Oncogenic Ras-Src-cortactin signaling rewires actin-generated forces to drive basement membrane rupture and initiate breast cancer invasion

33 HRas signaling, breast cancer invasion, basement membrane disruption, Src family kinases, 34 cortactin, Arp2/3 complex, myosin I, actin polymerization force, mechanotransduction, 3D 35 spheroid, tumor microenvironment stiffness 36 37 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint

38 Aberrant activation of Ras signaling drives pathological programs in various cancers 1. In breast 39 cancer, Ras hyperactivation occurs through dysregulation of multiple oncogenic pathways to 40 foster cancer progression 2. Cells with hyperactive Ras acquire metastatic properties 3, 4 and 41 must cross the basement membrane (BM) to infiltrate the tumor microenvironment (TME). The 42 BM scaffold acts as a cell-anchoring adhesion platform and a mechanical barrier that preserves 43 epithelial architecture and function 5, 6 . It consists of laminins, collagen IV, nidogen and 44 perlecan, which together form a dense, nanoporous scaffold with substantial mechanical 45 resistance to cell and ECM forces 7, 8. 46 Early mechanistic concepts of BM invasion focused on proteolytic degradation by matrix 47 metalloproteinases (MMPs), typically concentrated at invadopodia 9, 10. However, in clinical 48 trials MMP inhibitors neither blocked metastasis nor improved patients’ survival 11. More 49 recent studies have shown that cells can apply physical forces to break through the BM barrier: 50 actomyosin contractility acts together with proteases to weaken and disrupt the matrix 12. 51 Cancer spheroids form contractile protrusions that stretch and rupture the BM 13. In other cases, 52 epithelial cell clusters expand their volume and increase local contractility to compromise BM 53 integrity 14. Such findings have shifted attention toward the force generating and transmitting 54 cytoskeletal machinery as a major contributor to BM disruption. The underlying molecular 55 mechanisms of force generation are largely based on actin-myosin interactions: Non-muscle 56 myosin II motor protein generates high contractile forces by cooperative parallel actin filament 57 sliding, while myosin I/V/VII mono- or dimers act on filamentous actin to produce more locally 58 constrained actin-cell membrane tension and protrusive forces 15, 16. Besides myosin, cells use 59 actin polymerization forces for actin filament and network growth. This requires various actin 60 stabilizing factors, such as profilin or cortactin, as well as capping and geometry-shaping 61 factors. The Arp2/3 complex is important to bundle and branch actin filaments thereby shaping 62 the cellular cortex 17-19. 63 The mechanical properties of breast tissue undergo dramatic changes during cancer progression. 64 Stromal collagen cross-linking and fiber realignment progressively stiffen the extracellular 65 matrix (ECM) 20-22. TME stiffening steers cancer cell migration and promotes invasion through 66 mechanosensation and -transductive signaling cascades 23-25. The cellular mechanoresponse to 67 ECM stiffening involves reorganization of the actin cytoskeleton, the activation of the 68 contractile machinery, and finally significantly increased actin forces 26. This 69 mechanoadaptation enables cells to breach the BM 12. BM breakdown marks a critical step in 70 the transition from ductal carcinoma in situ to invasive carcinoma 27. Despite its importance, 71 the cellular mechanisms that control this pathological step remain incompletely defined. 72 To clarify these tumor-promoting processes, it is essential to understand how oncogenic 73 signaling integrates with mechanical TME cues 28. Oncogene activation can alter cellular 74 mechanosensation and force generation 29 to amplify oncogenic pathways 30-32. This 75 bidirectional crosstalk may play a decisive role in breast cancer, where increasing matrix 76 stiffness correlates with disease progression. However, the exact mechanisms by which HRas 77 oncogene cooperates with altered tissue mechanics to drive BM invasion remain unclear. 78 Here, we established breast spheroids based on MCF10A/ER:HRAS V12 cells 33 to dissect the 79 cellular events that drive early BM breach under oncogenic HRas signaling. This 3D cell model 80 allowed temporal control of HRas hyperactivation after the formation of basoapically polarized 81 BM-covered spheroids. We examined how HRas downstream signaling disrupts BM integrity 82 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint and how this effect depends on tumor microenvironment-like matrix stiffness. We investigated 83 a previously unrecognized oncogenic signaling axis that links HRas activity and altered ECM 84 mechanosensation with cell invasion. We analyzed the interplay of actin cortex remodeling, 85 myosin I- and Arp2/3-driven actin polymerization forces for mechanical BM disruption. 86 Finally, an in-silico evaluation was performed to correlate prognostic implications of the HRas-87 cortactin-Arp2/3 -axis for breast cancer patients. Together, this study aimed to define a new 88 mechanism by which Ras oncogene activation overrides BM barrier function and initiates early 89 invasive transition of breast epithelial cells. 90 91 92 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint

93 94 HRas hyperactivation induces invasive transition of healthy breast cell spheroids 95 To monitor basement membrane (BM) disruption and cell transmigration, we generated HRas-96 inducible breast spheroids derived from single MCF10A/ER:HRAS V12 cells 34, 35 (hereafter 97 referred as MCF10A/HRAS). These cells express an estrogen receptor (ER)-HRasV12 fusion 98 protein that can be pharmacologically activated by the ER ligand 4-hydroxytamoxifen (OHT). 99 After 10 days in culture (DiC) (Fig. 1A), MCF10A/HRAS spheroids (not induced with OHT) 100 featured a similar morphology as MC10A wildtype spheroids lacking the ER:HRAS V12 101 modification 5 (Fig. 1B): In both spheroid variants, a stable apicobasal polarity was established, 102 as indicated by inward-oriented Golgi organelles. Importantly, a continuous BM scaffold was 103 formed and maintained that surrounded the outer basal cell layer (Fig. 1B, compare arrowheads 104 in i and ii). 105 In contrast, MCF10A/HRasconst spheroids, derived from cells that constantly express an active 106 form of the HRas V12 protein, showed malfunctioned breast spheroid morphogenesis with 107 perturbed apical cell polarization, as indicated by randomly orientated Golgi protein. More 108 importantly, HRas const led to discontinuous BM scaffolds (Fig. 1B, arrowhead in iii). This 109 clearly demonstrated that HRas hyperactivation during early stages of breast gland development 110 disrupts basoapical polarization and BM formation. Consistent results of misshaped HRas-111 activated MCF10A spheroids have been reported 36. Thus, it implicated the necessity to induce 112 HRas in spheroids that gained proper polarity and BM scaffolds to analyze its oncogenic effects. 113 For this purpose, we made use of temporally controlled HRas activation in mature 114 MCF10A/HRas spheroids featuring fully developed BMs (10 DiC). At this point OHT 115 treatment (16 hours) reproducibly induced phosphorylation of the HRas downstream effector 116 ERK 35 (Fig. 1C). Quantification of pERK intensity levels revealed a significant increase of 117 ≈60% in OHT-treated spheroids compared to the EtOH control (Fig. 1C). The pERK increase 118 was most pronounced in the basal cell layer facing the BM-ECM interface. This activation 119 gradient was most likely caused by decreased drug accessibility towards the spheroid center. In 120 the following, OHT-treated spheroids are referred to as HRas on spheroids and the untreated 121 EtOH controls as HRasoff spheroids. Further, HRas activation frequently led to the invasive cell 122 transition of non-transformed breast spheroids that normally remain in a homeostatic state: 123 After 65 hours of HRas activation, 43% of spheroids showed invasive cell streams that breached 124 the BM reaching into the microenvironment. In contrast, HRas off controls remained polarized 125 without any visible morphological changes (Fig. 1D). 126 These results show that HRas hyperactivation alone induced an invasive transition in originally 127 non-transformed breast epithelial spheroids, even under microenvironmental conditions that 128 reflect the compliance of healthy breast tissue 37, 38 . This spheroid model thus provided a 129 controlled and reproducible platform to investigate the functional consequences of HRas 130 activation in vitro. 131 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint 132 Figure 1: HRas hyperactivation induces invasive transition of normal breast gland spheroids. 133 (A) Schematic of the spheroid morphogenesis assay. Single MCF10A/HRAS or MCF10A wild-type cells were 134 cultured in a collagen IV/laminin-rich EHS(Engelbreth-Holm-Swarm) hydrogel to generate basoapically polarized 135 spheroids after 10 days in culture (DiC). (B) Representative immunofluorescence micrographs show differences 136 in basoapical polarization of MCF10A spheroids at 10 DiC depending on HRas activation status. BM (collagen 137 IV, yellow), F-actin cytoskeleton (magenta), nuclei (DAPI, blue) and Golgi protein (GM130, green). (C) HRas 138 activation confirmed by pERK immunofluorescence after 1  hour OHT or EtOH treatment. Representative 139 immunofluorescence intensities of intracellular pERK protein (inverted grey scale) in MCF10A/HRAS spheroids 140 treated with OHT or EtOH for 16 hours. SAC: secondary antibody control. Right, quantification of mean pERK 141 intensity per spheroid (n ≥ 44; 3 independent experiments). Box: interquartile range; whiskers: 5th–95th 142 percentiles; red dots: median. (D) Phase-contrast images show the invasive transition of spheroids (10 DiC) with 143 cell transmigration into the EHS matrix after 65 hours of HRas activation with OHT. EtOH-treated HRasoff controls 144 remained non-invasive. Kolmogorov-Smirnov test was performed for the data in C; n.s.: p > 0.05; ****: p ≤ 0.0001. 145 Scale bars: 20 µm (B); 50 μm (C). Position of focal plane used for imaging and analyses is indicated by red bar. 146 147 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint HRas invasion bypasses tumor stiffness sensing, myosin II forces and BM proteolysis 148 After showing the pro-invasive consequences of HRas hyperactivation, we examined whether 149 tumor-like ECM stiffness, as fundamental mechanical cue, could enhance HRas invasion. 150 Previous work has shown that increased ECM rigidity promotes BM rupture in non-transformed 151 breast epithelial spheroids 5. We therefore established a quantitative assay (Fig. 2A) to measure 152 HRas-induced BM transmigration and cell invasion (Fig. 2B - C) as a function of 153 microenvironmental stiffening (see Supplementary Video 1). 154 HRasoff spheroids cultured on tumor-like stiff substrates exhibited pronounced invasion (72% 155 within 24 hours; median onset: 9 hours; Fig. 2D and F). In contrast, same spheroid group on 156 physiological normal-like ECM stiffness did not undergo invasion (cf. Fig. 1D). These results 157 demonstrate that HRas off spheroids are highly mechanosensitive and undergo invasion in 158 response to TME stiffening. Notably, the standard growth medium used in these initial 159 experiments induced basal activation of the ER:HRas fusion protein, thereby enhancing 160 invasion. This residual activation was attributable to estrogenic phenolic compounds (see 161 Supplementary Fig. S1). Consistently, MCF10A spheroids lacking the ER:HRas construct 162 displayed lower invasion under comparable conditions 5. To eliminate this interfering effect, 163 subsequent experiments were performed in steroid hormone-free conditions, which reduced the 164 TME stiffness-dependent invasion to 50% (median onset: 4.2 hours). This had no significant 165 impact on HRason spheroids (median onset: 9.5 hours). These data demonstrated that oncogenic 166 HRas is sufficient to drive invasive cell transition independently of TME stiffening, while 167 strongly synergizing with extracellular mechanical cues. 168 We asked whether HRas on spheroids convert TME stiffness into altered cellular forces that 169 could explain the rapid invasive cell transition. Because actomyosin contractility is a key driver 170 of BM stress and rupture 23, we inhibited non-muscle myosin II-ATPase activity. This treatment 171 did not affect the invasion outcome (cum. frequency: 100%, median onset: 5 hours; 172 Fig. 2C and D). To quantify cell-derived BM stress, we performed traction force microscopy 173 (TFM) and calculated strain energy (SE) as a readout of cellular traction transmitted through 174 the BM onto the underlying substrate 5. At the time of BM rupture, HRas off and HRas on 175 spheroids generated comparable SE (13.6 fJ, IQR: 8.2–20.2 fJ vs. 10.5 fJ, 5.5–19.3 fJ; Fig. 2E). 176 Myosin II inhibition reduced SE to near-background levels (1.0 fJ, 0.5–1.5 fJ), which 177 confirmed effective suppression of contractility by blebbistatin drug. Under these conditions, 178 SE values were close to background levels measured on cell-free substrates (Supplementary 179 Fig. S2). These findings demonstrated that HRas-driven invasion occurs without elevated 180 actomyosin-mediated cell contractions that could account for disruptive BM-stress. 181 Finally, we tested whether proteolytic BM degradation contributes to invasion by co-inhibition 182 of myosin II and major invasion-relevant MMPs (MT1-MMP, MMP-1, -2, -3, -7, -9) using the 183 broad-spectrum inhibitor marimastat 39. MMP inhibition did not affect invasion 184 (cf. Fig. 2C and D). Consistently, immunostaining showed comparable MT1-MMP localization 185 and signal intensity in the basal cell layer of both HRason/off spheroids (Supplementary Fig. S3). 186 Together, these data indicate that HRas-driven invasion is independent of MMP-mediated BM 187 degradation. 188 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint 189 Figure 2: HRas-driven invasion bypasses tumor stiffness sensing, myosin II forces and BM-proteolysis. 190 (A) Scheme of BM disruption and cell invasion assay. MCF10A/HRAS spheroids (10 DiC) were isolated from 191 EHS matrix and placed on elastomeric substrates (16 kPa, functionalized with EHS proteins) to count events of 192 local BM rupture and cell transmigration. Mechanical BM stress exertion by breast spheroids at time point of 193 invasion onset was measured by traction force microscopy (TFM): Surface-coupled fluorescent fiducial 194 microbeads were used to track tangential surface deformations from which strain energies were calculated as 195 measure for cell force-generated BM stress. (B) In spheroids, the outer basal cell layer is covered by a BM which 196 itself is in contact to the underlying substrate. Images show the BM integrity of a representative HRas on sample, 197 fixed and stained after adhering (1 hour) to the elastomeric substrate. Collagen IV (yellow), laminin-332 (cyan) 198 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint and F-actin cytoskeleton (magenta). Zoom in highlights in vivo -like layering of the endogenous BM. (C) 199 Representative sequence of phase-contrast images illustrates the first appearance of protrusive cell bodies (also 200 shown as zoom in), marking onset of BM disruption and cell transmigration. This was counted as a positive event 201 of invasion. (D) Cumulative distribution of BM disruption time, depending on HRas induction on 16 kPa substrates 202 (n ≥ 79 spheroids of ≥ 3 independent experiments). (E) Cumulative distribution of BM disruption time in spheroids 203 treated with blebbistatin for myosin II inhibition and additionally with marimastat for MMP inhibition after HRas 204 induction on stiff 16 kPa substrates (n ≥ 69 spheroids of ≥ 3 independent experiments). (F) Scatter plot shows 205 individual invasion onset time points for the sample conditions analyzed in (D and E) (median and 95% confidence 206 interval (CI)). (G) Calculated strain energies (SE) exerted by individual spheroids at onsets of BM disruption, 207 depending on HRas activation and actomyosin inhibition, (cf. D and E). Representative maps of cell-induced 208 traction stresses per condition from which SE were calculated. Scatter plot: median with 95% CI (n ≥ 48 from 3 209 independent experiments). Kruskal-Wallis test with Dunn’s multiple comparison test was performed for the data 210 in D and E; n.s.: p > 0.05; *: p ≤ 0.05; **: p ≤ 0.01; ***: p ≤ 0.001; ****: p ≤ 0.0001. Scale bars: 20 µm (B, C and 211 G). 212 213 214 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint HRas triggers BM disruption through local actin cortex thickening and reinforcement 215 Neither inhibition of myosin II nor combined blockade of MMPs reduced the invasiveness of 216 HRason spheroids, despite the clear lack of detectable pro-invasive contractile forces. This 217 indicated alternative, non-canonical force-generating mechanisms. We therefore examined the 218 F-actin cytoskeleton organization within the basal cell layer at the cell-BM-substrate (CBS) 219 interface where BM rupture occurred: Immunostaining revealed sequential cytoskeletal 220 changes after HRas induction (Fig. 3A). Within the first 1.5 hours, pre-invasive spheroids had 221 high coverage of actin-rich microspikes (MS) that were oriented towards the substrate and 222 previously described as ECM stiffness-sensing units 23. After 4.5 hours, cortical actin networks 223 thickened at BM contact sites. Figure 3A highlights such cortex thickening accompanied by 224 laterally oriented MS bundles (see i, arrow heads). After 5 hours, these regions showed cortical 225 actin reinforcement that frequently coincided with local BM scaffold loss (Fig. 3A, ii 226 arrowheads). Figure 3B shows a detailed view on late-stage invasion characterized by local BM 227 rupture (i) and cell transmigration into the microenvironment (ii + iii). 228 Of note, invasive cells lacked prominent F-actin stress fiber (SF) formation, which typically 229 characterize highly contractile and BM-stressing cells 23. This lack of SFs was consistent with 230 the low SE values measured before (cf. Fig. 2G). Ultimately, at the late stage of transmigration 231 into the ECM, spheroid cells showed pronounced actin thickening at BM rupture sites that were 232 accompanied by BM widening and hole formation (Fig. 3B, panels i and iii). These findings 233 identified a HRas-driven program of cortical actin reinforcement that coincided with, and likely 234 contributed to, non-proteolytic BM rupture and invasive cell transition. 235 236 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint 237 Figure 3: HRas hyperactivation triggers BM disruption through local actin cortex reinforcement 238 (A) Time-resolved immunofluorescence staining reveals cytoskeletal remodeling following HRas activation. 239 Spheroids were stained for F-actin (magenta) and BM (collagen IV, yellow) and imaged at the CBS-interface using 240 16 kPa elastomeric cell adhesion substrates. Pre-invasive HRason spheroids (1.5 h and 4.5 h) showed dense MS (i, 241 arrowheads). After 5 hours, pronounced cortical actin reinforcement appeared at sites of local BM loss (ii, 242 arrowheads). (B) Representative images highlight a late-stage event of BM cell transmigration: BM-collagen IV 243 XZ projection of an image stack shows a cell transmigration hole within the BM scaffold (i, asterisks). MIP (ii) 244 and a 3D image stack reconstruction show the F-actin cytoskeleton of a BM-transmigrating cell (iii). Scale bars: 245 10 µm (A) and 5 µm (B). Position of focal plane(s) used for imaging and analyses is indicated by red bar/rectangle. 246 247 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint HRas rewires actin cortex dynamics to mechanically stress the BM scaffold for invasion 248 After finding the spatial coincidence of cortical actin thickening and BM rupture, we next 249 analyzed how this characteristic cytoskeletal remodeling could be functionally linked with 250 invasion progression. Using 4D LCI, F-actin and BM dynamics were resolved in high 251 spatiotemporal manner after HRas activation. 252 Within the first four hours, HRas on spheroids exhibited a dynamic reshaping: basal cells at the 253 CBS-interface showed repeated pushing and pulling displacements of cortical actin patches 254 mechanically coupled to the BM. As representative example, a single cell retracted approx. 255 10 µm toward the spheroid center, while reinforcing the actin cortex (Fig. 4A, arrowheads). 256 This retraction phase was followed by local BM disruption (t = 5.5 hours). These qualitative 257 observations link the displacement of reinforced cortex regions to local mechanical deformation 258 and subsequent rupture of the BM cell migration barrier (see also Supplementary Video S2). 259 To quantify this HRas-induced BM deformation, an automated image-based BM segmentation 260 tool was developed (Fig. 4B). The analysis used laminin-332 fluorescence to define spheroid 261 outlines (masks) and calculate their compactness: spheroid compactness was defined as the 262 ratio of actual area to the area of a circle with identical perimeter. In other words, compactness 263 increases when spheroids adopt a more circular shape with smooth boundaries and decreases 264 with irregular, wrinkled deformations (see cartoon, Fig. 4C). 265 HRas activation caused a progressive and significant increase in compactness, reaching a 266 maximum prior to invasion onset (mean t = 4 hours). In contrast, non-invasive HRas off 267 spheroids maintained constant compactness over the entire time of analyses (t = 12 hours) 268 (Fig. 4B). Further stratification of HRas off spheroids into non-invasive and invasive groups 269 revealed that exclusively HRason spheroids exhibited increased compactness (Fig. 4C) whereas 270 the HRas off group exhibited only marginal changes (t = 4 hours, range: -0.05 to +0.05) 271 Supplementary Fig. S4A). This trend held true when measuring points of HRas on and non-272 invasive HRas off spheroid were normalized to relative time points (1/4, 2/4, 3/4 and 1): The 273 compactness elevated only in HRas on spheroids prior to invasion but not in HRas off controls 274 (Supplementary Fig. S4B). 275 We next tested whether the BM barrier itself was altered by the tension resulting from increased 276 compactness in HRas on spheroids. The mean fluorescence intensity of laminin-332 and 277 collagen IV were measured as proxies for BM overall protein density. Laminin-332 signals 278 were used to construct a 3D mask for the BM shell at the CBS-interface (Fig. 4D). 279 Quantification within this voxel mask showed significantly elevated intensities for both BM 280 components in HRas on spheroids after 2 hours of activation. Laminin-332 increased by 37% 281 (HRasoff: 460, s.d. 180 a.u.; HRason: 630, s.d. 220 a.u.), and collagen IV by 21% (HRasoff: 8,600, 282 s.d. 2.500 a.u.; HRas on: 10.400, s.d. 2.400 a.u.) (Fig. 4E and F). These data indicated a 283 densification of the BM scaffold. 284 These findings demonstrated that increased compactness and BM densification are specific 285 morphological adaptation of invasive HRason spheroids. They imply a functional link between 286 HRas activation, cortical actin reinforcement, and coordinated cell movements that 287 significantly compact both the spheroid cytoskeleton and the BM scaffold prior to cell 288 transmigration. This process most likely generated mechanical BM stress that contributes to its 289 disruption and facilitating invasive transition. 290 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint 291 Figure 4: Ras activates cellular cortical actin dynamics to mechanically stress the BM scaffold 292 (A) Time-lapse series of HRason spheroid motion on a 16 kPa substrate. The white asterisk marks a retractive actin 293 cortex movement; white arrowheads indicate reinforced cortical actin patches. Local BM disruption occurred at 294 5.5 hours after HRas activation (white line). BM is visualized with Alexa Fluor 488-conjugated laminin-332 295 antibody; F-actin is labeled with SiR-actin live-cell dye. Of note, a slight channel shift was caused by cell 296 movements during image acquisition without affecting data interpretation. For complete image series see 297 Supplementary Video 2. (B) Time-course quantification (0 – 12 hours) of spheroid compactness at the cell–BM–298 matrix interface. MIPs of confocal image stacks (laminin-332 signal) were used for shape analysis in HRas on and 299 HRasoff groups. (C) Scatter plot compares the changes of compactness for HRas on and HRas off spheroids 300 (t = 0 hours vs. t = individual time points of invasion onset) and for the non-invasive fraction of HRas on samples 301 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint (t =   0 hour vs. t= 12 hours). (n ≥ 15 spheroids from ≥ 3 experiments). Bars and whiskers: median and 95% CI. 302 The scheme illustrates the observed increase (smooth BM) or decrease (wrinkled BM) of spheroids compactness. 303 (D) Representative confocal stack used to reconstruct a 3D BM shell (10 µm height) from laminin-332 signals. 304 Within the volume of this 3D mask, signal intensities of laminin-332 and collagen IV were quantified (see 305

for detailed information). (E) Top-view on 3D-masked BM intensities (of laminin-332) in 306 representative HRason versus control spheroids. (F) Scatter plots summarize the quantitative signal intensities BM 307 protein intensities (laminin-332 and collagen IV) after 2 hours OHT-treatment (n  ≥  30 spheroids from ≥2 308 independent experiments; bars and whiskers: mean and 95% CI). Kruskal-Wallis test with Dunn’s multiple 309 comparison was performed for data in C. Shapiro–Wilk test of normality and unpaired t-test was performed for 310 data in F; n.s.: p > 0.05; *: p ≤ 0.05; **: p ≤ 0.01; ***: p ≤ 0.001; ****: p ≤ 0.0001. Scale bars: 5 µm (A); 20 µm 311 (B, D, E). Position of focal plane(s) used for imaging and analyses is indicated by red bar/rectangle. 312 313 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint HRas–Src–cortactin signaling drives cytoskeletal remodeling and BM invasion 314 To identify signaling events linking oncogenic HRas activity to cytoskeletal reorganization and 315 BM disruption, we profiled transcript levels and kinase activity downstream of HRas. A 316 targeted RT-qPCR array was used to assess cytoskeleton-related gene expression. Among the 317 analyzed transcripts, CDK5R1 (p35) and cortactin (CTTN) were upregulated (1.35-and 1.42-318 fold, respectively, Fig. 5A; Supplementary Table 1). Despite higher statistical significance for 319 CDK5R1, we prioritized cortactin based on its established role in actin organization 40, and its 320 consistency with the observed HRas-induced cytoskeletal phenotype. Because cortactin activity 321 is regulated by phosphorylation 41, we next assessed upstream kinase activation to identify 322 potential regulators of cortactin in HRason spheroids. 323 We profiled phosphotyrosine (PTK) and serine/threonine kinase (STK) activities using a 324 peptide-based microarray. Several kinases were differentially regulated in HRas on spheroids 325 compared to controls (Fig. 5B; Supplementary Table 2). EPHB4 and AURKA emerged as 326 candidate mediators of HRas-dependent invasion 42-45. However, inhibition of either kinase did 327 not affect invasion, with HRas on spheroids maintained their rapid invasion outcome (median 328 onset: 5 hours for EPHB4 and 4 hours for AURKA inhibition, Supplementary Fig.S5). 329 Integrated transcriptome and kinome analyses revealed increased Src expression and activity. 330 Src is a known effector of oncogenic Ras signaling 46-48 and regulates cortactin phosphorylation 331 41. We therefore examined the HRas–Src–cortactin axis in detail: Cortactin protein levels and 332 localization were compared between HRason and control spheroids. In both conditions, cortactin 333 localized predominantly to the basal cell layer within cortical actin networks adjacent to the 334 BM (Fig. 5C i and ii). In HRas on spheroids, cortactin accumulated in reinforced cortical actin 335 patches (Fig. 5C ii, white arrowhead). 336 Strikingly, these patches localized adjacent to sites of BM disruption (Fig. 5C ii, asterisk). In 337 contrast, HRasoff spheroids lacked this co-localization pattern and BM damage (Fig. 5C i). To 338 assess Src-dependent cortactin activation, we stained for pTyr421-cortactin 41: pTyr421-339 cortactin co-localized with thickened cortical actin (Fig. 6G). These data link Src activation to 340 cortactin phosphorylation at reinforced cortical actin sites. At later stages, HRason cells formed 341 prominent bundles of actin MS (cf. Fig. 2F). These protrusions penetrated the BM, forming 342 transmigration paths (Fig. 5D, asterisk). Cortactin was enriched within these BM-penetrating 343 structures (Fig. 5E and F). 344 Together, these findings revealed a spatiotemporal coupling of HRas activation, pTyr421-345 cortactin–mediated actin reinforcement, and BM disruption. 346 347 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint 348 Figure 5: HRas rewires Src and cortactin signaling to reinforce the actin cortex at BM rupture sites. 349 (A) Volcano plot compared the expression of 84 genes related to cytoskeletal regulation measured with a qRT-350 PCR array. Gene expression was analyzed after 4.5 hours of HRas activation on tumor stiffness (16 kPa). (B) 351 Profiling analysis of phospho-tyrosine-kinase (RTK) and serine-threonine-kinase (STK) activation depending on 352 HRas activation retrieved from kinase-substrate microarrays. Kinase rank: decreasing p-values of change in kinase 353 activity. Mean kinase statistic: color coded up (1)- and down (-1)-regulation. Kinase regulation was compared after 354 1 hour of OHT to EtOH treatment (16 kPa substrates). (C) Immunostained and fixed spheroids to compare the 355 actin cytoskeleton (F-actin, magenta), cortactin (green) and BM scaffold (collagen IV, yellow,) depending on 356 HRas activation. White arrowhead indicates reinforced cortical actin with cortactin accumulation; white asterisk 357 marks a site of BM rupture. (D) Detailed view of a HRason spheroid highlights a patch of reinforced actin, cortactin 358 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint incorporation and MS protrusion bundles within a BM tunnel for cell transmigration. The white asterisk marks the 359 ruptured BM site. (E) Image shows blue-outlined inset in (D) of actin-cortactin-protrusion (purple). Co-360 localization was calculated in Imaris (see Materials and Methods for detailed information). (F) BM surface 361 reconstruction (yellow) covering the orange-outlined inset in (D). On the right, top view along dashed line on the 362 protrusion bundle through the BM hole. (G) Representative image of pronounced pTyr421 localization within the 363 thickened actin cortex areas. (See Supplementary Fig. S6 for full image series with single channel display). Scale 364 bars: 20 µm (C), 5 µm (D, F and G). Position of focal plane(s) used for imaging and analyses is indicated by red 365 bar/rectangle. 366 367 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint HRas drives cortical triplet formation and myosin I-mediated BM rupture and invasion 368 HRas induced co-localization of reinforced cortical actin, phosphorylated cortactin, and local 369 BM rupture. We termed this pattern “cortical triplet” (CT). We developed an image analysis 370 pipeline with detection checkpoints to quantify CT formation and classify HRas-driven 371 invasion events (Fig. 6A and B). This analysis yielded CT coverage across the basal cell layer 372 (see 3D masks, Fig. 6A 4). 373 CT coverage was significantly increased in HRas on spheroids compared to controls (Fig. 6C). 374 Given that Src phosphorylates cortactin at Tyr421 to promote actin assembly 49 (see also 375 Fig. 5G), we inhibited Src activity. Src inhibition (PP2) reduced CT formation to control levels 376 (Fig. 6C). Quantitative analysis revealed a 1.6-fold increase in CT coverage in HRas on 377 spheroids, which was abolished upon Src inhibition (Fig. 6D). Consistently, Src inhibition 378 reduced both actin reinforcement and cortactin co-localization events (Supplementary Fig. S7). 379 We next tested whether reinforced cortical actin patches alter cortical tension to drive BM 380 rupture. To this end, reinforced cortical actin regions in living HRas on spheroids were locally 381 cut using laser-assisted nanosurgery. Local cortical tension was quantified by comparing 382 ablation-induced tangential retraction of actin fluorescence between thickened and thin cortical 383 regions (Fig. 6E and F). Reinforced regions retracted faster and over greater distances than 384 control regions (Fig. 6E; see also Supplemental Videos S3 and S4). Accordingly, reinforced 385 regions exhibited more pronounced ruptures (Fig. 6G). At 1 s post-ablation, retraction velocity 386 was 1.9-fold higher in reinforced regions, with 50% of values exceeding the maximum observed 387 in control (Fig. 6H and 6G, red lines). Faster actin retraction has been previously shown to 388 indicate elevated local cortical tension 50, which was also evident at reinforced actin sites in 389 HRas-induced spheroids. 390 Consistent with elevated cortical tension, we inhibited myosin I activity in HRas on spheroids. 391 Myosin I links cortical actin to the plasma membrane 51 and may transmit disruptive forces to 392 the BM. Myosin I inhibition markedly reduced invasion (50%; median onset: 9 hours), 393 approaching HRasoff levels (Fig. 6I and J). 394 We next tested whether CT formation directly contributes to invasion. Src inhibition (8 hours) 395 reduced invasion by >60% and delayed onset (~11 hours), comparable to HRas off controls 396 (Fig. 6I, grey box and J). However, overall invasion incidence remained partially reduced after 397 24 hours (87%). 398 Cortactin co-localizes with CTs (Fig. 5D) and promotes Arp2/3-mediated actin polymerization 399 52, 53. To assess the role of actin polymerization, we inhibited the Arp2/3 complex. Arp2/3 400 inhibition attenuated invasion, with delayed onset to 12 hours and reduced incidence to 65% 401 after 24 hours, similar to HRasoff controls. 402 Together, these data identify Src- and cortactin-dependent Arp2/3-mediated actin 403 polymerization, coupled to myosin I, as the primary mechanical driver of BM disruption in 404 HRas-activated spheroids, explaining the negligible impact of myosin II inhibition 405 406 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint 407 408 Figure 6: Cortical triplets and myosin I are functionally linked with BM rupture and cell invasion. 409 (A) Schematic workflow illustrates the automatic detection of cortical triplets (CTs) within the basal cell layer of 410 spheroids (fixed and stained after 3.5 hours of OHT or EtOH treatment, 16 kPa substrate): Sequential analyses of 411 Confocal image z-stacks (0) for the (1) detection of reinforced cortical actin sites being (2) co-localized with 412 cortactin signal. Co-localization negative objects were excluded. (3) Positive evaluated objects were checked for 413 co-appearance of faint/absent BM signal. Red outline: co-localization mask. (4) Objects that fulfilled all three 414 criteria were classified as CT. 3D masks serve to calculate the overall CT coverage in a BM volume of 25 µm 415 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint height. (B) Micrographs of a representative spheroid show steps 1 to 4 of the CT detection routine (for detailed 416 information on object detection see Supplemental Methods). (C) Representative 3D masks show CT-coverage of 417 the basal cell layer at the CBS-interface, depending on HRas activation and Src inhibition. (D) Scatter plot 418 summarizes the CT-coverage of individual spheroids (n ≥ 46 for each sample condition, from 3 independent 419 experiments) displayed with median and 95% CI. Representative cortex laser ablation of HRason spheroids at non-420 reinforced control sites (E) and CT-associated reinforced actin sites (F). Actin loss was monitored in 12 µm sectors 421 centered on the ablation site. Kymographs show fluorescence intensity normalized to pre-ablation levels. A 422 decrease in relative actin fluorescence intensity <0.5 was defined as rupture. (G) Binary kymographs display 423 sectors with median relative fluorescence intensity <0.5 (black; n = 24 control and n = 30 reinforced sites from 3 424 independent experiments). Red lines indicate median rupture width at 1 s post-ablation; corresponding retraction 425 velocities are shown in (H) (scatter with median ± 95% CI). (I) Cumulative distribution of BM disruption time, 426 depending on HRas activation and combined with myosin I, Src kinase or Arp2/3 inhibition (n ≥ 33 spheroids, per 427 condition from ≥ 2 independent experiments). (J) Individual events of BM disruption onset over time for the 428 samples analyzed in (I) displayed with median and 95% CI. Kruskal-Wallis test with Dunn’s multiple comparison 429 test was performed for the data in (D and F); Mann-Whitney test was performed for the data in (H): n.s.: p > 0.05; 430 *: p ≤ 0.05; **: p ≤ 0.01; ***: p ≤ 0.001; ****: p ≤ 0.0001. Scale bars: 20 µm. Position of focal planes used for 431 imaging and analyses is indicated by red rectangle. 432 433 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint HRas–cortactin–Arp2/3 axis predicts poor outcome in HRas-activated breast cancer 434 Given the strong link between cortactin overexpression and phosphorylation, its cellular 435 redistribution, and HRas-driven invasion, we evaluated the clinical relevance of this signaling 436 axis using breast cancer data from “The Cancer Genome Atlas (TCGA, National Institute of 437 Health)”. We stratified tumors by HRAS mutations and copy number variation, either “wild-438 type / loss” (HRas-/-, n=667) or “mutated / amplified” (HRas+/+, n=95). The latter was indicative 439 of elevated HRas activity in HRason spheroids. In the HRAS+/+ subgroup, high cortactin (CTTN) 440 gene expression correlated with worse recurrence-free survival (RFS) (Fig. 7B). In contrast, in 441 the HRas -/- subgroup, elevated CTTN levels were associated with favorable RFS (Fig. 7A). 442 Applying the same stratification to ARP2/3 complex subunits revealed that high expression of 443 the three subunits ARPC1B, ARPC2 and ARPC4 significantly predicted poorer RFS in 444 HRas+/+ tumors, whereas no adverse effect appeared in the HRas -/- group 445 (Supplementary Fig. S8). Since the functional Arp2/3 complex requires all of its seven subunits 446 54, 55, these findings point to Arp2/3-mediated actin polymerization within the cell cortex as a 447 key driver of invasiveness in HRas hyperactivated breast cancer. Collectively, the data 448 highlighted a clinically relevant role for the HRas–cortactin–Arp2/3 axis in breast cancer 449 invasion and prognosis. 450 451 452 Figure 7: High cortactin and HRas gene expression predict poor survival of breast cancer patients. 453 Univariate Kaplan-Meier curves of recurrence-free survival (RFS) of breast cancer patients stratified by HRas 454 status in wild-type / loss: HRas-/- (A) or mutated / amplified: HRas+/+ (B) and cortactin (CTTN) mRNA expression 455 (low or high). 456 457 458 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint

459 This study elucidated the poorly understood downstream signaling cascades that drive cellular 460 programs of HRas-induced breast cancer invasion. It is characterized by locally restricted 461 breakdown of the BM barrier, followed by cell transmigration into the TME. Our findings 462 underline the remarkable context specificity of Ras GTPases, which act as critical molecular 463 switches for cell fate decisions. Thereby, different Ras family members orchestrate diverse 464 signaling pathways that control homeostatic cell existence but also malignant progression 29, 56. 465 However, the underlying molecular mechanisms by which HRas coordinates the transition of 466 cells from in situ to invasive breast carcinomas remained unclear. 467 To address early steps of metastatic breast cancer progression, we used MCF10A-based breast 468 cell spheroids as simplified models of breast gland microtissue architecture. Without oncogenic 469 HRas, these spheroids are well-established and appreciated for studying both normal breast 470 gland morphogenesis and tumorigenic cell transition 57. A significant advantage of MCF10A 471 spheroids is their ability to build and maintain basoapically polarized architectures 58. MCF10A 472 spheroids endogenously produce and actively assemble a basement membrane (BM) scaffold 473 that forms a physiologically meaningful mechanical force barrier 8. Previous work further 474 demonstrated that invasive breast cells stress and disrupt the BM barrier, physically and 475 proteolytically, to transmigrate into the surrounding TME 5, 14. 476 Since our spheroid model lacks any accessory myoepithelial cells or TME-associated stromal 477 cells, our study strictly focused on elucidating bidirectional cell-ECM signal processing driven 478 by oncogenic HRas signaling. We used MCF10A-based breast spheroids with inducible HRas 479 downstream signaling 33. This 3D spheroid model resembled very early stages of oncogenic 480 HRas transformation and revealed how morphologically normal, yet Ras hyperactivated 481 epithelial cells initiate invasive behavior in response to matrix stiffening. 482 Tumor-associated ECM stiffening is a key driver of epithelial invasion, including in non-483 transformed breast spheroids 5, 59, 60 . Remarkably, we found that oncogenic HRas activation 484 partially bypasses this tumor-specific ECM cue, inducing effective invasive transitions in 485 physiologically compliant, hence, healthy-like environment (cf. Fig. 1D). This finding contrasts 486 with the inherently tumor-suppressive effects of normal matrix compliance on breast epithelial 487 cells 5, 14, 61. However, our finding is consistent with prior reports of HRas-driven invasion under 488 compliant conditions 62, highlighting its capacity to override invasion-protective 489 mechanotransduction cues. Conversely, tumor-like ECM stiffness strongly amplified HRas-490 driven invasion and accelerated its onset (cf. Fig. 2D). 491 Previous studies showed that Ras hyperactivation during early morphogenesis induces EMT 492 over several days in poorly polarized spheroids 62, 63. In contrast, HRas activation in fully 493 polarized spheroids triggered a rapid invasive transition within hours. To our knowledge, such 494 rapid HRas-driven invasion dynamics have not been previously reported. These findings 495 highlight the potent effect of temporally controlled HRas activation in differentiated, non-496 tumorigenic epithelium. In vivo, ECM stiffening occurs not only during malignant progression 497 but also in benign fibrocystic remodeling 64. These findings raise the possibility that oncogenic 498 HRas activity may promote early invasion in otherwise histologically unsuspicious breast tissue 499 65-67. 500 501 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint At the cellular level, our findings indicated altered mechanotransductive responses associated 502 with HRas activation. HRas on cells developed actin-rich microspikes (MS) at the BM-ECM 503 interface (cf. Fig. 3A), consistent with their described role as dynamic mechanosensory units in 504 MCF10A spheroids: without oncogenic HRas signaling, MS have been shown to convert into 505 contractile stress fibers (SF) that associate with an EMT-like phenotype, increased physical 506 BM-stress and disruption 23. In contrast, HRason cells retained MS morphology without forming 507 SFs. This new finding suggests the engagement of an alternative, SF-independent 508 mechanotransductive pathway. However, the exact mechanism of HRas-modulated ECM-509 sensing remained elusive and needs further in-depth investigation. 510 These observations prompted us to further dissect the cellular strategies underlying HRas-511 mediated invasion: SF-lack coincided with the absence of increased non-muscle myosin II-512 driven cell contractility – a cellular force mode typically linked to mechanical BM stress, 513 disruption and cell invasion 23, 68. Although such forces are usually amplified in response to 514 pathological ECM stiffening 14, 38, TFM revealed only marginal contractile activity in HRas on 515 spheroids even on tumor stiffness, insufficient to explain their high invasiveness. This 516 surprising finding highlighted the limited role of SF-mediated contractility and bulk traction 517 forces in HRas-induced BM invasion and suggests the involvement of a non-canonical, non-518 contractile BM breaching mechanism. 519 In addition to actomyosin contractility, matrix metalloproteinase (MMP)–mediated BM 520 degradation hallmarks epithelial invasion 69, 70. However, pharmacological inhibition of key 521 pro-invasive MMPs did not impair HRas-driven invasion 39, 69 and expression of the cell 522 membrane-anchored MT1-MMP (MMP14) remained unchanged upon HRas activation. This 523 contrasts with the MMP dependency observed under oncogenic EGFR signaling in the MCF10 524 wild-type spheroid model 5. Recent work reported that oncogenic Ras induces steady gradual 525 epithelial disintegration via hepsin-mediated laminin-332 degradation in soft matrix 526 environments 71. The highly effective invasion onset of HRas on spheroids on stiff substrates 527 suggested that serine proteases like hepsin could have contributed but are not the primary 528 mechanism. Our data demonstrated the contextual activation of distinct HRas invasion 529 programs. 530 In search of the still unknown invasion mechanism, we examined the CBS-interface and 531 discovered prominent cortical triplets (CTs) in HRason cells that most likely started the invasive 532 transition. These actin-enriched patches were hallmarks of invasion and characterized by 533 reinforced cortical actin, high cortactin accumulation, and localized BM disruption. Consistent 534 with cortactin incorporation at the basolateral cell cortex, CTTN gene expression was 535 substantially upregulated by HRas signaling. Cortactin is known to promote breast cancer cell 536 migration 72, and invadopodia formation 9, 73. Typical invadopodia are finger-shaped actin-rich 537 cell protrusions with MMP14-dependent proteolytic activity that widen collagen pores to 538 facilitate invasion 10, 74. However, CT-positive HRas on cells did not form such structures, and 539 invasion proceeded independent of MMP activity. Instead, HRason cells assembled CT patches 540 at basolateral cell surfaces. 541 542 Our data supported an HRas-driven invasion mechanism in which reinforced cortical actin 543 provides mechanical input for BM penetration, independent of MMP proteolysis and 544 invadopodia activity. Reinforced cortical actin coincided with dynamic pushing and pulling 545 motions, spheroid compaction, and progressive BM densification (cf. Fig. 4), shape changes 546 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint indicative of mechanical stress accumulation at the BM. While previous work demonstrated 547 that invasive breast spheroids can exert substantial BM-stressing actomyosin II forces 5, here 548 TFM revealed only low-level contractile stresses with minimal substrate deformations (strain 549 energy ~10 fJ) across all conditions and independent of HRas status. These values fell within 550 the range observed in non-invasive breast spheroids (1–32 fJ) (7) and were therefore 551 insufficient to account for the pronounced invasive phenotype. This suggests that HRas-driven 552 invasion relies on a distinct mode of force generation. Consistently, focal disruption of 553 reinforced cortical actin structures using laser-assisted nanosurgery demonstrated locally 554 elevated cortical tension at these sites (cf. Fig. 6H). Based on the curvature-dependent coupling 555 between cortical tension and outward pressure 75 this elevated tension is expected to generate 556 localized outward-directed forces at the CBS interface. Notably, the contractile forces generated 557 by the actin cortex are balanced by the BM. Therefore, little or no lateral stress is transmitted 558 to the substrate which is in line with our TFM results. 559 We identified Arp2/3-mediated actin polymerization and myosin I activity as key drivers of 560 localized pro-invasive forces. Mechanistically, HRas engages Src kinase to promote CT 561 formation. Src kinase, through phosphorylation of cortactin at Tyr421, likely links HRas 562 signaling to Arp2/3-dependent cytoskeletal remodeling at sites of BM disruption, where 563 cortactin co-localizes with reinforced cortical actin 49, 53, 76. 564 Arp2/3-driven actin branching promotes filament network densification, mechanical 565 reinforcement 77, and enhanced protrusive capacity of migrating cells 78 and has been shown to 566 mediate BM disruption independently of MMPs or actomyosin II in C. elegans embryogenesis 567 79. Recent work further demonstrated that myosin I synergizes with Arp2/3 to amplify force 568 generation in branched actin networks, likely by reorganizing filament architecture into a 569 mechanically more efficient configuration 80. Consistently, inhibition of either pathway reduced 570 invasion of HRason cells to levels comparable to HRas off controls, identifying Src, Arp2/3, and 571 myosin I as key effectors of HRas-driven BM disruption and potential therapeutic targets. 572 Whether elevated cortical tension and actin polymerization-driven force generation act 573 interdependently or represent parallel mechanical inputs to BM disruption remains to be 574 determined. 575 Together, these data reveal a coordinated downstream signaling cascade driving BM disruption: 576 (1) oncogenic HRas activates Src kinase, which (2) phosphorylates the actin binding protein 577 cortactin at Tyr421, leading to (3) Arp2/3 stabilization, (4) enhanced cortical actin branching 578 and polymerization forces. Ultimately, (5) these myosin I-synergized forces foster BM 579 disruption and invasive transition. Although additional candidates such as ephrin-B4 and aurora 580 kinase A were identified by kinase profiling, their inhibition had no functional outcome, 581 underscoring the specificity of the HRas–Src–cortactin–Arp2/3-myosin I axis for cell invasion. 582 These findings highlight that oncogenic Ras signaling integrates diverse pathways with context-583 dependent cellular consequences 81, 82. 584 To evaluate the translational relevance of our mechanistic findings, we assessed clinical breast 585 cancer datasets. Given the well-established context specificity of HRas signaling, it was critical 586 to determine whether this invasion-related pathway also manifests in patient tumors and 587 correlates with clinical outcome. HRas overexpression is recognized as an independent marker 588 of poor prognosis 83 and activating HRas mutations are associated with adverse survival in 589 breast cancer patients. We therefore stratified tumor cases based on HRas copy number and 590 mutation status and analyzed co-expression patterns with cortactin, a marker of metastatic 591 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint potential at both the transcript and protein levels as well 84, 85. Although the low prevalence of 592 HRas mutations in breast cancer 2 limited the size of available patient cohorts, our in silico 593 analysis revealed a significant association between high co-expression of HRas-cortactin and 594 reduced recurrence-free survival (RFS), which fits our in vitro observation of HRas-induced 595 cortactin transcription. Conversely, tumors with high cortactin but low HRas expression 596 indicated improved RFS. This suggests a directionally dependent functional interaction 597 between these two effectors in tumor progression. Further supporting evidence is that the 598 amplification of ARP2/3 complex subunits was associated with poor prognosis in the HRas-599 mutated patient subgroup. Among the seven ARP2/3 subunits 55, 86, two structural core elements 600 (ARPC2 and ARPC4) 87 and one key mediator of actin nucleation and assembling (ARPC1B) 601 88, 89 , each correlated significantly with reduced recurrence-free survival. These findings 602 strengthen the link between the HRas–cortactin–Arp2/3 signaling axis and breast cancer 603 progression, supporting its relevance for patient outcome. 604 605

of the study 606 While 3D spheroid models enabled functional dissection of the HRas–Src–cortactin–Arp2/3 607 axis, our study is inherently based on a cell culture model, whose simplified architecture cannot 608 fully recapitulate the complexity of HRas-mutated or -hyperactivated tumors in vivo. This 609 reflects a fundamental technical limitation, as precise spatiotemporal control of HRas activity- 610 required to dissect early oncogenic events- cannot currently be achieved in vivo: Importantly, 611 breast cell spheroids enabled causal dissection of early oncogenic processes and their associated 612 cytoskeletal dynamics at a resolution not experimentally accessible using for instance rodent 613 models. Nevertheless, the absence of accessory myoepithelial cells and other TME-associated 614 stromal components precludes to resemble reciprocal epithelial–stromal feedback loops that 615 may influence epithelial cell behavior 90. Future studies incorporating such heterotypic 616 interactions will be important to further refine our understanding of HRas-driven breast cancer 617 progression. Accordingly, validation in complementary in vivo systems, such as inducible 618 HRason/off rodent xenografts, will be an important step to extend and validate these mechanistic 619 and prognostic insights. 620 621 Concluding remarks 622 Our findings reveal that oncogenic HRas signaling rewires mechanosensing of tumor 623 microenvironmental stiffness, cell mechanics and actin forces to drive BM disruption and early 624 cell invasion in phenotypical non-malignant breast epithelial cells. We show that this early-625 stage invasive transition is mechanistically dependent on Src kinase phosphorylated Tyr421 626 cortactin, Arp2/3, actin polymerization and myosin I forces. Intriguingly, these cellular 627 switches could represent promising therapeutic targets to counteract HRas-driven invasion at 628 early stages. While BM-covered breast spheroids simplify in vivo complexity, the correlation 629 of HRas downstream effector expression with reduced recurrence-free survival in breast cancer 630 cohorts supports the clinical relevance of this newly discovered oncogenic signaling pathway 631 (see Fig. 8). Finally, these results deepen our understanding of how spontaneous oncogene 632 activation perturbs mechanobiological homeostasis during early breast cancer progression. 633 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint 634 Figure 8: Oncogenic HRas rewires myosin I and actin polymerization forces to drive BM invasion. 635 (1) Oncogenic HRas activation in breast epithelial cells reprograms epithelial responses to TME stiffening, 636 resulting in cell compaction, BM densification and elevated cortical tension. (2) HRas activated cells initiate 637 invasion via a Src–cortactin signaling axis: Src phosphorylates cortactin at Tyr421, which recruits and stabilizes 638 the Arp2/3 complex at the basolateral cortex to establish a reinforced actin-rich invasion front. Arp2/3-dependent 639 actin polymerization together with myosin I-generated forces deform and ruptures the BM barrier, enabling cell 640 transmigration. HRas-dependent upregulation of cortactin expression fosters this invasion program. 641 Pharmacological inhibition of Src, Arp2/3, or myosin I restore a non-invasive phenotype. Clinically, elevated 642 expression of HRas effectors (cortactin and ARP2/3 subunits) correlates with reduced recurrence-free survival in 643 HRas-amplified or -mutated breast cancers. Schematics are not to scale. Created with BioRender.com 644 645 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint

646 Cell Maintenance 647 MCF10A wildtype cells (purchased from ATCC), MCF10A_ER:HRas G12V and 648 MCF10A_HRasG12V cells described in 34 and kindly provided by Buzz Baum 35 were maintained 649 in culture dishes under standard culture conditions (37 °C, 5% CO 2) in DMEM/F12 growth 650 medium (ThermoFisher Scientific) containing 5% horse serum (ThermoFisher Scientific) or 651 steroid hormone free horse serum (c.c.pro) (in all experiments from Fig. 3 onwards), 0.5 µg/mL 652 hydrocortisone, 100 ng/mL cholera toxin, 20 ng/mL EGF, 10 µg/mL insulin (all Sigma 653 Aldrich), 100 U/mL penicillin and 100 µg/mL streptomycin (both ThermoFisher Scientific). 654 For spheroid morphogenesis, assay media with adapted compositions were used (see below). 655 Spheroid morphogenesis and isolation from EHS matrix 656 Single cells were seeded on top of growth factor reduced EHS matrix (Geltrex, ThermoFisher 657 Scientific) and cultured for ten days as described in 23. MCF10A wildtype cells were cultivated 658 from day one to day nine and MCF10A cells with HRas constructs from day one to day six in 659 assay medium with 5 ng/mL EGF and 1% horse serum. Assay medium without EGF was used 660 for wildtype cells from day nine to day ten and for MCF10A cells with HRas constructs from 661 day six to day ten. For further analyses, spheroids were isolated from EHS matrix and washed 662 with ice-cold PBS and incubated in 2 mL ice-cold cell recovery solution (CRS) (BD 663 Biosciences) for 30 min (4 °C) to depolymerize the EHS matrix. For inhibition experiments, 664 inhibitors were added 30 min prior to isolation. Next, spheroids were washed with fresh EGF-665 free assay medium, picked under a stereo microscope and seeded onto 35 mm cell culture 666 dishes, either with glass bottoms or glass bottoms with a layer of silicone elastomer (see below). 667 Spheroids adhered to these EHS-coated substrates for 15 min (37 °C, 5% CO 2) and were 668 subsequently covered with 2 mL EGF-free assay medium. The time point of adding media was 669 defined as assay start. 670 Preparation of elastomeric substrates and functionalization with EHS coating 671 Spheroids were transferred on an 80 µm thick layer of cross-linked PDMS silicone elastomer 672 substrate (Sylgard 184, Dow Corning) with a Young’s modulus of 16 kPa. Preparation of these 673 substrates was done as described in 91. In brief, layer thickness was set by spin coating on 674 100 µm thin cover slips (Cover Slip, Ø22 mm, #0, Menzel-Gläser). Silicone-coated cover slips 675 were glued to the bottom of 3.5 cm Petri dishes to cover predrilled 1.8 cm holes and cross linked 676 for 16 hours at 60 °C. Young’s modulus silicone elastomers was determined as described 677 previously 92. For TFM, fluorescent beads (FluoSpheres, carboxylate-modified, 0.2 µm, red, 678 ThermoFisher Scientific) were immobilized on top of elastomeric substrates, as described 679 elsewhere 93. Before spheroid transfer, elastomeric substrates and glass substrates (Cover Slip, 680 24 x 24 mm, HP, Menzel-Gläser) were functionalized for adhesion with 600 µL of non-gelling 681 EHS protein solution (20 µg/mL) in ice-cold PBS for 18 hours at 4 °C. 682 683 684 Biochemical treatments 685 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint Spheroids were transferred on elastomeric substrates and incubated with EGF-free assay 686 medium after 15 min. For activation of HRas, a final concentration of 1 µM OHT (H7904, 687 Sigma-Aldrich; dissolved in EtOH) was used (final EtOH concentration = 0.002%). For 688 inhibition of cellular myosin II activity, a final concentration of 25 µM blebbistatin (B0560, 689 Sigma-Aldrich; dissolved in DMSO) was used (final DMSO concentration = 0.29%). For 690 inhibition of matrix metalloproteinases, a final concentration of 20 µM marimastat (M2699, 691 Sigma-Aldrich; dissolved in DMSO) was used (final DMSO concentration = 0.03%). For 692 inhibition of ephrin type-B receptor 4, a final concentration of 5 µM NVP-BHG712 (HY-693 13258A, MedChemExpress; dissolved in DMSO) was used (final DMSO 694 concentration = 0.01%). For inhibition of aurora kinase A, a final concentration of 5 µM MK-695 5108 (HY-13252, MedChemExpress; dissolved in DMSO) was used (final DMSO 696 concentration = 0.025%). For inhibition of myosin I, a final concentration of 20 µM 697 Pentachloropseudilin (HY-115669, MedChemExpress; dissolved in DMSO) was used (final 698 DMSO concentration = 0.007%). For inhibition of Src kinase, a final concentration of 5 µM 699 PP2 (1407, Tocris; dissolved in EtOH) was used (final EtOH concentration = 0.05%). For 700 inhibition of Arp2/3 complex, a final concentration of 40 µM CK-666 (3950, Tocris; dissolved 701 in EtOH) was used (final EtOH concentration = 0.04%). 702 Immunofluorescent staining 703 Spheroids were fixed for 20 min with 3.7% paraformaldehyde in cytoskeleton-buffer (CB: 704 5 mM EGTA, 5 mM glucose, 10 mM MES, 5 mM MgCl2, 150 mM NaCl, 1 g/L streptomycin; 705 all Sigma-Aldrich), washed for 5 min with 20 mM glycine in CB (Sigma-Aldrich), and 706 permeabilized with 1% Triton-X 100 in CB (Sigma-Aldrich) at RT. After one washing step 707 with CB, non-specific antibody binding was blocked with 5% skim milk powder (Sigma-708 Aldrich) and 1% AffiniPure F(ab’) 2 fragment goat anti-mouse IgG (115-006-006, Jackson 709 ImmunoResearch, West Grove, PA, USA) in CB for 2 hours at RT. Spheroids were incubated 710 overnight at 4 °C with primary antibodies anti-collagen IV (1:500, abcam, ab6586), anti-711 GM130 (1:500, BD Biosciences, 610822, clone 35), Alexa Fluor 488-conjugated anti-laminin-712 5 antibody (1:1000, Sigma-Aldrich, MAB19562X; clone D4B5), anti-cortactin (p80/85) (1:300, 713 Sigma-Aldrich, 05-180-I, clone 4F11), anti-phospho-Tyr421-cortactin (1:300, ThermoFisher, 714 44-854G), anti-pERK (Phospho-p44/42, Thr202/Tyr204) (1:200, Cell Signaling, 9101S), anti-715 MT1-MMP (1:200, Abnova, MAB12762, clone 133CT15.10.5.1) diluted in 1% skim milk 716 powder in CB followed by three washing steps with CB. 717 Conjugated secondary antibodies chicken anti-rabbit IgG Alexa Fluor™ 488 (ThermoFisher 718 Scientific, A21441), goat anti-rabbit IgG Alexa Fluor™ 488 (ThermoFisher Scientific, 719 A11008), goat anti-rabbit IgG Alexa Fluor™ 488 (ThermoFisher Scientific, A31556), donkey 720 anti-rabbit IgG Alexa Fluor™ 546 (ThermoFisher Scientific, A10040), donkey anti-mouse IgG 721 Alexa Fluor™ 546 (ThermoFisher Scientific, A10036), were diluted (1:1000) in 1% skim milk 722 powder in CB and incubated for 1 h at RT followed by two washing steps with CB. Phalloidin-723 Atto 633 (Sigma-Aldrich, 68825) or 488 (Sigma-Aldrich, 49409) labeling was done parallel to 724 secondary antibody administration (1:1000). Nuclei were stained with 1:1000 DAPI in CB 725 (ThermoFisher Scientific, NucBlue™ Fixed Cell ReadyProbes™) or 1:1000 DRAQ5™ in CB 726 (ThermoFisher Scientific, 62251) for 10 min at RT. To enable reliable signal quantification, 727 fixation, staining, and imaging were performed in an identical manner across experiments, 728 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint maintaining sequence and procedures for each step. For live cell imaging, spheroids were 729 stained in assay medium with a final concentration of 10 nM SiR-actin (Spirochrome, SC001, 730 solved in DMSO, final DMSO concentration = 1x10 -5%) for 2.5 hours, isolated from EHS 731 matrix and incubated with 1:100 Alexa Fluor 488-conjugated anti-laminin-5 antibody (Sigma-732 Aldrich, MAB19562X; clone D4B5) for 1 hour at 4 °C. Spheroids were then transferred onto 733 silicone elastomer substrates. 734 Confocal microscopy and image processing 735 Fixed and stained spheroids were imaged at room temperature, and live-cell imaging was 736 performed at 37 °C and 5% CO₂ using an inverted confocal laser scanning microscope 737 (LSM880 with Airyscan detector; Carl Zeiss). The system was equipped with a 40× LD C-738 Apochromat water immersion objective (NA 1.1; Zeiss) and standard laser lines (405, 488, 561, 739 and 633 nm) with appropriate filter sets. Images were acquired using (Fast) Airyscan detection 740 and processed by Airyscan processing in ZEN 2.3 Black (Carl Zeiss). Three-dimensional 741 visualizations were generated using PyVista (93) and Imaris 10.2 (Oxford Instruments). 742 Laser-assisted nanosurgery of cortical actin structures 743 Scanner-based laser ablation was performed using a nanosecond-pulsed 355 nm laser (UGA-744 42 Firefly/Caliburn DPSL-355/42/CLS; Rapp OptoElectronic) coupled to the LSM described 745 above. A mean output power of 8.4 mW was used. A 20x air objective (NA 0.8, Carl Zeiss) 746 was used for all ablation experiments. Spheroids were stained with live actin dye SiR-actin (see 747 section Immunocytochemistry) for 3 hours (parallel to OHT treatment). Visibly thickened 748 cortical regions were used as samples for reinforced actin sites while thin regions served as 749 control, non-reinforced sites. Ablation was performed using an unfilled rectangular region of 750 interest (42 x 3 pixels; 8.11 x 0.58 µm at the given image scale) oriented perpendicular to the 751 actin cortex (step size 5, 10 repetitions; sequence mode SysCon, Rapp OptoElectronic). Object 752 illumination and therefore ablation took 187 ms. Images were acquired every 200 ms in Fast-753 Airyscan mode. Imaging started ≥1.5 s prior to ablation for baseline calculation (see 754 Supplemental Information). 755 Live-cell imaging for invasion assay and TFM 756 Experiments were carried out at 37 °C and 5% CO 2 using an inverted microscope (Axio 757 Observer, Zeiss), equipped with an Axiocam 712 mono camera (Zeiss) and an EC Plan-758 Neofluar 40x oil immersion objective (PH3, NA 1.3, Zeiss). Cells were imaged in phase-759 contrast and fluorescent beads with a LED module (Colibri 7, Zeiss) at 548 nm and excitation 760 and emission filter settings of 538 – 562 nm and 570 – 640 nm, respectively. Time increment 761 of image acquisition was 20 min (image pixel size = 0.17 µm). Tangential substrate 762 deformations were visualized by tracking fluorescent beads to calculate strain energy, the 763 elastic energy of the cell-deformed substrate (see Supplemental Information). 764 Digital image processing of confocal laser scanning microscopy micrographs 765 All routines for analyses were developed in Python 3.12, including quantification of pERK, 766 MT1-MMP, BM and CT signal, of compactness and of actin after laser ablation, and are 767 described in detail in Supplemental Methods. 768 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint Kinome profiling analysis 769 Kinase profiles were determined using the PamChip® peptide tyrosine kinase (PTK) microarray 770 system or the PamChip ® Ser/Thr Kinase assay (STK) on PamStation ®12 (PamGene 771 International, ´s-Hertogenbosch, The Netherlands). Each PTK-PamChip ® array contained 196 772 individual phospho-sites(s) that are peptide sequences derived from substrates for tyrosine 773 kinases. Each STK-PamChip® array contained 144 individual phospho-site(s) that are peptide 774 sequences derived from substrates for Ser/Thr kinases. Each peptide on the PAMChip is a 775 15-amino acid sequence representing a putative endogenous phosphorylation site, which 776 functions as a kinase substrate. Sample preparation and phosphorylation detection are described 777 in detail in the Supplemental Methods section. 778 RT-qPCR analysis 779 Spheroids on 16 kPa elastomeric substrates, incubated 270 min with EtOH or OHT, were 780 mechanically pipetted in solution with assay medium (without EGF). Total RNA from 781 spheroids was isolated through column chromatography according to the manufacturer’s 782 protocol (QIAGEN, miRNeasy Tissue/Cells Advanced Mini Kit, 217604) and quantified using 783 a NanoDrop spectrophotometer (ThermoFisher). Total RNA samples were reverse transcribed 784 using RT2 First Strand Kit (QIAGEN, 330401) according to the manufacturer’s protocol. 785 Generated cDNA was used for qRT-PCR array with cytoskeleton regulators (QIAGEN, PAHS-786 088Z). PCR measurements were conducted on a QIAquant 96 station (QIAGEN) and analyzed 787 according to the array manufacturer’s instructions and their GeneGlobe online tool. A C T cut-788 off of 35 was used and normalization was done with two genes based on minimal difference of 789 geometric means across samples (<1 CT). 790 TCGA data and survival statistics 791 Publicly available breast cancer data from The Cancer Genome Atlas (TCGA) network (Breast 792 Invasive Carcinoma (Firehose Legacy) (1,108 samples)) were used to determine the clinical 793 impact of CTTN and ARP expression in context of genetic HRAS alterations comprising 794 transcriptomic (RNASeqV2 data: CTTN, ACTR2, ACTR3, ARPC1A, ARPC1B, ARPC2 , 795 ARPC3, ARPC4, ARPC5, ARPC5L ), genomic (HRAS mutation status and HRAS-associated 796 copy number variation) and clinical follow-up data of overall n=1,082 primary breast cancer 797 samples (https://gdac.broadinstitute.org/runs/stddata__2016_01_28/data/BRCA/20160128/). 798 Data was accessed by using the cBio Cancer Genomics Portal (http://cbioportal.org) 94 and 799 analyzed by using SPSS software version 29.0.0.0 (SPSS Inc., Chicago, USA). Survival curves 800 for recurrence free survival (RFS) were stratified by genetic HRAS alterations (copy number 801 variation and mutations), calculated using the Kaplan-Meier method with log-rank. RFS was 802 measured from surgery until relapse (local/distant) and was censored for patients without 803 evidence of tumor recurrence at the last follow-up date. 804 Figure preparation 805 Figure 8 was created in BioRender. Platz-baudin, E. (2026) https://BioRender.com/pl9dmm8 806 807 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint

808 1. G. Yin et al., Targeting small GTPases: emerging grasps on previously untamable 809 targets, pioneered by KRAS. Signal Transduction and Targeted Therapy 8, 212 (2023). 810 2. M. Galiè, RAS as Supporting Actor in Breast Cancer. Front Oncol 9, (2019). 811 3. M. J. Hoenerhoff et al., BMI1 cooperates with H-RAS to induce an aggressive breast 812 cancer phenotype with brain metastases. Oncogene 28, 3022-3032 (2009). 813 4. K. L. Wright et al., Ras Signaling Is a Key Determinant for Metastatic Dissemination 814 and Poor Survival of Luminal Breast Cancer Patients. Cancer Research 75, 4960-4972 815 (2015). 816 5. A. Gaiko-Shcherbak et al., Cell Force-Driven Basement Membrane Disruption Fuels 817 EGF- and Stiffness-Induced Invasive Cell Dissemination from Benign Breast Gland 818 Acini. Int J Mol Sci 22, (2021). 819 6. R. Sekiguchi, K. M. Yamada, in Current Topics in Developmental Biology, E. S. 820 Litscher, P. M. Wassarman, Eds. (Academic Press, 2018), vol. 130, pp. 143-191. 821 7. R. Jayadev, D. R. Sherwood, Basement membranes. Current Biology 27, R207-R211 822 (2017). 823 8. G. Fabris et al., Nanoscale Topography and Poroelastic Properties of Model Tissue 824 Breast Gland Basement Membranes. Biophysical Journal 115, 1770-1782 (2018). 825 9. S. Linder, P. Cervero, R. Eddy, J. Condeelis, Mechanisms and roles of podosomes and 826 invadopodia. Nature Reviews Molecular Cell Biology 24, 86-106 (2023). 827 10. D. A. Murphy, S. A. Courtneidge, The 'ins' and 'outs' of podosomes and invadopodia: 828 characteristics, formation and function. Nature Reviews Molecular Cell Biology 12, 413-829 426 (2011). 830 11. L. M. Coussens, B. Fingleton, L. M. Matrisian, Matrix Metalloproteinase Inhibitors and 831 Cancer—Trials and Tribulations. Science 295, 2387-2392 (2002). 832 12. J. L. Chang, O. Chaudhuri, Beyond proteases: Basement membrane mechanics and 833 cancer invasion. J Cell Biol 218, 2456-2469 (2019). 834 13. S. S. Nazari, A. D. Doyle, C. K. E. Bleck, K. M. Yamada, Long Prehensile Protrusions 835 Can Facilitate Cancer Cell Invasion through the Basement Membrane. Cells 12, 2474 836 (2023). 837 14. J. Chang et al., Cell volume expansion and local contractility drive collective invasion 838 of the basement membrane in breast cancer. Nature Materials, (2023). 839 15. D. E. Rassier, A. Månsson, Mechanisms of myosin II force generation: insights from 840 novel experimental techniques and approaches. Physiological Reviews 105, 1-93 (2025). 841 16. J. Robert-Paganin, O. Pylypenko, C. Kikuti, H. L. Sweeney, A. Houdusse, Force 842 Generation by Myosin Motors: A Structural Perspective. Chemical Reviews 120, 5-35 843 (2020). 844 17. L. Blanchoin, R. Boujemaa-Paterski, C. Sykes, J. Plastino, Actin Dynamics, 845 Architecture, and Mechanics in Cell Motility. Physiological Reviews 94, 235-263 846 (2014). 847 18. Roger J. Daly, Cortactin signalling and dynamic actin networks. Biochemical Journal 848 382, 13-25 (2004). 849 19. R. J. Davey, P. D. J. Moens, Profilin: many facets of a small protein. Biophysical Reviews 850 12, 827-849 (2020). 851 20. T. Koorman et al., Spatial collagen stiffening promotes collective breast cancer cell 852 invasion by reinforcing extracellular matrix alignment. Oncogene 41, 2458-2469 (2022). 853 21. P. P. Provenzano, D. R. Inman, K. W. Eliceiri, P. J. Keely, Matrix density-induced 854 mechanoregulation of breast cell phenotype, signaling and gene expression through a 855 FAK–ERK linkage. Oncogene 28, 4326-4343 (2009). 856 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint 22. M. W. Pickup et al., Stromally Derived Lysyl Oxidase Promotes Metastasis of 857 Transforming Growth Factor-β–Deficient Mouse Mammary Carcinomas. Cancer 858 Research 73, 5336-5346 (2013). 859 23. J. Eschenbruch et al., From Microspikes to Stress Fibers: Actin Remodeling in Breast 860 Acini Drives Myosin II-Mediated Basement Membrane Invasion. Cells 10, (2021). 861 24. O. Chaudhuri et al., Extracellular matrix stiffness and composition jointly regulate the 862 induction of malignant phenotypes in mammary epithelium. Nature Materials 13, 970-863 978 (2014). 864 25. V. Gkretsi, T. Stylianopoulos, Cell Adhesion and Matrix Stiffness: Coordinating Cancer 865 Cell Invasion and Metastasis. Front Oncol 8, (2018). 866 26. X. Li, J. Wang, Mechanical tumor microenvironment and transduction: cytoskeleton 867 mediates cancer cell invasion and metastasis. International Journal of Biological 868 Sciences 16, 2014-2028 (2020). 869 27. S. F. Ghannam, C. S. Rutland, C. Allegrucci, N. P. Mongan, E. Rakha, Defining invasion 870 in breast cancer: the role of basement membrane. Journal of Clinical Pathology 76, 11 871 (2023). 872 28. F. Broders-Bondon, T. H. Nguyen Ho-Bouldoires, M.-E. Fernandez-Sanchez, E. Farge, 873 Mechanotransduction in tumor progression: The dark side of the force. Journal of Cell 874 Biology 217, 1571-1587 (2018). 875 29. A. Nyga, S. Ganguli, H. K. Matthews, B. Baum, The role of RAS oncogenes in 876 controlling epithelial mechanics. Trends in Cell Biology 33, 60-69 (2023). 877 30. S. Piccolo, T. Panciera, P. Contessotto, M. Cordenonsi, YAP/TAZ as master regulators 878 in cancer: modulation, function and therapeutic approaches. Nature Cancer 4, 9-26 879 (2023). 880 31. J. Jung et al., p53 mitigates the effects of oncogenic HRAS in urothelial cells via the 881 repression of MCOLN1. iScience 24, 102701 (2021). 882 32. W. Huang et al., Regulatory networks in mechanotransduction reveal key genes in 883 promoting cancer cell stemness and proliferation. Oncogene 38, 6818-6834 (2019). 884 33. A. Nyga et al., Oncogenic RAS instructs morphological transformation of human 885 epithelia via differential tissue mechanics. Science Advances 7, (2021). 886 34. M. Molina-Arcas, D. C. Hancock, C. Sheridan, M. S. Kumar, J. Downward, Coordinate 887 Direct Input of Both KRAS and IGF1 Receptor to Activation of PI3 Kinase in KRAS-888 Mutant Lung Cancer. Cancer Discovery 3, 548-563 (2013). 889 35. H. K. Matthews et al., Oncogenic Signaling Alters Cell Shape and Mechanics to 890 Facilitate Cell Division under Confinement. Developmental Cell 52, 563-573 (2020). 891 36. S. Ganguli et al., Oncogenic Ras deregulates cell-substrate interactions during mitotic 892 rounding and respreading to alter cell division orientation. Current Biology 33, 2728-893 2741.e3 (2023). 894 37. F. Friedland et al., ECM-transmitted shear stress induces apoptotic cell extrusion in early 895 breast gland development. Frontiers in Cell and Developmental Biology 10, (2022). 896 38. M. J. Paszek et al., Tensional homeostasis and the malignant phenotype. Cancer Cell 8, 897 241-254 (2005). 898 39. J. Cathcart, A. Pulkoski-Gross, J. Cao, Targeting matrix metalloproteinases in cancer: 899 Bringing new life to old ideas. Genes & Diseases 2, 26-34 (2015). 900 40. K. C. Kirkbride, B. H. Sung, S. Sinha, A. M. Weaver, Cortactin: A multifunctional 901 regulator of cellular invasiveness. Cell Adhesion & Migration 5, 187-198 (2011). 902 41. M. Schnoor, T. E. Stradal, K. Rottner, Cortactin: Cell Functions of A Multifaceted Actin-903 Binding Protein. Trends in Cell Biology 28, 79-98 (2018). 904 42. Z. Xiao et al., EphB4 promotes or suppresses Ras/MEK/ERK pathway in a context-905 dependent manner. Cancer Biology & Therapy 13, 630-637 (2012). 906 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint 43. N. L. Lartey, M. van der Ent, R. Alonzo, D. Chen, P. D. King, A temporally-restricted 907 pattern of endothelial cell collagen 4 alpha 1 expression during embryonic development 908 determined with a novel knockin Col4a1-P2A-eGFP mouse line. genesis 62, e23539 909 (2024). 910 44. S. A. Wander et al., The Genomic Landscape of Intrinsic and Acquired Resistance to 911 Cyclin-Dependent Kinase 4/6 Inhibitors in Patients with Hormone Receptor–Positive 912 Metastatic Breast Cancer. Cancer Discovery 10, 1174-1193 (2020). 913 45. M. Peindl et al., EMT, Stemness, and Drug Resistance in Biological Context: A 3D 914 Tumor Tissue/In Silico Platform for Analysis of Combinatorial Treatment in NSCLC 915 with Aggressive KRAS-Biomarker Signatures. Cancers 14, (2022). 916 46. L. C. Kim, L. Song, E. B. Haura, Src kinases as therapeutic targets for cancer. Nature 917 Reviews Clinical Oncology 6, 587-595 (2009). 918 47. J. Luo et al., SRC kinase-mediated signaling pathways and targeted therapies in breast 919 cancer. Breast Cancer Research 24, 99 (2022). 920 48. C. L. C. Poon, A. M. Brumby, H. E. Richardson, Src Cooperates with Oncogenic Ras in 921 Tumourigenesis via the JNK and PI3K Pathways in Drosophila epithelial Tissue. Int J 922 Mol Sci 19, 1585 (2018). 923 49. S. Tehrani, N. Tomasevic, S. Weed, R. Sakowicz, J. A. Cooper, Src phosphorylation of 924 cortactin enhances actin assembly. Proceedings of the National Academy of Sciences 925 104, 11933-11938 (2007). 926 50. A. Saha et al., Determining Physical Properties of the Cell Cortex. Biophysical Journal 927 110, 1421-1429 (2016). 928 51. R. Nambiar, R. E. McConnell, M. J. Tyska, Control of cell membrane tension by myosin-929 I. Proceedings of the National Academy of Sciences 106, 11972-11977 (2009). 930 52. T. Liu et al., Cortactin stabilizes actin branches by bridging activated Arp2/3 to its 931 nucleated actin filament. Nature Structural & Molecular Biology 31, 801-809 (2024). 932 53. T. Uruno et al., Activation of Arp2/3 complex-mediated actin polymerization by 933 cortactin. Nature Cell Biology 3, 259-266 (2001). 934 54. M. D. Welch, A. H. DePace, S. Verma, A. Iwamatsu, T. J. Mitchison, The Human 935 Arp2/3 Complex Is Composed of Evolutionarily Conserved Subunits and Is Localized 936 to Cellular Regions of Dynamic Actin Filament Assembly. Journal of Cell Biology 138, 937 375-384 (1997). 938 55. I. Rouiller et al., The structural basis of actin filament branching by the Arp2/3 complex. 939 Journal of Cell Biology 180, 887-895 (2008). 940 56. D. Bar-Sagi, A. Hall, Ras and Rho GTPases: A Family Reunion. Cell 103, 227-238 941 (2000). 942 57. C. Hebner, V. M. Weaver, J. Debnath, Modeling Morphogenesis and Oncogenesis in 943 Three-Dimensional Breast Epithelial Cultures. Annual Review of Pathology: 944 Mechanisms of Disease 3, 313-339 (2008). 945 58. J. Debnath, S. K. Muthuswamy, J. S. Brugge, Morphogenesis and oncogenesis of MCF-946 10A mammary epithelial acini grown in three-dimensional basement membrane 947 cultures. Methods 30, 256-268 (2003). 948 59. F. Kai, H. Laklai, V. M. Weaver, Force Matters: Biomechanical Regulation of Cell 949 Invasion and Migration in Disease. Trends in Cell Biology 26, 486-497 (2016). 950 60. K. Beslmüller, R. R. de Mercado, G. H. Koenderink, E. H. J. Danen, T. Schmidt, Matrix 951 Stiffness Affects Spheroid Invasion, Collagen Remodeling, and Effective Reach of 952 Stress into ECM. Organoids 4, 11 (2025). 953 61. S. P. Carey, K. E. Martin, C. A. Reinhart-King, Three-dimensional collagen matrix 954 induces a mechanosensitive invasive epithelial phenotype. Scientific Reports 7, 42088 955 (2017). 956 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint 62. C. Plunkett et al., H-Ras Transformation of Mammary Epithelial Cells Induces ERK-957 Mediated Spreading on Low Stiffness Matrix. Advanced Healthcare Materials 9, 958 1901366 (2020). 959 63. S. C. Wei et al., Matrix stiffness drives epithelial–mesenchymal transition and tumour 960 metastasis through a TWIST1–G3BP2 mechanotransduction pathway. Nature Cell 961 Biology 17, 678-688 (2015). 962 64. W. A. Berg et al., Quantitative Maximum Shear-Wave Stiffness of Breast Masses as a 963 Predictor of Histopathologic Severity. American Journal of Roentgenology 205, 448-964 455 (2015). 965 65. Z. Abu-Rahmeh, W. Nseir, I. Naroditzky, Invasive ductal carcinoma within 966 fibroadenoma and lung metastases. Int J Gen Med 5, 19-21 (2012). 967 66. Y.-T. Wu et al., Breast cancer arising within fibroadenoma: collective analysis of case 968 reports in the literature and hints on treatment policy. World Journal of Surgical 969 Oncology 12, 335 (2014). 970 67. M. Y. Ho, V. Ng, J. Hing, Invasive carcinoma in recurrent breast fibroadenoma: a case 971 report and literature review. Annals of Breast Surgery 9, (2024). 972 68. K. Weißenbruch, R. Mayor, Actomyosin forces in cell migration: Moving beyond cell 973 body retraction. BioEssays 46, 2400055 (2024). 974 69. K. Hotary, X.-Y. Li, E. Allen, S. L. Stevens, S. J. Weiss, A cancer cell metalloprotease 975 triad regulates the basement membrane transmigration program. Genes & Development 976 20, 2673-2686 (2006). 977 70. G. Khalili-Tanha, E. S. Radisky, D. C. Radisky, A. Shoari, Matrix metalloproteinase-978 driven epithelial-mesenchymal transition: implications in health and disease. Journal of 979 Translational Medicine 23, 436 (2025). 980 71. T. A. Tervonen et al., Oncogenic Ras Disrupts Epithelial Integrity by Activating the 981 Transmembrane Serine Protease Hepsin. Cancer Research 81, 1513-1527 (2021). 982 72. S. Mezi, L. Todi, E. Orsi, A. Angeloni, P. Mancini, Involvement of the Src-cortactin 983 pathway in migration induced by IGF-1 and EGF in human breast cancer cells. Int J 984 Oncol 41, 2128-2138 (2012). 985 73. M. Schoumacher, R. D. Goldman, D. Louvard, D. M. Vignjevic, Actin, microtubules, 986 and vimentin intermediate filaments cooperate for elongation of invadopodia. Journal of 987 Cell Biology 189, 541-556 (2010). 988 74. R. Ferrari et al., MT1-MMP directs force-producing proteolytic contacts that drive tumor 989 cell invasion. Nature Communications 10, 4886 (2019). 990 75. A. W. Adamson, Physical chemistry of surfaces . (Wiley, New York, NY, ed. 6, 1997). 991 76. A. M. Weaver et al., Cortactin promotes and stabilizes Arp2/3-induced actin filament 992 network formation. Current Biology 11, 370-374 (2001). 993 77. P. Bieling et al., Force Feedback Controls Motor Activity and Mechanical Properties of 994 Self-Assembling Branched Actin Networks. Cell 164, 115-127 (2016). 995 78. J. Mueller et al., Load Adaptation of Lamellipodial Actin Networks. Cell 171, 188-996 200.e16 (2017). 997 79. L. C. Kelley et al., Adaptive F-Actin Polymerization and Localized ATP Production 998 Drive Basement Membrane Invasion in the Absence of MMPs. Developmental Cell 48, 999 313-328.e8 (2019). 1000 80. M. Q. Xu et al., Myosin-I synergizes with Arp2/3 complex to enhance the pushing forces 1001 of branched actin networks. Science Advances 10, (2024). 1002 81. T. Santra et al., An Integrated Global Analysis of Compartmentalized HRAS Signaling. 1003 Cell Reports 26, 3100-3115.e7 (2019). 1004 82. A. E. Karnoub, R. A. Weinberg, Ras oncogenes: split personalities. Nature Reviews 1005 Molecular Cell Biology 9, 517-531 (2008). 1006 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint 83. T. Clair, W. R. Miller, Y. S. Chochung, Prognostic-Significance of the Expression of a 1007 Ras Protein with a Molecular-Weight of 21,000 by Human-Breast Cancer. Cancer 1008 Research 47, 5290-5293 (1987). 1009 84. C. J. Ormandy, E. A. Musgrove, R. Hui, R. J. Daly, R. L. Sutherland, Cyclin D1, EMS1 1010 and 11q13 amplification in breast cancer. Breast Cancer Res Tr 78, 323-335 (2003). 1011 85. Y. Li et al., Cortactin Potentiates Bone Metastasis of Breast Cancer Cells1. Cancer 1012 Research 61, 6906-6911 (2001). 1013 86. S. Z. Chou, M. Chatterjee, T. D. Pollard, Mechanism of actin filament branch formation 1014 by Arp2/3 complex revealed by a high-resolution cryo-EM structure of the branch 1015 junction. P Natl Acad Sci USA 119, (2022). 1016 87. H. Gournier, E. D. Goley, H. Niederstrasser, T. Trinh, M. D. Welch, Reconstitution of 1017 human Arp2/3 complex reveals critical roles of individual subunits in complex structure 1018 and activity. Molecular Cell 8, 1041-1052 (2001). 1019 88. J. V. G. Abella et al., Isoform diversity in the Arp2/3 complex determines actin filament 1020 dynamics. Nature Cell Biology 18, 76-86 (2016). 1021 89. R. K. Vadlamudi, F. Li, C. J. Barnes, R. Bagheri‐Yarmand, R. Kumar, p41‐Arc subunit 1022 of human Arp2/3 complex is a p21‐activated kinase‐1‐interacting substrate. EMBO 1023 reports 5, 154-160 (2004). 1024 90. M. El-Tanani et al., Unraveling the tumor microenvironment: Insights into cancer 1025 metastasis and therapeutic strategies. Cancer Letters 591, 216894 (2024). 1026 91. C. M. Cesa et al., Micropatterned silicone elastomer substrates for high resolution 1027 analysis of cellular force patterns. Review of Scientific Instruments 78, (2007). 1028 92. A. Ulbricht et al., Cellular Mechanotransduction Relies on Tension-Induced and 1029 Chaperone-Assisted Autophagy. Current Biology 23, 430-435 (2013). 1030 93. D. Ahrens et al., A Combined AFM and Lateral Stretch Device Enables 1031 Microindentation Analyses of Living Cells at High Strains. Methods and Protocols 2, 1032 (2019). 1033 94. E. Cerami et al., The cBio Cancer Genomics Portal: An Open Platform for Exploring 1034 Multidimensional Cancer Genomics Data (vol 2, pg 401, 2012). Cancer Discovery 2, 1035 960-960 (2012). 1036 95. R. Merkel, N. Kirchgeßner, C. M. Cesa, B. Hoffmann, Cell Force Microscopy on Elastic 1037 Layers of Finite Thickness. Biophysical Journal 93, 3314-3323 (2007). 1038 96. S. Houben, N. Kirchgeßner, R. Merkel, in DAGM 2010, the 32nd Annual Symposium of 1039 the German Association for Pattern Recognition, M. Goesele, S. Roth, A. Kuijper, B. 1040 Schiele, K. Schindler, Eds. (Springer Berlin, Darmstadt, 2010), vol. 6376, pp. 71-80. 1041 97. N. Otsu, A Threshold Selection Method from Gray-Level Histograms. IEEE 1042 Transactions on Systems, Man, and Cybernetics 9, 62-66 (1979). 1043 98. G. W. Zack, W. E. Rogers, S. A. Latt, Automatic measurement of sister chromatid 1044 exchange frequency. Journal of Histochemistry & Cytochemistry 25, 741-753 (1977). 1045 99. D. H. Douglas, T. K. Peucker, Algorithms for the Reduction of the Number of Points 1046 Required to Represent a Digitized Line or its Caricature. Cartographica 10, 112-122 1047 (1973). 1048 100. C. S. Chirumamilla et al., in T-Cell Motility: Methods and Protocols, N. K. Verma, Ed. 1049 (Humana Press, New York, NY, 2019), vol. 1930, chap. Profiling Activity of Cellular 1050 Kinases in Migrating T-Cells, pp. 99-113. 1051 1052 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint Resource availability 1053 Lead contact 1054 Requests for further information and resources should be directed to and will be fulfilled by the 1055 lead contact, Erik Noetzel ([email protected]). 1056

availability 1057 All unique/stable reagents generated in this study are available from the lead contact without 1058 restriction. 1059 Data and code availability 1060 All data needed to evaluate the conclusions in the paper are present in the paper and/or 1061 the supplemental information. Any additional information required to reanalyze the data 1062 reported in this paper is available from the lead contact upon request, Erik Noetzel ( e.noetzel-1063 [email protected]). 1064 Acknowledgments 1065 We thank Helen Matthews (Sir Henry Dale Fellow School of Biosciences, University of 1066 Sheffield UK) and Buzz Baum (MRC Laboratory of Molecular Biology, Cambridge UK) for 1067 kindly providing the HRas cells and Frederik Rastfeld (IBI-2 Research Center Jülich) for his 1068 practical support. This work was supported by the Corona Foundation (S199/10084/2021), and 1069 by the Deutsche Forschungsgemeinschaft (DFG) (SFB TRR219 – Project-ID 322900939; 1070 subproject M07 and INST 222/1598-1 FUGG; Projekt: 566187929) to E.P.C.v.d.V. Funded by 1071 the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) 1072 363055819/GRK2415 to EN. 1073 Author contributions 1074 Conceptualization, EN, EPB; methodology, EPB, JE, YH, GD, RW, MR, RM; formal analysis, 1075 EPB, GD, RW, EPCvdV, MR; funding acquisition, EN; project administration, EN, RM; 1076 investigation, EPB, JE, YH, EPCvdV, MR, RW; supervision, EN, RM; validation, EPB; 1077 visualization, EPB, EN, EPCvdV, MR; writing – original draft, EPB, EN, MR; writing – review 1078 & editing, EPB, RM, EN, MR. 1079 Declaration of interests 1080 The authors declare no competing interests. 1081 Declaration of generative AI and AI-assisted technologies in the writing process 1082 During the preparation of this work the authors used ChatGPT5.2 to check grammar, spelling 1083 and clarification of phrases in parts of the manuscript. After using this tool, the authors reviewed 1084 and edited the content as needed and take full responsibility for the content of the published 1085 article. 1086 1087 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint Supplemental Information 1088 Supplemental Video S1: HRas on spheroid in invasion assay on 16 kPa. Phase contrast image 1089 series of an HRas-activated spheroid. After 12 hours the first cell invades the surrounding. 1090 Corresponds to Figure 2C. 1091 Supplemental Video S2: BM breach of HRas on spheroid on 16 kPa. Series of confocal images 1092 (F-actin in magenta and BM/laminin-332 in cyan) with a time increment of 15 min. Cells of the 1093 spheroid retracted during the initial 4 hours, followed by a pushing phase that culminated in 1094 BM disruption at 6 hours. Corresponds to Figure 4A. 1095 Supplemental Video S3: Laser ablation of a non-reinforced cell cortex (control sites) at an 1096 HRason spheroid on 16 kPa. Series of confocal images (F-actin in grey) with a time increment 1097 of 200 ms. In the video, ablation started after relative setpoint t = 0 ms. Corresponds to 1098 Figure 6E. 1099 Supplemental Video S4: Laser ablation of a reinforced cell cortex at an HRas on spheroid on 1100 16 kPa. Series of confocal images (F-actin in grey) with a time increment of 200 ms. In the 1101 video, ablation started after relative setpoint t = 0 ms. Corresponds to Figure 6F. 1102 1103 Supplemental Methods 1104 Traction force microscopy 1105 Maps of cell-induced traction stresses were calculated by regularized least-square fitting to the 1106 mechanical response of an elastic layer of 80 µm thickness on rigid substrates 95. From these 1107 maps, strain energy was calculated as a scalar measure of overall mechanical activity. These 1108 calculations were done as described in previous work 95, 96 and integrated over an area of 1109 38,007 µm2 to cover the entire region of influence of a spheroid. The required algorithms were 1110 implemented in MatLab (R2015a, The MathWorks Inc.). Spheroids were analyzed for at least 1111 24 hours. All displacements (strains) and forces (stresses) were calculated with reference to the 1112 first image of the series (t = 0 hours). Changes in strain energy were therefore determined with 1113 respect to that reference state. For baseline correction, calculated strain energy values were 1114 subtracted by the mean of 3 time points of at least two cell-free positions per experiment (see 1115 Supplementary Fig. S2). 1116 1117 Python routines for image processing 1118 For quantification of pERK signals , images of pERK staining with nuclei and F-actin co-1119 staining (pixel size=0.09 µm) were processed using a Gaussian filter (σ=1). To extract the outer 1120 spheroid shape, a maximum intensity projection (MIP) through the three channels was 1121 generated that used the highest intensity value at each pixel. Using the threshold selection 1122

developed by Otsu 97 and scaling of the resulting threshold by a factor of 0.5, a binary 1123 spheroid mask was generated of the MIP image. The resulting mask was refined using a 1124 morphological closing operation (cross-shaped structuring element of 3x3 pixels, applied 1125 iteratively 50 times), hole filling to eliminate internal gaps, a morphological opening operation 1126 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint (cross-shaped structuring element of 3x3 pixels, applied iteratively 20 times) and selection of 1127 the largest connected component. 1128 To identify individual cell masks within the spheroid, the nuclei channel was segmented using 1129 the Otsu thresholding method (scaled by a factor of 0.5). Morphological operations were then 1130 applied to clean the segmentation: opening with a square-shaped structuring element (5 pixels 1131 in size), removal of objects smaller than 2500 pixels, closing with the same rectangular 1132 structuring element and hole filling. A watershed algorithm was applied to all objects, to 1133 separate touching or clustered nuclei. This involved computing the distance transform of the 1134 binary nuclei mask and identifying local minima as markers for the watershed. The boundaries 1135 generated by the watershed segmentation were used to delineate individual cell regions. The 1136 mean intensity of pERK was calculated of individual cells inside the spheroid shape and 1137 averaged per spheroid. 1138 For quantification of MT1-MMP signals , images of MT1-MMP staining were segmented 1139 using the threshold selection method developed by Otsu 97. The mean intensity was calculated 1140 of pixels above the threshold. 1141 For compactness measurements , a maximum intensity projection (MIP) of z-planes 1142 (smoothed by a Gaussian filter; σ=3) was generated from the laminin-332 channel (pixel 1143 size=0.22 µm) for each spheroid. The minimum grayscale value of the MIP image was 1144 subtracted. Next, a threshold was computed using the triangle algorithm 98. This threshold was 1145 scaled by a factor of 1.2 and applied to separate signal from background. The resulting mask 1146 was refined using a morphological closing operation with a disk-shaped structuring element 1147 (15 pixels in radius), a morphological opening operation with a disk-shaped structuring element 1148 (5 pixels in radius), followed by hole filling to eliminate any internal gaps. If multiple objects 1149 remained after these steps, the largest connected component was selected and used as the final 1150 spheroid mask for calculation of compactness by 1151

 = 4 ∗   (1) as the ratio of the measured area to the area of a circle with the same perimeter. 1152 For quantification of basement membrane signals , individual z-planes were analyzed 1153 separately. The image of the laminin-332 channel (pixel size=0.11 µm) was smoothed using a 1154 Gaussian filter (σ=2). An Otsu threshold 97 was applied to separate signal from background. 1155 The resulting binary mask was refined using a binary opening operation with a disk-shaped 1156 structuring element (3 pixels in radius). For laminin-332 and collagen IV average intensities 1157 were calculated within the mask. Subsequently, intensities from different z-planes were 1158 averaged again. 1159 For analysis of cortical triplets (CT) , individual z-planes were analyzed separately (pixel 1160 size=0.07 µm). The spheroid border was defined in the cortactin image channel as follows. 1161 Using the threshold selection method developed by Otsu 97, and scaling the resulting threshold 1162 by a factor of 0.5, background was separated from signal to generate a binary spheroid mask. 1163 The resulting mask was refined using a morphological closing operation (disk-shaped 1164 structuring element of 10 pixels in radius), hole filling to eliminate internal gaps, selecting 1165 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint largest connected component, a morphological opening operation (disk-shaped structuring 1166 element of 10 pixels in radius) and a morphological dilation operation (disk-shaped structuring 1167 element of 10 pixels in radius). For identification of localized, reinforced actin sites, the 1168 distance transform of the spheroid mask was calculated and only the outermost 54 pixels (equal 1169 to 4 µm at given image scale) were considered. In the following, this section is referred to as 1170 outer strip. Next, intensity thresholds were computed separately for each image plane based on 1171 the intensity distribution of the entire plane by 1172    +    −    ∗  (2) where x=0.25 for F-actin and cortactin channels and x=0.33 for collagen IV channel. In the F-1173 actin channel (smoothed by a Gaussian filter; σ=5) bright objects (reinforced actin) were 1174 segmented by intensity thresholding (Eq. 2). Only objects within the outer strip were considered 1175 and objects smaller than 67 pixels were rejected. Next, we calculated the average intensity of 1176 cortactin within each identified F-actin object. If it exceeded the cortactin threshold (Eq. 2), this 1177 spot was identified as co-localized in F-actin and cortactin. Otherwise, it was rejected. In the 1178 final step, we searched for weakened BM at co-localized objects. Since the BM, and thus the 1179 collagen IV signal, surrounds the spheroid, the object was enlarged radially outwards by 1180 morphological dilation (size: 14 pixels, equal to 1 µm at given image scale). To do so, the center 1181 of mass of each object was determined in the previously calculated distance transform. Since 1182 the lowest values in the distance transform are located in the direction of the spheroid boundary, 1183 radial outwards extension was only performed for values that were below the center of mass 1184 value. Enlarged objects were rejected if the mean grey value of collagen IV was above the 1185 threshold. Conversely, the cortical triplet (CT) condition (F-actin and cortactin co-localization 1186 together with low BM signal) was fulfilled if the mean grey value of collagen IV was below the 1187 threshold. For each spheroid, CT frequency (CT positive pixels within outer strip to all pixels 1188 within outer strip) across all image planes was calculated. 1189 For image visualizations, co-localization was calculated in Imaris (software version 10.2, 1190 Oxford Instruments) in “Threshold Selection Mode”. From the z-stack source image, one plane 1191 was randomly selected and thresholds were calculated by 1192    +    −    ∗ 0.03 (3) For quantification of actin cortex retractions after laser-assisted rupture (image pixel 1193 size = 0.19 µm), the cell edge contour was drawn manually in ImageJ (1.50b) and the resulting 1194 coordinates were replaced by a simplified polygonal path determined by an iterative point 1195 selection algorithm 99 with a tolerance of 2 px. From the remaining coordinate pairs, 20 1196 equidistant reference points were interpolated along the smoothed path. As ablation was always 1197 performed in the image center but with angles adjusted to match the local cell edge orientation, 1198 ten rectangular-shaped sectors (polygon of four points; 3 px width and 15 px height) were 1199 arranged on each side of the center ablation sector in a row. In the next step, each segment was 1200 translated along its longitudinal axis such that the midpoint of the baseline contacted the 1201 approximated cell edge contour. The mean fluorescence intensity of each sector was calculated 1202 over time. The baseline per sector was calculated from the mean intensity values prior to 1203 ablation and set to 1. 1204 1205 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint Sample preparation and phosphorylation detection for Kinome profiling analysis 1206 Spheroids on 16 kPa elastomeric substrates, incubated 90 min with EtOH or OHT, were 1207 mechanically pipetted in solution with assay media and washed once in ice-cold PBS after 1208 respective treatments, , and lysed for 15 min on ice using M-PER Mammalian Extraction Buffer 1209 containing Halt Phosphatase Inhibitor and EDTA-free Halt Protease Inhibitor Cocktail (1:100 1210 each; ThermoFischer Scientific). Three biological replicates were used per condition. Lysates 1211 were centrifuged for 15 min at 16,000x g at 4 °C in a pre-cooled centrifuge. Supernatants were 1212 aliquoted, snap-frozen in liquid nitrogen and stored at −80 °C. Protein quantification was 1213 performed with BCA Assay (ThermoFischer Scientific) according to the manufacturer’s 1214 instructions. For experiments, new aliquots were thawed. 1215 For the PTK assay, 9.0 µg of protein was applied per array (n=3 per condition) and the assay 1216 was carried out using the standard protocol supplied by Pamgene International B.V. All 1217 reagents used for PTK activity profiling were supplied by Pamgene. Initially, to prepare the 1218 PTK Basic Mix, the freshly thawed supernatant was added to 4 µl of 10x protein PTK reaction 1219 buffer (PK), 0.4 µl of 100x bovine serum albumin (BSA), 0.4 µl of 1 M dithiothreitol (DTT) 1220 solution, 4 µl of 10x PTK additive, 4 µl of 4 mM ATP and 0.6 µl of monoclonal anti-1221 phosphotyrosine FITC-conjugate detection antibody (clone PY20). Total volume of the PTK 1222 Basic Mix was adjusted to 40 µl by adding distilled water (H2O). Before loading the PTK Basic 1223 Mix on the array, a blocking step was performed applying 30 µl of 2% BSA to the middle of 1224 every array and washing with PTK solution for PamChip® preprocessing. Next, 40 µl of PTK 1225 Basic Mix were applied to each array of the PamChips®. Then, the microarray assays were run 1226 for 94 cycles. An image was recorded by a CCD camera PamStation ®12 at kinetic read cycles 1227 32–93 at 10, 50 and 200 ms and at end-level read cycle at 10, 20, 50, 100 and 200 ms. 1228 For the STK assay, 2.0 µg of protein and 400 µM ATP were applied per array (n=3 per 1229 condition) together with an antibody mix to detect phosphorylated Ser/Thr. After incubation for 1230 an hour (30 °C) during which the sample was pumped back and forth through the porous 1231

to maximize binding kinetics and minimize assay time, a second FITC-conjugated 1232 antibody is used to detect the phosphorylation signal. Imaging was done using a LED imaging 1233 system. Spot intensity at each time point was quantified (and corrected for local background) 1234 using the BioNavigator software version 6.3 (PamGene). Upstream Kinase Analysis (UKA) 100, 1235 a functional scoring method (PamGene) was used to rank kinases based on combined specificity 1236 scores (based on peptides linked to a specific kinase, derived from six databases) and sensitivity 1237 scores (based on treatment-control differences) (BioNavigator version 6.3). 1238 1239 1240 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint Supplemental Figures 1241 1242 Supplementary Figure S1: Steroid hormone-free media attenuates OHT-independent invasiveness of 1243 HRasoff spheroids. Cumulative distribution of BM transmigration time on stiff 16 kPa substrates (n ≥ 44 spheroids 1244 from ≥ 3 independent experiments). Data is shown here for comparison and corresponds to data in Fig 2D (standard 1245 media) and to data in Fig 6I (steroid-hormone-free media). 1246 1247 1248 Supplementary Figure S2: Substrate specific strain energy detected by traction force microscopy. Baseline 1249 of SE measured at cell free positions. For each condition, SE values were recorded at 5, 10, and 20 hours from at 1250 least 2 spheroid-free control positions per experiment and averaged to obtain experiment-specific baseline values. 1251 These mean baseline values are shown here (from 3 independent experiments per condition), indicated according 1252 to spheroid treatment. 1253 1254 1255 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint Supplementary Figure S3: MMP localization and cellular content in HRas on breast spheroids. (A) 1256 Representative images of MT1-MMP (inversed grey scale) staining of MCF10A/HRas spheroids incubated with 1257 1 µM OHT or EtOH (HRas off control) for 16 hours, and the secondary antibody control (SAC) to measure 1258 unspecific background signals. (B) Quantification of fluorescence intensities of MT1-MMP staining (n = 60 and 1259 n = 30 for SAC from three and two individual staining experiments, respectively). Scatter plot includes median 1260 and 95% CI. Mann-Whitney-U-test was performed for the data (n.s.: p > 0.05). Scale bars: 20 µm. Position of 1261 focal plane used for imaging and analyses is indicated by red bar. 1262 1263 1264 Supplementary Figure S4: Comparative spheroid compactness depending on HRas activation. (A) Marginal 1265 changes of compactness for non-invasive HRas off at 4 hours (= mean invasion onset of HRas on group) control 1266 spheroids (n = 21 spheroids from 3 independent experiments). Shown with median and 95% CI. (B) Relative time 1267 curves show the change of compactness. To ease comparison data points were normalized: For non-invasive 1268 HRasoff spheroids, the overall measurements points at 3, 6, 9 and 12 hours were transformed to quarters 1/4, 2/4, 1269 3/4 and 1, respectively. Each value was subtracted from the starting compactness (t = 0 hours). For invasive HRason 1270 spheroids the period until individual invasion onsets were quartered and subtracted from the respective starting 1271 compactness values (0 hours). Line represents the mean with 95% CI error bands (n ≥ 15 spheroids from ≥ 3 1272 independent experiments). 1273 1274 1275 Supplementary Figure S5: HRas-driven cell invasion is not dependent on EPHB4 or AURKA signaling. 1276 Cumulative distribution of BM transmigration time dependent on HRas induction and EPHB4-kinase and 1277 AURKA-kinase inhibition. Both experiments were done once. Higher inhibitor concentration showed cell 1278 toxicity. 1279 1280 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint 1281 Supplementary Figure S6: Co-localization of pTyr421 cortactin with reinforced cortical actin structures. 1282 Representative micrographs of fixed and immunostained HRas on spheroid on 16 kPa substrate after 4 hours OHT 1283 treatment. Congruent pTyr42-cortactin (grey) and cortactin (green) signal distribution, accumulated at sites of 1284 thickened, reinforced actin (magenta), see white arrowheads. The zoom-in illustrates locally restricted phospho-1285 cortactin-actin accumulation. Scale bar: 20 µm. Cartoon: Indicates The position of the optical image plane (red 1286 bar). 1287 1288 1289 Supplementary Figure S7: HRas activation increases actin reinforcement and cortactin co-localization. 1290 Coverage of reinforced actin patches [%] in (A) and additionally co-localized cortactin [%] in (B) of individual 1291 spheroids was plotted with median and 95% CI (n ≥ 46 for each sample condition, from 3 independent 1292 experiments). Kruskal-Wallis test with Dunn’s multiple comparison test was performed for the data (n.s.: p > 0.05; 1293 *: p ≤ 0.05; **: p ≤ 0.01; ***: p ≤ 0.001; ****: p ≤ 0.0001). 1294 1295 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint 1296 Supplementary Figure S8: High ARP2/3 subunit and HRas gene expression predict poor survival of breast 1297 cancer patients. Univariate Kaplan-Meier curves show the recurrence-free survival (RFS) of breast cancer 1298 patients stratified by HRAS status (deletion or WT, left row; amplification or mutation, right row) RFS is shown 1299 depending on actin-related protein (ARP) subunits (HGNC group ID: 39) gene expression states (low vs. high). 1300 1301 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint Supplemental Tables 1302 Supplementary table 1: Raw data of mRNA expression values of 84 genes related to cytoskeletal regulation 1303 measured using a qRT-PCR array (QIAGEN, PAHS-088Z). Gene expression was analyzed in HRas on spheroids 1304 after 4.5 hours of OHT treatment compared to HRas off spheroids, both on tumor stiffness (16 kPa). 1305 Gene Fold Regulation Log2(Fold Regulation) p-value (compared to HRasoff group) -Log10(p-value) ACTR2 1.11 0.15 0.81 0.09 ACTR3 1.15 0.2 0.38 0.42 ARAP1 -1.19 -0.25 0.44 0.36 ARFIP2 1.12 0.16 0.65 0.19 ARHGAP6 -1.02 -0.03 0.79 0.10 ARHGDIB -1.02 -0.03 0.88 0.05 ARHGEF11 -1.30 -0.38 0.18 0.74 ARPC1B -1.05 -0.07 0.54 0.27 ARPC2 1.06 0.08 0.72 0.14 ARPC3 1.05 0.07 0.74 0.13 ARPC4 1.04 0.06 0.80 0.10 ARPC5 -1.01 -0.01 0.89 0.05 AURKA -1.26 -0.34 0.55 0.26 AURKB -1.56 -0.64 0.54 0.26 AURKC -1.38 -0.47 0.39 0.41 BAIAP2 -1.17 -0.22 0.58 0.24 CALD1 1.08 0.11 0.66 0.18 CALM1 1.08 0.11 0.78 0.11 CASK 1.11 0.15 0.85 0.07 CCNA1 1.35 0.43 0.64 0.19 CCNB2 1.18 0.24 0.99 0.00 CDC42 1.02 0.03 0.99 0.00 CDC42BPA 1.06 0.08 0.73 0.13 CDC42EP2 1.03 0.04 0.89 0.05 CDC42EP3 1.11 0.15 0.17 0.78 CDK5 -1.10 -0.14 0.51 0.30 CDK5R1 1.35 0.43 0.04 1.37 CFL1 1.06 0.08 0.81 0.09 CIT -1.62 -0.69 0.22 0.66 CLASP1 1.16 0.21 0.59 0.23 CLASP2 1.08 0.11 0.94 0.03 CLIP1 -1.07 -0.09 0.89 0.05 CLIP2 / / / / CRK -1.04 -0.06 1.00 0.00 CTTN 1.42 0.51 0.09 1.04 CYFIP1 -1.18 -0.23 0.49 0.31 CYFIP2 -1.06 -0.09 0.45 0.35 DIAPH1 -1.06 -0.09 0.87 0.06 DSTN -1.04 -0.06 0.86 0.07 EZR 1.02 0.03 0.97 0.02 FNBP1L -1.07 -0.1 0.58 0.24 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint FSCN2 -1.43 -0.51 0.26 0.58 GSN -1.00 0 0.97 0.01 IQGAP1 -1.03 -0.04 0.83 0.08 IQGAP2 -1.45 -0.54 0.27 0.57 LIMK1 1.14 0.19 0.22 0.65 LIMK2 -1.24 -0.32 0.67 0.18 LLGL1 1.02 0.03 0.93 0.03 MACF1 1.00 0 0.94 0.03 MAP3K11 -1.04 -0.06 0.87 0.06 MAP4 1.30 0.38 0.61 0.22 MAPK13 1.04 0.06 0.80 0.10 MAPRE1 1.08 0.11 0.76 0.12 MAPRE2 -1.06 -0.09 0.88 0.06 MAPT -1.12 -0.17 0.78 0.11 MARK2 -1.23 -0.3 0.49 0.31 MID1 -1.24 -0.3 0.46 0.33 MSN -1.02 -0.03 0.83 0.08 MYLK -1.21 -0.27 0.50 0.30 MYLK2 -1.20 -0.27 0.55 0.26 NCK1 -1.10 -0.14 0.53 0.28 NCK2 1.04 0.06 0.61 0.22 PAK1 -1.14 -0.2 0.70 0.15 PAK4 1.03 0.04 0.92 0.04 PFN2 1.10 0.14 0.70 0.15 PHLDB2 1.30 0.38 0.45 0.35 PIKFYVE -1.05 -0.07 0.72 0.14 PPP1R12A 1.15 0.2 0.73 0.13 PPP1R12B -1.24 -0.3 0.22 0.66 PPP3CA 1.10 0.14 0.66 0.18 PPP3CB 1.22 0.29 0.46 0.34 RAC1 1.07 0.1 0.82 0.09 RACGAP1 -1.12 -0.17 0.72 0.14 RDX 1.19 0.25 0.77 0.12 RHOA -1.15 -0.2 0.54 0.26 ROCK1 -1.02 -0.03 0.90 0.05 SSH1 -1.18 -0.23 0.18 0.75 SSH2 -1.04 -0.06 0.84 0.07 STMN1 -1.21 -0.27 0.29 0.54 TIAM1 1.28 0.36 0.52 0.28 VASP 1.13 0.18 0.51 0.29 WAS 1.35 0.43 0.55 0.26 WASF1 1.12 0.16 0.80 0.10 WASL 1.04 0.06 0.91 0.04 1306 1307 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint Supplementary table 2: Full data set of kinase profiling analysis of phospho-tyrosine-kinase (RTK) and serine-1308 threonine-kinase (STK) activation depending on HRas activation retrieved from kinase-substrate microarrays. 1309 Kinase rank: decreasing p-values of change in kinase activity. Mean kinase statistic: color coded up (1)- and down 1310 (-1)-regulation. Kinase regulation in HRsson spheroids (after 1 hour of OHT) was compared with HRasoff controls, 1311 all cultivated on 16 kPa substrates. PamChip® peptide tyrosine kinase (PTK) microarray system or the PamChip® 1312 Ser/Thr Kinase assay (STK). 1313 PTK STK Kinase Uniprot ID Kinase Name Kinase rank/Median Final score Mean Kinase Statistic Kinase Uniprot ID Kinase Name Kinase rank/Median Final score Mean Kinase Statistic P54760 EphB4 2.27868 1.57606 P31751 Akt2/PKB[beta] 2.01484 0.48170 Q06187 BTK 1.83851 0.84904 Q13464 ROCK1 1.98851 0.74470 P29317 EphA2 1.79134 0.78815 P68400 CK2[alpha]1 1.91966 0.72202 P07949 Ret 1.71652 0.91063 Q9UBS0 p70S6K[beta] 1.86851 0.47515 P07333 FmS/CSFR 1.68445 0.80322 P31749 Akt1/PKB[alpha] 1.71381 0.45414 P42681 TXK 1.56786 0.71669 O14965 AurA/Aur2 1.57551 0.67705 P10721 Kit 1.51902 0.71311 O75582 MSK1 (RPS6KA5) 1.57355 0.52124 Q06418 Tyro3/Sky 1.50048 0.76366 P49674 CK1[epsilon] 1.52719 0.57413 Q08881 ITK 1.46223 0.73304 Q15131 CDK10 1.30515 0.45603 P16591 Fer 1.44254 0.65002 P51812 RSK2 1.16236 0.38970 P12931 Src 1.30047 0.71359 Q15349 RSK1/p90RSK 1.16236 0.45373 P54753 EphB3 1.28231 1.01164 P10398 ARAF 1.08933 -0.50885 P07332 Fes 1.27638 0.65748 P15056 BRAF 1.06932 -0.32671 P08581 Met 1.02885 0.56433 Q96GD4 AurB/Aur1 1.06060 0.48304 P51817 PRKX 1.05329 0.38842 Q16566 CaMK4 1.01226 0.42324 1314 Supplemental References 1315 95. R. Merkel, N. Kirchgeßner, C. M. Cesa, B. Hoffmann, Cell Force Microscopy on Elastic 1316 Layers of Finite Thickness. Biophysical Journal 93, 3314-3323 (2007). 1317 96. S. Houben, N. Kirchgeßner, R. Merkel, in DAGM 2010, the 32nd Annual Symposium of 1318 the German Association for Pattern Recognition, M. Goesele, S. Roth, A. Kuijper, B. 1319 Schiele, K. Schindler, Eds. (Springer Berlin, Darmstadt, 2010), vol. 6376, pp. 71-80. 1320 97. N. Otsu, A Threshold Selection Method from Gray-Level Histograms. IEEE 1321 Transactions on Systems, Man, and Cybernetics 9, 62-66 (1979). 1322 98. G. W. Zack, W. E. Rogers, S. A. Latt, Automatic measurement of sister chromatid 1323 exchange frequency. Journal of Histochemistry & Cytochemistry 25, 741-753 (1977). 1324 99. D. H. Douglas, T. K. Peucker, Algorithms for the Reduction of the Number of Points 1325 Required to Represent a Digitized Line or its Caricature. Cartographica 10, 112-122 1326 (1973). 1327 100. C. S. Chirumamilla et al., in T-Cell Motility: Methods and Protocols, N. K. Verma, Ed. 1328 (Humana Press, New York, NY, 2019), vol. 1930, chap. Profiling Activity of Cellular 1329 Kinases in Migrating T-Cells, pp. 99-113. 1330 1331 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 18, 2026. ; https://doi.org/10.64898/2026.04.15.717430doi: bioRxiv preprint