In vitro and in vivo evaluation of paclitaxel-induced release of apoptotic 1 biomarker ccCK18 to guide treatment optimization in ovarian cancer

Background: Cytokeratins hold potential as biomarkers due to their epithelial specificity, abundance and 33 cleavage by caspases during apoptosis. We evaluated paclitaxel-induced circulating caspase-cleaved (cc) 34 cytokeratin 18 (CK18) as potential apoptotic cell-death marker to guide treatment optimization in ovarian 35 cancer. Methods: Six ovarian cancer cell lines (SK-OV-3, SK-OV-3lucIP1, Caov-3, NIH:OVCAR-3, PA-1 and PM-LGSOC- 37 01) were exposed in vitro to paclitaxel (PTX, 0 to 1000 nM) for 24h. Extracellular levels of ccCK18 were 38 measured until 5 days after drug exposure. Cell count and ccCK18 release data were analyzed using a phase- 39 nonspecific pharmacodynamic model implemented in NONMEM®. PA-1 and SK-OV-3lucIP1 xenografted 40 female SCID/Beige mice received a placebo or single dose of 50 mg/kg PTX intraperitoneally. Response to 41 PTX was evaluated in vivo using tumor volume and released ccCK18 levels. 42 Results: In vitro, the correlation between cell count and released ccCK18 levels was present in all cell lines 43 (Spearman ’s rank correlation coefficient > 0.64). Tumor volume and ccCK18 longitudinal dynamics were 44 markedly different for controls and PTX-treated PA-1 xenografts with changes in ccCK18 release preceding 45 changes in tumor volume. For SK-OV-3lucIP1 xenografts, no differences were found between controls and 46 PTX-treated mice. Conclusions: An association between PTX-induced ccCK18 release and cell count was demonstrated in vitro. later on treatment needs further refinement.


Conclusions:
An association between PTX-induced ccCK18 release and cell count was demonstrated in vitro. 48 The in vivo study supported the presence of an early-apoptotic peak in ccCK18 levels compared to a later 49 observed effect on tumor volume in PTX-sensitive xenografts. Given the heterogeneous character of ovarian 50 cancer, application and implementation of ccCK18 in a clinical setting to optimize or personalize cancer 51 treatment needs further refinement. 52 53 KEYWORDS: 54 ovarian cancer, paclitaxel, caspase-cleaved cytokeratin 18, apoptosis, treatment optimization 55 II.

Background 56
As a member of the intermediate filament protein family, cytokeratin 18 (CK18) belongs to the type 1 acidic 57 keratins and is expressed in a variety of single-layered and simple epithelia [1]. Besides its role in tumor cell 58 behavior and numerous cellular processes such as apoptosis and cell cycle progression, CK18 is also involved 59 in providing the intracellular scaffolding that structures the cytoplasm, supports normal mitochondrial 60 structures and resists stress applied from the outside of the cell [2,3]. Furthermore, this keratin is also used 61 as an epithelial marker in tumor pathology [4]. 62 In contrast to its filament partner cytokeratin 8 (CK8), CK18 is a substrate for caspase cleavage during 63 where N refers to the cell number, t is time, kng is the net growth rate constant and Emax refers to the 132 maximum PTX-induced treatment effect. Conc refers to the concentration of PTX to which cells were 133 exposed and C50 denotes the concentration at which half of the maximum effect is reached. 134 There was also a term base_count included in this model to take into account the estimated cell numbers 135 present before treatment with PTX. Population parameters and error variance were estimated using the 136 first-order (FO) estimation routine. 137 g. Phase-nonspecific pharmacodynamic model for released ccCK18 levels 138 The ccCK18 release data were also analyzed using a phase-nonspecific pharmacodynamic model with 139 NONMEM® (version 7.3.0, ICON, Hanover, MD, USA). The term kprod was used here as the net production 140 rate constant of extracellular ccCK18. Parameters such as the maximum amount of released marker (Amax) 141 and the Hill coefficient γ, to describe the steepness of the relationship between PTX concentration and 142 response, were added. A refers here to the extracellular amount of ccCK18. 143 Also here the base_amount term was included in the model to estimate the initial amount of released 145 ccCK18 before exposure to PTX. Population parameters and error variance were estimated using the first-146 order (FO) method. 147

IN VIVO STUDY 148
a. PA-1 and SK-OV-3LucIP1 xenograft models 149 The human ovarian cancer cell lines PA-1 and SK-OV-3LucIP1 cells were used to xenograft SCID/Beige mice 150 (mice were commercially obtained from Envigo, The Netherlands). SK-OV-3LucIP1 cells were cultured in 151 ATCC-formulated Dulbecco's Modified Eagle's Medium (DMEM) and PA-1 cells in ATCC-formulated Eagle's 152 Minimum Essential medium (EMEM). All growth media were completed by adding 10% fetal bovine serum 153 (FBS) and antibiotics (penicillin/streptomycin). Cells were incubated at 37 °C with 5% CO2 in air. 7 to 9-week-154 old female SCID/Beige mice were unilaterally subperitoneally injected with 1×10 6 cancer cells (1:1 serum free 155 medium:Matrigel (Corning, The Netherlands). In total, 20 PA-1-xenografted mice and 30 SK-OV-3LucIP1-156 xenografted mice where studied. The control groups were based on 10 PA-1-and 15 SK-OV-3LucIP1-157 xenografted mice, respectively. The maximum caging density was 6 from the same experimental group. All 158 mice were maintained on a regular diurnal lighting cycle (12:12 light:dark) with ad libitum access to food and 159 water. Mice were monitored once every day. 160 b. Single dose study 161 Mice were randomly assigned to the placebo or treatment group after subperitoneal tumor cell injection. 162 PTX (Abraxane®, Celgene, US) was intraperitoneally (ip, 50 mg/kg) administered to the treatment group 2.5 163 weeks after tumor cell injection whereas the control group received an intraperitoneal injection of 0.9% 164 NaCl (placebo) in an equal volume at the same time. Whole blood was collected via cardial puncture in 165 K3EDTA-treated Sarstedt (B.V.B.A. Berchem, Belgium) Microvette® 200 tubes, prior to obtain plasma. 166 Samples were collected up to 2 weeks after PTX administration to measure ccCK18 levels. Blood collection 167 via cardial puncture was considered a terminal procedure before mice were euthanized using an isoflurane 168 overdose prior to cervical dislocation. Every individual mouse was considered an experimental unit in this 169 study. The animals were excluded from the study if no tumor was present 2.5 weeks after tumor cell 170 Compared to the control solution, PTX concentrations of 1 to 1 000 nM were found to affect cell counts over 189 time for all tested cell lines. Table 1 shows the estimated parameters, their relative standard errors (RSE) and 190 95% confidence intervals (CIs) for the phase-nonspecific models on cell count. 191 All timepoints, up to 120h post treatment, were taken into account in this model for all cell lines, except for 192 SK-OV-3. As for the SK-OV-3 profiles over time only a clear difference between PTX concentrations was 193 noticed starting 96h post treatment, only the last two timepoints were included in the model. 194 From Table 1 we see that all parameters are estimated with good precision except for C50 in SK-OV-3 and SK-195 OV-3lucIP1 cells as the large confidence intervals (CIs) are indicative for a high level of uncertainty. The 196 estimated C50s for SK-OV-3lucIP1 and SK-OV-3 were 85.2 nM and 6 080 nM, respectively, suggesting that SK-197 OV-3lucIP1 is less sensitive compared to the other studied cell lines but with an estimated C50 within the 198 tested concentration range whereas SK-OV-3 is rather insensitive to PTX with an estimated C50 outside the 199 tested concentration range. To avoid difficulties with the estimation fitted, a net growth rate parameter was 200 included for each cell line instead of separate rate constants for natural cell growth and cell death. The 201 estimated net growth rate constant was lowest in SK-OV-3 (0.0036 h -1 ) and highest in SK-OV-3lucIP1 (0.0164 202 h -1 ). For SK-OV-3, we were not able to separately estimate Emax and C50 due to the limitations of the data, 203 hence Emax was fixed to the mean Emax of all five other cell lines. The estimated Emax parameter was lowest in 204 PM-LGSOC-01 (0.0151 h -1 ) compared to similar estimates of about 0.025 h -1 for the other cell types. We also 205 found that the estimated base_count parameter, representing the amount of cells that were present prior to 206 exposure to PTX, was different across cell lines. Based on these estimates, cell division and growth is 207 The estimated values for the fitted phase-nonspecific pharmacodynamic model on released ccCK18 levels 214 are presented in Table 2, accompanied by the percent relative standard error (RSE) and 95% confidence 215 intervals (CIs) for all parameters. This table illustrates a slower release rate for ccCK18 (Kprod) in PM-LGSOC-216 01 (0.0155 h -1 ) and PA-1 (0.0177 h -1 ) cells compared to the other cell lines. The estimated Emax, the effect of 217 PTX on release of ccCK18 was highest in PM-LGSOC-01 (0.0428 h -1 ) followed by PA-1 (0.02 h -1 ), SK-OV-3 218 (0.0128 h -1 ), NIH:OVCAR-3 (0.0087 h -1 ), Caov-3 (0.008 h -1 ) and SK-OV-3lucIP1 (0.0076 h -1 ). An almost 11-fold 219 difference in C50 value was observed between NIH:OVCAR-3 (8. proportionally larger increases in ccCK18 for PM-LGSOC-01 and SK-OV-3. This is seen from Figure 2 where 233 most data points lie above the identity line. In contrast, reductions in cell count, at higher PTX levels, are 234 accompanied with less than proportional increases in ccCK18 for NIH:OVCAR-3. As observed from Figure 2, 235 at later timepoints, ccCK18 levels start decreasing or reach a plateau. 236

IN VIVO STUDY 237 a. Relationship between tumor volume and released ccCK18 levels 238
Due to the observed differences in sensitivity to PTX and ccCK18 release in vitro, xenografts based on PA-1 239 and SK-OV-3lucIP1 cells were selected to be studied in vivo. Figure  As many chemotherapeutic agents cause apoptosis, the potential of ccCK18 as a quantitative 257 pharmacodynamic biomarker and hence, a helpful decision tool in treatment optimization was evaluated 258 here. In this study, next to observed differences in sensitivity to PTX, the correlation between cell count and 259 apoptotic cell death marker ccCK18 was shown in ovarian cancer cell lines SK-OV-3lucIP1, PA-1, NIH:OVCAR-260 3, PM-LGSOC-01, SK-OV-3 and Caov-3. Based on the in vitro results, SK-OV-3lucIP1 and PA-1 xenografts were 261 studied in vivo representing, respectively, a model less sensitive and sensitive to PTX. In vivo, a correlation 262 between released ccCK18 levels and tumor volume was observed in PA-1 xenografts whereas no clear effect 263 of PTX was found on tumor volume or ccCK18 levels over time in the SK-OV-3lucIP1 xenografted mice. 264 The C50 values observed, based on the in vitro model, are hard to compare with previously published work 265 due to differences in experimental conditions (such as drug concentration and exposure time). Compared to 266 our C50 of about 6 000 nM for the SK-OV-3 cells, Au et al. [17] found a C50 equal to 5 nM for SK-OV-3 exposing 267 the cells 24h to PTX and measuring the drug effect after 96h. Possible reasons for the variability in observed 268 C50 value for SK-OV-3 in both studies could be related to cell cycle, confluency and/or passage effects [18]. It 269 is also important to highlight that in this experiment C50 was estimated considering a 1 week period whereas 270 commonly only one single timepoint is used. In addition, the difference in C50 values reported between cell 271 types can also be explained by heterogeneous responses to anticancer treatments [19]. 272 Brandt and colleagues (2010)[7] concluded that circulating levels of ccCK18 might be a useful biomarker to 273 monitor treatment response in patients with gastrointestinal cancers. A correlation between intact and 274 caspase-cleaved forms of CK18 levels and clinical response to therapy in breast cancer was observed by 275 Olofsson et al. (2007)[8]. In patients with disseminated testicular germ cell cancer, de Haas et al. (2008)  276 observed changes in total and caspase-cleaved CK18 during chemotherapy [9]. As a result, measuring soluble 277 keratin protein fragments in the clinic is believed to be of great value for the early detection of tumor 278 progression, metastasis formation and for a quick evaluation of the therapeutic response in epithelial 279 malignancies [20]. Outside the field of oncology, serum CK18 levels have also been studied as cell death 280 markers in (alcoholic) hepatitis [21,22] and in patients with cirrhosis[23] or drug-induced liver injury [24] . 281 In our in vitro study, at later time points, deviating biomarker release patterns were observed which can be 282 explained by the loss of M30 reactivity as reported in apoptotic cells [25] or by the different ongoing 283 processes during cell death [26]. As PTX induces cell death processes, Lieu at al. [27] described the 284 apoptogenic mechanisms after PTX treatment to be mitotic arrest and microtubule damage, the first one 285 induced at PTX concentrations below 200 nM and the last one at higher PTX concentrations. As a result, high 286 concentrations of PTX are able to induce apoptosis independently of mitotic arrest at any phase of the cell 287 cycle. 288 Next to differences in CK18 expression at the cell level [28], also variation in population doubling time 289 between these different cell lines may explain our observed differences in ccCK18 release post therapy. As a 290 result, a higher number of cell divisions will be accompanied by a longer duration of drug present to a larger 291 number of cells. Next to an extended drug effect, the previously mentioned phenomenon might also cause 292 an increase in the specific cell fraction subjected to PTX toxicity [17]. In addition, the cell fraction passing 293 through the mitotic phase will also play a crucial role with regard to the release of ccCK18, since PTX-induced 294 cell cycle arrest and apoptosis are mainly occurring in the mitotic phase [29]. We did not discriminate 295 between the G1, S, G2 and M phase in this study in accordance with the clinical setting in which equally no 296 cellular analysis of the complete tumor is possible before treating the patient. 297 Based on our in vivo observations, it can be concluded that the release of ccCK18 into the blood circulation 298 happens early, before the effect of PTX on tumor volume was observed in PA-1 xenografts. This is also in 299 agreement with the early-stage apoptosis reflected by the M30 ELISA as M30 does not measure late 300 apoptosis stages including generation of secondary CK18 caspase cleavage products [30]. Induction of 301 apoptosis by chemotherapeutic agents is typically characterized by slow kinetics over 24 hours. Regarding 302 paclitaxel, as this compound is retained in tumors for over 5 days [31], CK18 is typically cleaved after more 303 than 12 hours of drug exposure [32]. Levels of ccCK18 are likely a better marker to inform on PTX-induced 304 tumor cell death for the PA-1 group compared to the SK-OV-3lucIP1 group. Nevertheless, the small number 305 of animals used did not allow to draw firm conclusions on the potential of ccCK18 to provide significant 306 information regarding chemotherapeutic responses. 307 Given the complexity of a disease like ovarian cancer and the variety of cell death processes, we believe 308 another potential explanation for our in vivo findings, i.e. to consider ccCK18 a better marker for effects of 309 PTX on PA-1 cells compared to SK-OV-3lucIP1 cells, might be related to the involvement of Akt [33]. The 310 serine-threonine kinase Akt is known to exert anti-apoptotic effects through several downstream targets. It 311 is also known that Akt is cleaved during mitochondrial-mediated apoptosis in a caspase-dependent manner. 312 The observation that cleavage of Akt occurs during apoptosis suggests that either a level of baseline Akt 313 signaling is vital for cell survival or that Akt activation occurs during apoptosis and acts as a 'brake' on this 314 process [34]. Since it is known that Akt inactivation sensitizes human ovarian cancer cells to paclitaxel, Kim 315 and colleagues (2007)[33] explored the difference in Akt phosphorylation levels between PTX-sensitive PA-1 316 cells and PTX-resistant SK-OV-3 cells. In their study, a higher level of phosphorylated Akt in SK-OV-3 cells 317 compared to PA-1 cells was shown. As a result, the link between Akt activity, PTX resistance and paclitaxel-318 induced apoptosis was hypothesized [35]. Our in vivo findings are also in line with this as we concluded that 319 SK-OV-3lucIP1 is associated with less ccCK18 release as a result of less paclitaxel-induced caspase cleavage of 320 CK18 compared to PA-1. Taking into account the higher Akt activity in SK-OV-3 cells observed by Kim et al. 321 (2007)[33], mediation of survival signals by Akt might preserve SK-OV-3lucIP1 cells from apoptotic pathways. 322

VI.
Conclusion 323 In conclusion, our in vitro study showed evidence of an association between PTX-induced cell toxicity and 324 released ccCK18 levels. Based on our in vivo study with PTX-sensitive xenografts, we found that response to 325 PTX was immediately shown as a peak in released ccCK18 whereas an observed decrease in tumor volume 326 was rather delayed. Overall, these experiments indicate the potential of ccCK18 to inform on drug-induced 327  count and released ccCK18 levels. These graphs are based on the predicted data from the different phase-457 nonspecific pharmacodynamic models. On the x-axis cell count is represented as the following ratio: 458 prediction for control (0 nM) over prediction for a specific PTX level, M30 on the y-axis represents ccCK18 459 using the ratio calculated from prediction for a specific PTX level over prediction for control (0 nM). The grey line in all graphs illustrates the line of identity. From left to right, the symbols are directly related to the 461 specified time points studied (0, 24, 48, 72, 96, 120 and 144h). This figure was made using GraphPad Prism 8 462 (GraphPad Software, Inc., USA). 463