Independent Tuning of Stiffness and Pore Size in 3D Rat Tail Collagen I Matrices

13 The interplay between the extracellular matrix (ECM) mechanical properties and the tumor 14 microenvironment is increasingly recognized as a critical factor in cancer progression. Three-15 dimensional (3D) culture systems have emerged as essential platforms for in-vitro cell-based 16 applications, offering microenvironments that are more physiologically relevant compared to 17 traditional two-dimensional (2D) cultures. However, independently controlling the topological 18 and mechanical features of 3D matrices remains challenging due to the interdependence of these 19 parameters. In this study, we demonstrate a method for independently tuning pore size and 20 stiffness in collagen I (Coll I) networks and examine their effects on breast cancer and epithelial 21 cell morphology and cluster formation. Collagen concentration was used to modulate bulk 22 stiffness, while polymerization temperature was adjusted to control pore size. Using this 23 approach, we developed a 3D Coll I matrix with tuned stiffnesses from 80, 228 and 360 Pa while 24 simultaneously holding pore size constant (2.5 µm). Similarly, we developed a low- (1.5 mg/mL) 25 and high- (3.5 mg/mL) concentration collagen hydrogel with varying pore sizes from 2.5 µm to 26 3.1 µm and 2.0 µm to 2.4 µm, respectively, without altering stiffness (80 Pa and 350 Pa). 27 Integrating a breast epithelial cell line, MCF-10A, and metastatic breast cancer cell line, MDA-28 MB-231, we demonstrate matrix stiffness and pore size independently and differentially regulate 29 cell morphology and cluster formation. Our results establish a robust method for decoupling 30 stiffness and pore size in Coll I matrices enabling more precise investigations into how ECM 31 mechanical properties influence metastatic and epithelial cell behavior. 32

Collagen, ECM (extracellular matrix), Mechanical Properties, Cell morphology 33 34 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint Statement of Significance 35 This study presents a robust method to independently tune stiffness and pore size in 3D collagen 36 I matrices, overcoming a key challenge in extracellular matrix modeling. By decoupling these 37 parameters through collagen concentration and polymerization temperature, the platform enables 38 more accurate investigation of how ECM mechanical properties influence metastatic and 39 epithelial cell behavior. Our finding reveals that matrix stiffness and pore size independently and 40 differentially regulate cell morphology and cluster formation, demonstrating the distinct cellular 41 responses to specific ECM properties and underscoring the importance of the tumor 42 microenvironment in cancer biology and tissue engineering. 43 44 45 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint 1. Introduction 46 The extracellular matrix (ECM) plays a central role in both development and disease processes. 47 Through surface receptors, cells sense and respond to the ECM’s chemical composition and 48 physical properties. Dynamic changes in the ECM’s structure and composition can alter its 49 chemical and physical cues, influencing key cellular behaviors such as migration, alignment, 50 proliferation, and morphology. In two-dimensional (2D) continuous substrates, environmental 51 factors such as stiffness, size, density, and the spatial distribution of adhesion sites can be 52 precisely controlled to study their effects on cell migration. However, these properties become 53 more complex and less predictable in three-dimensional (3D) porous hydrogels [1]. The most 54 influential parameters affecting cancer cell behavior in 3D environments are matrix stiffness, and 55 confinement, often characterized by pore size. 56 Recent research has shown that changes in stiffness not only accompany disease progression but 57 may actively contribute to it [2]. Increased ECM stiffness is a known regulator of tumor 58 proliferation, immune cell infiltration, epithelial to mesenchymal transition and drug delivery [3, 59 4]. Many of the current studies on matrix stiffness rely on 2D synthetic surfaces. However, cells 60 exhibit different morphologies, modes of migration, proliferation rates, and nascent ECM 61 secretion on 3D when compared to 2D substrates [5, 6]. 62 In a similar manner, pore sizes of the ECM have been shown to be an essential factor in 63 determining cell tracking, cell phenotype, cell behavior and protein secretion [7]. For instance, if 64 the pore is too small compared to the nuclear dimensions, it will restrict the locomotion, 65 diffusion of nutrients, and removal of waste metabolites. If the pore is larger than its nuclear 66 dimensions, the physical barriers will no longer exist, and the cell will undergo invasion [8]. 67 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint However, if pore size is too large compared to the nuclear dimensions, it will reduce the surface 68 area for cell attachment [9]. 69 Recent studies in 3D matrices have demonstrated that increased matrix stiffness can enhance cell 70 spreading, tumor outgrowth, ligand densities, intracellular stiffness, and increased motor activity 71 [10-13]. Other work has shown that by decreasing the pore size of the matric, thereby increasing 72 confinement, impedes proliferation, viability, cell morphology and multicellular cluster 73 formation [14-17]. However, these studies do not independently control matrix stiffness and pore 74 sizes, due to the interdependence of these parameters. This interrelationship makes it difficult to 75 delineate whether the observed cellular behaviors are driven primarily by stiffness, pore size, or a 76 combination of both. As a result, these conclusions regarded as isolated effects of stiffness or 77 pore size may be confounded, creating a gap in understanding, as few systems allow for 78 independent control of stiffness and pore size. The lack of tunable 3D matrices that can decouple 79 these parameters remains a significant barrier to fully elucidating their individual roles in cell 80 behavior and disease progression. 81 Ideally, dynamic processes can be studied in engineered cell culture systems in which stiffness 82 changes, and pore size changes can be controlled, independent of one another. Recent strategies 83 to modulate 3-D matrix properties have included modifying matrices with alginate [18], using 84 synthetic polymers with tunable cross-linking densities [19-22], and creating inter-penetrating 85 networks comprised of natural proteins and other hydrogels [23-26]. Although these 86 modifications can generate 3-D scaffolds with tunable mechanical properties, they also change 87 the fundamental structural properties of the culture system. 88 The dimensions and organization of collagen fibrils can be modestly changed with various 89 ranges of collagen concentration and polymerization temperature that are consistent with cell 90 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint viability. Previous studies have demonstrated that increasing collagen concentration 91 simultaneously increases stiffness (storage moduli, G’) and decreases average pore size [27, 28]. 92 Additionally, other reports have demonstrated that decreasing the collagen polymerization 93 temperature to 22°C resulted in increased pore size [29, 30]. This is not surprising due to the 94 dependency of collagen lateral growth on temperature, thus reducing the gelation temperature 95 increases the fiber diameter, increasing pore size. In conclusion, adjusting gelation temperature 96 can optimize the control of pore size independent from chemical parameters. 97 Herein, we utilize 3D collagen I matrices polymerized at both low (Room Temperature, 22°C) 98 and high temperature (37°C) while simultaneously varying bulk collagen concentrations (1.5, 2.5 99 and 3.5 mg/mL). We observe 3D Coll I matrix architecture via confocal reflectance microscopy 100 (CRM) technique and quantify the pore size via 3D Euclidean distance map (EDT) based bubble 101 analysis. In parallel, we perform mechanical characterization via rheology to determine bulk 102 storage modulus (G’) of the collagen gels. 103 Our results demonstrate that we obtain collagen gels with tunable stiffness and pore size, 104 independent from one another, eliminating the need to integrate chemical crosslinkers or 105 synthetic matrices. Together, our study describes a simple and robust method for independently 106 tuning 3D collagen stiffness and pore size and elucidates the critical role these biophysical 107 parameters play in modulating breast cell behavior and invasion potential. 108 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint 2. Results 109 Schematic 1. Independent Tuning of 3D Collagen Hydrogels. Schematic illustrating the integration of collagen110 concentration and polymerization temperature to independently decouple stiffness and pore size in 3D Coll I111 matrices. 112 113 2.1 Temperature and Concentration Dependent Collagen Assembly 114 The lateral growth of collagen fibrils is temperature dependent, therefore adjusting the gelation 115 temperature and collagen concentration allows for control over the collagen architecture 116 en l I .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint independent of chemical crosslinkers (Schematic 1). Neutralized collagen solutions were 117 polymerized in microwells (200 µL volume) at 37°C (referred to as “37°C”) or by gradually 118 increasing the temperature from room temperature to 37°C (referred to as “Room Temp”) (Figure 119 1A). Confocal reflectance microscopy (CRM) was used to quantify the structure of the collagen 120 network as this imaging requires no additional sample preparation or drying—which inevitably 121 perturb network structure—therefore providing more accurate structural information of the 122 samples in their hydrated state [31]. Representative CRM images in Figure 1B show the 123 microarchitecture of Coll I matrices as a function of the bulk collagen concentration and the 124 polymerization temperature. As expected, fiber density increased with collagen concentration 125 and fiber diameter increased at lower polymerization temperature and appear to be independent 126 from concentration. 127 128 129 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint 130 Figure 1. Fabrication and confocal reflectance microscopy of collagen hydrogels. (A) Schematic representation131 of the 3D collagen neutralization and gelation process. (B) Maximum projection images of collagen hydrogels132 captured using confocal reflectance microscopy (CRM) at varying temperatures and concentrations. Experiments133 were performed in three independent experiments. Scale bars represent 50/i1μ m. 134 135 on els nts .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint 2.2 Topological Characterization of 3D Coll I Matrices 136 An 3D image segmentation analysis was applied as described in experimental methods. Briefly, a 137 CRM image stack (50 µm height, 0.2 µm step size) is denoised and thresholded before applying 138 a 3D Euclidean Distance Transform. The Euclidean Distance Map (EDM) represents the shortest 139 distance of each non-fibril pixel to the neighboring collagen fibrils. Local maxima are 140 determined from the EDM which represent the radii of the largest sphere that would fit into each 141 respective 3D pore of the collagen scaffold (Supplemental Figure 2). The pore diameters are 142 calculated (2*radii) and the actual pore size is given by the median of all pore diameters of the 143 collagen scaffold sample. The distribution of pore size diameters is visualized to quantify 144 scaffold microarchitecture and assess structural heterogeneity (Figure 2A). While the median 145 pore diameter reflects the central tendency, histograms reveal the full range of pore sizes, 146 highlighting variability as well as rare large or small pores that may influence scaffold 147 performance. 3D pore size analysis of the network pore diameter confirmed that Room Temp 148 Coll I matrices were composed of larger pores and thicker fibers compared to 37°C Coll I 149 matrices (Figure 2B-C). However, because the resolution of the microscope approaches 0.2µm, 150 measured fiber diameter should be interpreted qualitatively, rather than quantitatively. Features 151 below or near the resolution limit may be overestimated due to point spread and optical 152 limitations. Furthermore, CRM image analysis verifies that the porous structure of the collagen 153 scaffolds is homogenous in each condition, but varies between polymerization temperature 154 (Figure 2D, Supplemental Figure 1). 155 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint 156 Figure 2. Room temperature collagen polymerization increases median pore size in 3D collagen hydrogels. (A) 157 Representative average histogram distributions of pore size diameters from 3D pore size analysis. (B) Comparison 158 of average median pore diameters in each gel under varying collagen concentration and polymerization temperature. 159 Experiments were performed in three independent experiments. Two-Way ANOVA followed by Tukey’s post-hoc 160 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint test, not significant (ns) P/i1 >/i1 0.05, *P/i1≤/i1 0.05, **P/i1≤/i1 0.01, ***P/i1≤/i1 0.001, ****P/i1≤/i1 0.0001. Error bars 161 represent SEM. (C) Line graph represents the mean fibril diameter based on un-paired t-test, not significant 162 (ns) P/i1 >/i1 0.05, *P/i1≤/i1 0.05, **P/i1≤/i1 0.01, ***P/i1≤/i1 0.001, ****P/i1≤/i1 0.0001. Experiments were performed in 163 three independent experiments. Error bars represent SEM. (D) CRM image analysis showing the cross-sectional 164 distribution of collagen fibers from 1.5 mg/mL collagen. Additional cross-sectional distributions (2.5 and 3.5 165 mg/mL) can be found in Supplemental Figure 3. 166 167 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint 2.3 Experimental Analysis of Collagen Hydrogel Mechanical Properties 168 We investigated how different fabrication techniques influenced the mechanical properties of the 169 3D Coll I matrices. Consistent with the prior studies of 3D collagen matrices, we measured the 170 storage modulus (G’) of the structurally distinct gels using bulk rheology [32-34]. Briefly, 171 coverslips were affixed to the rheometer base, and a 25mm parallel plate geometry was used to 172 perform a time sweep at a fixed frequency of 0.1 Hz and constant strain of 0.1% for 10 minutes 173 at room temperature. The storage modulus was calculated as the average of the final 50 points. 174 Interestingly, we observed no significant change in G’ between polymerization temperatures 175 when collagen concentration was held constant. However, increasing collagen concentration 176 from 1.5 to 2.5 and 3.5 mg/mL led to a marked increase in G’: at 37°C values rose from 80 to 177 230 and 360 Pa, and at room temperature, from 80 to 140 and 340 Pa, respectively. These results 178 indicate that the G’ is primarily governed by collagen concentration, with polymerization 179 temperature exerting a secondary influence. Overall, this data suggests that collagen 180 concentration is the dominant parameter controlling the mechanical stiffness of these 3D Coll I 181 matrices. 182 183 184 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint 185 Figure 3. Rheological mechanical characterization of 3D collagen hydrogels. (A) Schematic depiction of 186 rheometer set up for 3D collagen hydrogels utilizing 25mm parallel plate geometry. (B) Mean storage moduli for 187 37°C and Room Temp collagen hydrogels obtained via rheological measurements (N= 4 - 6 collagen gels). Two-188 Way ANOVA followed by Tukey’s post-hoc test, not significant (ns) P/i1 >/i1 0.05, *P/i1≤/i1 0.05, **P/i1≤/i1 0.01, 189 ***P/i1≤/i1 0.001, ****P/i1≤/i1 0.0001. Error bars represent SEM. 190 191 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint 2.4 Independent modulation of pore size and stiffness via temperature and collagen 192 concentration. 193 Independent tuning of pore size and bulk stiffness in 3D collagen hydrogels was achieved by 194 varying collagen concentration and polymerization temperature (see Schematic 1). These two 195 parameters allow decoupling of mechanical stiffness from microstructural features such as pore 196 size. Bulk stiffness was modulated by: (1) increasing collagen concentration (e.g., from 1.5 to 197 2.5/i1 mg/mL), which elevated collagen content and increased the storage modulus; and (2) 198 decreasing the polymerization temperature (from 37/i1 °C to room temperature), which increased 199 pore size. This approach enables tuning of stiffness without significantly affecting average pore 200 size (Table 1). Using this method, we generated three collagen hydrogels with similar median 201 pore sizes (~2.5/i1μ m) but varying stiffnesses of 80, 228, and 360/i1 Pa (Table 1, Figure 4A). To 202 independently tune pore size, we held collagen concentration constant to maintain stiffness and 203 reduced polymerization temperature to increase median pore size (Table 2). Specifically, by 204 using low- (1.5/i1 mg/mL) and high- (3.5/i1 mg/mL) collagen concentrations, we achieved distinct 205 pore sizes—ranging from 2.0 to 3.1/i1μ m—while maintaining stiffness at 80 and 350/i1 Pa, 206 respectively (Table 2, Figure 4B–C). This strategy enables independent control over the 207 mechanical and structural properties of 3D Coll I matrices without relying on chemical 208 crosslinkers or synthetic polymer additives. 209 210 211 212 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint 213 Figure 4. Independent tuned pore size and stiffness in 3D Coll I matrices. Average normalized pore size 214 distributions of (A) independently tuned stiffness with no change to pore distribution. Independently tuned pore size 215 at both (B) low- (1.5 mg/mL) and (C) high- (3.5 mg/mL) collagen concentration. 216 217 Table 1. Independent Tuning of Stiffness in 3D Collagen Hydrogels. 218 Tune Stiffness Collagen Concentration [mg/mL] Polymerization Temperature Storage Modulus [Pa] ± Std Dev. Median Pore Size [µm] 1.5 37°C 80 /g3399 37 2.5 /g3399 0.23 2.5 Room Temp 228 /g3399 90 2.5 /g3399 0.13 3.5 Room Temp 360 /g3399 140 2.4 /g3399 0.09 219 Table 2. Independent Tuning of Pore Size in 3D Collagen Hydrogels. 220 Tune Pore Size Collagen Concentration [mg/mL] Polymerization Temperature Storage Modulus [Pa] ± Std Dev. Median Pore Size [µm] 1.5 37°C 80 /g3399 37 2.5 /g3399 0.23 1.5 Room Temp 81 /g3399 39 3.1 /g3399 0.39 3.5 37°C 340 /g3399 37 2.0 /g3399 0.05 3.5 Room Temp 360 /g3399 140 2.4 /g33990.09 221 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint 2.5 Impact of Coll I stiffness and pore size on cell morphology and cluster formation 222 To investigate how extracellular matrix (ECM) properties influence cell behavior, we used the 223 3D Coll I matrices with controlled topology and elasticity. Our study distinctively determined 224 cell behavior under varying pore size or stiffness while controlling the other parameters (pore 225 size or stiffness). Based on the matrix characterization earlier, we evaluated Coll I matrices with 226 fixed pore size (~2.5 µm) and varying stiffnesses of 80, 228, and 360 Pa (Table 1). To isolate 227 pore size, we chose additional matrices with constant stiffness (~ 80Pa at 1.5 mg/mL) with pore 228 sizes ranging from 2.5 µm to 3.1 µm. Similarly, matrices at high collagen concentration (3.5 229 mg/mL, ~350 Pa) were prepared with pore sizes ranging from 2.0 µm to 2.4 µm. 230 231 For cellular analysis, two breast cell lines have been chosen with distinctively different migration 232 characteristics and phenotypes: MDA-MB-231, displays mesenchymal migration, while the non-233 tumorigenic epithelial cell line, MCF-10A, migrates collectively. These selected breast cell lines 234 are frequently used as a standard tool to assess biological and biophysical behavior or cells. 235 Single cells were embedded within the Coll I matrix, polymerized, and cultured for 4 days. The 236 Coll I matrices were subsequently analyzed for cluster formation and single cells were analyzed 237 for cell area and cell aspect ratios. 238 239 As shown in Figure 5A-D, MDA-MB-231 cells exhibited a reduction in both cell areas and cell 240 aspect ratios with increasing matrix stiffness. In contrast, MCF-10A maintained smaller areas 241 and aspect ratios than MDA-MB-231 cells across all stiffnesses (Figure 5A-C). MCF-10A cells 242 are known to form large clusters, which is also observed in these 3D Coll I matrices (Figure 5B). 243 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint While MCF10A cluster formation was found to subtly decrease with increasing Coll I matrix 244 stiffness, cell area remains largely unchanged. However, their aspect ratios and circularity did 245 increase with stiffness (Figure 5C, Supplemental Figure 4). 246 The effect of pore size on cellular behavior was also examined over 4 days under both low (1.5 247 mg/mL) and high (3.5 mg/mL) collagen concentrations (Figure 6A). Across all conditions, MCF-248 10A cells consistently formed more clusters than MDA-MB-231 (Figure 6A, E). At low collagen 249 concentration (1.5 mg/mL, ~80Pa) MDA-MB-231 cells decreased their area and aspect ratios 250 and increased their circularity at when pore sizes decrease from 3.1 µm to 2.5 µm (Figure 6C-D, 251 Supplemental Figure 4). In contrast, MCF-10A cells showed an increased area and aspect ratio 252 and decreased circularity under the same conditions. At high collagen concentration (3.5 mg/mL, 253 ~350Pa), MDA-MB-231 cells reduced their area with a decrease in pore size (2.0 µm to 2.4 µm), 254 while MCF-10A cells displayed an opposite trend with their cell area and circularity increasing 255 alongside pore size. 256 257 258 259 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint 260 Figure 5. Impact of Coll I matrix stiffness on cell morphology and cluster formation. (A) CRM images of 261 MDA-MB-231 breast cancer cells and MCF-10A breast epithelial cells cultivated within 3D Coll I matrices with 262 stiffnesses of 80, 228 and 360 Pa for 4 days. Scale bars represent 100 µm. (B) Quantitative analysis of cluster 263 formation of MDA-MB-231 cells and MCF-10A cells, whereby cells without contact to other cells were counted as 264 single cells. Experiments were performed in three independent experiments (MDA-MB-231 n = 455-1125 cells, 265 MCF-10A n = 387-687 cells). One-Way ANOV A followed by Tukey’s post-hoc test, not significant (ns) P/i1 >/i1 0.05, 266 *P/i1≤/i1 0.05, **P/i1≤/i1 0.01, ***P/i1≤/i1 0.001, ****P/i1≤/i1 0.0001. Error bars represent SEM. Quantitative analysis of 267 MDA-MB-231 breast cancer and MCF-10A breast epithelial (C) cell area and (D) aspect ratio as a function of 268 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint stiffness. Significance was tested between samples of varying stiffness and similar pore size. Experiments were 269 performed in three independent experiments. One-Way ANOVA followed by Tukey’s post-hoc test, not significant 270 (ns) P/i1 >/i1 0.05, *P/i1≤/i1 0.05, **P/i1≤/i1 0.01, ***P/i1≤/i1 0.001, ****P/i1≤/i1 0.0001. Error bars represent SEM. 271 272 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint 273 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint Figure 6. Impact of Coll I pore size on cell morphology and cluster formation. (A) CRM images of MDA-MB-274 231 breast cancer cells and MCF-10A breast epithelial cells (green) cultivated within 3D Coll I matrices (grey) of 275 both low collagen concentration (1.5 mg/mL) and varying median pore size from 2.5 µm to 3.1 µm and high 276 collagen concentration (3.5 mg/mL) with varying median pore size from 2.0 µm to 2.4 µm for 4 days. Scale bars 277 represent 100 µm. Quantitative analysis of cluster formation of MDA-MB-231 cells and MCF-10A cells at both low 278 collagen concentration (B) and high collagen concentration (C). Cells without contact to other cells were counted as 279 single cells (grey). Experiments were performed in three independent experiments (MDA-MB-231 n = 455-1125 280 cells, MCF-10A n = 387-687 cells). One-Way ANOVA followed by Tukey’s post-hoc test, not significant 281 (ns) P/i1 >/i1 0.05, *P/i1≤/i1 0.05, **P/i1≤/i1 0.01, ***P/i1≤/i1 0.001, ****P/i1≤/i1 0.0001. Error bars represent SEM. 282 Quantitative analysis of MDA-MB-231 breast cancer and MCF-10A breast epithelial at cell area and aspect ratio as 283 a function of pore size for both low collagen concentration (D-F) and high collagen concentration (E-G). 284 Significance was tested between samples of varying pore size and similar stiffness. Experiments were performed in 285 three independent experiments. One-Way ANOV A followed by Tukey’s post-hoc test, not significant 286 (ns) P/i1 >/i1 0.05, *P/i1≤/i1 0.05, **P/i1≤/i1 0.01, ***P/i1≤/i1 0.001, ****P/i1≤/i1 0.0001. Error bars represent SEM. 287 288 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint 3. Discussion 289 290 The invasion of a broad range of cells is strongly regulated by the biophysical properties of the 291 ECM, including the microstructure and elasticity. However, the ability to recapitulate these 292 native ECM architectures for 3D matrices is lacking due to their dependent nature on one 293 another. This problem is further compounded where 3D Coll I matrices are mechanically tuned 294 using tunable cross-linking densities that alter the fundamental structural and chemical properties 295 of the 3D system. The current study determines a method to decouple stiffness and pore size 296 from one another in 3D Coll I matrices, without the need for chemical crosslinkers or polymer 297 additives. Our findings show that bulk stiffness can be tuned from 80 Pa to 360 Pa by varying 298 collagen concentration, while maintaining a constant median pore size of 2.5 μ m through 299 polymerization temperature control. Conversely, at a fixed collagen concentration (e.g., 1.5 300 mg/mL) or stiffness (e.g., 80 Pa), adjusting the polymerization temperature allows pore size to be 301 modulated from 2.5 μ m to 3.1 μ m. Cellular integration further amplifies the independent effect 302 of stiffness and pore size on both metastatic and non-tumorigenic morphological responses. 303 Together, these results highlight the utility of controlling polymerization temperature and protein 304 concentration to independently modulate 3D matrix stiffness and pore size, enabling further 305 precision to study cellular responses to mechanical cues in 3D environments. 306 307

to alter polymerization temperature for pore size control have been explored previously 308 in the literature. Several reports have described methods to alter fiber diameter, or pore size, of 309 3D Coll I matrices using decreased polymerization temperatures [29, 35]. While these studies 310 demonstrate that lower polymerization temperatures can increase pore size, they do not isolate 311 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint stiffness from pore size without introducing chemical crosslinkers or other polymers into the 312 Coll I network. Additionally, many of these approaches simultaneously alter pH and ionic 313 strength outside of physiological conditions, which likely alter collagen electrostatic interactions 314 and further confound mechanical and structural properties. In contrast, we identify a range of 315 collagen concentrations and polymerization temperatures where stiffness can be tuned 316 independently from pore size under biologically relevant conditions. It is important to note that 317 cancer to healthy breast ECM ranges from 200 Pa to 4000 Pa and our 3D Coll I gels range from 318 80 to 350 Pa, a 4-fold increase in the bulk stiffness to replicate the native healthy and cancer 319 breast ECM [36]. We also define additional conditions to maintain stiffness, enabling pore size 320 modulation. At these parameters, we also observe a distribution of pore sizes ranging from 321 1/i1μ m to 7/i1μ m, with independently tuned median pore sizes between 2.0 and 3.1/i1μ m. Given 322 the inherent heterogeneity and occasional unpredictability of 3D Coll I, it is important to note 323 that variation in polymerization temperature may introduce increased variability. Although the 324 structural stability of the 3D Coll I architecture was assessed over 10 days (data not shown), 325 long-term stability was not quantified. In conclusion, these results accurately recapitulate a 326 biomimetic microenvironment, consistent with other reports [37]. 327 While prior studies have primarily relied on indirect metrics such as 2D imaging or porosity to 328 assess the matrix structure, our method uniquely quantifies the 3D pore size directly, offering 329 greater precision when characterizing the pericellular matrix dimensions. Average pore size or 330 mesh size is a key parameter to physically characterize the structure of biopolymeric filamentous 331 networks, as it directly influences cellular behavior like adhesion, polarization, and migration [8, 332 38]. Due to the randomness of biological networks and the heterogeneity of collagen matrices, 333 the extraction of physical parameters via confocal microscopy is not straightforward and remains 334 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint a non-trivial task. To overcome this, we applied a 3D Euclidean distance transformation to 335 segmented collagen networks to quantify the largest diameter that can fit within each pore. This 336

more accurately approximates the true 3D special constraints experienced by a cell 337 migration through the matrix. It is important to acknowledge the limitations of CRM, which only 338 detects reflected light and preferentially visualizes horizontal fibers. Thus, CRM is known for a 339 blind spot, where fibers at an angle steeper than a cutoff angle, are not detected [39]. As a result, 340 3D Coll I networks with fewer fibers can result in an overestimation of the pore size network 341 [40]. Cellular integration into the 3D Coll I matrices is a critical indicator on how biophysical 342 properties, like stiffness and pore size, influence cellular behavior. Cell integration encompasses 343 multiple steps, including protrusion extension, adhesion formation and matrix remodeling, all of 344 which are directly influenced by the local pore architecture. 345 346 Stiffness was modulated independently of pore size, and MDA-MB-231 cells responded by 347 decreasing their cell area and aspect ratio with increasing stiffness. Previous studies have shown 348 that these cells migrate more slowly in stiffer matrices, which may indicate that they require 349 more time to deform and remodel the surrounding extracellular matrix [41]. Interestingly, MDA-350 MB-231 cells exerted approximately 10-fold higher traction forces on soft gels compared to stiff 351 ones, a surprising finding given that many cell types typically become more contractile on stiffer 352 substrates[42]. Additionally, it is well established that traction forces tend to decrease with 353 reduced cell spreading area [43-46]. This suggests that MDA-MB-231 cells may adopt a 354 different migration or mechanosensing strategy in response to increased stiffness, possibly 355 shifting toward amoeboid migration, which relies more on cortical tension and cell deformability 356 than on traction forces. Amoeboid migration in the MDA-MB-231 cell line can be characterized 357 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint by a rounder cell morphology. One study investigating MDA-MB-231 cells under varying 358 collagen densities, which also increases stiffness and decreases pore size, reported a reduction in 359 both cell area and migration speed, along with the emergence of amoeboid-like, rounded cell 360 morphologies [47]. Cells exhibiting bleb-like morphologies secreted more matrix-361 metalloproteinases (MMPs), facilitating collagen degradation. Interestingly, they observe a 362 significant decrease in matrix degradation, suggesting a possible shift from mesenchymal to 363 ameboid migration mode. 364 Pointcloix et. al. further reported that the roundedness of MDA-MB-231 generated tensional 365 forces through β1-integrin adhesion and the absence of lamellipodial extensions [48]. Similarly, 366 other studies have shown that an increased matrix stiffness can reduce invadopodia formation 367 thereby limiting metastatic invasion, compares to softer 3D environments [49]. Integrin-mediate 368 singling also plays a critical role in activating downstream pathways, including Rho/ROCK 369 pathway, which are known regulators of actin dynamics and cellular contractility. It is plausible 370 that these signaling mechanisms contribute to the observed behavior in low-stiffness conditions, 371 although further studies would be needed to elucidate their role in this system. Together, these 372 findings suggest that the reduced cell areas observed in MDA-MB-231 cells may be contributing 373 to decreased traction forces, consistent with ameboid migration patterns. 374 375 MCF10A cells were less likely to form large clusters in stiffer 3D Coll I matrices, suggesting that 376 the increased stiffness may limit cell-cell cohesion. This behavior contrasts with softer matrices, 377 where a more permissive mechanical environments may support collective behavior, allowing 378 cells to aggregate and form tumor-like structures. A previous study has shown that this decrease 379 in cluster formation in stiff matrices shifts from non-tumorigenic MCF10A cells led to an 380 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint increase in invasive sites presenting higher levels of vimentin, a key regulator in 381 mechanosensation [50]. However, while no significant change in cell area was observed in 382 MCF10A cells, aspect ratios decreased in stiffer matrices, indicating a shift towards a more 383 rounded, ameboid morphology. 384 385 Based on the data, increasing the collagen concentration (low vs. high) while tuning 386 independently tuning pore size differentially influenced the morphological response from MDA-387 MB-231 and MCF-10A cells. For MDA-MB-231 cells in low collagen concentration, increasing 388 the pore size led to an increase in aspect ratio and cell area. This suggests that in environments 389 with less collagen density and larger pores size, MDA-MB-231 cells can better extend their 390 protrusions due to reduced steric hindrance, facilitating a more elongated shape, mesenchymal 391 phenotype indicative of invasion behavior. Conversely, in a high collagen concentration, 392 increasing pore size resulted in decreased cell areas and aspect ratios. One possible explanation 393 is that the larger pores in an otherwise dense collagen matrix create fiber-sparse regions, causing 394 cells to retract and decrease aspect ratios due to insufficient anchoring. In contrast, smaller pores 395 within high concentration gels may provide enhanced fibril alignment and bundling for guidance 396 that supports their elongation. Additionally, under these 3D conditions, the acquired amoeboid 397 shape demonstrates cell adaptivity to higher matrix density. This is evidenced by a shift from the 398 elongated cell morphology (low concentration) to a rounder shape (high concentration), enabling 399 cells to navigate tight spaces [47, 51]. 400 In contrast, the morphological response from MCF-10A cells was less sensitive to changes in 401 pore size across both high and low collagen concentrations. The ability of MCF-10A to form 402 large, multicellular clusters appears to be more strongly determined by pore size. This 403 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint observation is consistent with other reports, where MCF10A cells proliferate to form spheroid-404 like clusters [52] after 4 days in culture that reduced pore size is associated with decreased 405 single-cell spreading. It is also known that MCF10A display stable cadherin junctions and high 406 E-Cadherin expression, regulating strong cell-cell adhesion and reducing the potential of the 407 epithelial to mesenchymal transition (EMT) [53]. Similarly to the MDA-MB-231 cells, MCF-408 10A also exhibited reduced cell area and aspect ratios with increasing collagen content. This 409 suggests that denser matrices impose greater physical constrains on cell spreading and elongation 410 across both malignant and non-tumorigenic cell lines. 411 412 In summary, we provide the first study, to our knowledge, to independently decouple stiffness 413 and pore size from 3D Coll I matrices utilizing collagen concentration and polymerization 414 temperature. Additionally, we utilized a 3D EDT analysis to accurately recapitulate the 415 pericellular pore sizes found within the 3D Coll I matrices, rather than a discrete analysis or 416 overestimation from 2D analysis. We show that mesenchymal cell line and non-tumorigenic cell 417 line respond differently to cluster formation and cellular morphology under varying stiffnesses 418 and pore sizes. It is important to note that these changes may vary between cell lines. As such, 419 this developed method could serve as an important tool to accurately recapitulate the TME in 3D 420 and could be used to inform future novel personalized medicine and tissue engineering screening 421 platforms. 422 4. Conclusion 423 In this study, we developed a novel method for independently decoupling pore sizes and stiffness 424 from one another in 3D Coll I matrices. We employed various parameters to verify the 425 robustness of our results. Our findings in the metastatic cell line, MDA-MB-231, indicate that 426 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint modulating both pore size and stiffness regulates the invasive morphology. In contrast, the non-427 tumorigenic cell line, MCF-10A, showed limited responsiveness to changes in stiffness and pore 428 size. These results signify the importance of decoupled mechanical and architectural features to 429 elucidate the independent effects to cell morphology. Overall, our results provide a reliable 430 platform for studying independently controlled mechanical stimulation in a 3D ECM biomimetic 431 platform. 432 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint 5. Experimental Methods 433 5.1 Preparation of 3D Coll I Matrices 434 3D collagen matrices were formed using rat tail type I collagen (Advanced Biomatrix, 4 mg/mL). 435 Final collagen concentrations of 1.5, 2.5 and 3.5 mg/mL were adjusted by mixing neutralization 436 solution, sterile water and phosphate buffer to achieve constant pH (~7.4) and constant ionic 437 strength (~170mM). Coll I solutions were prepared and kept on ice (4°C) to prevent 438 polymerization. Subsequently, 200µL of collagen solution was transferred onto imaging or 439 rheological characterization glass slide, as needed. To initiate polymerization, “37°C” gels were 440 placed at 37°C for 1 hour and the “Room Temp” gels were placed at room temperature (~22°C) 441 for 30 minutes followed by 37°C, 5% CO /g2870 and 95% humidity for 1 hour. To maintain proper 442 hydration, 1X phosphate buffer or cell culture media was added to gels after 30 minutes of 443 polymerization. 444 5.2 Rheological Preparation of 3D Coll I Matrices 445 Glutaraldehyde-treated glass coverslips were using during rheological testing to minimize slip, 446 Circular 25 mm glass coverslips were first cleaned and silanized by incubating them in a 1% 447 solution of 3-aminopropyl-trimethoxysilane (Alfa Aesar, Haverhill, MA) for 10 minutes. They 448 were then treated with 0.5% glutaraldehyde (Amresco, Solon, OH), rinsed thoroughly with 449 deionized water, and allowed to dry. The collagen gel solution was prepared on ice as described 450 previously. A volume of 500 µL of the collagen solution was applied to each treated coverslip 451 and allowed to polymerize. Rheological testing was performed immediately after polymerization. 452 5.3 Rheological Assessment on 3D Coll I Matrices 453 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint Coll I viscoelastic were measured by using an AR-200 rheometer with a built-in temperature and 454 gap calibration. Frequency and strain sweeps were conducted in oscillatory shear on fully 455 crosslinked hydrogels to determine the linear viscoelastic region (LVR). Changes in stiffness in 456 response to concentration and polymerization temperature were conducted in time sweeps of 10 457 minutes with readings every (constant: 0.1% strain, 1Hz). A 25 mm serrated parallel plate 458 geometry was used with a gap of 1mm, with a target normal force set to 0.25N. The average 459 storage moduli (G’) is taken as the average of the last 50 points. 460 5.4 Confocal Reflectance Microscopy for 3D Collagen Architecture Analysis 461 Confocal laser scanning microscope in reflectance mode was used to obtain images of fibrillar 462 collagen microstructure for quantitative assessment. All samples were prepared in Bio-Labs 16 463 culture well with removable chambered coverglass #1.5. Briefly, collagen hydrogels were 464 neutralized and allowed to self-assemble at desired polymerization temperature as previously 465 described. Immediately after polymerization, images were collected in reflectance mode with a 466 (Leica Stellaris SP8) using a 63X oil immersion objective (N.A = 1.40). Each z-stack began at 467 least 10 µm above the cover glass and consisted of 50 slices with 0.2 µm spacing between slices 468 (10 µm total thickness). Three stacks were taken per sample. 469 5.5 Collagen Architecture Quantification 470 A 3D pore-size analysis method was adapted as described by Fischer at al. [54]. Critical 471 metadata from the image stack, such as voxel size, were directly obtained from the Leica 472 microscope. In effort to handle scattering and absorption due to sample heights and large travel 473 lengths of exciting and emitted light, image segmentation of fibril and non-fibril volumes was 474 performed for each image in the stack. Before segmentation, a total variation denoising and a 475 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint gaussian filter was applied to reduce noise while preserving edges. Subsequently, adaptive local 476 thresholding was applied to identify fibrils with brighter signals from sparse areas. The 477 segmentation resulted in a 3D binary volume where the black area represents pores and white 478 representer collagen fibers. A 3D Euclidean Distance Transform (EDT) is applied to the non-479 fibril part of the 3D binary volume. The Euclidean Distance Map (EDM) represents the shortest 480 distance of each non-fibril pixel to the neighboring collagen fibrils (Supplemental Figure 2). A 481 gaussian filter is applied to the EDM and pores along the edge are excluded. Local maxima are 482 determined from the filtered EDM. The minimum detectable pore size is defined by the 483 numerical aperture of the objective lens. The maxima represent the radii of the largest sphere that 484 would fit into each respective 3D pore of the collagen scaffold. The pore diameters are calculated 485 (2*radii) and the actual pore size, ζ , is given by the median of all pore diameters of the collagen 486 scaffold sample, as given in Eq. 1 487 /g2022 /g3404 /g1865/g1857/g1856/g1861/g1853/g1866/g4666/g1856 /g3043/g4667 ( 1 ) 488 To estimate the fiber diameters, the home-built image processing procedure was utilized as 489 described by Franke et al. [14] equipped with an erosion algorithm and autocorrelation analysis, 490 respectively. Briefly, this method automates the image segmentation into pore and fibril 491 segments for each xy-image of the image stack and evaluates the mean fibril diameter via 492 autocorrelation. 493 494 5.6 Breast Epithelial and Metastatic Breast Cancer Cell Culture 495 Human neuroblastoma cell line MCF10A and MDA-MB-231 were chosen from our cell 496 depository. MDA-MB-231 cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) 497 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint supplemented with 10% fetal bovine serum (Gibco, Grand Island, NY , USA) and antibiotics 498 (penicillin and streptomycin) (Gibco, Grand Island, NY , USA). MCF-10A cells were cultured in 499 20 ng/mL recombinant human EGF (Thermo Fischer), 0.5 mg/mL hydrocortisone (Sigma), 100 500 ng/mL cholera toxin (Sigma), 10 µg/mL insulin (Sigma), Horse Serum (Invitrogen) and 501 antibiotics (penicillin and streptomycin) (Gibco, Grand Island, NY , USA). 502 5.7 Morphological Assessment of Breast Cell Response to Mechanical Stimulation 503 Cell invasion and morphology were analyzed after 4 days of culture. For analysis, cells were 504 fixed in 4% paraformaldehyde for 10 minutes at room temperature and rinsed 3 times with PBS, 505 waiting 5 minutes for each wash. Afterwards, cells were permeabilized with 0.1% Triton X-100 506 for 10 minutes at room temperature and rinsed 3 times with PBS, waiting 5 minutes for each 507 wash. For analysis of cell invasion and morphology, cells were stained with Alexa Fluor 488 508 Phalloidin (Invitrogen, Germany) overnight at 4°C and rinsed 3 times with PBS, waiting 5 509 minutes for each wash. Cells were imaged using a Leica Stellaris SP8 equipped with a 10x 510

(N.A = 0.40). The images were 1024x1024 pixels, and z-stacks were taken of 150 μ m 511 thick sections. Cell morphology was visualized using Alexa Fluor 488 Phalloidin signal. Aspect 512 ratio (major axis/minor axis) was determined using ImageJ freehand tool and shape descriptors. 513 Roundness (4*area/(π /g1499 major axis^2 )) was determined using ImageJ free hand tool. Cluster 514 formation was manually counted, whereby cells without contact with other cells were counted as 515 single cells. For morphological and cluster formation analysis, at least 9 stacks were taken per 516 sample, and 3 independent experiments were analyzed per condition. Representative higher-517 resolution images were captured with 40X (N.A = 0.95) oil immersion objective. 518 5.8 Statistics 519 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint Statistical analysis was carried out using Prism 10.0 GraphPad software was used for graph 520 generation and statistical analyses. The significance level was set to p < 0.05 for all conditions. 521 The number of independent biological replicates, sample sizes analyzed, and statistical tests used 522 are stated in the figure legends. 523 524 Author Contributions 525 Data curation, T.V .; Formal analysis, T.V .; Investigation, Q.W and H.S.Z.; Resources, Q.W and 526 H.S.Z.; Supervision, Q.W and H.S.Z.; Writing—original draft, H.S.Z., Q.W., and T.V .; Writing—527 review and editing, H.S.Z., Q.W., and T.V . All authors have read and agreed to the published 528 version of the manuscript. 529

530 The authors would like to thank Dr. Catherine Whittington and Athenia Jones for gifting MDA-531 MB-231 cell line and Dr. Michele Vitolo for gifting the MCF-10A cell line. This work was 532 supported by NIH R01GM157590. 533 Funding Source 534 This research did not receive any specific grant from funding agencies in the public, commercial, 535 or non-profit sectors. 536 Conflicts of Interest 537 The authors declare no conflicts of interest. 538 539 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint 540

541 1. Geiger, F., et al., Fiber stiffness, pore size and adhesion control migratory phenotype of 542 MDA-MB-231 cells in collagen gels. PLoS One, 2019. 14(11): p. e0225215. 543 2. Stowers, R.S., S.C. Allen, and L.J. Suggs, Dynamic phototuning of 3D hydrogel stiffness. 544 Proc Natl Acad Sci U S A, 2015. 112(7): p. 1953-8. 545 3. Mai, Z., et al., Modulating extracellular matrix stiffness: a strategic approach to boost 546 cancer immunotherapy. Cell Death Dis, 2024. 15(5): p. 307. 547 4. Rice, A.J., et al., Matrix stiffness induces epithelial-mesenchymal transition and promotes 548 chemoresistance in pancreatic cancer cells. Oncogenesis, 2017. 6(7): p. e352. 549 5. Pelham, R.J., Jr. and Y . Wang, Cell locomotion and focal adhesions are regulated by 550 substrate flexibility. Proc Natl Acad Sci U S A, 1997. 94(25): p. 13661-5. 551 6. Gardel, M.L., et al., Traction stress in focal adhesions correlates biphasically with actin 552 retrograde flow speed. J Cell Biol, 2008. 183(6): p. 999-1005. 553 7. Oliviero, O., M. Ventre, and P.A. Netti, Functional porous hydrogels to study 554 angiogenesis under the effect of controlled release of vascular endothelial growth factor. 555 Acta Biomater, 2012. 8(9): p. 3294-301. 556 8. Wolf, K., et al., Physical limits of cell migration: control by ECM space and nuclear 557 deformation and tuning by proteolysis and traction force. J Cell Biol, 2013. 201(7): p. 558 1069-84. 559 9. Saif Ur Rahman, M., et al., Matrix mechanophysical factor: pore size governs the cell 560 behavior in cancer. Advances in Physics: X, 2022. 8(1). 561 10. Mason, B.N., et al., Tuning three-dimensional collagen matrix stiffness independently of 562 collagen concentration modulates endothelial cell behavior. Acta Biomater, 2013. 9(1): p. 563 4635-44. 564 11. Karamichos, D., R.A. Brown, and V . Mudera, Collagen stiffness regulates cellular 565 contraction and matrix remodeling gene expression. J Biomed Mater Res A, 2007. 83(3): 566 p. 887-94. 567 12. Kim, J.E., et al., Characterization of the mechanical properties of cancer cells in 3D 568 m atrices in response to collagen concentration and cytoskeletal inhibitors. Integr Biol 569 (Camb), 2018. 10(4): p. 232-241. 570 13. Riching, K.M., et al., 3D collagen alignment limits protrusions to enhance breast cancer 571 cell persistence. Biophys J, 2014. 107(11): p. 2546-58. 572 14. Franke, K., et al., Topologically defined composites of collagen types I and V as in vitro 573 cell culture scaffolds. Acta Biomater, 2014. 10(6): p. 2693-702. 574 15. Sapudom, J., et al., The phenotype of cancer cell invasion controlled by fibril diameter 575 and pore size of 3D collagen networks. Biomaterials, 2015. 52: p. 367-75. 576 16. Zhang, J., et al., Pore architecture and cell viability on freeze dried 3D recombinant 577 human collagen-peptide (RHC)-chitosan scaffolds. Mater Sci Eng C Mater Biol Appl, 578 2015. 49: p. 174-182. 579 17. Murphy, C.M., M.G. Haugh, and F.J. O'Brien, The effect of mean pore size on cell 580 attachment, proliferation and migration in collagen-glycosaminoglycan scaffolds for 581 bone tissue engineering. Biomaterials, 2010. 31(3): p. 461-6. 582 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint 18. Ort, C., et al., Bioprintable, Stiffness-Tunable Collagen-Alginate Microgels for Increased 583 Throughput 3D Cell Culture Studies. ACS Biomater Sci Eng, 2021. 7(6): p. 2814-2822. 584 19. Davidenko, N., et al., Control of crosslinking for tailoring collagen-based scaffolds 585 stability and mechanics. Acta Biomater, 2015. 25: p. 131-142. 586 20. Suesca, E., et al., Multifactor analysis on the effect of collagen concentration, cross-587 linking and fiber/pore orientation on chemical, microstructural, mechanical and 588 biological properties of collagen type I scaffolds. Mater Sci Eng C Mater Biol Appl, 589 2017. 77: p. 333-341. 590 21. Lang, N.R., et al., Biphasic response of cell invasion to matrix stiffness in three-591 dimensional biopolymer networks. Acta Biomater, 2015. 13: p. 61-7. 592 22. Grier, W.K., E.M. Iyoha, and B.A.C. Harley, The influence of pore size and stiffness on 593 tenocyte bioactivity and transcriptomic stability in collagen-GAG scaffolds. J Mech 594 Behav Biomed Mater, 2017. 65: p. 295-305. 595 23. Berger, A.J., et al., Decoupling the effects of stiffness and fiber density on cellular 596 behaviors via an interpenetrating network of gelatin-methacrylate and collagen. 597 Biomaterials, 2017. 141: p. 125-135. 598 24. Suri, S. and C.E. Schmidt, Photopatterned collagen-hyaluronic acid interpenetrating 599 polymer network hydrogels. Acta Biomater, 2009. 5(7): p. 2385-97. 600 25. Suo, H., et al., Interpenetrating polymer network hydrogels composed of chitosan and 601 photocrosslinkable gelatin with enhanced mechanical properties for tissue engineering. 602 Mater Sci Eng C Mater Biol Appl, 2018. 92: p. 612-620. 603 26. Wei, Z., et al., Fiber Microarchitecture in Interpenetrating Collagen-Alginate Hydrogel 604 with Tunable Mechanical Plasticity Regulates Tumor Cell Migration. Adv Healthc Mater, 605 2023. 12(29): p. e2301586. 606 27. Miron-Mendoza, M., J. Seemann, and F. Grinnell, The differential regulation of cell 607 motile activity through matrix stiffness and porosity in three dimensional collagen 608 matrices. Biomaterials, 2010. 31(25): p. 6425-35. 609 28. Critser, P.J., et al., Collagen matrix physical properties modulate endothelial colony 610 forming cell-derived vessels in vivo. Microvasc Res, 2010. 80(1): p. 23-30. 611 2 9. Yang, Y .L., S. Motte, and L.J. Kaufman, Pore size variable type I collagen gels and their 612 interaction with glioma cells. Biomaterials, 2010. 31(21): p. 5678-88. 613 30. Xie, J., et al., Collagen Gels with Different Fibrillar Microarchitectures Elicit Different 614 Cellular Responses. ACS Appl Mater Interfaces, 2017. 9(23): p. 19630-19637. 615 31. Wolf, K., et al., Collagen-based cell migration models in vitro and in vivo. Semin Cell 616 Dev Biol, 2009. 20(8): p. 931-41. 617 32. Lai, G., Y . Li, and G. Li, Effect of concentration and temperature on the rheological 618 behavior of collagen solution. Int J Biol Macromol, 2008. 42(3): p. 285-91. 619 33. Suo, H., et al., Low-temperature 3D printing of collagen and chitosan composite for 620 tissue engineering. Mater Sci Eng C Mater Biol Appl, 2021. 123: p. 111963. 621 34. Yang, Y .L., L.M. Leone, and L.J. Kaufman, Elastic moduli of collagen gels can be 622 predicted from two-dimensional confocal microscopy. Biophys J, 2009. 97(7): p. 2051-60. 623 35. Sung, K.E., et al., Control of 3-dimensional collagen matrix polymerization for 624 reproducible human mammary fibroblast cell culture in microfluidic devices. 625 Biomaterials, 2009. 30(27): p. 4833-41. 626 36. Stowers, R.S., et al., Extracellular Matrix Stiffening Induces a Malignant Phenotypic 627 Transition in Breast Epithelial Cells. Cell Mol Bioeng, 2017. 10(1): p. 114-123. 628 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint 37. Koorman, T., et al., Spatial collagen stiffening promotes collective breast cancer cell 629 invasion by reinforcing extracellular matrix alignment. Oncogene, 2022. 41(17): p. 2458-630 2469. 631 38. Tozluoglu, M., et al., Matrix geometry determines optimal cancer cell migration strategy 632 and modulates response to interventions. Nat Cell Biol, 2013. 15(7): p. 751-62. 633 39. Jawerth, L.M., et al., A blind spot in confocal reflection microscopy: the dependence of 634 fiber brightness on fiber orientation in imaging biopolymer networks. Biophys J, 2010. 635 98(3): p. L1-3. 636 40. Lang, N.R., et al., Estimating the 3D pore size distribution of biopolymer networks from 637 directionally biased data. Biophys J, 2013. 105(9): p. 1967-75. 638 41. Tien, J., et al., Matrix Pore Size Governs Escape of Human Breast Cancer Cells from a 639 Microtumor to an Empty Cavity. iScience, 2020. 23(11): p. 101673. 640 42. Koch, T.M., et al., 3D Traction forces in cancer cell invasion. PLoS One, 2012. 7(3): p. 641 e33476. 642 43. Reinhart-King, C.A., M. Dembo, and D.A. Hammer, Endothelial Cell Traction Forces on 643 RGD-Derivatized Polyacrylamide Substrata. Langmuir, 2002. 19(5): p. 1573-1579. 644 44. Li, F., et al., Cell shape regulates collagen type I expression in human tendon fibroblasts. 645 Cell Motil Cytoskeleton, 2008. 65(4): p. 332-41. 646 45. Wang, N., et al., Micropatterning tractional forces in living cells. Cell Motil 647 Cytoskeleton, 2002. 52(2): p. 97-106. 648 46. Rape, A.D., W.H. Guo, and Y .L. Wang, The regulation of traction force in relation to cell 649 shape and focal adhesions. Biomaterials, 2011. 32(8): p. 2043-51. 650 47. Rodriguez-Cruz, D., et al., Three-dimensional cell culture conditions promoted the 651 Mesenchymal-Amoeboid Transition in the Triple-Negative Breast Cancer cell line MDA-652 MB-231. Front Cell Dev Biol, 2024. 12: p. 1435708. 653 48. Poincloux, R., et al., Contractility of the cell rear drives invasion of breast tumor cells in 654 3D Matrigel. Proc Natl Acad Sci U S A, 2011. 108(5): p. 1943-8. 655 49. Zaman, M.H., et al., M igration of tumor cells in 3D matrices is governed by matrix 656 stiffness along with cell-matrix adhesion and proteolysis. Proc Natl Acad Sci U S A, 657 2006. 103(29): p. 10889-94. 658 50. Stowers, R.S., et al., Matrix stiffness induces a tumorigenic phenotype in mammary 659 epithelium through changes in chromatin accessibility. Nat Biomed Eng, 2019. 3(12): p. 660 1009-1019. 661 51. Holle, A.W., et al., Cancer Cells Invade Confined Microchannels via a Self-Directed 662 Mesenchymal-to-Amoeboid Transition. Nano Lett, 2019. 19(4): p. 2280-2290. 663 52. Carey, S.P., K.E. Martin, and C.A. Reinhart-King, Three-dimensional collagen matrix 664 induces a mechanosensitive invasive epithelial phenotype. Sci Rep, 2017. 7: p. 42088. 665 53. Park, K.S., M.J. Dubon, and B.M. Gumbiner, N-cadherin mediates the migration of 666 MCF-10A cells undergoing bone morphogenetic protein 4-mediated epithelial 667 mesenchymal transition. Tumour Biol, 2015. 36(5): p. 3549-56. 668 54. Fischer, T., A. Hayn, and C.T. Mierke, Fast and reliable advanced two-step pore-size 669 analysis of biomimetic 3D extracellular matrix scaffolds. Sci Rep, 2019. 9(1): p. 8352. 670 671 .CC-BY-NC-ND 4.0 International licenseavailable 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 made The copyright holder for this preprintthis version posted October 4, 2025. ; https://doi.org/10.1101/2025.10.02.680089doi: bioRxiv preprint