textured microparticles can be used to induce bone regeneration in vivo , confirming their .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 37 osteogenic commitment [15]. To translate these findings, 2PP was used to fabricate GLI1 expression gradients with subcellular resolution [89]. While previous models relied on chemical inducers and genetic modification [90, 91], our platform offers a material-independent, non-genetic alternative with improved scalability and physiological relevance . Li et al. engineered 'sender' cells to express SHH using a 4 -hydroxytamoxifen-inducible system and created open-loop receptor cells by modifying PTCH1 alleles with a doxycycline (Dox)- inducible promoter [90]. While this 2D system offered valuable insights into HH spatial patterning, its scalability and reproducibility potentially suffered from its high cost [92], and the possibility that Dox may exert off-target effects [93]. Johnson et al. developed an SHH morphogen gradient by overlaying a genetically modified epit helial layer producing SHH onto mesenchymal tissues embedded in a hyaluronic acid -collagen gel [91]. However, the model's reliance on genetic modulation may limit physiological relevance and scalability. By demonstrating spatial control of GLI1 expression via engineered topographical gradients, evidence is provided herein that mechanical cues can orchestrate cell signalling with precision previously thought possible only through biochemical means, independent of the material used for fabrication. This allows for versatile fabrication using either solvent evaporation oil -in-water emulsion (for scalable, high -throughput microparticle production) or 2PP (for precise, customisable spatial patterning of localised cell behaviour and tissue gradients) . It has been demonstrated that GLI1 activity reflects SHH responses in a proportional manner, thereby reflecting subtle changes in SHH concentration without amplification [94]. Our approach enables the modulation of cellular responses by engineered topographical cues in isolation of external biochemical additives. The observed GLI1 gradient spanned approximately .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 38 290 μm, aligning with the reported range of SHH morphogen activity of 200 µm [90] and closely parallels the documented SHH signalling range of 300 μm in the limb bud [95]. This spatial control demonstrates that topographical features can function as a precise and reproducible platform for in vitro modelling of spatially patterned stem cell behaviour, with broad relevance to tissue engineering, developmental biology, and drug screening. Concluding remarks We present a versatile, modular microparticle-based platform that enables precise, additive-free control of osteogenic differentiation in hMSCs through engineered 3D topographies. This topography -driven induction , achieved without biochemical additives or genetic manipulation , represents a significant step towards streamlining cell-instructive platform design while avoiding regulatory complexities associated with exogenous growth factors. Transcriptomic profiling underscores the biological potency of topographical programming, and the use of high-resolution 2PP lithography enabled precise enginee ring of bioinstructive microtopographical cues, supporting reproducible, spatially controlled cell responses . By decoupling topograph y from soluble signalling, this platform provides a scalable , physiologically relevant system for probing mechanotransduction and differentiation. Beyond osteogenesis, this approach holds promise for broader applications in modelling other developmental or disease contexts where spatial organisation and mechanical inputs play instructive roles, such as vasculogenesis and organoid patterning. The platform is modular and adaptable to different culture formats , which are key characteristics required for scalable and reproducible deployment in translational settings. Collectively, this lays the groundwork for next -generation bioinstructive systems in regenerative medicine, developmental biology, and high-content screening. .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 39

Experimental model and subject details This study used primary human bone marrow -derived mesenchymal stromal cells obtained from five independent donors representing diverse demographic backgrounds. Three donor lots (designated as donors 1, 2 and 3) were obtained from RoosterBio (RoosterVial ™-hBM-1M, MSC -003, RoosterBio Inc., USA), while two additional donor lots (designated as donors 4 and 5) were obtained from Lonza (PT - 2501, Lonza, Germany). An overview of donor characteristics is provided in Table S1.

Fabrication of smooth and dimpled microparticles Poly(D,L-lactic acid) (PLA) microparticles (Ashland Viatel DL 09 E, Mn 56.5 kDa, Mw 111 kDa, IV 0.8 –1.0 dL/g) were prepared using a solvent evaporation oil -in-water emulsion technique [15, 17]. For fabricating smooth microparticles, a 20% ( w/v) solution of PLA in dichloromethane (DCM; ≥99.8%, Thermo Fisher Scientific, USA) was homogenised (Silverson Machines Ltd., UK) at 3800 rpm for 5 min. The homogenised organic phase was then emulsified into 100 mL of an aqueous continuous phase containing 1 % (w/v) poly(vinyl acetate-co-alcohol) (PVA; MW 13 – 23 kDa, #348406, Sigma -Aldrich). The resulting emulsion was stirred at room temperature to facilitate solvent evaporation. Microparticles were collected by centrifugation and washed with deionised water to remove resid ual PVA. After washing, microparticles were sieved using strainers (40–70 µm) (Greiner bio-one) and then freeze-dried for storage. For the fabrication of dimpled microparticles, the addition of fusidic acid (FA; 98%, #5552333, Thermo Fisher Scientific, USA) into the organic phase was used to create the topographical patterns. A 30% ( w/v) ratio of FA to PLA was used, resulting in a .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 40 total FA/PLA concentration of 10% ( w/v) in DCM. FA -loaded microparticles were incubated in phosphate -buffered saline (PBS , Gibco) at 37 °C for 7 days for FA release, as previously detailed [15]. Microparticle size analysis The particle size distribution of the fabricated microparticles was measured using a laser diffraction particle size analyser (Mastersizer 3000, Malvern Instruments Ltd, UK). Particle size distributions were calculated automatically using the optical Mie model [96] within the Mastersizer 3000 software (v3.71). Each analysis was performed at least three times. The polydispersity index (PDI) of size distribution was determined by dividing the square of the standard deviation by the square of the mean diameter. Brunauer-Emmett-Teller (BET) surface area measurements The surface area of fabricated microparticles was determined as previously described [15]. Krypton (Kr) sorption isotherms were conducted using a Micromeritics ASAP 2420 (Micromeritics, USA) at −196 °C. Approximately 500 mg of each sample was degassed under high vacuum (1.77 Torr) [97], sorption isotherms were measured over a relative pressure range of 0.10 to 0.65. The specific surface area was calculated from 0.05 to 0.30 relative pressure range using the BET model. Micropore volume was calculated at 2 nm, and limited mesopore volume from 2-20 nm using pore size vs cumulative pore volume using the Derjaguin– Broekhoff–de Boer model [98]. Atomic force microscopy (AFM) topographical analysis Atomic force microscopy (AFM) images were acquired on a Bioscope Catalyst AFM (Bruker) mounted on an Eclipse Ti -U (Nikon) inverted optical microscope, with a .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 41 Nanoscope IV controller, Nanoscope v9.1 software (Bruker), and a ScanAsyst ™-Air (Bruker) AFM probe. These silicon nitride probes with an aluminium coating have a nominal spring constant of 0.4 N m -1 (Bruker AFM Probes, USA). Imaging was performed using ScanAsyst Peak-Force Tapping mode in air with a scan size of 7 μm at a scan rate of 0.988 Hz. The instrument is periodically calibrated using a grating with 180 nm deep and 10 mm² depressions. Images were processed by applying flattening of first-order using Nanoscope Analysis software (v3.0). Primary human mesenchymal stromal cell culture Primary human bone marrow -derived mesenchymal stromal cells obtained from five independent donors representing diverse demographic backgrounds were used. Three donor lots (designated as donors 1, 2 and 3) were obtained from RoosterBio (RoosterVial™-hBM-1M, MSC-003, RoosterBio Inc., USA), while two additional donor lots (designated as donors 4 and 5) were obtained from Lonza (PT -2501, Lonza, Germany). An overview of the donor characteristics is provided in Table S1. Cells were cultured in Dulbecco's modified Eagle's medium (#21969-035, Gibco) supplemented with 1% (w/v) l-glutamine (Gibco), 1% (w/v) penicillin-streptomycin (Gibco) and either 10% (v/v) fetal bovine serum (FBS; Gibco) for routine passaging or 2% FBS (referred to as serum-reduced medium). Each donor batch was maintained as an independent stock and cells were used between passages three and six. Tri-lineage differentiation potential hMSCs was confirmed using StemPro™ differentiation kits (Gibco, UK). Microparticles preparation for cell seeding Microparticles were placed in CELLSTAR® cell-repellent surface 96-well plates (Greiner Bio-One) and sterilised with UV light at 254 nm for 30 min at 4 × 104 mJ. Mass of smooth and dimpled microparticles were calculated to achieve a consistent surface .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 42 area for cell attachment. Following sterilisation, the microparticles were conditioned in serum-reduced medium (2% FBS) for 1 h. Cells were seeded onto microparticles at a density of 1 × 104 cells/cm² and placed on a plate shaker for 15 min to ensure even distribution. For 2D controls, cells were seeded in tissue culture-treated 96-well plates (CytoOne®, Starlab). Cell viability and proliferation Cell viability was assessed three days post -seeding using the Viability/Cytotoxicity Assay Kit for Animal Live & Dead Cells (30002 -T, Biotium, UK) according to the manufacturer's instructions. Briefly, 1 μM calcein -acetoxymethyl (calcein-AM) and 4 μM ethidium homodimer III (EthD-III) were added to each well. Imaging was performed using a ZEISS Cell discoverer 7 imaging system (ZEISS, Germany). Proliferation was assessed by measuring DNA concentration from cell lysates using the Quant -iT™ PicoGreen® dsDNA Assay Kit (P7589, Invitrogen, USA), following manufacturer’s instructions. Cells were lysed with CelLytic ™ M lysis buffer (C2978, Sigma-Aldrich) with two additional freeze -thaw cycles. Fluorescence was measured at λexc/λem 480/520 nm using a Varioskan™ LUX multimode microplate reader (Thermo Fisher Scientific, USA). DNA content was measured by comparing to a standard curve generated from the supplied standards. Scanning electron microscopy (SEM) Microparticles were directly mounted onto double-sided copper tape and placed on an aluminium pin stub. For imaging cell -microparticle aggregates, cells were fixed after three days of culture using 2.5% ( v/v) glutaraldehyde (G6257, Sigma -Aldrich). Fixed samples were then dehydrated with a graded ethanol series (10, 25, 50, 80, and 100%; Fisher Chemical). The dehydrated cell aggregates were mounted onto double -sided .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 43 copper tape (Agar Scientific) and placed on an aluminium pin stub (AGG301, Agar Scientific). Prior to imaging, samples were sputter -coated with an 80% gold/20% palladium alloy using a Q150 S/E/ES Plus sputter coater (Quorum Technologies, UK) under vacuum at 40 mA for 4 min. Scanning electron microscopy (SEM) images were acquired using a Tescan Vega 3 (Tescan, UK) at 5 and 10 kV, as detailed in the figure captions. Dimple sizes were characterised using ImageJ software (v1.53q) by measuring the diameters of a minimum of 250 dimpled microparticles, across ten independent SEM images from five fabricated batches. Immunocytochemistry Cells were fixed with 3.7% (v/v) formaldehyde (Thermo Fisher Scientific, USA) in PBS and permeabilised with 0.1% ( v/v) Triton X-100 (A16046, Fisher Chemicals) in PBS. To block non -specific binding, samples were incubated for 1 h in 1% ( w/v) bovine serum albumin (BSA; SLCK4263, Sigma -Aldrich) in PBS, supplemented with 10% (v/v) normal goat serum (G9023, Sigma -Aldrich) or normal donkey serum (#D9663, Sigma-Aldrich), based on the secondary antibody. Following blocking, cells were incubated with the primary antibody overnight at 4 °C, then the secondary antibody for 2 h. For F-actin staining, cells were stained with ActinGreen ™ 488 ReadyProbes™ Reagent (R37110, Invitrogen) and nuclei were counter-stained with NucBlue™ Fixed Cell ReadyProbes ™ Reagent (R37606, Invitrogen). Cells were observed using a Zeiss LSM 880 inverted AiryScan confocal microscope (Carl Zeiss, Germany). The following primary antibodies were used, goat anti-GLI-1 affinity purified polyclonal antibody (1:100; AF3455, R&D systems, RRID: AB_2247710), rabbit anti-osteocalcin (OCN) polyclonal antibody (1:75; AB10911, Millipore, RRID: AB_1587337), goat anti- type I collagen polyclonal antibody (1:200; #1310 -01, SouthernBiotech, RRID: AB_2753206) and monoclonal anti -Vinculin (1:70; V4505, Sigma -Aldrich, RRID: .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 44 AB_477617). All secondary antibodies were obtained from Invitrogen and used at a 1:500 dilution, including donkey anti -goat IgG (H+L) cross -adsorbed, Alexa Fluor ™ 594 (A-11058, RRID: AB_2534105), goat anti-rabbit IgG (H+L) cross-adsorbed, Alexa Fluor™ 488 (A-11008, RRID: AB_143165), donkey anti -goat IgG (H+L) highly cross- adsorbed, Alexa Fluor ™ Plus 488 (A32814, RRID: AB_2762838) , and goat anti - mouse IgG (H + L) cross -adsorbed, Alexa Fluor ™ 647 (A -21235, RRID: AB_2535804). Nuclei were counter -stained with NucBlue™ Fixed Cell ReadyProbes™ (R37606, Invitrogen). Gene expression analysis using real-time PCR Total RNA samples were extracted using the RNAqueous ™-Micro Kit (AM1931, Thermo Fisher Scientific) with minor modifications to prevent dissolution of microparticles in ethanol. The concentration of RNA in each sample was measured using a NanoDrop ™ 1000 Spectrophotometer (Thermo Fisher Scientific). Reverse transcription of RNA was achieved using the iScript ™ Select cDNA Synthesis Kit (#1708896, Bio-Rad, USA) following the manufacturer's protocol. A GS00482 Thermal Cycler (G-STORM, UK) was used following the reaction conditions listed in Table S6. Transcript levels were determined using SsoFast™ EvaGreen® Supermix (#1725201, Bio-Rad), quantified using CFX384 Touch Real -Time PCR Detection System (Bio - Rad) and following the reaction conditions listed in Table S6. Primer sequences are provided in Table 2, and were used at a concentration of 500 nM with an annealing temperature of 64 °C. No template controls (NTC) for each primer set and no reverse- transcriptase controls (NRT) for each sample were included. Table 2:Primer sequences used for quantitative real-time PCR analysis Target Gene GenBank Accession No. Primer Direction Primer Sequence (5’–3’) Amplicon Length (bp) TBP NM_001172085.2 Forward GGCCGCCGGCTGTTTAACTT 130 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 45 Reverse GGCTGTGGGGTCAGTCCAGT YWHAZ NM_003406.4 Forward CGCTGGTGATGACAAGAAAGGGAT 116 Reverse GGGCCAGACCCAGTCTGATAGG GLI1 NM_001160045.2 Forward TAAGCCCGGCACCCCTTCTC 93 Reverse CTCGCCCCTCACCTCCCTTC SMO NM_005631.5 Forward CGCTACAACGTGTGCCTGGG 124 Reverse CATTCCGGAGGCCCGACCA PTCH1 NM_000264.5 Forward CCCGCTGCACACACACAGAG 100 Reverse CTTGTGCTCCTCGGCAACCC RUNX2 NM_001024630.4 Forward AACCACAAGTGCGGTGCAAACT 90 Reverse GGCTGGTAGTGACCTGCGG SPP1 NM_001040058.2 Forward AGCAGCAGGAGGAGGCAGAG 90 Reverse TTCCTTGGTCGGCGTTTGGC IBSP NM_001251829.2 Forward GGGCAAGGGCACCTCGAAGA 123 Reverse CATTGGCGCCCGTGTATTCGT SP7 NM_001173467.3 Forward TGGCGTCCTCCCTGCTTGAG 110 Reverse TGTTGAGTCCCGCAGAGGGC BGLAP NM_199173.6 Forward GAGAGCCCTCACACTCCTCGC 138 Reverse TTCACTACCTCGCTGCCCTCC Abbreviations: TBP, TATA-box binding protein; YWHAZ, tyrosine 3 -monooxygenase/tryptophan 5-monooxygenase activation protein zeta; GLI1, glioma -associated oncogene homolog 1; PTCH1, patched1; SMO, smoothened; RUNX2, runt-related transcription factor2; SPP1, secreted phosphoprotein-1; IBSP, integrin -binding sialoprotein; SP7, specificity protein -7, BGLAP, bone gamma-carboxyglutamate protein. Purmorphamine ( #130-104-465, StemMACS ™ Purmorphamine, Miltenyi Biotec) treatment was used as a positive 2D control for GLI1, SMO, PTCH1 and RUNX2 expression, dissolved in dimethyl sulfoxide (DMSO; 5 mM), and used as 2 μM in serum -reduced medium (2% FBS). Corresponding 2D vehicle-only controls were prepared as 0.06% ( v/v) DMSO in serum -reduced medium (2% FBS). Cells in 2D culture were treated with osteoinductive media containing dexamethasone (A1007201, StemPro, Gibco, UK) as positive 2D controls for osteoblastic markers , except for RUNX2, as dexamethasone has been rep orted to promote osteogenesis indirectly by inhibiting chondrogenesis rather than directly inducing RUNX2 expression [2]. Corresponding negative 2D controls were cultured in serum -reduced medi um. Relative expression levels of each gene of interest were calculated using the 2 −ΔΔCt .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 46

[99], normalised to the geometric mean of two housekeeping genes, TATA- box binding protein ( TBP) and tyrosine 3 -monooxygenase/tryptophan 5 - monooxygenase activation protein zeta ( YWHAZ), known for their stable expression in hMSCs during osteogenic differentiation [100, 101]. Results were further normalised to corresponding untreated 2D controls. Treatment with KAAD-cyclopamine HMSCs were cultured in serum -reduced medium for 24 h, then treated with 300 nM KAAD-cyclopamine (ab142146, Abcam), an SMO antagonist. Corresponding 2D and 3D vehicle -only controls, were prepared as 0.06% ( v/v) DMSO in serum -reduced medium , shown to maintain the surface topographical features intact [17], for 3 and 14 days, with the treatment media refreshed every 2 days. Cells treated with KAAD - cyclopamine will be referred to as KAAD-cyclopamine -treated. RNA-Seq and library preparation hMSCs (N= 3) were cultured on smooth and dimpled microparticles as previously outlined. Total RNA was extracted on days 3 and 14 post-seeding, as detailed above. Total RNA quality was assessed using the Agilent 4200 TapeStation system with RNA ScreenTape assays (G2991BA, Agilent Technologies Inc), and the RNA Integrity Number (RIN) was determined using TapeStation Analysis Software (v5.1; Agilent Technologies). Unmapped paired -end sequences from NovaSeq 6000 sequencer were tested by FastQC ( Available from: http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Sequence adapters were removed, and reads were quality trimmed using Trimmomatic_0.39 [102]. Reads were mapped against the reference human genome (hg38) and counts per gene were calculated using annotation from GENCODE 44 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 47 (http://www.gencodegenes.org/) using STAR_2.7.7a [103]. RNA samples from donor 2 (MSCs seeded on untreated dimpled microparticles at day 3) were excluded because the read counts assigned to genes were below our threshold of 1 × 10⁷ [104] (Figure S 4). Normalised read counts were averaged for each gene across the conditions of interest; genes with a mean normalised read count < 10 were excluded from further analysis. Normalisation, Principal Components Analysis (PCA), and differentially expressed genes (DEGs) was calculated with DESeq2_1.40.2 [105]. Adjusted p-values (padj) were corrected for multiple testing (Benjamini and Hochberg method). Heatmaps were drawn with ComplexHeatmap v2.16.0 [106]. Hierarchical clustering was performed on the means of each sample group (log2) that had been z- transformed (for each gene the mean set to zero, standard deviation to 1). A threshold of log2 fold change > 1 (upregulated) or < −1 (downregulated) and an adjusted p-value (padj) < 0.05 was applied to identify significantly differentially expressed genes. Gene ontology enrichment was studi ed using clusterProfiler v4.8.3 [107] and Enrichr v3.2 [108]. Gene enrichment was studied using ReactomePA 1.44.0 [109]. For Ingenuity Pathway Analysis (IPA), a more stringent filter was applied, excluding genes with a mean normalised read count < 50 to enhance confidence in pathway predictions. Canonical pathways and upstream regulators at day 14 post-seeding were identified/predicted using IPA software (QIAGEN). DEGs with a log 2 fold change > 2 or < −2 and padj 1.3, which corresponds to padj < 0.05, calculated using Fisher's Exact Test to determine statistical significance of enrichment [110]. The activation z -score was employed to predict th e activation or inhibition of pathways by comparing the observed gene expression patterns against curated information in the IPA Knowledge Base. Pathways were considered significantly .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 48 impacted based on z -score threshold of > 2, indicating predicted activation, or < -2, indicating predicted inhibition. To explore IGF-II regulatory mechanisms, the Interaction Network tool in IPA was applied to DEGs identified in dimpled versus smooth cultures. Networks were trimmed using IPA’s built -in filtering functions to exclude low -confidence nodes and genes that did not meet differential expression thresholds or were absent from the input dataset. The IGF -II-centred interaction network was graphically simplified using BioRender to enhance visual clarity. The IPA- generated network is shown in Figure S3, with simplified visualisation in Figure 6A. Multiplex fluorescent western blotting After 14 days in culture, cells were lysed using CelLytic ™ M lysis buffer (C2978, Sigma-Aldrich) containing 1 mM phenylmethyl sulfonyl fluoride (PMSF; #36978, Thermo Fisher Scientific) and 1 mM ethylenediaminetetraacetic acid (EDTA; #46-034- CI, Corning) at pH 8.0, and 1 μL protease inhibitor cocktail (P8340, Sigma -Aldrich). Following incubation on ice for 20 min with continuous shaking, cell lysates were centrifuged for 15 min at 13,000  × g. Total protein concentrations were determined using the Pierce ™ Bicinchoninic acid (BCA) Protein Assay Kit (#23227, Thermo Fisher Scientific). Equal amounts of protein were prepared in NuPAGE (4X) lithium dodecyl sulphate sample buffer (NP0007, Invitrogen) with 2 -mercaptoethanol as a reducing agent (M3148, Sigma -Aldrich) and denatured at 95 °C for 5 min. S amples were then subjected to SDS -PAGE on a pre -cast NuPAGE ™ 4–12% Bis -Tris gel (NP0335B0X, Invitrogen) and transferred to a nitrocellulose membrane using the Trans-Blot Turbo Transfer system (#1704270, Bio-Rad). Membranes were blocked in PBS with 5% non -fat milk and 0.1% Tween -20 for 1 h on a rolling shaker at room temperature and incubated with primary antibodies overnight at 4 °C. Goat anti -IGF-I polyclonal antibody (1:5000; AF-291-SP, R&D Systems, RRID: AB_2122119), mouse .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 49 anti-IGF-II monoclonal antibody (1:1000; MAB2921, R&D Systems, RRID: AB_2233454) and goat anti -Glyceraldehyde-3-Phosphate Dehydrogenase (GAPDH) polyclonal antibody (1:1000, AF5718, R&D Systems, RRID: AB_2278695) were used simultaneously for primary detection. Secondary detection was performed using IRDye 800CW donkey anti -goat IgG (#926 -32214, LI -COR Biosciences, RRID: AB_621846) for IGF-I and GAPDH, detected in the blue channel, and IRDye 680RD donkey anti-mouse IgG (#926-68072, LI-COR Biosciences, RRID: AB_10953628) for IGF-II, detected in the red channel. Fluorescence signals were visualised in a ChemiDoc™ MP Imaging System using Image Lab software (v6.1.0, Bio-Rad). Two-photon polymerisation (2PP) lithography Computer-aided design (CAD) files were created using Autodesk Fusion 360 (v.2.0.19941) and Materialise Magics (v27.0), then exported as Standard Tessellation Language (STL) files. Single hemispherical microstructures (27.5 µm height) with defined dimple di ameters (2 or 7 µm) served as building blocks for the assembly of array configurations. Structures were imported into the DeScribe software (v2.7, Nanoscribe GmbH), arranged along the x and y axes to generate the required layouts, and converted to GWL form at (Table S7). Direct laser writing was performed using NanoWrite (v1.10.5) on the Photonic Professional GT system (Nanoscribe GmbH, Germany). To initiate the fabrication process, a droplet of IP-Visio photoresin (Nanoscribe GmbH, Germany) was deposited onto the fused silica substrate coated with indium doped tin oxide (ITO). The substrate was then secured to the sample holder and mounted in a holder compatible with the piezoelectric stage. Arrays were written in a bottom -up sequence, with the first layer adhered directly to the substrate surface using a 25x magnification/0.8 NA microscope objective (0.8 DIC Imm Korr, Carl Zeiss AG). .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 50 Uncured resin was removed by alternating washes with propylene glycol monomethyl ether acetate (PGMEA; #484431, Sigma -Aldrich) followed by isopropyl alcohol (IPA; I9030, Sigma-Aldrich). This cycle was repeated until all uncured resin was completely removed, and printed constructs were then allowed to air dry for 3-5 min. Post-processing the printed arrays and assessment of GLI1 expression A two-step process was used to quench resin autofluorescence. First, UV bleaching was performed by exposing printed arrays to UV at 365 nm for 3 h at 4 × 104 mJ (CL- 1000, Analytik Jena, US). The sample stage was positioned 5 cm from the UV source and encased in aluminium foil to increase the efficiency of UV exposure using its reflective properties. Afterwards, TrueBlack ® Lipofuscin Autofluorescence Quencher (#23007, Biotium) was applied according to the manufacturer's protocol. Briefly, a 1X TrueBlack solution in 70% ethanol was added to cover the samples for 7 min, followed by a thorough, gentle wash with 1X PBS to remove excess solution. Treated arrays were sterilised and conditioned before cell seeding, as described above. GLI1 expression was detected by immunostaining after 7 days in culture (See: Immunocytochemistry). Confocal fluorescence microscopy images were processed using ImageJ software (v1.53q). Although TrueBlack ® and UV bleaching have been reported to effectively quench major autofluorescent structures while preserving immunofluorescence signals [111, 112] autofluorescence in confocal imaging remained a significant challenge in dual -topography arrays. Therefore, for the 3D arrays, a sliding paraboloid background s ubtraction (rolling ball radius = 0.6 px) was applied to the red channel in the sum -projected confocal images. This method was selected to accommodate local intensity variations and complex geometries [113], ensuring accurate background estimation without over-subtraction near high-intensity features. The rolling ball radius was determined based on the scale of background .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 51 noise relative to image features and was applied consistently across all red fluorescence channel images. Using different methods for 3D and 2D samples was necessary to ensure optimal background correction for each sample type . For corresponding 2D controls, ImageJ’s Math>Subtract method (8 px) was applied to the red channel in the sum-projected confocal images. Statistical analysis Statistical analysis was performed using GraphPad Prism 9.3.1 (GraphPad Software Inc., USA). Data distribution was first assessed using Shapiro -Wilk and Kolmogorov- Smirnov normality tests. Parametric one -way or two -way ANOVA with Tukey's or Dunnett's post-hoc tests were used where appropriate. For experiments with N= 3 or datasets that did not meet normality assumptions, non-parametric Freidman test with Dunn’s multiple comparisons test for more than two groups. Data are presented as mean ± standard deviation (SD), with p < 0.05 considered the threshold for statistical significance.

FG is supported by a scholarship from Kuwait University. MA acknowledges support by the Academy of Medical Sciences Springboard Scheme [SBF008 \1057]. We acknowledge Dr Rachel Saunders (University of Manchester) for assistance with 2PP printing, Dr. Steven Marsden for assistance with AFM microscopy, Dr David Spiller for assistance with confocal microscopy, Dr. Jamie Tibble and Mrs Shahla Khan for assistance with Mastersizer. The Bioimaging Core Facility AFMs used in this study were purchased with grants from BBSRC, Wellcome, Walgreen Boots Alliance and the University of Manchester Strategic Fund. This work was also supported by the Henry Royce Institute for Advanced Materials, funded through EPSRC grants EP/R00661X/1, EP/S019367/1, EP/P025021/1 and EP/P025498/. .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 52 Supplementary information Supplementary material (attached)

1 Muntz, I., et al. (2022) The role of cell –matrix interactions in connective tissue mechanics. Physical biology 19, 021001 2 Della Bella, E., et al. (2021) Dexamethasone Induces Changes in Osteogenic Differentiation of Human Mesenchymal Stromal Cells via SOX9 and PPARG, but Not RUNX2. Int J Mol Sci 22 3 Iantomasi, T., et al. (2023) Oxidative Stress and Inflammation in Osteoporosis: Molecular Mechanisms Involved and the Relationship with microRNAs. Int J Mol Sci 24 4 Chen, Z., et al. (2024) Progress in biomaterials inspired by the extracellular matrix. Giant 19, 100323 5 Yang, L. , et al. (2020) Topography induced stiffness alteration of stem cells influences osteogenic differentiation. Biomater Sci 8, 2638-2652 6 Böker, K.O. , et al. (2020) Laser ablated periodic nanostructures on titanium and steel implants influence adhesion and osteogenic differentiation of mesenchymal stem cells.

13, 3526 7 Zhang, G. , et al. (2024) Titanium nanoparticles released from orthopedic implants induce muscle fibrosis via activation of SNAI2. Journal of Nanobiotechnology 22, 522 8 Zeng, Y., et al. (2024) Osteogenic differentiation of bone mesenchymal stem cells on linearly aligned triangular micropatterns. Journal of Materials Chemistry B 12, 8420-8430 9 Montaño -Machado, V. , et al. (2016) A comparison of adsorbed and grafted fibronectin coatings under static and dynamic conditions. Physical Chemistry Chemical Physics 18, 24704-24712 10 Ohba, S. (2020) Hedgehog Signaling in Skeletal Development: Roles of Indian Hedgehog and the Mode of Its Action. Int J Mol Sci 21 11 Pan, A., et al. (2013) A review of hedgehog signaling in cranial bone development. Front Physiol 4, 61 12 Johnson, G.P. , et al. (2021) Primary cilium -mediated MSC mechanotransduction is dependent on Gpr161 regulation of hedgehog signalling. Bone 145, 115846 13 Ghuloum, F.I. , et al. (2022) From mesenchymal niches to engineered in vitro model systems: Exploring and exploiting biomechanical regulation of vertebrate hedgehog signalling.

Today Bio 17, 100502 14 Onodera, S. , et al. (2020) Hedgehog Activation Regulates Human Osteoblastogenesis. Stem Cell Reports 15, 125-139 15 Amer, M.H. , et al. (2021) Designing topographically textured microparticles for induction and modulation of osteogenesis in mesenchymal stem cell engineering. Biomaterials 266, 120450 16 Yang, Y., et al. (2022) Gaussian curvature–driven direction of cell fate toward osteogenesis with triply periodic minimal surface scaffolds. Proceedings of the National Academy of Sciences 119, e2206684119 17 Ghuloum, F.I. , et al. (2023) Towards modular engineering of cell signalling: Topographically-textured microparticles induce osteogenesis via activation of canonical hedgehog signalling. Biomater Adv 154, 213652 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 53 18 Shivani, S. , et al. (2022) Programmed Topographic Substrates for Studying Roughness Gradient-Dependent Cell Migration Using Two-Photon Polymerization. Front Cell Dev Biol 10, 825791 19 van Altena, P.F.J. and Accardo, A. (2023) Micro 3D Printing Elastomeric IP -PDMS Using Two-Photon Polymerisation: A Comparative Analysis of Mechanical and Feature Resolution Properties. Polymers 15, 1816 20 Alvarez-Paino, M., et al. (2019) Polymer Microparticles with Defined Surface Chemistry and Topography Mediate the Formation of Stem Cell Aggregates and Cardiomyocyte Function. ACS Applied Materials & Interfaces 11, 34560-34574 21 Sarmadi, M. , et al. (2020) Modeling, design, and machine learning -based framework for optimal injectability of microparticle-based drug formulations. Science Advances 6, eabb6594 22 Smith, D. , et al. (2019) Microparticles for Suspension Culture of Mammalian Cells. ACS Applied Bio Materials 2, 2791-2801 23 Jain, K. , et al. (2025) Immobile Integrin Signaling Transit and Relay Nodes Organize Mechanosignaling through Force-Dependent Phosphorylation in Focal Adhesions. ACS Nano 19, 2070-2088 24 Geoghegan, I.P., et al. (2021) Estrogen withdrawal alters cytoskeletal and primary ciliary dynamics resulting in increased Hedgehog and osteoclastogenic paracrine signalling in osteocytes. Sci Rep 11, 9272 25 Suchors, C. and Kim, J. (2022) Canonical Hedgehog Pathway and Noncanonical GLI Transcription Factor Activation in Cancer. Cells 11 26 Béguin, E.P. , et al. (2020) Flow -induced Reorganization of Laminin -integrin Networks Within the Endothelial Basement Membrane Uncovered by Proteomics. Mol Cell Proteomics 19, 1179-1192 27 Romaus-Sanjurjo, D., et al. (2022) Overexpressing eukaryotic elongation factor 1 alpha (eEF1A) proteins to promote corticospinal axon repair after injury. Cell Death Discovery 8, 390 28 Duncan, B.W., et al. (2021) Molecular mechanisms of L1 and NCAM adhesion molecules in synaptic pruning, plasticity, and stabilization. Frontiers in Cell and Developmental Biology 9, 625340 29 Ahn, H., et al. (2023) Hierarchical Topography with Tunable Micro- and Nanoarchitectonics for Highly Enhanced Cardiomyocyte Maturation via Multi -Scale Mechanotransduction. Advanced Healthcare Materials 12, 2202371 30 Mohindra, P. and Desai, T.A. (2021) Micro -and nanoscale biophysical cues for cardiovascular disease therapy. Nanomedicine: Nanotechnology, Biology and Medicine 34, 102365 31 Huang, P. , et al. (2023) Improving hard metal implant and soft tissue integration by modulating the “inflammatory-fibrous complex” response. Bioactive Materials 20, 42-52 32 Spada, S. , et al. (2021) Fibronectin as a multiregulatory molecule crucial in tumor matrisome: from structural and functional features to clinical practice in oncology. Journal of Experimental & Clinical Cancer Research 40, 102 33 Wehrli, B.M. , et al. (2003) Sox9, a master regulator of chondrogenesis, distinguishes mesenchymal chondrosarcoma from other small blue round cell tumors. Hum Pathol 34, 263- 269 34 Hu, D.P. , et al. (2017) Cartilage to bone transformation during fracture healing is coordinated by the invading vasculature and induction of the core pluripotency genes. Development 144, 221-234 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 54 35 Lin, Y. -C., et al. (2021) SCUBE3 loss -of-function causes a recognizable recessive developmental disorder due to defective bone morphogenetic protein signaling. The American Journal of Human Genetics 108, 115-133 36 Collette, N.M., et al. (2016) Sostdc1 deficiency accelerates fracture healing by promoting the expansion of periosteal mesenchymal stem cells. Bone 88, 20-30 37 Xie, Y., et al. (2020) FGF/FGFR signaling in health and disease. Signal Transduction and Targeted Therapy 5, 181 38 Zhang, J. , et al. (2020) SERPINE2 promotes esophageal squamous cell carcinoma metastasis by activating BMP4. Cancer Letters 469, 390-398 39 Izu, Y., et al. (2016) Collagens VI and XII form complexes mediating osteoblast interactions during osteogenesis. Cell Tissue Res 364, 623-635 40 Wang, K. , et al. (2024) COL6A3 enhances the osteogenic differentiation potential of BMSCs by promoting mitophagy in the osteoporotic microenvironment. Molecular Biology Reports 51, 206 41 Leung, V.Y., et al. (2011) SOX9 governs differentiation stage-specific gene expression in growth plate chondrocytes via direct concomitant transactivation and repression. PLoS Genet 7, e1002356 42 Reyes Alcaraz, V. , et al. (2024) A Narrative Review of the Roles of Chondromodulin -I (Cnmd) in Adult Cartilage Tissue. Int J Mol Sci 25 43 Simonds, M.M. , et al. (2020) Juvenile idiopathic arthritis fibroblast -like synoviocytes influence chondrocytes to alter BMP antagonist expression demonstrating an interaction between the two prominent cell types involved in endochondral bone formation. Pediatric Rheumatology 18, 89 44 Callaghan, B. , et al. (2022) Genome -wide transcriptome profiling of human trabecular meshwork cells treated with TGF-β2. Scientific Reports 12, 9564 45 Krämer, A. , et al. (2014) Causal analysis approaches in Ingenuity Pathway Analysis. Bioinformatics 30, 523-530 46 Cohen, L. and Walt, D.R. (2019) Highly Sensitive and Multiplexed Protein Measurements. Chemical Reviews 119, 293-321 47 Suh, H.-S., et al. (2013) Insulin-like growth factor 1 and 2 (IGF1, IGF2) expression in human microglia: differential regulation by inflammatory mediators. Journal of Neuroinflammation 10, 805 48 Bunea, A.-I., et al. (2021) Micro 3D Printing by Two-Photon Polymerization: Configurations and Parameters for the Nanoscribe System. Micro 1, 164-180 49 Jin, M., et al. (2024) Distraction force promotes the osteogenic differentiation of Gli1+ cells in facial sutures via primary cilia-mediated Hedgehog signaling pathway. Stem Cell Research & Therapy 15, 198 50 Hasan, M. and Blair, S. (2022) Maximizing transmittance in two-photon 3D printed materials for micro-optics in the visible. Opt. Mater. Express 12, 895-906 51 Rosenbohm, J. , et al. (2022) A multi -material platform for imaging of single cell -cell junctions under tensile load fabricated with two-photon polymerization. Biomed Microdevices 24, 33 52 Sagner, A. and Briscoe, J. (2017) Morphogen interpretation: concentration, time, competence, and signaling dynamics. WIREs Developmental Biology 6, e271 53 Ferraz, M.P. (2024) An Overview on the Big Players in Bone Tissue Engineering: Biomaterials, Scaffolds and Cells. Int J Mol Sci 25 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 55 54 Jeon, H. , et al. (2014) A mini -review: Cell response to microscale, nanoscale, and hierarchical patterning of surface structure. Journal of Biomedical Materials Research Part B: Applied Biomaterials 102, 1580-1594 55 Chen, D. , et al. (2021) Early time -point cell morphology classifiers successfully predict human bone marrow stromal cell differentiation modulated by fiber density in nanofiber scaffolds. Biomaterials 274, 120812 56 Kilian, K.A., et al. (2010) Geometric cues for directing the differentiation of mesenchymal stem cells. Proceedings of the National Academy of Sciences 107, 4872-4877 57 Huang, G., et al. (2024) The mechanism of ITGB4 in tumor migration and invasion. Front Oncol 14, 1421902 58 Dumbauld, D.W. , et al. (2010) Focal adhesion kinase -dependent regulation of adhesive forces involves vinculin recruitment to focal adhesions. Biology of the Cell 102, 203-213 59 Ghasemi, F., et al. (2024) Regeneration of actin filament branches from the same Arp2/3 complex. Sci Adv 10, eadj7681 60 Li, B. , et al. (2021) Phosphoproteomics identifies potential downstream targets of the integrin α2β1 inhibitor BTT-3033 in prostate stromal cells. Ann Transl Med 9, 1380 61 Hu, A., et al. (2022) Cholesterylation of Smoothened is a calcium-accelerated autoreaction involving an intramolecular ester intermediate. Cell Res 32, 288-301 62 Lei, M. , et al. (2023) Cell-cell and cell-matrix adhesion regulated by Piezo1 is critical for stiffness-dependent DRG neuron aggregation. Cell Rep 42, 113522 63 Pardo-Pastor, C., et al. (2018) Piezo2 channel regulates RhoA and actin cytoskeleton to promote cell mechanobiological responses. Proceedings of the National Academy of Sciences 115, 1925-1930 64 Kumari, A., et al. (2022) Unique lingual expression of the Hedgehog pathway antagonist Hedgehog-interacting protein in filiform papillae during homeostasis and ectopic expression in fungiform papillae during Hedgehog signaling inhibition. Dev Dyn 251, 1175-1195 65 Holmes, G., et al. (2021) Single-cell analysis identifies a key role for Hhip in murine coronal suture development. Nat Commun 12, 7132 66 Yoshida, T. and Delafontaine, P. (2020) Mechanisms of IGF -1-Mediated Regulation of Skeletal Muscle Hypertrophy and Atrophy. Cells 9 67 Cho, E. -S., et al. (2012) Constitutive activation of smoothened leads to impaired developments of postnatal bone in mice. Molecules and Cells 34, 399-405 68 Zhang, N., et al. (2023) Identification of distinct subpopulations of Gli1-lineage cells in the mouse mandible. J Anat 243, 90-99 69 Hojo, H. , et al. (2013) Hedgehog-Gli Activators Direct Osteo -chondrogenic Function of Bone Morphogenetic Protein toward Osteogenesis in the Perichondrium*. Journal of Biological Chemistry 288, 9924-9932 70 Tan, Z. , et al. (2018) Synergistic co -regulation and competition by a SOX9 -GLI-FOXA phasic transcriptional network coordinate chondrocyte differentiation transitions. PLoS Genet 14, e1007346 71 Xu, W., et al. (2025) The Hedgehog-GLI1 Pathway Regulates Osteogenic Differentiation of Human Cervical Posterior Longitudinal Ligament Cells by BMP Signalling Pathway. J Cell Mol Med 29, e70393 72 Galea, G.L., et al. (2021) Making and shaping endochondral and intramembranous bones. Dev Dyn 250, 414-449 73 Matsushita, Y., et al. (2023) Bone marrow endosteal stem cells dictate active osteogenesis and aggressive tumorigenesis. Nature Communications 14, 2383 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 56 74 Ebrahimighaei, R., et al. (2022) Combined role for YAP-TEAD and YAP-RUNX2 signalling in substrate-stiffness regulation of cardiac fibroblast proliferation. Biochim Biophys Acta Mol Cell Res 1869, 119329 75 Cheng, S.L. , et al. (2003) MSX2 promotes osteogenesis and suppresses adipogenic differentiation of multipotent mesenchymal progenitors. J Biol Chem 278, 45969-45977 76 Jia, S., et al. (2017) Small-molecule Wnt agonists correct cleft palates in Pax9 mutant mice in utero. Development 144, 3819-3828 77 Yoshida, C.A., et al. (2012) SP7 inhibits osteoblast differentiation at a late stage in mice. PLoS One 7, e32364 78 Duchamp de Lageneste, O. (2018) Periosteum contains skeletal stem cells with high bone regenerative potential controlled by Periostin. Nat. Commun. 9 79 Li, C., et al. (2016) Tenascin C affects mineralization of SaOS2 osteoblast-like cells through matrix vesicles. Drug Discov Ther 10, 82-87 80 Morgan, J.M., et al. (2011) Regulation of tenascin expression in bone. J Cell Biochem 112, 3354-3363 81 Miwa, H.E. , et al. (2010) Isoform -specific O -Glycosylation of Osteopontin and Bone Sialoprotein by Polypeptide N -Acetylgalactosaminyltransferase-1*. Journal of Biological Chemistry 285, 1208-1219 82 Lopes, J. , et al. (2013) Type VIII collagen mediates vessel wall remodeling after arterial injury and fibrous cap formation in atherosclerosis. The American journal of pathology 182, 2241-2253 83 Ko, F.C. and Sumner, D.R. (2021) How faithfully does intramembranous bone regeneration recapitulate embryonic skeletal development? Developmental Dynamics 250, 377-392 84 Fults, D.W. (2004) IGF2 enhances sonic hedgehog-induced medulloblastomafrom nestin- expressing neu;ral progenitors in mice. Cancer Research 64, 1283-1284 85 Long, F. , et al. (2004) Ihh signaling is directly required for the osteoblast lineage in the endochondral skeleton. Development 131, 1309-1318 86 Shi, Y. , et al. (2015) Hedgehog signaling activates a positive feedback mechanism involving insulin-like growth factors to induce osteoblast differentiation. Proc Natl Acad Sci U S A 112, 4678-4683 87 Pricci, F. , et al. (1996) Insulin-like growth factors I and II stimulate extracellular matrix production in human glomerular mesangial cells. Comparison with transforming growth factor- beta. Endocrinology 137, 879-885 88 Jeong, E.Y., et al. (2013) Enhancement of IGF-2-induced neurite outgrowth by IGF-binding protein-2 and osteoglycin in SH-SY5Y human neuroblastoma cells. Neuroscience Letters 548, 249-254 89 Prina, E., et al. (2020) Bioinspired Precision Engineering of Three -Dimensional Epithelial Stem Cell Microniches. Advanced Biosystems 4, 2000016 90 Li, P., et al. (2018) Morphogen gradient reconstitution reveals Hedgehog pathway design principles. Science 360, 543-548 91 Johnson, B.P. , et al. (2021) A Microphysiological Approach to Evaluate Effectors of Intercellular Hedgehog Signaling in Development. Front Cell Dev Biol 9, 621442 92 Donocoff, R.S., et al. (2020) Optimization of tamoxifen-induced Cre activity and its effect on immune cell populations. Scientific Reports 10, 15244 93 De Boeck, J. and Verfaillie, C. (2021) Doxycycline inducible overexpression systems: how to induce your gene of interest without inducing misinterpretations. Mol Biol Cell 32, 1517- 1522 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint 57 94 Stamataki, D. , et al. (2005) A gradient of Gli activity mediates graded Sonic Hedgehog signaling in the neural tube. Genes Dev 19, 626-641 95 Briscoe, J. and Thérond, P.P. (2013) The mechanisms of Hedgehog signalling and its roles in development and disease. Nature Reviews Molecular Cell Biology 14, 416-429 96 Virden, A. (2017) Particle Sizing by Laser Diffraction. Malvern Instruments Limited: Malvern, UK 97 Chao, K.-j., et al. (2005) Thin films of mesoporous silica: characterization and applications. Comptes Rendus Chimie 8, 727-739 98 Chiu, C.-Y., et al. (2006) Mesoporous silica powders and films—Pore size characterization by krypton adsorption. Microporous and mesoporous materials 91, 244-253 99 Livak, K.J. and Schmittgen, T.D. (2001) Analysis of relative gene expression data using real-time quantitative PCR and the 2− ΔΔCT method. methods 25, 402-408 100 Jeon, R. -H., et al. (2019) PPIA, HPRT1, and YWHAZ Genes Are Suitable for Normalization of mRNA Expression in Long -Term Expanded Human Mesenchymal Stem Cells. BioMed Research International 2019, 3093545 101 Ferreira, D.B., et al. (2024) RPLP0/TBP are the most stable reference genes for human dental pulp stem cells under osteogenic differentiation. World J Stem Cells 16, 656-669 102 Bolger, A.M. , et al. (2014) Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics 30, 2114-2120 103 Dobin, A., et al. (2013) STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29, 15- 21 104 Liu, Y. , et al. (2013) RNA-seq differential expression studies: more sequence or more replication? Bioinformatics 30, 301-304 105 Love, M.I., et al. (2014) Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol 15, 550 106 Gu, Z. , et al. (2016) Complex heatmaps reveal patterns and correlations in multidimensional genomic data. Bioinformatics 32, 2847-2849 107 Wu, T., et al. (2021) clusterProfiler 4.0: A universal enrichment tool for interpreting omics data. Innovation (Camb) 2, 100141 108 Kuleshov, M.V., et al. (2016) Enrichr: a comprehensive gene set enrichment analysis web server 2016 update. Nucleic Acids Res 44, W90-97 109 Yu, G. and He, Q.Y. (2016) ReactomePA: an R/Bioconductor package for reactome pathway analysis and visualization. Mol Biosyst 12, 477-479 110 Pazoki, R., et al. (2021) Genetic analysis in European ancestry individuals identifies 517 loci associated with liver enzymes. Nature Communications 12, 2579 111 Zhang, Z. , et al. (2023) Management of autofluorescence in formaldehyde -fixed myocardium: choosing the right treatment. Eur J Histochem 67 112 Sharaf, A. , et al. (2023) Suppression of auto -fluorescence from high -resolution 3D polymeric architectures fabricated via two-photon polymerization for cell biology applications. Micro and Nano Engineering 19, 100188 113 Kiernan, L. , et al. (2020) Adaptive Background Correction of Crystal Image Datasets: Towards Automated Process Control. Sensing and Imaging 21, 48 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 14, 2025. ; https://doi.org/10.1101/2025.07.11.664383doi: bioRxiv preprint