Spatiotemporal dynamics of Mcl-1 abundance and its influence on apoptosis susceptibility

29 The Bcl-2 protein family defines cellular competence for mitochondrial outer membrane 30 permeabilization (MOMP) and apoptotic cell death. In proliferating cells, the Bcl -2 31 family member Mcl -1 accumulates across the cell cycle and confers trans -mitotic 32 resistance to extrinsic apoptosis. We show here that Mcl-1, but not Bcl-xL, additionally 33 undergoes a coordinated redistribution from the cytosol to mitochondria, concomitant 34 with its over -proportional accumulation late in the cell cycle. Live -cell monitoring of 35 Mcl-1 dynamics at single -cell resolution, combined with mathematical modelling, 36 enabled us to quantify that Mcl -1 redistribution substantially contributes to elevating 37 MOMP thresholds. Furthermore, we found that Mcl -1 accumulation and redistribution 38 act concomitantly but independently to increase MOMP thresholds as cells approach 39 mitosis and this elevated resistance is reset in daughter cells after division. Notably, 40 heterogeneities in Mcl-1 abundance and subcellular distribution are pronounced even 41 among isogenic cells within the same cell -cycle phase, and thus contribute to 42 substantial cell-to-cell variability in MOMP susceptibility. Analysis of colorectal cancer 43 tissue samples showed that variability in Mcl -1 expression and distribution is likewise 44 prominent between cells in patient tumors and were predicted to drive intra-tumour 45 heterogeneity in responses to treatments that induce MOMP. Overall, we demonstrate 46 how changes in Mcl-1 amounts and localisation integrate with cell-cycle progression to 47 modulate apoptotic susceptibility, thereby shaping cell-fate outcomes and contributing 48 to cell-to-cell heterogeneities in death decision making. 49 50 Key Words: Apoptosis, Mcl-1, MOMP, cell cycle, systems biology 51 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint

52 Apoptosis is a highly regulated form of programmed cell death, which is required during 53 development, for elimination of damaged or infected cells and for maintaining tissue 54 homeostasis (Nagata & Tanaka, 2017) . Consequently, dysregulated apoptosis can 55 contribute to degenerative disorders and proliferative diseases, such as cancer 56 (Hanahan, 2022; Vitale et al., 2023) . Many different stressors a nd types of cellular 57 damage can initiate apoptosis signalling through the extrinsic or intrinsic apoptosis 58 pathways, both of which converge at the mitochondria, where sufficient apoptotic 59 signals lead to mitochondrial outer membrane permeabilization (MOMP). MOMP 60 represents a central checkpoint that triggers the execution phase of apoptosis and is 61 both orchestrated and tightly regulated through the interactions of members of the 62 Bcl-2 protein family (Czabotar & Garcia -Saez, 2023; Singh et al., 2019) . The pro -63 apoptotic members Bax and Bak can oligomerize to form pores in the outer 64 mitochondrial membrane, whereas anti-apoptotic inhibitors such as Bcl -2, Bcl-xL and 65 Mcl-1 inhibit Bax and Bak. Members of the group of pro -apoptotic BH3-only proteins 66 either antagonize the anti -apoptotic Bcl -2 family proteins or additionally directly 67 activate Bax and Bak (Czabotar & Garcia -Saez, 2023; Singh et al., 2019) . Elevated 68 expression of anti-apoptotic Bcl-2 family members consequently promotes apoptosis 69 resistance by neutralizing BH3-only proteins and keeping Bax and Bak in check, and 70 overexpression or gene amplification of Bcl -2, Bcl-xL and Mcl-1 is frequently found in 71 various cancers (Beroukhim et al., 2010; Tsujimoto et al., 1984) . Despite their 72 overlapping protective functions, these proteins differ markedly in their regulation and 73 stability. Bcl -2 and Bcl -xL are relatively stable and provide sustained apoptosis 74 protection (Goodall et al., 2016) . In contrast, Mcl -1 is turned over far more rapidly 75 (half-life = 20 – 40 min, (Nijhawan et al., 2003; Schwickart et al., 2010) ), allowing its 76 protein levels to be rapidly adjusted by changes in transcription, translation, and 77 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint proteasomal degradation (Senichkin et al., 2020). Mcl-1 is predominantly localized to 78 the MOM, but Mcl-1 is also reported to be found in the cytoplasm, at the endoplasmic 79 reticulum, and in the nucleus (Fu et al., 2022; Kale et al., 2018; Nakajima et al., 2014). 80 At the MOM, Mcl -1 exerts its canonical anti -apoptotic function by sequestering pro -81 apoptotic pore formers, but also appears to support mitochondrial bioenergetics and 82 cellular metabolic fitness (Brinkmann et al., 2025; Wright et al., 2024) . In contrast to 83 Bcl-2 and Bcl-xL, Mcl-1 expression is additionally regulated by cell cycle progression 84 (Harley et al., 2010; Pollak et al., 2021) with elevated Mcl -1 amounts conferring 85 transient, trans-mitotic resistance to extrinsic apoptosis (Pollak et al., 2021). 86 Here, we combined quantitative, high-resolution imaging at single cell and subcellular 87 scales together with mathematical modelling to elucidate how Mcl -1 expression 88 dynamics and subcellular distribution define mitochondrial apoptosis thresholds. 89 Furthermore, we show how cell -to-cell differences in apoptosis susceptibility emerge 90 from natural intercellular heterogeneities in Mcl -1 expression and distribution, and 91 demonstrate that both factors independently contribute to apoptosis resistance. 92 93

94 Mcl-1 accumulates with progression in cell cycle and redistributes to 95 mitochondria. 96 Previous biochemical analyses in cell populations and single-cell microscopic analyses 97 showed that Mcl-1 levels are regulated across the cell cycle (Harley et al., 2010; Pollak 98 et al., 2021) . Moreover , maximum intensity projection -imaging indicated that Mcl -1 99 might redistribute towards the mitochondria during cell cycle progression, although 100 these measurements lacked authentic spatial resolution (Pollak et al., 2021). Here, we 101 therefore set out to quantitatively determine genuine relative changes in Mcl-1 amounts 102 and localisation at single cell resolution within and across major cellular compartments, 103 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint including the nucl eus, the cytosol and mitochondria. NCI -H460 cells expressing 104 geminin-GFP as a cell cycle indicator were labelled with MitoTracker red and DAPI as 105 segmentation markers for mitochondria and nuclei, respectively, and stained for Mcl-1 106 (Figure 1A). Single layer segmentation allowed us to separate nuclear, cytosolic and 107 mitochondrial regions ( Figure 1B, see also methods section). Mcl -1 concentrations 108 were highest in mitochondrial regions, corresponding to results for Bcl -xL, a related 109 Bcl-2 family member ( Figure 1C). The presence of non -mitochondrial pools of Mcl -1 110 was confirmed also by its signals substantially exceeding those obtained by using 111 secondary antibody-only staining, as well as by biochemical fractionation (Figure S1A-112 C). Separating and comparing cells from early and late cell cycle stages based on 113 absent or very high geminin reporter signals (Figure S1D) showed that Mcl -1 114 concentrations in all compartments increased from early cell cycle (G1) to late cell cycle 115 stages (late S/G 2), with the highest concentration increases observed at the 116 mitochondria ( Figure 1D, E ). In contrast, subcellular Bcl -xL concentrations did not 117 increase in any of the compartments ( Figure 1F ). To compare subcellular protein 118 distributions of Mcl-1, we calculated the ratios between cytoplasmic and mitochondrial 119 mean intensities for each individual cell. While ratios for Bcl -xL remained constant, 120 Mcl-1 very prominently redistributed towards mitochondria (Figure 1G). While the size 121 of nuclear, cytosolic and mitochondrial areas all increased across the cell cycle (Figure 122 S1E), Mcl-1 concentrations and distributions did not correlate with and therefore could 123 not be explained by these changes ( Figure S1F-H). Taken together, these analyses 124 thus demonstrate that cellular Mcl -1 concentrations increase with cell cycle 125 progression and additionally that Mcl -1 disproportionally accumulates at the 126 mitochondria (Figure 1H). 127 128 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint Mcl-1 redistribution occurs concomitantly with , but independently of , Mcl-1 129 accumulation. 130 To study whether or not increases in Mcl -1 across the cell cycle and its redistribution 131 towards mitochondria are temporally and mechanistically coupled, we applied a 132 combinatorial cell staining and pseudotime analysis. To infer the dynamics of changes 133 in Mcl-1 expression and redistribution, we combined geminin -GFP signals and DAPI 134 intensities, the latter indicating DNA content. As snapshot data, both provide 135 information on the cell cycle position of individual cells from S -phase into G 2 and on 136 towards entry into mitosis. These positions were then linked to cellular Mcl -1 staining 137 data. Pooling and normalizing independent experiments (Figure 2A) provided us with 138 information from a total of 4500 individual cells. Mcl-1 amounts and Mcl-1 redistribution 139 to the mitochondria concomitantly increased along both the geminin -GFP and DAPI 140 dimensions (Figure 2B,C). Due to the short duration of mitosis, M phase cells were 141 quantitatively underrepresented in this analysis . We therefore separately collected 142 imaging data including cells in M phase and newly divided G1 sibling cells (Figure 2D). 143 Cells in M phase expressed Mcl -1 at significantly higher concentrations than in S/G 2 144 phases (Figure 2E). Cells right after or during division initially retained the high Mcl -1 145 concentration observed in M phase ( Figure 2E ). Cells also maintained the 146 mitochondrial accumulation of Mcl -1 during mitosis and in newly divided sibling cells 147 (Figure 2F ). This suggests that daughter cells initially inherit both increased 148 expression and mitochondrial accumulation of Mcl -1 from the mother cells, before 149 Mcl-1 amounts and distributions revert back to values found in typical G 1 cells. Mcl-1 150 expression and redistribution are therefore also regulated concurrently during and 151 across mitosis. 152 We next assessed if accumulation and redistribution depended on each other. 153 NCI-H460 cells with inducible Mcl-1 depletion (Fig. 2G) failed to accumulate notable 154 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint amounts of Mcl -1 from G 1 to S/G 2 phases (Fig.2H). However, the low amounts of 155 remaining Mcl-1 still redistributed towards the mitochondria ( Figure 2I). This finding 156 excludes that the high accumulation of Mcl-1 in late cell cycle phases alone is already 157 sufficient to drive its over-proportional accumulation at the mitochondria. 158 Taken together, we therefore conclude that both Mcl -1 expression levels and its 159 subcellular distribution are regulated in parallel with cell cycle progression, yet 160 independently of each other. 161 162 Mcl-1 is rapidly exchanged between mitochondria and cytoplasm. 163 The detection of Mcl -1 amounts at the mitochondria, but also in the cytoplasm and 164 nucleus, and the evidence for its spatiotemporal redistribution raise the question on 165 how mobile Mcl-1 is between these compartments. 166 To study exchange kinetics between mitochondria and cytoplasm, we fluorescently 167 tagged endogenous Mcl-1 with mScarlet at the N-terminus via CRISPR-Cas9 knock-in 168 in NCI-H460 cells (H460 -S-Mcl-1). H460-S-Mcl-1 cells expressed mScarlet -Mcl-1 at 169 the expected molecular weight ( Figure 3A,B ). Sequencing and PCR amplification 170 further confirmed the correct insertion of mScarlet DNA into the Mcl -1 locus 171 (Figure S2A), and mScarlet -Mcl-1 fluorescence dropped when targeting Mcl -1 172 expression by siRNA ( Figure S 2B). Furthermore, mScarlet-Mcl-1 expression 173 increased and mScarlet -Mcl-1 redistributed towards the mitochondria along 174 progression in the cell cycle ( Figure 3C,D, Figure S2C ). mScarlet-Mcl-1 maintained 175 its antiapoptotic function, as shown by comparable time-to-death profiles between NCI-176 H460 WT and H460-S-Mcl-1 cells upon Fc-scTRAIL treatment, and by eliminating the 177 previously described transmitotic resistance to TRAIL treatment (Pollak et al., 2021) 178 by pharmacological inhibition of mScarlet-Mcl-1 (Figure S2D,E). Taken together, these 179 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint findings demonstrate that the mScarlet -Mcl1 fusion protein authentically reflects the 180 behaviour and anti-apoptotic potency of the intrinsic untagged protein. 181 Next, we measured the exchange kinetics of mScarlet -Mcl-1 (mitochondrial binding 182 and unbinding) by Fluorescence Loss In Photobleaching (FLIP) in single cells 183 (Figure 3E). Neighbouring cells served as control cells to correct signals for 184 photobleaching in adjacent regions by light scatter ( Figure S2F-H). Corrected FLIP 185 measurement data ( Figure S 2I) then allowed us to plot fluorescence decay in the 186 cytoplasm and the mitochondria. Cytoplasmic mScarlet-Mcl-1 intensities declined fast, 187 as expected from free diffusion between the bleaching and the measurement areas 188 (Figure 3F). Surprisingly, also mitochondrial mScarlet-Mcl-1 intensity decayed swiftly, 189 suggesting a rather high mobility also for the pool of Mcl -1 associated with 190 mitochondria ( Figure 3F). To quantify the exchange kinetics, we defined a 191 mathematical model based on ordinary differential equations to describe the FLIP 192

by estimates of the underlying kinetic parameters (Figure 3G). As anticipated, 193 the bleaching rate 𝑘𝑏𝑙𝑒𝑎𝑐ℎ was uniquely identifiable from the experimental data, as it 194 was independent of the other parameters (Figure S2J). In contrast, the exchange rates 195 𝑘𝑜𝑛 and 𝑘𝑜𝑓𝑓 depended linearly on each other (Figure 3H), so that any combination of 196 𝑘𝑜𝑛 and 𝑘𝑜𝑓𝑓 along the linear optimum satisfied the experimental data (Figure 3I). For 197 simplicity, we set 𝑘𝑜𝑛 to zero to determine the minimal 𝑘𝑜𝑓𝑓 as an explicit number for 198 the model. Importantly, inhibiting protein translation or degradation by cycloheximide 199 or bortezomib did not affect the range of minimal 𝑘𝑜𝑓𝑓 values determined from 200 individual cells in these experiments (Figure 3J ). Overall, these results therefore 201 demonstrate that Mcl-1 exchanges between the cytoplasmic and mitochondrial pools, 202 and that these exchanges proceed at rates faster than those for protein production or 203 degradation (Figure 3K). 204 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint To determine if this exchange encompasses the entire Mcl -1 pool, we examined 205 mitochondrial Mcl-1 retention after selective plasma membrane permeabilization with 206 digitonin (Niklas et al., 2011) . After permeabilization, large amounts of Mcl-1 diffused 207 out of the cell, yet less than would be expected from an entirely soluble cytosolic protein 208 (compare to GAPDH) (Figure S3A). In comparison to Mcl-1, TOM20 signals were far 209 better retained, as expected from a membrane integrated protein (Figure S3A). Longer 210 incubation times with digitonin only marginally increased Mcl -1 loss, showing that the 211 remaining Mcl-1 pool is stably retained inside of cells (Figure S3B). Observing the loss 212 of Scarlet-tagged Mcl-1 upon digitonin addition in real time provided similar results. A 213 substantial pool of Mcl-1 was lost immediately upon digitonin-based permeabilization, 214 and ratiometric analysis showed that the remaining Mcl -1 was predominantly 215 associated with mitochondria ( Figure S3C,D). These data suggest, that a fraction of 216 Mcl-1 is quickly washed out of the cell after permeabilization, and that a pool of 217 membrane-associated or integrated Mcl -1 remains stably retained even after longer 218 digitonin incubation . When purifying mitochondria for carbonate extraction 219 experiments, Mcl -1 was not found in supernatants of isolated mitochondria, further 220 confirming that any remaining Mcl-1 is stably associated with or integrated into the 221 outer mitochondrial membrane ( Figure S3E,F). Carbonate extraction upon elevated 222 pH more easily transferred subpools of Mcl -1 into the supernatant than membrane 223 integrated TOM20, and substantial amounts of Mcl-1 remained tightly associated with 224 the mitochondrial fraction (Figure S3F). 225 Taken together, these results therefore show that Mcl-1 exists in two pools, one of 226 which can rapidly exchange between mitochondria (or mitochondria-associated 227 regions) and the cytosol, the other of which being tightly membrane associated or 228 integrated at the mitochondria. 229 230 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint Quantitative estimates of altered MOMP thresholds indicate that peak resistance 231 requires both redistribution and accumulation of Mcl-1. 232 Since Mcl -1 accumulation and redistribution occur in parallel during cell cycle 233 progression, we used deterministic mathematical modelling to quantify their relative 234 contribution to altering MOMP susceptibility throughout the cell cycle. We developed a 235 first model to describe spatiotemporal Mcl-1 regulation (Figure 4A), to then couple this 236 with a separate model component that allowed to quantify MOMP thresholds. 237 Progression through the cell cycle was integrated by inferring the duration of the 238 individual phases, the proportion of cells in the respective phases and the overall 239 duration of the cell cycle in H460 cells ( Figure S4A). Furthermore, cellular volume 240 changes with progression in the cell cycle were taken into account (Figure S4B). The 241 resulting model could then be used to simulate cell cycle-dependent accumulation of 242 Mcl-1 and its subcellular redistribution ( Figure 4B,C ). To quantitatively verify the 243 accuracy of the model, we simulated a population of 2000 unsynchronized cells 244 (Figure S4C ) from which population snapshots could directly be compared to 245 experiments. The simulation results for cell cycle-dependent Mcl-1 redistribution and 246 accumulation excellently corresponded to experimental data ( Figure 4D,E). We then 247 used t he model to mathematically uncouple the otherwise biologically concomitant 248 accumulation of Mcl -1 and its redistribution to mitochondria, so that their respective 249 contributions to MOMP thresholds could be estimated independently of one another . 250 In contrast to the full model that reflected both Mcl -1 redistribution and accumulation 251 as observed experimentally (Figure 4F), the “Only Redistribution” model variant kept 252 Mcl-1 amounts constant to isolate the sole effect of Mcl-1 redistribution (Figure 4G). 253 Conversely, the “Only Accumulation” model variant kept the affinity of Mcl -1 towards 254 the mitochondria constant, thereby separating the accumulation of Mcl -1 throughout 255 the cell cycle from effects arising from altered subcellular redistribution ( Figure 4H). 256 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint Mitochondrial Mcl-1 concentrations were very similar between all variants during G 1, 257 but differed substantially towards the end of the cell cycle ( Figure 4I ). Both 258 redistribution and the accumulation of Mcl -1 significantly contribute d to elevating 259 mitochondrial Mcl -1 concentrations, with both acting together to achieve peak 260 mitochondrial Mcl-1 concentrations in M phase ( Figure 4J). The mitochondrial Mcl-1 261 concentrations from the model variants then served as input for a simplified, 262 experimentally validated mathematical model capable of estimating MOMP 263 susceptibility (Hantusch et al., 2018) . In this model, t -Bid served as a representative 264 and tuneable pro-apoptotic input signal, whereas MOMP susceptibility was assessed 265 by the amount of Bax and Bak in pores (interactome visualised in Figure S4D-G). In 266 comparison to only Mcl-1 redistribution or accumulation, substantially more tBid was 267 required for Bax and Bak oligomerisation when both Mcl -1 accumulation and 268 redistribution were taken into account (Figure 4K). Defining a threshold of 10% of Bax 269 and Bak in pores as a minimum for efficient MOMP, both processes jointly elevated 270 MOMP resistance approx. 2-3 fold from G1 to M phase (Figure 4K,L). 271 Taken together, these simulation results suggest that both accumulation and 272 redistribution contribute substantially to apoptosis resistance and that both processes 273 are required to achieve peak MOMP resistance late in the cell cycle. 274 275 Mcl-1 subcellular localization and expression levels independently contribute to 276 regulating apoptosis susceptibility. 277 Next, we set out to study if Mcl -1 redistribution and Mcl -1 accumulation indeed 278 independently contribut e to elevated apoptosis resistance in experiments. We first 279 measured the heterogeneity of Mcl-1 expression levels and subcellular distributions in 280 H460-S-Mcl-1 cells. To make these cells dependent on endogenous Mcl-1 for survival, 281 we inhibited both Bcl-2 and Bcl -xL (10 µM ABT -199, 10 µM WEHI -539), and then 282 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint analysed the relationship between single -cell Mcl-1 abundance and localization, and 283 the kinetics of cell death upon Mcl-1 inhibition (1 µM S63845) (Figure 5A). 284 To sufficiently capture population heterogeneity, we first measured Mcl -1 expression 285 and distribution in a total of 420 cells and then applied our mathematical model to 286 estimate MOMP resistances of these cells ( Figure 5B). As would be expected from 287 independent contributors to cell death resistance, MOMP thresholds increased with 288 Mcl-1 expression and also with its redistribution towards the mitochondria (Figure 5B). 289 Experimentally measured apoptosis sensitivities (time to death (TTD) after Mcl -1 290 inhibition) indeed corresponded very well to this prediction (Figure 5C), and both Mcl-291 1 expression and mitochondrial accumulation correlated with survival times 292 (Figure S5A,B). For a more thorough and quantitative analysis of the effects of Mcl-1 293 expression and distribution on survival times, we compared cells with similar Mcl -1 294 expression but different distributions, or vice versa ( Figure 5D,E). While cells with a 295 high Mcl-1cyto/mito ratio and low Mcl -1 expression died fastest, either a redistribution 296 towards mitochondria or an increase in overall Mcl -1 expression was sufficient to 297 significantly extend survival times ( Figure 5F). Cells with high Mcl -1 expression and 298 high accumulation at the mitochondria survived the longest (Figure 5F, Figure S5C). 299 We further examined whether other factors , such as cell size or cell positioning , 300 additionally influenced the measured survival times. However, these parameters 301 showed only minimal explanatory power, compared to the distribution of Mcl -1 and 302 overall Mcl-1 expression (Figure S5D). 303 To quantify the relative contribution s of Mcl -1 distribution and expression on the 304 survival times, we performed univariate and bivariate linear regressions ( Figure 5G). 305 We found, that the bivariate model was significantly better in predicting survival 306 durations (TTD) than the individual univariate models (Figure 5H), which confirms that 307 both parameters contribute significantly to apoptosis sensitivity. Importantly, the 308 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint absolute values of the regression coefficients for overall Mcl -1 amounts and its 309 distribution ratios were within a similar range, demonstrating from experimental data 310 that both parameters contribute equally strong to apoptosis resistance (Figure 5I). 311 Taken together, these experimental findings demonstrate that the Mcl -1 distribution 312 ratio and total Mcl-1 expression levels act independently of one another and contribute 313 to apoptosis resistance to a comparable degree. 314 315 Patient tumor heterogeneity in Mcl-1 distributions contributes to high cell-to-cell 316 variability in expected MOMP resistance. 317 Our results show that proliferating cancer cells cycle through phases of differential 318 apoptosis resistance, driven by changes in both the abundance and subcellular 319 localisation of Mcl -1. This dynamic regulation creates substantial cell -to-cell 320 heterogeneity in susceptibility to MOMP. We next asked whether similar heterogeneity 321 could also be observed in human tumour tissues. 322 To address this, we evaluated treatment-naïve stage III colorectal cancer tissue 323 samples that had previously been analysed by multiplex immunofluorescence imaging 324 of apoptosis signalling proteins (Lindner et al., 2022) . In brief, stained samples from 325 four patients underwent cell segmentation to achieve single-cell resolution, followed by 326 cell type assignment into stromal, immune and tumour compartments. In the present 327 study, we extended this analysis to include subcellular segmentation into cytoplasmic 328 and mitochondrial regions, quantitative assessment of Mcl -1 expression and 329 distribution, and model -based calculation of MOMP resistance for individual cells 330 (Fig.6A). Individual cells were segmented based on nuclear DNA and cytoplasmic S6-331 kinase staining. Mitochondria were segmented using a combination of B ak and Smac 332 signals (Figure 6B). Tumour cells were identified using the epithelial markers AE1 and 333 PCK26 (Figure 6C, Figure S6A). As expected, mitochondrial compartments exhibited 334 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint elevated amounts of Mcl -1 when compared with nuclear and cytoplasmic regions 335 (Figure S6B). Moreover, Mcl-1 distribution ratios in tumour cells closely matched those 336 observed in human cancer cells (Figure S6C). Both total Mcl-1 expression levels and 337 its subcellular distribution displayed pronounced heterogeneity across all four patient 338 samples ( Figure 6D ), comparable to the heterogeneity observed in our cell line 339 experiments. 340 We next investigated the extent to which heterogeneity in Mcl -1 distribution impacts 341 predicted MOMP resistance in tumour cells. First, we estimated MOMP resistance 342 based solely on total Mcl -1 abundance, analogous to classical pathological 343 stratification into low, medium, and high express ing samples , but at single -cell 344 resolution ( Figure 6E ). When simulating the apoptosis sensitivity of these cells, 345 increasing Mcl-1 abundance expectedly correlated with increased MOMP resistance 346 (Figure 6F). Importantly, incorporating the measured subcellular distribution of Mcl -1 347 into the model markedly increased the heterogeneity in predicted MOMP resistance 348 within each expression group ( Figure 6G ). Consequently, cell populations that 349 appeared well separated based on Mcl -1 abundance alone now substantial ly 350 overlapped in their predicted MOMP resistance (Figure 6H, I; Figure S6D). Extending 351 this analysis beyond the discrete low, medium, and high expression groups to all 352 analysed cells produced a very similar pattern to that observed in cell lines: both Mcl-1 353 abundance and its distribution between subcellular compartments contribute to 354 determining the overall MOMP threshold (Figure 6J, Figure S6E). In other words, cells 355 with similar total Mcl -1 levels are expected to differ markedly in their apopto sis 356 sensitivity due to differences in the subcellular distribution of Mcl-1. 357 Taken together, these findings demonstrate that both Mcl -1 expression levels and 358 intracellular localisation influence the predicted apoptosis sensitivity of individual 359 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint tumour cells, with both the amounts and distributions independently contributing to high 360 heterogeneity in expected apoptosis resistance within human tumour tissue. 361 362

363 In this study, we show that Mcl-1 expression and its redistribution between cytoplasm 364 and mitochondria are coordinated with cell cycle progression. Both processes function 365 as parallel, independent and equally potent rheostats that increase apoptosis 366 resistance as cells progress through the cell cycle. 367 While cell cycle dependent regulation of Mcl-1 abundance has been reported before at 368 population and single-cell scales (Harley et al., 2010; Pollak et al., 2021), we here, for 369 the first time, identify Mcl-1 redistribution as an independent process and quantitatively 370 define its impact on apoptosis resistance. Our results indicate that Mcl-1 exists in both 371 soluble pools and mitochondria-associated pools, the latter including tightly bound or 372 membrane-integrated fractions. The existence of different Mcl-1 pools is supported by 373 reports identifying Mcl-1 fractions in the nucleus and cytosol (Fu et al., 2022; Kale et 374 al., 2018; Nakajima et al., 2014) and by evidence for fractions of Mcl -1 engaging in 375 different interactions at and within mitochondria. For example, p ools of Mcl -1 were 376 reported to localise to the mitochondrial matrix, where they affect mitochondrial 377 respiration and fusion (Perciavalle et al., 2012). Mcl-1 can also associate with ACSL1 378 in the outer mitochondrial membrane to regulate long-chain fatty acid oxidation (Wright 379 et al., 2024), and Mcl-1 obviously can engage other Bcl-2 family members in the outer 380 mitochondrial membrane (Kale et al., 2018) . Given the rapid exchange of the mobile 381 Mcl-1 pool between mitochondria and cytosol identified here, it is tempting to speculate 382 that a substantial pool of Mcl-1 is membrane associated rather than integrated, thereby 383 permitting swift exchange. Dynamic shuttling of other Bcl -2 family members, most 384 notably during Bcl-xL dependent Bax and Bak retrotranslocation from mitochondria into 385 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint the cytosol, has been reported before (Edlich et al., 2011; Todt et al., 2015). However, 386 the cell cycle dependent regulation of Mcl-1 localization appears distinct, as we did not 387 observe a comparable regulation for Bcl-xL. Moreover, Mcl -1 exchange kinetics are 388 remarkably fast, exceeding both the reported rates of Bcl -xL mediated 389 retrotranslocation and Mcl-1 protein turnover (Edlich et al., 2011; Nijhawan et al., 2003; 390 Slomp et al., 2021; Todt et al., 2015). 391 The accumulation of Mcl -1 on mitochondria from S phase and peaking in M phase 392 depends on both increased Mcl -1 abundance and its net redistribution towards 393 mitochondria. Cells also increase in volume and mitochondrial content as they proceed 394 through the cell cycle (Posakony et al., 1977) , which could contribute to higher local 395 Mcl-1 levels . In both our simulations and measurements, we corrected for these 396 factors, yet still observed a net accumulation of Mcl -1 on mitochondria. This indicates 397 that redistribution must be driven by other processes that promote translocation to 398 mitochondria, reduce retrotranslocation, or both. 399 We show that daughter cells initially inherit the peak Mcl -1 amounts achieved in their 400 mother cell during M phase, before these Mcl -1 amounts are reset back to the lower 401 levels typical of G 1 phase cells. This may appear to contradict previous studies 402 reporting that Mcl-1 degradation begins during M phase (Allan et al., 2018; Clarke et 403 al., 2018; Haschka et al., 2015; Wertz et al., 2011) . However, those studies largely 404 focused on scenarios of prolonged mitotic arrest and made use of chemically 405 synchronized cell populations. It can be conceived that under prolonged arrest, 406 degradation mechanisms are activated that are absent, or activated later, during or 407 after unperturbed mitosis. Moreover, chemical cell synchronization can substantially 408 influence cell cycle - and proliferation-associated processes (Cooper, 2003; Ligasová 409 & Koberna, 2021; Min et al., 2020) , potentially interfering with the dynamics studied 410 here. Still, we cannot formally exclude the possibility that enhanced Mcl-1 degradation 411 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint is at least initiated in M phase, without notably affecting protein amounts yet. Since 412 cytokinesis strongly reduces daughter cell volumes compared with the mother cell, 413 Mcl-1 concentrations could possibly be maintained initially despite ongoing 414 degradation. Importantly, also the distribution of Mcl-1 is inherited to daughter cells, so 415 that newly divided cells are expected to initially exhibit high apoptosis resistance. 416 Indeed, our prior work on transmitotic cell fates during extrinsic apoptosis showed that 417 daughter cells inheriting high Mcl -1 levels and active caspase -8 from mother cells 418 require at least 30 to 60 min to regain MOMP competency (Pollak et al., 2021). 419 While both Mcl -1 amounts and localisation are regulated across the cell cycle, we 420 observed substantial cell-to-cell heterogeneity in both parameters, even among cells 421 from the same cell cycle stage. Because we could show that localisation is as important 422 for cell fate decision making as overall Mcl-1 amounts, this represents an independent 423 and equally significant layer contributing to cell -to-cell variability in apoptosis 424 sensitivity. Although non-genetic heterogeneity in cell death susceptibility is naturally 425 revealed in single cell studies (Rehm et al., 2002; Spencer et al., 2009) , it is now 426 increasingly recognized as a critical contributor that can allow some cells to evade 427 death and potentially drive cancer relapse and disease progression (Ichim et al., 2015; 428 Russo et al., 2024). Our analyses of patient tumour samples confirm that heterogeneity 429 in Mcl-1 expression and subcellular distribution also occurs among cells from individual 430 tumours. Moreover, our estimates of apoptosis thresholds indicate that both 431 parameters strongly influence expected cell death resistance, even blurring distinctions 432 between cell populations that would be clearly separable based on Mcl -1 expression 433 alone. 434 In our study, we focussed on identifying the dynamics of spatiotemporal Mcl -1 435 regulation and their relevance for functional consequences affecting cell fates. Our 436 evidence that Mcl-1 spatial regulation significantly affects cellular life/death decisions, 437 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint experimentally shown to be as important as overall Mcl -1 accumulation, provides a 438 justification for future studies aimed at dissecting the likely complex and multifactorial 439 mechanisms controlling Mcl-1 subcellular localization. These mechanisms might be 440 linked to Mcl-1 itself, its interaction partners, or features of mitochondria that change 441 over the cell cycle and specifically affect Mcl-1 rather than Bcl-xL. Because Mcl-1 pools 442 with distinct mobilities exist, understanding these underlying processes could also 443 enable the development of pharmacological strategies that selectively affect the 444 subcellular localization of specific Mcl -1 pools. This might contribute to novel 445 therapeutic avenues by which systemic toxicities encountered upon global Mcl -1 446 inhibition could be avoided (Bolomsky et al., 2020; Kelly & Strasser, 2020). 447 448 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint

449 Reagents and antibodies 450 Primary antibodies were as follows: mouse anti -α-Tubulin (DM1A, Cell Signaling 451 #3873, 1:1000 western blotting (WB)) , rabbit anti -Calnexin (Cell Signaling #2433, 452 1:1000 WB), rabbit anti-Cox IV (3E11, Cell Signaling #4850, 1:1000 WB), mouse anti-453 GAPDH (D4C6R, Cell Signaling # 97166, 1:1000 WB), mouse anti-Lamin A/C (Cell 454 Signaling #4777, 1:1000 WB), rabbit anti -Mcl-1 (D2W9E, Cell Signaling # 94296, 455 1:1000 WB, 1:800 IF, 1:50 flow cytometry (FC)), mouse anti -RFP (ChromoTek 6g6, 456 1:1000 WB), rabbit anti-Phospho-Histone H3 (D2C2, Cell Signaling #3377 and #8552, 457 1:50 FC), rabbit anti-Tom20 (Cell Signaling #42406, 1:1000 WB). 458 Secondary antibodies were as follows: goat anti -rabbit IgG – Alexa Fluor 647 (Life 459 Technologies Corporation), goat anti-rabbit IgG – Alexa Fluor 488 (Life Technologies 460 Corporation), goat anti-rabbit IgG – Peroxidase (Dianova GmbH), goat anti-mouse IgG 461 – Peroxidase (Dianova GmbH), goat anti-rabbit IgG - IRDye®680RD (LI-COR GmbH), 462 goat anti-mouse IgG - IRDye®800CW (LI-COR GmbH). 463 Fc-scTRAIL was produced as described previously (Hutt et al., 2017) . DMSO was 464 purchased from Carl Roth, ABT-199 from Active Biochem, S63845 and WEHI-539 from 465 APExBIO and bortezomib from UBPBio. MioTracker ™ Red CMXRos, MitoTracker ™ 466 Green FM and DAPI were obtained from Invitrogen, cycloheximide was purchased 467 from Sigma. Puromycin was obtained from AppliChem GmbH. 468 469 Plasmids and transfections 470 Plasmid EZ-Tet-pLKO-Blast (Addgene #85973) was used to express siRNA against 471 Mcl-1 (CCAGTATACTTCTTAGAAAGT), cloned as a hairpin sequence for stabilization 472 (CTAGC-CCAGTATACTTCTTAGAAAGT-TACTAGT-473 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint ACTTTCTAAGAAGTATACTGG-TTTTT-G). This plasmid was lentivirally transduced 474 in NCI-H460 cells. 475 Plasmids pSpCas9(BB) -2A-Puro human MCL -1 and pUC Scarlet -MCL-1 were both 476 received from Stephen Tait (Cao et al., 2022) and used to endogenously tag Mcl-1 with 477 mScarlet in NCI-H460 cells. Silencer®Select siRNA targeting Mcl -1 (ambion #s8583; 478 CCAGUAUACUUCUUAGAAATT) was transfected using Lipofectamine®RNAiMAX 479 (Thermo Fisher Scientific) according to the manufacturer’s protocol. 480 481 Tagging endogenous Mcl-1 with mScarlet 482 NCI-H460 cells were seeded in a 10 cm petri dish and transfected at roughly 30% 483 confluence. pSP-Cas9 and pUC-mScarlet plasmid DNA (kindly received from Stephen 484 Tait, published in (Cao et al., 2022) , was prepared with the Lipofectamine 3000 485 transfection kit (Thermo Fisher) according to the manufacturer's protocol and added 486 onto the cells. 24 hours after transfection, the medium was changed to medium 487 containing 0.5 µg/ml puromycin for the selection of transfection -positive cells. Cells 488 were sorted at the FACS for their mScarlet fluorescence and expanded as single cell 489 clones. The mScarlet intensity of the expanded clones was measured via flow 490 cytometry with excitation at 561 nm and a 586/15 nm emission filter. The molecular 491 size of the mScarlet -Mcl-1 construct was evaluated via immunoblot for Mcl -1 and 492 mScarlet. The subcellular distribution of mScarlet -Mcl-1 was validated using live cell 493 imaging at the LSM 980 microscope. To validate the mScarlet insertion, genomic DNA 494 was amplified using the quick extract DNA kit ( Biozym, #QE0905T) according to the 495 manufacturer's protocol. The mScarlet -insertion was amplified via PCR and the PCR 496 products were successfully sequenced and confirmed a correct knock-in. The following 497 primers were used to amplify the Mcl -1 locus: Mcl -1-UTR-for: 498 CACTTCCGCTTCCTTCCAGT; Mcl -1-Ex1-rev2: CCGCGTTTCTTTTGAGGCCA. 499 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint PCR was done using the PCR -NEB Q5 Kit. PCR samples were loaded on an 1% 500 agarose gel and DNA was stained via EtBr in the gel. Respective bands were cut out 501 of the gel and sequenced. 502 503 Cell culture 504 Cells were cultivated with RPMI 1640 medium, supplemented with 10% (v/v) FBS in 505 cell culture flasks (37°C, 5% CO2). Cells were seeded and passaged with a passaging 506 ratio between 1:3 and 1:10. 507 508 Immunofluorescence imaging and image analysis 509 To analyse subcellular protein concentrations, cells grown on coverslips were 510 incubated with MitoTracker Red CMXRos (100 nM) for 90 minutes prior to fixation with 511 4% PFA, permeabilization with 0.3% Triton -X-100 and immunostaining. The nuclei 512 were stained using DAPI. Images were acquired on a Zeiss Axio Observer SD Spinning 513 Disk microscope equipped with a PlanApochromat 40×/1.4 NA oil objective and an 514 Axiocam 503 Mono CCD camera. Geminin was excited with a 488 nm diode laser 515 using a 525/50 nm emission filter, the Alexa Fluor 647 dye was excited with a 638 nm 516 diode laser using the 690/50 nm emission filter and DAPI was excited with a 405 nm 517 diode laser using the 450/50 nm emission filter. MitoTracker Red was excited with a 518 561 nm diode laser using a 575/50 nm emission filter. All images were taken as Z -519 stacks with 0.5 µm distance between each Z-layer. For quantification, a single Z-layer 520 close to the surface of the coverslip was analysed using CellProfiler 4.2.7 (Stirling et 521 al., 2021). In the CellProfiler analysis, nuclei were recognized through DAPI staining 522 and the respective cell bodies were expanded around each nucleus based on the Mcl-1 523 staining. Based on the MitoTracker signal, the mitochondrial area was defined for each 524 cell individually. The cytoplasm represents the inverse of the mitochondria, the nucleus 525 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint was defined as a separate object. To quantify subcellular protein expression, the 526 respective IF staining intensities were measured. The distinction between early and 527 late cell cycle stages was set stringently at the 0.15 and 0.85 percentile of geminin 528 mean intensities inside the nucleus, cells in between were not considered for cell cycle 529 specific analysis. For the specific analysis of M cells and newly divided sibling cells, 530 DNA- and cell morphology were manually evaluated and categorized. 531 532 Time-lapse imaging and image analysis 533 For analysis of transmitotic apoptosis resistance, cells were plated on 35 mm glass-534 bottom dishes (CellView Cell Culture Dish, Greiner Bio One) in Phenol Red-free RPMI 535 1640 containing 10% FBS. Images were acquired at 37°C and 5% CO2 on a Zeiss Cell 536 Observer microscope equipped with an Axiocam MRm CCD camera and a Plan -537 Apochromat 20×/0.8 objective. Medium containing Fc-scTRAIL alone or in combination 538 with S6385 was added and cells were imaged for 24 h in 15 minute intervals. The time 539 until death (t death) after treatment was measured for individual cells. Thereby, the 540 population of analysed cells was subdivided into cells that underwent mitosis (f0+f1) 541 or not (f0). 542 For analysis of Mcl-1 expression and distribution, cells containing endogenous-tagged 543 mScarlet-Mcl-1 were plated on 35 mm glass -bottom dishes (CellView Cell Culture 544 Dish, Greiner Bio One) in Phenol Red -free RPMI 1640 containing 10% FBS. Before 545 imaging, cells were incubated with Phenol Red -free RPMI 1640 containing 10% FBS 546 and 100 nM MitoTracker Green and 1 µg/ml Hoechst 33342 for 30 minutes at 37°C 547 and 5% CO2. Images were acquired at 37°C and 5% CO2 on a confocal laser scanning 548 microscope (LSM 980 Airyscan 2) equipped with a Plan-Apochromat 63×/1.40 Oil DIC 549 M27 objective. mScarlet-Mcl-1 was excited with a 561 nm laser using a 573 -627 nm 550 emission filter, MitoTracker Green was excited with a 488 nm laser using a 490 -512 551 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint emission filter. If added to the experiment, Hoechst was excited with a 405 nm laser 552 using a 380 -548 emission filter. Cell segmentation was done as described for 553 immunofluorescence images, using the Hoechst staining for nuclei, mScarlet -Mcl-1 554 signal for the cell body and MitoTracker Green for mitochondria. For the Fluorescence 555 Loss in Photobleaching (FLIP) experiment s, measurement regions and control cells 556 were manually cropped and analysed via CellProfiler as describe above. 557 558 Flow cytometry 559 Cells were harvested and centrifuged at 300 g and 4°C for 5 minutes. The supernatant 560 was aspirated and the pellet was resuspended in 100 µl 4% PFA in PBS for 20 minutes 561 at RT. After PFA fixation, the cells were centrifuged again at the same conditions as 562 before and resuspended in 100 µl permeabilization solution 2 (BD Biosciences, 563 Germany) for 20 minutes at RT. Followed by another centrifugation step, the cells were 564 resuspended in 100 µl medium for at least 30 minutes at RT, to block unspecific binding 565 sites. After centrifugation, desired target antigens were then primed by incubation with 566 the respective primary antibody in medium for 90 minutes at RT, including an isotype 567 control antibody. After another centrifugation step, cells were washed twice in PBS for 568 5 minutes each, including centrifugation in between. To detect the bound primary 569 antibodies, the cells were incubated with fluorescen tly-tagged secondary antibodies. 570 Incubation with the secondary antibody was done in RPMI 1640 medium for 45 minutes 571 at RT. Follow ing a final centrifugation step, the cells were resuspended in PB S 572 supplemented with 0.02% (w/v) sodium azide and transferred into a 96 -well plate for 573 flow cytometry (MACSQuant VYB, Miltenyi Biotec, Germany). 574 575 Western blotting 576 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint Cells were lysed in solubilization buffer [50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM 577 EDTA, 1% (v/v) TritonX -100 plus Complete Protease Inhibitors (Roche) and 1 mM 578 DTT], incubated on ice for 20 min and centrifuged at 21,000 g and 4°C. Protein 579 concentrations were determined by Bradford assay (Roti® -Quant). Afterwards, 580 samples were incubated in Laemmli buffer for 10 minutes at 98°C. Proteins were 581 separated using BoltTM 4 – 12% Bis-Tris precast gels (Thermo Fisher) and blotted on 582 a nitrocellulose membrane using the iBlot2® (Thermo Fisher). Unspecific binding sites 583 on the membrane were blocked by incubation with western-blocking-reagent (Roche) 584 1:10 in TBS-T for at least 60 minutes at RT. Incubation with primary antibodies in PBS 585 supplemented with 0.02% (w/v) sodium azide was done at 4°C overnight. Followed by 586 three washing steps with TBS -T, the membrane was incubated with the a HRP -587 conjugated secondary antibody for 45 minutes at RT in TBS-T and 0.5X blocking 588 reagent. After three more washing steps with TBS -T, the membrane was incubated 589 with immobilon forte substrate (manufacturer) for 5 minutes at RT in the dark, followed 590 by measurement at an Amersham™ imager 600 (GE Healthcare, USA). Alternatively, 591 a fluorescently-labelled secondary antibody was used and imaged at a Li-Cor 9120 592 Odyssey® imager (LI-COR Biosciences GmbH, Germany). 593 594 Subcellular fractionation 595 NCI-H460 were harvested and homogenized in ice cold homogenization buffer [225 596 mM mannitol, 75 mM sucrose, 0.1 mM EGTA/EDTA, 30 mM Tris -HCl pH 7.4 ] 597 supplemented with cOmplete protease inhibitors. All subsequent centrifugation steps 598 were performed at 4°C. To isolate the nuclear fraction, the homogenate suspension 599 was centrifuged at 600 g for 5 min. The resulting pellet, containing nuclei and cell 600 debris, was resuspended in homogenization buffer with 100 µg/ml digitonin for further 601 purification and centrifuged again at 600 g for 5 min. The pellet was then resuspended 602 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint in PBS, washed, and centrifuged once more at 600 g for 5 min to obtain the final 603 nuclear pellet. The purified nuclei were stored at -20°C for subsequent analyses. The 604 supernatant from the first centrifugation was centrifuged again at 600 g for 5 min to 605 remove residual nuclei. The resulting supernatant was then centrifuged at 7,000 g for 606 15 min to obtain the mitochondrial fraction. The mitochondrial pellet was resuspended 607 in 2 ml of ice-cold PBS, centrifuged again at 7,000 g and subsequently at 10,000 g for 608 10 min. The final mitochondrial pellet was stored at -20°C for further analysis. The 609 supernatant from the 7,000 g centrifugation was further centrifuged at 20,000 g for 30 610 min. The remaining supernatant was then centrifuged at 100,000 g for 1 h. The 611 resulting pellet contained the endoplasmic reticulum (ER), while the supernatant 612 represented the cytosolic fraction. The cytosolic fraction was additionally precipitated 613 by adding trichloroacetic acid (TCA) to a final concentration of 15%, followed by 614 incubation on ice for 30 min. After centrifugation at 18,000 g for 10 min, the resulting 615 pellet was washed by gently adding 100% ethanol. This step was repeated once. 616 Proteins from the cytosolic fraction were dissolved in Laemmli buffer, heated at 95°C 617 for 10 min, and analysed by Western Blot. For all other fractions, protein concentrations 618 were first determined using the Bradford assay before further processing. 619 620 Sodium carbonate extraction 621 Cells were harvested, centrifuged at 400 g for 5 min. and the pellet was resuspended 622 in 5 ml homogenization buffer containing cOmplete protease inhibitors and kept on ice 623 for 2 min. The cells were then mechanically homogenized with a glass pestle by gently 624 moving it up and down. Successful homogenization was confirmed under a 625 microscope. To remove intact cells, the suspension was centrifuged at 3,200 g and 626 4°C for 5 min. The supernatant was collected, and the centrifugation step was repeated 627 until no sediment was observed. The supernatant was divided into four aliquots and 628 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint centrifuged at 18,000 g and 4°C for 10 min. The resulting pellet was resuspended in 1 629 ml homogenization buffer and centrifuged again at 18,000 g and 4°C for 10 min. The 630 pellets containing the crude mitochondria fraction were resuspended in 100 mM 631 Na2CO3 with pH values of 10, 11.25 and 12.5, or in PBS as a no treatment control, and 632 incubated on ice for 30 min. The samples were centrifuged at 100,000 g and 4°C for 633 60 min. The insoluble fraction in the pellet was dissolved in Laemmli buffer and heated 634 at 98°C for 10 min. The soluble fraction in the supernatant was precipitated by adding 635 TCA. The final samples were analysed using immunoblotting. 636 637 Analysis and quantification of Mcl-1 shuttling kinetics via FLIP 638 In each cell, a bleaching area at one edge of the cell, and a measurement area at the 639 opposing edge of the same cell was defined. After two minutes of equilibration, 640 mScarlet-Mcl-1 was bleached within the bleaching area using 100% power of the 561 641 nm laser every 10 seconds. Simultaneously, the decay of mScarlet intensity in the 642 measurement area was measured every 10 seconds. The measurement area was 643 segmented into mitochondrial and cytoplasmic compartments based on MitoTracker 644 green staining. To correct for unwanted bleaching through light dispersion, a control 645 cell next to the FLIP cell was measured in the same way. A one -phase exponential 646 decay function 𝑓(𝑡) was fitted to the mScarlet-Mcl-1 decay intensities 𝐼(𝑡) of the control 647 cell, to quantify unspecific bleaching outside of the bleaching area. Thus, the raw data 648 𝐼(𝑡) was corrected by dividing it by 𝑓(𝑡) and scaling to 𝑓(0), which led to the corrected 649 control decay 𝑐(𝑡) = 𝐼(𝑡) 𝑓(𝑡) ∙ 𝑓(0). In the same way, the data from the FLIP cell was 650 corrected using the corrective function 𝑓(𝑡) from the control cell on the FLIP data. To 651 obtain kinetic parameters of Mcl-1 (𝑘𝑜𝑛 or 𝑘𝑜𝑓𝑓), a simplified ODE model was defined 652 that describe d the FLIP experiment. The model consist ed of a cytoplasmic and a 653 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint mitochondrial compartment and a theoretical external compartment. Mcl -1 can be 654 either localized at the mitochondria or in the cytoplasm, or existing in the external 655 compartment as “bleached Mcl -1”. In the model, the bleaching was simulated by 656 depletion of cytoplasmic Mcl-1 into the external compartment. Mcl -1 was assumed to 657 bind to the mitochondria and be released back into the cytoplasm based on mass -658 action kinetics. This model was mathematically written as a system of three coupled 659 ODEs: 660 Equation (I) – ODE of cytoplasmic Mcl-1 in the FLIP model: 661 𝑑 𝑑𝑡𝑀𝑐𝑙1𝐶𝑦𝑡𝑜(𝑡) = 𝑘𝑜𝑓𝑓 ∙ 𝑀𝑐𝑙1𝑀𝑖𝑡𝑜(𝑡) − (𝑘𝑏𝑙𝑒𝑎𝑐ℎ + 𝑘𝑜𝑛) ∙ 𝑀𝑐𝑙1𝐶𝑦𝑡𝑜(𝑡) 662 Equation (II) – ODE of mitochondrial Mcl-1 in the FLIP model: 663 𝑑 𝑑𝑡𝑀𝑐𝑙1𝑀𝑖𝑡𝑜(𝑡) = −𝑘𝑜𝑓𝑓 ∙ 𝑀𝑐𝑙1𝑀𝑖𝑡𝑜(𝑡) + 𝑘𝑜𝑛 ∙ 𝑀𝑐𝑙1𝐶𝑦𝑡𝑜(𝑡) 664 Equation (III) – ODE of bleached Mcl-1 in the FLIP model: 665 𝑑 𝑑𝑡𝑀𝑐𝑙1𝐵𝑙𝑒𝑎𝑐ℎ𝑒𝑑(𝑡) = 𝑘𝑏𝑙𝑒𝑎𝑐ℎ ∙ 𝑀𝑐𝑙1𝐶𝑦𝑡𝑜(𝑡) 666 Using the SymPy library in python, this ODE system was solved analytically to 667 efficiently optimize the kinetic parameters of the system to the experimental data. 668 Therefore, the initial conditions 𝑀𝑐𝑙1𝐶𝑦𝑡𝑜(0) and 𝑀𝑐𝑙1𝑀𝑖𝑡𝑜(0) were taken from the 669 experimental data and 𝑀𝑐𝑙1𝐵𝑙𝑒𝑎𝑐ℎ𝑒𝑑(0) was set to zero. For each optimization step, the 670 ODE system was evaluated every 10 s for 10 minutes, similar to the experiment. Then, 671 an exponential decay function was fitted through these modelled decay curves of 672 𝑀𝑐𝑙1𝐶𝑦𝑡𝑜 and 𝑀𝑐𝑙1𝑀𝑖𝑡𝑜 and compared to the measured decay curves in the respective 673 compartments. The loss was defined as the absolute difference in the decay 674 coefficients between the ODE model and experimental data. Using the DIviding 675 RECTangles (DIRECT) optimization algorithm from the SciPy library, a pre -defined 676 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint parameter space of the kinetic parameters was iteratively screened for a global 677 parameter optimum that fitted the experimental data. 678 679 Plasma membrane permeabilization using digitonin 680 Cells were stained with MitoTracker for 90 minutes. Prior to fixation, cells were treated 681 with 100 µg/ml Digitonin in PBS for 1 min at RT, and washed of immediately after with 682 PBS. After permeabilization, the cells were kept in PBS for different time points before 683 fixation and immunofluorescence staining. 684 685 In silico cell cycle model of Mcl-1 expression and distribution 686 The ODE model consist ed of a volumetric mitochondrial and cytoplasmic 687 compartment, that contain mitochondrial or cytoplasmic Mcl-1, respectively. Therefore, 688 it was implemented within the python library Tellurium in Antimony (Choi et al., 2018; 689 Smith et al., 2009). 690 Mcl-1 was assumed to shuttle between the cytoplasmic and the mitochondrial 691 compartments obeying the experimentally defined distribution ratio and the shuttling 692 rates defined by FLIP measurements: 693 Equation (IV) – ODE of cytoplasmic Mcl-1 in the cell cycle model: 694 𝑑 𝑑𝑡𝑀𝑐𝑙1𝐶𝑦𝑡𝑜(𝑡) = 𝑘𝑜𝑓𝑓 ∙ 𝑀𝑐𝑙1𝑀𝑖𝑡𝑜(𝑡) − 𝑘𝑜𝑛 ∙ 𝑀𝑐𝑙1𝐶𝑦𝑡𝑜(𝑡) 695 Equation (V) – ODE of mitochondrial Mcl-1 in the cell cycle model: 696 𝑑 𝑑𝑡𝑀𝑐𝑙1𝑀𝑖𝑡𝑜(𝑡) = −𝑘𝑜𝑓𝑓 ∙ 𝑀𝑐𝑙1𝑀𝑖𝑡𝑜(𝑡) + 𝑘𝑜𝑛 ∙ 𝑀𝑐𝑙1𝐶𝑦𝑡𝑜(𝑡) 697 To include the cell cycle, the lengths of the cell cycle phases (G1, combined S/G2 and 698 M) were determined using flow cytometry and respective cell cycle phase markers. By 699 dividing the total cell cycle length of 18 h by the respective percentages of cells in each 700 phase, the duration of each phase was estimated. By using growth rates from literature 701 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint (Cadart et al., 2022) , the volume of the compartments was dynamically defined to 702 exponentially double over the time of one cell cycle and was reset to starting values 703 upon division. Since 𝑀𝑐𝑙1𝐶𝑦𝑡𝑜 and 𝑀𝑐𝑙1𝑀𝑖𝑡𝑜 were defined as concentrations, volume 704 growth also influenced these values in the ODE system: 705 Equation (VI) – Dynamic change in mitochondrial volume: 706 𝑀𝑖𝑡𝑜𝑉𝑜𝑙𝑢𝑚𝑒(𝑡) = 𝑀𝑖𝑡𝑜𝑉𝑜𝑙𝑢𝑚𝑒(0)𝑔𝑟𝑜𝑤𝑡ℎ𝑟𝑎𝑡𝑒∙𝑡 707 Equation (VII) – Dynamic change in cytoplasmic volume: 708 𝐶𝑦𝑡𝑜𝑉𝑜𝑙𝑢𝑚𝑒(𝑡) = 𝐶𝑦𝑡𝑜𝑉𝑜𝑙𝑢𝑚𝑒(0)𝑔𝑟𝑜𝑤𝑡ℎ𝑟𝑎𝑡𝑒∙𝑡 709 Mcl-1 turnover was implemented in the model by a continuous production of 710 cytoplasmic Mcl-1 that exponentially increases with cell growth, since Mcl -1 mRNA 711 concentrations was found to remain constant with cell cycle progression (Pollak et al., 712 2021). Mcl-1 degradation was defined based on mass-action law, with the degradation 713 rate calculated from a Mcl-1 half-life of 30 minutes (Nijhawan et al., 2003; Slomp et al., 714 2021). 715 Equation (VIII) – Turnover addition to Equation (IV): 716 𝑑 𝑑𝑡𝑀𝑐𝑙1𝐶𝑦𝑡𝑜(𝑡) = 𝑝𝑟𝑜𝑑𝑢𝑐𝑡𝑖𝑜𝑛𝑔𝑟𝑜𝑤𝑡ℎ𝑟𝑎𝑡𝑒∙𝑡 − 𝑀𝑐𝑙1𝐶𝑦𝑡𝑜(𝑡) ∙ 𝑑𝑒𝑔𝑟𝑎𝑑𝑎𝑡𝑖𝑜𝑛𝑟𝑎𝑡𝑒 717 To account for the cell cycle dependent accumulation of Mcl-1, the degradation rate 718 was exponentially decreased with cell cycle progression, with the rate empirically fitted 719 to experimental data: 720 Equation (IX) – Dynamic decrease in degradation rate: 721 𝑑𝑒𝑔𝑟𝑎𝑑𝑎𝑡𝑖𝑜𝑛𝑟𝑎𝑡𝑒 = 𝑑𝑒𝑔𝑟𝑎𝑑𝑎𝑡𝑖𝑜𝑛𝑟𝑎𝑡𝑒 ∙ (2 − 𝑒𝑑𝑒𝑐𝑟𝑒𝑎𝑠𝑒𝑟𝑎𝑡𝑒∙𝑡) 722 To account for the cell cycle dependent change in Mcl-1 distributions, the ratio between 723 𝑘𝑜𝑛 and 𝑘𝑜𝑓𝑓 was adjusted over time. Since 𝑘𝑜𝑛 and 𝑘𝑜𝑓𝑓 are directly dependent on 724 each other through the ratio between cytoplasmic and mitochondrial Mcl-1, 𝑘𝑜𝑛 can be 725 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint described as 𝑘𝑜𝑛 = 𝑘𝑜𝑓𝑓 𝑟𝑎𝑡𝑖𝑜. In the model, the ratio value was dynamically changed to 726 match the experimental observations in early and late cell cycle stages. 727 The overall model was solved using the roadrunner library in python (Somogyi et al., 728 2015). This allowed to simulate any single cell or cell ensembles over any number of 729 cell cycles based on the following initial parameters: Mcl -1 expression level, Mcl -1 730 distribution, Cell Size, Cell Cycle Stage. 731 For predictive perturbation simulations, Mcl-1 accumulation or redistribution were 732 independently deactivated. Mcl -1 accumulation was deactivated by keeping the 733 degradation rate constant over the whole cell cycle, which leaves Mcl -1 turnover in a 734 steady equilibrium and constant concentration. Mcl-1 redistribution was deactivated by 735 keeping the ratio value constant over the whole cell cycle. 736 737 Time to death experiments and analysis 738 H460-S-Mcl1 cells were seeded and stained with 100 nM MitoTracker Green and 739 imaged as mentioned above. Multiple fields of view with at least 10 cells each were 740 defined and imaged to capture Mcl-1 expression and distribution. After the initial image, 741 the cells were treated on stage with a BH3 mimetic combination of 10 µM ABT-199, 10 742 µM WEHI-539 and 1 µM S63845 in medium. Following the treatment, the cells were 743 continuously imaged every 5 minutes to determine the time until each cell 744 morphologically died (membrane blebbing). Single cell parameters such as mScarlet-745 Mcl-1 intensities, compartment sizes or cell positions were analysed from the initial 746 image. For each cell, the Mcl -1 expression and Mcl-1 distribution were used as input 747 for the above-mentioned apoptosis susceptibility model to simulate an individual t -Bid 748 threshold. Further, correlations of Mcl -1 expression and distribution were analysed 749 using Pearson’s correlation coefficient. To determine the extreme 10% per corner, the 750 distances of each cell to each corner were calculated and the 10% of cells that were 751 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint closest to each corner were assigned to that respective group. Univariate and bivariate 752 regressions and the resulting R 2 values and regression coefficients were analysed in 753 python using the Scikit-learn library. Variance inflation factor analysis was done using 754 the Statsmodels library in python. 755 756 Colorectal cancer tissue samples and multiplex imaging 757 Formalin-fixed, paraffin-embedded primary tumor tissue sections were obtained from 758 four chemotherapy -naïve, resected stage III CRC patients collected from Queen’s 759 University Belfast (UK). Three samples from the core of each tumour were assembled 760 on tissue microarrays (TMAs). Multiplexed immunofluorescence iterative staining of 761 the TMAs was performed as previously described (Lindner et al., 2022) using the Cell 762 DIVE™ technology (Leica Microsystems; formerly GE Healthcare). This involved 763 iterative staining and imaging of the same tissue section with multiple antibodies and 764 is achieved by mild dye oxidation between successive staining and imaging rounds. In 765 total, there were 13 staining rounds using cell segmentation, cell identity and apoptosis 766 signalling antibodies as described in (Lindner et al., 2022) , and DAPI was imaged in 767 each round. The Leica Bond (Leica Biosystems) was used for antibody staining. 768 Commercially acquired antibodies underwent multi-step process of validation and dye 769 conjugation as previously described (Lindner et al., 2022). Exposure times were set to 770 fixed values for all images of a given marker. 771 772 Single cell segmentation and analysis in tumour samples 773 Raw images of stained tumour tissues were obtained as described (Lindner et al., 774 2022). In the present study , nuclei and surrounding cytoplasmic areas of single cells 775 from each TMA were segmented using the Cellpose library in python. This library 776 contains pretrained models that recognize nuclei and whole cell. The cyto2 pretrained 777 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint model was used together with the DAPI staining for nuclei and S6 kinase staining for 778 the cytoplasm. The segmented masks were imported into CellProfiler, where a 779 combined image of Bak and Smac staining was used to segment mitochondria within 780 the cytoplasmic mask. Thereby, the rescaled B ak signal was weighted 2/3 and the 781 rescaled Smac signal was weighted 1/3 of the resulting image. Mitochondrial structures 782 were segmented using adaptive thresholding , so that tiny structures were size -783 excluded to minimize false-positive mitochondrial masking. 784 Classification of cells into tumo ur, immune or stroma cells was done using CD3 785 (immune marker) and AE1/PCK26 (tumo ur marker s) intensity distributions. Cells 786 negative for all markers were assigned as stroma. All analysis of Mcl-1 amounts and 787 distributions were conducted in the tumour cells. 788 789 Statistical analysis 790 Statistical analysis was performed using PR ISM 10 (GraphPad Software) and the 791 following libraries in python: Statsmodels, Scikit-learn, SciPy, NumPy. Unpaired t-test 792 with Welch’s correction was used in Figures 1D,F,G; S1E; 3D; S2D,E. One-Way 793 ANOVA with Dunnett’s T3 multiple comparisons was used in Figures 2E,F; S3B, 4 J,L; 794 5F; S6B,C. Mann-Whitney test was used in Figures 2H,I. Linear regression was 795 performed, and nested models were compared using ANOVA and the F-test in Figures 796 5H,I. 797 798 Visualisations 799 Schematics in figures 3K and S2A were created with BioRender.com 800 801 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint Acknowledgments 802 The authors acknowledge Prof Stephen Tait (Cancer Research UK Scotland Institute, 803 Univ. of Glasgow) and Dr Joel Riley ( Medical University of Innsbruck ) for providing 804 crucial plasmids and helpful discussions, and Cristina Jaus (University of Stuttgart) for 805 technical assistance. The authors also acknowledge the Technology Platform “Cellular 806 Analytics“ of the Stuttgart Research Center Systems Biology for their support & 807 assistance, Ms Elizabeth McDonough (GEHC Technology and Innovation center ) for 808 running Cell DIVE analysis of CRC stage III samples, and Dr. Sanghee Cho for data 809 processing. 810 811 Data availability 812 Data are available from the authors. 813 814 Code availability 815 All codes developed for this study are available at: 816 https://doi.org/10.5281/zenodo.18185366 817 818 Funding 819 This research was funded by the Deutsche Forschungsgemeinschaft (DFG) under 820 DFG grant INST 38/655 -1 (ID 471011418) – TRR 353, DFG grant MO 3226/4 -1 and 821 through Germany’s Excellence Strategy, DFG grant EXC 2075 (ID 390740016) 822 awarded to MM . This work was also supported by a US -Northern Ireland -Ireland 823 Tripartite grant funded by Research Ireland and the Health Research Board to JHMP 824 (16/US/3301), the National Cancer Institute (Systems Modeling of Tumor 825 Heterogeneity and Therapy Response in Colorectal Cancer; R01CA208179) to FG, 826 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint and Health and Social Care Northern Ireland (STL/5715/15) to DBL, and by the Health 827 Research Board (ERA-TRANSCAN-2022-002) to JHMP. 828 829 Author Contributions 830 FK: data acquisition, analysis and interpretation, draft writing; NP, AB: data acquisition, 831 analysis and interpretation, draft revision; BK: data acquisition, draft revision; FG, DBL: 832 draft revision, funding acquisition; JP: draft revision, supervision, funding acquisition ; 833 MR: data interpretation, draft writing, supervision, funding acquisition. 834 835 Ethics declarations 836 Raw data related to Fig.6 and Fig.S6 were used in line with FAIR data principles from 837 a previously published study (Lindner et al., 2022) . Tissues were supplied by the 838 Queen’s University Belfast Department of Pathology with written consent provided by 839 all patients and institutional ethical approval granted. Ethical approval for processing 840 of samples and clinical data was also obtained by the Beaumont Hospital Research 841 and Ethics Committee. 842 Competing interests 843 The authors declare no competing interests. 844 845 846 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint

847 Allan, L. A., Skowyra, A., Rogers, K. I., Zeller, D., & Clarke, P. R. (2018). Atypical 848 APC/C‐dependent degradation of Mcl‐1 provides an apoptotic timer during 849 mitotic arrest. The EMBO Journal, 37(17). 850 https://doi.org/10.15252/embj.201796831 851 Beroukhim, R., Mermel, C. H., Porter, D., Wei, G., Raychaudhuri, S., Donovan, J., 852 Barretina, J., Boehm, J. S., Dobson, J., Urashima, M., McHenry, K. T., 853 Pinchback, R. M., Ligon, A. H., Cho, Y . J., Haery, L., Greulich, H., Reich, M., 854 Winckler, W., Lawrence, M. S., … Meyerson, M. (2010). The landscape of 855 somatic copy-number alteration across human cancers. Nature, 463(7283), 899–856 905. https://doi.org/10.1038/nature08822 857 Bolomsky, A., Vogler, M., Köse, M. C., Heckman, C. A., Ehx, G., Ludwig, H., & Caers, 858 J. (2020). MCL-1 inhibitors, fast-lane development of a new class of anti-cancer 859 agents. In Journal of Hematology and Oncology (Vol. 13, Number 1). BioMed 860 Central Ltd. https://doi.org/10.1186/s13045-020-01007-9 861 Brinkmann, K., McArthur, K., Malelang, S., Gibson, L., Tee, A., Elahee Doomun, S. 862 N., Rowe, C. L., Arandjelovic, P., Marchingo, J. M., D’Silva, D., Bachem, A., 863 Monard, S., Whelan, L. G., Dewson, G., Putoczki, T. L., Bouillet, P., Fu, N. Y., 864 Brown, K. K., Kueh, A. J., … Strasser, A. (2025). Relative importance of the anti-865 apoptotic versus apoptosis-unrelated functions of MCL-1 in vivo. Science, 866 389(6764), 1003–1011. https://doi.org/10.1126/science.adw1836 867 Cadart, C., Venkova, L., Piel, M., & Cosentino Lagomarsino, M. (2022). Volume 868 growth in animal cells is cell cycle dependent and shows additive fluctuations. 869 ELife, 11. https://doi.org/10.7554/eLife.70816 870 Cao, K., Riley, J. S., Heilig, R., Montes-Gómez, A. E., Vringer, E., Berthenet, K., 871 Cloix, C., Elmasry, Y., Spiller, D. G., Ichim, G., Campbell, K. J., Gilmore, A. P ., & 872 Tait, S. W. G. (2022). Mitochondrial dynamics regulate genome stability via 873 control of caspase-dependent DNA damage. Developmental Cell, 57(10), 1211-874 1225.e6. https://doi.org/10.1016/j.devcel.2022.03.019 875 Choi, K., Medley, J. K., König, M., Stocking, K., Smith, L., Gu, S., & Sauro, H. M. 876 (2018). Tellurium: An extensible python-based modeling environment for systems 877 and synthetic biology. BioSystems, 171, 74–79. 878 https://doi.org/10.1016/j.biosystems.2018.07.006 879 Clarke, P. R., Allan, L. A., & Skowyra, A. (2018). Timed degradation of Mcl-1 controls 880 mitotic cell death. Molecular and Cellular Oncology, 5(6). 881 https://doi.org/10.1080/23723556.2018.1516450 882 Cooper, S. (2003). Rethinking synchronization of mammalian cells for cell cycle 883 analysis. Cellular and Molecular Life Sciences, 60(6), 1099–1106. 884 https://doi.org/10.1007/s00018-003-2253-2 885 Czabotar, P . E., & Garcia-Saez, A. J. (2023). Mechanisms of BCL-2 family proteins in 886 mitochondrial apoptosis. In Nature Reviews Molecular Cell Biology (Vol. 24, 887 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint Number 10, pp. 732–748). Nature Research. https://doi.org/10.1038/s41580-888 023-00629-4 889 Edlich, F., Banerjee, S., Suzuki, M., Cleland, M. M., Arnoult, D., Wang, C., Neutzner, 890 A., Tjandra, N., & Youle, R. J. (2011). Bcl-xL retrotranslocates Bax from the 891 mitochondria into the cytosol. Cell, 145(1), 104–116. 892 https://doi.org/10.1016/j.cell.2011.02.034 893 Fu, D., Pfannenstiel, L., Demelash, A., Phoon, Y. P., Mayell, C., Cabrera, C., Liu, C., 894 Zhao, J., Dermawan, J., Patil, D., DeVecchio, J., Kalady, M., Souers, A. J., 895 Phillips, D. C., Li, X., & Gastman, B. (2022). MCL1 nuclear translocation induces 896 chemoresistance in colorectal carcinoma. Cell Death and Disease, 13(1). 897 https://doi.org/10.1038/s41419-021-04334-y 898 Goodall, K. J., Finch-Edmondson, M. L., Van Vuuren, J., Yeoh, G. C., Gentle, I. E., 899 Vince, J. E., Ekert, P. G., Vaux, D. L., & Callus, B. A. (2016). Cycloheximide can 900 induce bax/bak dependent myeloid cell death independently of multiple BH3-only 901 proteins. PLoS ONE, 11(11). https://doi.org/10.1371/journal.pone.0164003 902 Hanahan, D. (2022). Hallmarks of Cancer: New Dimensions. In Cancer Discovery 903 (Vol. 12, Number 1, pp. 31–46). American Association for Cancer Research Inc. 904 https://doi.org/10.1158/2159-8290.CD-21-1059 905 Hantusch, A., Das, K. K., Garciá-Saéz, A. J., Brunner, T., & Rehm, M. (2018). Bax 906 retrotranslocation potentiates Bcl-xL’s antiapoptotic activity and is essential for 907 switch-like transitions between MOMP competency and resistance. Cell Death 908 and Disease, 9(4). https://doi.org/10.1038/s41419-018-0464-6 909 Harley, M. E., Allan, L. A., Sanderson, H. S., & Clarke, P. R. (2010). Phosphorylation 910 of Mcl-1 by CDK1-cyclin B1 initiates its Cdc20-dependent destruction during 911 mitotic arrest. EMBO Journal, 29(14), 2407–2420. 912 https://doi.org/10.1038/emboj.2010.112 913 Haschka, M. D., Soratroi, C., Kirschnek, S., Häcker, G., Hilbe, R., Geley, S., 914 Villunger, A., & Fava, L. L. (2015). The NOXA-MCL1-BIM axis defines lifespan 915 on extended mitotic arrest. Nature Communications, 6, 1–13. 916 https://doi.org/10.1038/ncomms7891 917 Hutt, M., Marquardt, L., Seifert, O., Siegemund, M., Müller, I., Kulms, D., Pfizenmaier, 918 K., & Kontermann, R. E. (2017). Superior properties of Fc-comprising scTRAIL 919 fusion proteins. Molecular Cancer Therapeutics, 16(12), 2792–2802. 920 https://doi.org/10.1158/1535-7163.MCT-17-0551 921 Ichim, G., Lopez, J., Ahmed, S. U., Muthalagu, N., Giampazolias, E., Delgado, M. E., 922 Haller, M., Riley, J. S., Mason, S. M., Athineos, D., Parsons, M. J., vandeKooij, 923 B., Bouchier-Hayes, L., Chalmers, A. J., Rooswinkel, R. W., Oberst, A., Blyth, K., 924 Rehm, M., Murphy, D. J., & Tait, S. W. G. (2015). Limited Mitochondrial 925 Permeabilization Causes DNA Damage and Genomic Instability in the Absence 926 of Cell Death. Molecular Cell, 57(5), 860–872. 927 https://doi.org/10.1016/j.molcel.2015.01.018 928 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint Kale, J., Osterlund, E. J., & Andrews, D. W. (2018). BCL-2 family proteins: Changing 929 partners in the dance towards death. Cell Death and Differentiation, 25(1), 65–930 80. https://doi.org/10.1038/cdd.2017.186 931 Kelly, G. L., & Strasser, A. (2020). Toward Targeting Antiapoptotic MCL-1 for Cancer 932 Therapy. Annual Review of Cancer Biology, 4(Volume 4, 2020), 299–313. 933 https://doi.org/https://doi.org/10.1146/annurev-cancerbio-030419-033510 934 Ligasová, A., & Koberna, K. (2021). Strengths and weaknesses of cell 935 synchronization protocols based on inhibition of dna synthesis. In International 936 Journal of Molecular Sciences (Vol. 22, Number 19). MDPI. 937 https://doi.org/10.3390/ijms221910759 938 Lindner, A. U., Salvucci, M., McDonough, E., Cho, S., Stachtea, X., O’Connell, E. P., 939 Corwin, A. D., Santamaria-Pang, A., Carberry, S., Fichtner, M., Van 940 Schaeybroeck, S., Laurent-Puig, P., Burke, J. P., McNamara, D. A., Lawler, M., 941 Sood, A., Graf, J. F., Rehm, M., Dunne, P. D., … Prehn, J. H. M. (2022). An atlas 942 of inter- and intra-tumor heterogeneity of apoptosis competency in colorectal 943 cancer tissue at single-cell resolution. Cell Death and Differentiation, 29(4), 806–944 817. https://doi.org/10.1038/s41418-021-00895-9 945 Min, M., Rong, Y., Tian, C., & Spencer, S. L. (2020). Temporal integration of mitogen 946 history in mother cells controls proliferation of daughter cells. Science, 947 368(6496), 1261–1265. https://doi.org/10.1126/science.aay8241 948 Nagata, S., & Tanaka, M. (2017). Programmed cell death and the immune system. 949 Nature Reviews Immunology 2017 17:5, 17(5), 333–340. 950 https://doi.org/10.1038/nri.2016.153 951 Nakajima, W., Hicks, M. A., Tanaka, N., Krystal, G. W., & Harada, H. (2014). Noxa 952 determines localization and stability of MCL-1 and consequently ABT-737 953 sensitivity in small cell lung cancer. Cell Death and Disease, 5(2), e1052-10. 954 https://doi.org/10.1038/cddis.2014.6 955 Nijhawan, D., Fang, M., Traer, E., Zhong, Q., Gao, W., Du, F., & Wang, X. (2003). 956 Elimination of Mcl-1 is required for the initiation of apoptosis following ultraviolet 957 irradiation. Genes and Development, 17(12), 1475–1486. 958 https://doi.org/10.1101/gad.1093903 959 Niklas, J., Melnyk, A., Yuan, Y ., & Heinzle, E. (2011). Selective permeabilization for 960 the high-throughput measurement of compartmented enzyme activities in 961 mammalian cells. Analytical Biochemistry, 416(2), 218–227. 962 https://doi.org/10.1016/j.ab.2011.05.039 963 Perciavalle, R. M., Stewart, D. P., Koss, B., Lynch, J., Milasta, S., Bathina, M., 964 Temirov, J., Cleland, M. M., Pelletier, S., Schuetz, J. D., Youle, R. J., Green, D. 965 R., & Opferman, J. T. (2012). Anti-apoptotic MCL-1 localizes to the mitochondrial 966 matrix and couples mitochondrial fusion to respiration. Nature Cell Biology, 967 14(6), 575–583. https://doi.org/10.1038/ncb2488 968 Pollak, N., Lindner, A., Imig, D., Kuritz, K., Fritze, J. S., Decker, L., Heinrich, I., 969 Stadager, J., Eisler, S., Stöhr, D., Allgöwer, F., Scheurich, P., & Rehm, M. (2021). 970 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint Cell cycle progression and transmitotic apoptosis resistance promote escape 971 from extrinsic apoptosis. Journal of Cell Science, 134(24), jcs258966. 972 https://doi.org/10.1242/jcs.258966 973 Posakony, J. W., England, J. M., & Attardi, G. (1977). Mitochondrial growth and 974 division during the cell cycle in HeLa cells. Journal of Cell Biology, 74(2), 468–975 491. https://doi.org/10.1083/jcb.74.2.468 976 Rehm, M., Düßmann, H., Jänicke, R. U., Tavaré, J. M., Kögel, D., & Prehn, J. H. M. 977 (2002). Single-cell fluorescence resonance energy transfer analysis 978 demonstrates that caspase activation during apoptosis is a rapid process: Role 979 of caspase-3. Journal of Biological Chemistry, 277(27), 24506–24514. 980 https://doi.org/10.1074/jbc.M110789200 981 Russo, M., Chen, M., Mariella, E., Peng, H., Rehman, S. K., Sancho, E., Sogari, A., 982 Toh, T. S., Balaban, N. Q., Batlle, E., Bernards, R., Garnett, M. J., Hangauer, M., 983 Leucci, E., Marine, J. C., O’Brien, C. A., Oren, Y., Patton, E. E., Robert, C., … 984 Bardelli, A. (2024). Cancer drug-tolerant persister cells: from biological questions 985 to clinical opportunities. Nature Reviews Cancer, 24(10), 694–717. 986 https://doi.org/10.1038/s41568-024-00737-z 987 Schwickart, M., Huang, X., Lill, J. R., Liu, J., Ferrando, R., French, D. M., Maecker, 988 H., O’Rourke, K., Bazan, F., Eastham-Anderson, J., Yue, P., Dornan, D., Huang, 989 D. C. S., & Dixit, V. M. (2010). Deubiquitinase USP9X stabilizes MCL1 and 990 promotes tumour cell survival. Nature, 463(7277), 103–107. 991 https://doi.org/10.1038/nature08646 992 Senichkin, V. V., Streletskaia, A. Y ., Gorbunova, A. S., Zhivotovsky, B., & Kopeina, G. 993 S. (2020). Saga of Mcl-1: regulation from transcription to degradation. In Cell 994 Death and Differentiation (Vol. 27, Number 2, pp. 405–419). Springer Nature. 995 https://doi.org/10.1038/s41418-019-0486-3 996 Singh, R., Letai, A., & Sarosiek, K. (2019). Regulation of apoptosis in health and 997 disease: the balancing act of BCL-2 family proteins. In Nature Reviews 998 Molecular Cell Biology (Vol. 20, Number 3, pp. 175–193). Nature Publishing 999 Group. https://doi.org/10.1038/s41580-018-0089-8 1000 Slomp, A., Moesbergen, L. M., Eldering, E., Kersten, M. J., Minnema, M. C., & 1001 Peperzak, V. (2021). Phosphatase PP2A enhances MCL-1 protein half-life in 1002 multiple myeloma cells. Cell Death and Disease, 12(3). 1003 https://doi.org/10.1038/s41419-020-03351-7 1004 Smith, L. P., Bergmann, F. T., Chandran, D., & Sauro, H. M. (2009). Antimony: A 1005 modular model definition language. Bioinformatics, 25(18), 2452–2454. 1006 https://doi.org/10.1093/bioinformatics/btp401 1007 Somogyi, E. T., Bouteiller, J. M., Glazier, J. A., König, M., Medley, J. K., Swat, M. H., 1008 & Sauro, H. M. (2015). LibRoadRunner: A high performance SBML simulation 1009 and analysis library. Bioinformatics, 31(20), 3315–3321. 1010 https://doi.org/10.1093/bioinformatics/btv363 1011 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint Spencer, S. L., Gaudet, S., Albeck, J. G., Burke, J. M., & Sorger, P. K. (2009). Non-1012 genetic origins of cell-to-cell variability in TRAIL-induced apoptosis. Nature, 1013 459(7245), 428–432. https://doi.org/10.1038/nature08012 1014 Stirling, D. R., Swain-Bowden, M. J., Lucas, A. M., Carpenter, A. E., Cimini, B. A., & 1015 Goodman, A. (2021). CellProfiler 4: improvements in speed, utility and usability. 1016 BMC Bioinformatics, 22(1). https://doi.org/10.1186/s12859-021-04344-9 1017 Todt, F., Cakir, Z., Reichenbach, F., Emschermann, F., Lauterwasser, J., Kaiser, A., 1018 Ichim, G., Tait, S. W., Frank, S., Langer, H. F., & Edlich, F. (2015). Differential 1019 retrotranslocation of mitochondrial Bax and Bak. The EMBO Journal, 34(1), 67–1020 80. https://doi.org/10.15252/embj.201488806 1021 Tsujimoto, Y., Finger, L. R., Yunis, J., Nowell, P. C., & Croce, C. M. (1984). Cloning of 1022 the Chromosome Breakpoint of Neoplastic B Cells with the t(14;18) 1023 Chromosome Translocation. Science, 226(4678), 1097–1099. 1024 https://doi.org/10.1126/science.6093263 1025 Vitale, I., Pietrocola, F., Guilbaud, E., Aaronson, S. A., Abrams, J. M., Adam, D., 1026 Agostini, M., Agostinis, P., Alnemri, E. S., Altucci, L., Amelio, I., Andrews, D. W., 1027 Aqeilan, R. I., Arama, E., Baehrecke, E. H., Balachandran, S., Bano, D., Barlev, 1028 N. A., Bartek, J., … Galluzzi, L. (2023). Apoptotic cell death in disease—Current 1029 understanding of the NCCD 2023. Cell Death & Differentiation 2023 30:5, 30(5), 1030 1097–1154. https://doi.org/10.1038/s41418-023-01153-w 1031 Wertz, I. E., Kusam, S., Lam, C., Okamoto, T., Sandoval, W., Anderson, D. J., 1032 Helgason, E., Ernst, J. A., Eby, M., Liu, J., Belmont, L. D., Kaminker, J. S., 1033 O’Rourke, K. M., Pujara, K., Kohli, P. B., Johnson, A. R., Chiu, M. L., Lill, J. R., 1034 Jackson, P. K., … Dixit, V. M. (2011). Sensitivity to antitubulin chemotherapeutics 1035 is regulated by MCL1 and FBW7. Nature, 471(7336), 110–114. 1036 https://doi.org/10.1038/nature09779 1037 Wright, T., Turnis, M. E., Grace, C. R., Li, X., Brakefield, L. A., Wang, Y. D., Xu, H., 1038 Kaminska, E., Climer, L. K., Mukiza, T. O., Chang, C. L., Moldoveanu, T., & 1039 Opferman, J. T. (2024). Anti-apoptotic MCL-1 promotes long-chain fatty acid 1040 oxidation through interaction with ACSL1. Molecular Cell, 84(7), 1338-1353.e8. 1041 https://doi.org/10.1016/j.molcel.2024.02.035 1042 1043 1044 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint Figure legends 1045 Figure 1: Mcl-1 accumulates with progression in cell cycle and redistributes to 1046 mitochondria. 1047 A) NCI-H460/geminin cells were incubated with MitoTracker (red), fixed and 1048 stained for Mcl-1 (white) and DNA (blue). Scale bar = 20 µm. 1049 B) Segmentation of m itochondrial, cytoplasmic and nuclear area for a 1050 representative cell. 1051 C) Mean Mcl-1 or Bcl -xL intensities were background -corrected and measured 1052 within the segmented compartments. n = 243 cells for Mcl-1 and n = 347 cells 1053 for Bcl-xL. 1054 D) Comparison of subcellular Mcl -1 concentrations between cells from early and 1055 late in the cell cycle, based on geminin expression intensities. n = 35 cells per 1056 group. 1057 E) Relative increases in Mcl-1 concentrations in the respective compartments with 1058 cell cycle progression. 1059 F) Comparison of subcellular Bcl -xL concentrations between cells from early and 1060 late in the cell cycle, based on geminin expression intensities. n = 44 cells per 1061 group. 1062 G) Protein distributions of Mcl-1 and Bcl-xL early and late in the cell cycle, displayed 1063 as their cytoplasmic to mitochondrial ratios. n = 35 cells per group for Mcl-1 and 1064 44 cells per group for Bcl-xL. 1065 H) Scheme depicting the simultaneous increase in Mcl -1 expression and its 1066 redistribution to the mitochondria with progression in cell cycle. 1067 (C,D,F,G) show one representative experiment from 3 independent biological repeats. 1068 Each dot represents a single cell. Error bars represent mean ± standard deviation. p-1069 values from an unpaired t-test with Welch’s correction are depicted. 1070 1071 Figure 2: Mcl-1 redistribution occurs concomitantly with but independently of 1072 Mcl-1 accumulation. 1073 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint A) Data from 11 individual experiments ( NCI-H460/geminin cells were incubated 1074 with MitoTracker, fixed and stained for Mcl -1 and DNA) were pooled. The 1075 pooled, normalized distributions are shown. 1076 B) Hexbin plot of geminin vs. DAPI intensities as a proxy for cell cycle progression. 1077 The colour code indicates the Mcl-1 expression levels. 1078 C) Hexbin plot ting as in (B), with the colour indicat ing the Mcl -1 distribution 1079 between cytoplasm and mitochondria. 1080 D) NCI-H460/geminin cells were incubated with MitoTracker, fixed, and stained for 1081 Mcl-1 and DAPI, with a focus on M phase cells and newly divided sibling cells 1082 (very early G1). Scale bar = 20 µm. 1083 E) Mcl-1 mean intensities in different cell cycle stages were quantified. n > 80 cells 1084 per group. 1085 F) Mcl-1 distribution ratios in different cell cycle stages were quantified. 1086 G) Immunoblot showing the DOX-inducible downregulation of Mcl-1 expression in 1087 NCI-H460/sh-Mcl-1 cells. 1088 H) Mcl-1 mean intensities from untreated cells and cells with DOX -induced Mcl-1 1089 downregulation are depicted and compared between G1 and late S/G2 phases. 1090 I) Mcl-1 distribution ratios from untreated cells and cells with DOX -induced Mcl-1 1091 downregulation are depicted and compared between G1 and late S/G2 phases. 1092 (E, F) show one representative experiment from two independent biological repeats. 1093 Each dot represents a single cell. Error bars represent mean ± standard deviation. p-1094 values from a One-Way ANOVA with Dunnett’s T3 multiple comparisons are depicted. 1095 (H, I) show one representative experiment from three independent biological repeats. 1096 Each dot represents a single cell. Error bars represent median ± interquartile range. p-1097 values from Mann-Whitney tests are depicted. 1098 1099 Figure 3: Mcl-1 is rapidly exchanged between mitochondria and cytoplasm. 1100 A) Immunoblots showing WT Mcl -1 (41 kDa) or mScarlet -Mcl-1 (70 kDa) 1101 expression. 1102 B) Immunoblot of mScarlet-Mcl-1 (70 kDa) by RFP-directed antibody detection. 1103 C) Fluorescence of mScarlet -Mcl-1 was measured for the different cell cycle 1104 stages, separated as shown in Figure S2C. Intensities were normalized to G 1 1105 and displayed as fold expression relative to G1. One representative experiment 1106 from two independent repeats is shown. 1107 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint D) H460-S-Mcl-1 cells were imaged for mScarlet -Mcl-1 distribution ratios. Cell 1108 cycle stages were discriminated by DNA content using NucBlue dye. n = 29 cells 1109 per group, one representative experiment from two independent repeats is 1110 shown. 1111 E) Exemplary overview of fluorescence loss in photobleaching (FLIP) imaging. 1112 H460-S-Mcl1 cells were stained with MitoTracker green. mScarlet -Mcl-1 (red) 1113 was bleached in the bleaching area and measured in the segmented 1114 measurement areas. Scale bar = 20 µm. 1115 F) Fluorescence decay of mScarlet -Mcl-1 fluorescence intensity in the 1116 measurement areas. The black line s depict plateaus followed by exponential 1117 decays fitted to the experimental data. 1118 G) Scheme and underlying ODEs of the mathematical model that describes the 1119 FLIP experiments. 1120 H) Graph that shows the loss value (difference of the model to the experimental 1121 data) depending on the respective combination of 𝑘𝑜𝑛 and 𝑘𝑜𝑓𝑓 in the FLIP 1122 model. The red line depicts a linear function fitted through the optimal 1123 combinations, with the equation displayed in the graph. 1124 I) Examples of combinations of 𝑘𝑜𝑛 and 𝑘𝑜𝑓𝑓 along the linear optimum. The model 1125

(black) replicate the experimental data. 1126 J) The lowest possible 𝑘𝑜𝑓𝑓 values from 33 FLIP analyses. Cells were either left 1127 untreated, subjected to 25 µg/ml cycloheximide (CHX) alone, or to a 1128 combination of 25 µg/ml cycloheximide and 65 µM bortezomib (BTZ). 1129 K) Scheme visualizing the different timescales of turnover and mitochondrial -1130 cytoplasmic exchange. 1131 1132 Figure 4: Quantitative estimates of altered MOMP thresholds indicate that peak 1133 resistance requires both redistribution and accumulation of Mcl-1. 1134 A) Scheme depicting the mitochondrial and cytoplasmic compartment s as the 1135 foundation of the simulation approach. Mcl -1 exchanges between the two 1136 compartments following mass-action kinetics. 1137 B) Simulation of Mcl-1 amounts over time. 1138 C) Simulation of subcellular Mcl-1 distribution over time. 1139 D) Comparison of experimentally measured Mcl-1 distribution ratios and simulated 1140 Mcl-1 distribution ratios between early and late cell cycle phases. 1141 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint E) Comparison of experimentally m easured Mcl -1 levels and simulated Mcl -1 1142 expression levels in different cell cycle stages. 1143 F) Top: Scheme depicting the regulation of Mcl -1 across the cell cycle. Bottom: 1144 Mitochondrial (blue) and cytoplasmic (orange) Mcl -1 concentrations simulated 1145 by the ODE model over two consecutive cell cycles. 1146 G) Top: Scheme depicting the “Only Redistribution” model of Mcl -1. Bottom: 1147 Mitochondrial (blue) and cytoplasmic (orange) Mcl -1 concentrations simulated 1148 by the ODE model over two consecutive cell cycles. 1149 H) Top: Scheme depicting the “Only Accumulation” model of Mcl -1. Bottom: 1150 Mitochondrial (blue) and cytoplasmic (orange) Mcl -1 concentrations simulated 1151 by the ODE model over two consecutive cell cycles. 1152 I) Mitochondrial Mcl -1 concentrations from the models shown in ( F-H) are 1153 summarized over two cell cycles. Early G 1 (blue) and M phase (red) are 1154 highlighted. For each model, 500 cells reflecting experimentally observed Mcl-1 1155 heterogeneity were simulated. 1156 J) Mitochondrial Mcl-1 concentrations in G1 were compared to Mcl-1 in M phase in 1157 the different models. Error bars represent mean ± standard deviation from the 1158 500 single cells simulated in (I) . p-values from a One -Way ANOVA with 1159 Dunnett’s T3 multiple comparisons post hoc test are depicted. 1160 K) Mitochondrial Mcl-1 concentrations from (J) were used as input for an apoptosis 1161 susceptibility model. It is shown how much t -Bid in each scenario is necessary 1162 to push simulated cells above a threshold of 10% of Bax and Bak in pores. 1163 L) The 10% t-Bid threshold values of each scenario are compared using One Way 1164 ANOVA with Dunnett’s T3 multiple comparisons. Error bars represent mean ± 1165 standard deviation from the 500 single cells simulated in (I). 1166 1167 Figure 5: Mcl-1 subcellular localization and expression levels independently 1168 contribute to regulating apoptosis susceptibility 1169 A) Experimental scheme. Living H460-S-Mcl-1 cells were measured on stage for 1170 their Mcl -1 expression and distribution. Cells were then treated with BH3 1171 mimetics, with Mcl-1 inhibition driving cell death. Individual times to death were 1172 recorded. 1173 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint B) Hexbin plot showing the predicted t -Bid threshold of 420 cells pooled from 3 1174 independent, batch -corrected experiments. t-Bid threshold s are color-coded 1175 and plotted against the respective Mcl-1 ratio and Mcl-1 total expression. 1176 C) Hexbinplot showing the time to death (TTD) of 420 cells pooled from 3 1177 independent, batch -corrected experiments. TTD is color -coded and plotted 1178 against the respective Mcl-1 ratio and Mcl-1 total expression. 1179 D) 10% of cells in each corner were grouped and the average time to death of each 1180 group is depicted as a colored area. 1181 E) Scheme depicting the Mcl-1 expression and distribution in the four groups. 1182 F) Individual cells of the four groups were compared using One Way ANOVA with 1183 Dunnett’s T3 multiple comparisons. 1184 G) The contribution of either the ratio or the total Mcl -1 to the times to death was 1185 determined by a comparative approach using univariate linear regressions and 1186 multivariate linear regression via ANOVA. 1187 H) The adjusted R 2 values of predicted vs . measured times to death from either 1188 the univariate regressions or the bivariate regression are depicted and 1189 compared using ANOVA and F-statistic. 1190 I) Absolute regression coefficients of the bivariate model are depicted. 1191 1192 Figure 6: In vivo heterogeneity in Mcl-1 distributions contributes to high cell-to-1193 cell variability in expected MOMP resistance 1194 A) Scheme depicting the workflow of segmentation and analysis of the tumo ur 1195 samples. 1196 B) IF images of the compartment markers DAPI (for nuclei), S6 -Kinase (for 1197 cytoplasm) and combined B ak/Smac signals (for mitochondria) with the 1198 respective masks. 1199 C) Classification of single cells into stroma, tumo ur or immune cells based on 1200 tumour markers AE1/PCK26 and immune marker CD3. Only tumour cells were 1201 used for further analysis. 1202 D) Mcl-1 distribution ratio s and Mcl-1 expression of single tumo ur cells from four 1203 different samples. 1204 E) Cells from each tumour sample were classified into three groups based on their 1205 single-cell Mcl-1 expression (low, medium or high). 1206 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint F) The mean Mcl -1 expression of the cells grouped in (E) was used as input for 1207 the MOMP susceptibility model to simulate t -Bid sensitivity and Bax/Bak pore 1208 formation. 1209 G) The mean Mcl-1 expression as well as the subcellular Mcl-1 distribution of the 1210 cells grouped in (E) was used as input for the apoptosis susceptibility model to 1211 simulate t-Bid sensitivity and Bax/Bak pore formation. 1212 H) The t-Bid threshold values were compared between the groups. Thresholds with 1213 only Mcl-1 expression levels as input are depicted as solid filled dots. 1214 Thresholds derived from combined Mcl -1 expression and distribution data as 1215 inputs are depicted as violin plots. 1216 I) The percentage overlap between the groups was calculated for each patient 1217 sample. 1218 J) Hexbinplot showing the simulated t -Bid threshold s of single tumo ur cells 1219 dependent on the Mcl-1 distribution ratio and total Mcl-1 expression. Thresholds 1220 increase along both axes. 1221 1222 1223 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 9, 2026. ; https://doi.org/10.64898/2026.02.06.704006doi: bioRxiv preprint