The squamous cell carcinoma antigen/SERPINB3 protects cervical cancer cells from chemoradiation by preventing lysoptosis

The endogenous lysosomal cysteine protease inhibitor SERPINB3 (squamous cell carcinoma antigen 1, SCCA1) is elevated in patients with cervical cancer and other malignancies. High serum SERPINB3 is prognostic for recurrence and death following chemoradiation therapy (CRT). Cervical cancer cells genetically lacking SERPINB3 are more sensitive to ionizing radiation (IR), suggesting this protease inhibitor plays a role in therapeutic response. Here we demonstrate that SERPINB3-decient cells have enhanced sensitivity to IR-induced cell death. Knock out of SERPINB3 sensitizes cells to a greater extent than cisplatin, the current standard of care. IR in SERPINB3 decient cervical carcinoma cells induces cell death morphologically consistent with necrosis, with biochemical and cellular features of lysoptosis. Moreover, rescue with wild-type SERPINB3 or a reactive site loop mutant indicates that protease inhibitory activity is required to protect cervical tumor cells from radiation-induced death. These data suggest targeting of SERPINB3 and lysoptosis to treat radioresistant cervical cancers.


Introduction
Cervical cancer remains a leading cause of cancer death for women worldwide 1 . Despite efforts to improve screening and human papilloma-virus (HPV) vaccination rates, overall mortality from the disease has not changed substantially in the last several decades 1,2 . De nitive chemoradiation is the standard of care for most women with cervical cancer, but is associated with recurrence rates as high as 30-70% 3,4 , implying that many cervical cancers display inherent or adaptive cellular resistance to these therapies.
For women who experience recurrent disease following CRT, less than 5% of women survive beyond ve years 3 . Thus, there is an unmet need to understand of the molecular mechanisms of resistance to radiation and chemotherapy in cervical cancer, with the goal of developing novel therapies for both upfront and salvage therapy.
We previously demonstrated that elevated serum squamous cell carcinoma antigen (SCCA) is independently prognostic for recurrence and survival following de nitive radiation therapy for cervical cancer 5,6 . Moreover, we found that failure of serum SCCA to normalize by the fourth week of treatment was an early indicator of failed response to therapy as indicated by positive post-therapy FDG-PET, and increased recurrence and death 6 . Clinical serum SCCA assays measure levels of SERPINB3 and SERPINB4 proteins, also known as SCCA1 and SCCA2, respectively. These proteins share high homology, and diverge in amino acid sequence primarily within the carboxy-terminal reactive site loop (RSL), which serves as a pseudo-substrate for speci c proteases 7 . Binding and cleavage of the RSL by a target protease results in a rapid conformational change in the SERPIN, preventing further processive proteolysis of the SERPIN and results in a covalently bound SERPIN-protease complex that is ultimately degraded 8 .
SERPINB3 is an intracellular cysteine protease inhibitor that is upregulated in many autoimmune diseases and cancer 5,11−16 . The Caenorhabditis elegans (C. elegans) homologue of SERPINB3, SRP-6, protects against lysosomal damage and organismal death by inhibiting cysteine protease activity induced by diverse stressors including hypotonic saline, hydrogen peroxide, DNA-damage and reactive oxygen species 17 . This phenotype can be rescued in the srp-6 null animal by driving expression of wildtype SRP-6, but not RSL-mutant SRP-6, suggesting that inhibition of cysteine proteases is necessary for cytoprotection. SERPINB4 demonstrates speci city for chymotrypsin-like serine proteases in vitro 9 , and has been shown to inhibit granzyme M-induced cell death when overexpressed in HeLa cells, but does not have other well-described cellular functions 10 .
While high levels of SERPINB3 in the serum is associated with poor response to RT in patients with cervical cancer, it is not clear if SERPINB3 impacts radiation response on a cellular level. We hypothesized that SERPINB3 serves as a radioprotective factor in cervical cancer cells. Indeed we found that cervical cancer cell lines engineered by CRISPR-Cas9-mediated gene editing of the SERPINB3 locus resulting in knock-out (B3-KO) were signi cantly more sensitive to ionizing radiation in vitro compared to isogenic control cells, as determined by clonogenic cell survival assay 11 . The molecular mechanism of protection against radiation is unknown. Murakami et al found that transient transfection of vector-driven SERPINB3 into 293T human embryonic kidney cells resulted in higher survival as measured by MTT assay up to 36 hours after treatment with radiation 18 . Examination of caspases important for apoptosis demonstrated decreased levels of caspase 9 cleavage and caspase 3 activity in cells transfected with SERPINB3. Based on these ndings the authors concluded that SERPINB3 suppresses radiation-induced apoptosis. While the regulation of caspases is important in apoptotic cell death pathways, exogenous stressors, such as irradiation, often induce multiple cellular death pathways within a population of cells. Additionally, radiation does not engage apoptosis effectively in most solid tumors due to common dysregulation of pro-apoptotic pathways 19 . Thus, precise molecular mechanism of protection by SERPINB3 against radiation remains unknown.
In the current study, we found that the mechanism of SERPINB3-mediated radioprotection in cervical cancer cells is inhibition of cell death. Moreover, we identi ed the mode of cell death induced by radiation in SERPINB3-KO cells as lysoptosis, a regulated lysosome-mediated necrosis pathway, with little to no evidence of apoptotic cell death, necroptosis, pyroptosis or ferroptosis. Additionally, the SERPINB3 RSL is required for protection against radiation, further supporting the mechanism of lysosomal protease inhibition. To our knowledge, this is the rst report that SERPINB3 protects cancer cells against radiationinduced necrosis, and does so in a lysosomal cysteine protease inhibitory manner. Moreover, we present the rst evidence that ionizing radiation induces a lysosome-mediated cell death pathway in cancer cells, particularly when a protective factor is eliminated, suggesting a potential vulnerability for targeting radioresistant cancers.

SERPINB3-KO cells are more sensitive to radiation and cisplatin-induced cell death
For these studies, a panel of cervical cancer cell lines was selected to provide the most clinically relevant models. Two cell lines with high levels of SERPINB3 (SW756 and HT3) and two with low to undetectable levels of SERPINB3 (SiHa and C33A), had varying sensitivity to radiation as determined by clonogenic cell survival, with the cell lines expressing SERPINB3 more radioresistant than those without SERPINB3 (Supplemental Fig. 1A, B). Importantly, SW756 with a single integrated copy of the HPV type 18 and SiHa with an integrated copy of HPV type 16, were selected to represent commonly HPV-associated cancers. Conversely, HT3 and C33A cell lines, which do not have HPV DNA, were selected to represent the approximately 10% of HPV negative cervical cancers (Supplemental Fig. 1A, B). Patients with HPV negative cervical cancers in some clinical series have worse outcomes following chemoradiation 20 , and in the case of head and neck squamous cell carcinoma, these cancers are more likely to be resistant to radiation treatment 21,22 . Thus, we felt it important to understand the impact of SERPINB3-KO on tumor cells of both HPV-positive and HPV-negative origins.
To determine if SERPINB3 protects cells against stress-induced death through a conserved mechanism similar to srp-6, we rst quanti ed cell death by exclusion of Sytox™ nucleic acid stain in cells treated with sham or increasing doses of radiation over a time course of several days. We found that B3-KO in both HT3 and SW756 cervical cancer cells had signi cantly higher percent cell death at every dose and every time point following treatment with IR, up to a 20% absolute difference in the SW756 background, and 50% in the HT3 background ( Fig. 1A-C). We then compared the effect of single gene knock out of SERPINB3 on radiosensitivity with or without cisplatin chemotherapy, the radiosensitizing agent used as standard of care for the treatment of patients with cervical cancer. We found that the effect of SERPINB3-KO on both cell death ( Fig. 1D-E) and clonogenic survival ( Fig. 1F-G) was greater than the radiosensitizing effect seen with cisplatin chemotherapy. Dose modifying factors (DMF, determined as the ratio of the dose resulting in 10% surviving fraction compared to control) was greater for single gene SERPINB3-KO compared to cisplatin in both SW756 (DMF 0.69 for B3-KO versus 1.01 for cisplatin) and HT3 (DMF 0.73 for B3-KO versus 0.96 for cisplatin), and cisplatin did not further sensitize B3-KO cells to radiation (DMF 0.69 vehicle versus 0.67 cisplatin for HT3, and 0.73 vehicle versus 0.75 cisplatin for SW756). SERPINB3-KO tumors are more sensitive to radiation in vivo To determine the role of SERPINB3 in a p53 mutant cervix tumor in vivo, we established subcutaneous ank xenografts of the HT3 cell line in female athymic nude mice, with either C or B3-KO clones. Not surprisingly, HT3-C tumors were signi cantly resistant to radiation, and after a single high-dose fraction of 10Gy we observed no tumor regression, and only one of ve HT3 C tumor xenografts with growth delay (Supplemental Fig. 1C). All but 1 of the tumors doubled in size by three weeks after irradiation ( Fig. 2A). A dose of 10Gy was selected based on in vitro dose response and is within the range of clinically relevant doses for cervix cancer. In contrast, all but one HT3-B3-KO xenografts showed initial tumor regression followed by sustained tumor growth delay (Supplemental Fig. 1D), and less than 50% of the tumors doubled over the course of the experiment. Additionally, four of ve tumors in the irradiated HT3-C group also developed ulcers requiring sacri ce compared to none of the irradiated HT3-B3-KO tumors (asterisks). Days to tumor doubling was signi cantly different for the HT3-B3-KO irradiated tumors but not for the HT3-C tumors (Fig. 2B).
SW756 xenograft tumors were more radiosensitive; a single fraction of 10Gy induced tumor growth delay in both C and B3-KO tumors (Supplemental Fig. 1E, F). Fewer radiated B3-KO tumors reached preradiation size compared to C tumors (Fig. 2C), and tumor doubling time was signi cantly longer in the B3-KO irradiated tumors but not the C irradiated tumors (Fig. 2D). Of note, radiation was administered to B3-KO tumors 10 days later than C tumors because the B3-KO tumors exhibited slower initial tumor growth rate (Supplemental Fig. 1E, F).
Increased radiation-induced cell death in B3-KO cells is not explained by cell cycle distribution or compromised repair of DNA double strand breaks Sensitivity to radiation-induced death can vary depending on the phase of cell cycle at the time of treatment. Therefore, we analyzed cell cycle distribution of C and B3-KO cells without RT and 48 hours after 4Gy by quantifying DNA content. Although some absolute differences were signi cant, we found no meaningful differences in cell cycle distribution either de novo or following 4Gy to explain the differential radiosensitivity ( Fig. 3A-D). Sub-G1, sham-treated S-phase (HT3 and SW756) and G2/M phase (SW756) populations were statistically different between the C and B3-KO cells; however, the differences were small in magnitude, and if meaningful, would suggest B3-KO cells should be more sensitive to RT, the opposite of our observations (Fig. 3B, D). For instance, SW756 B3-KO sham-treated cells had roughly 40% of cells in S-phase, compared to 32% in C cells; however, cells in S phase are usually more sensitive to the molecular effects of irradiation, thus one would expect the C cells would be more sensitive to radiationinduced death, which is not what we observe. In response to radiation, accumulation of cells in the G2/M phase was not different in C and B3-KO cells with the SW756 cell line background (HPV positive, p53 wild type). Cell cycle distribution did not change signi cantly following RT for either HT3 C or B3-KO cells (HPV negative, p53 mutant), consistent with aberrant cell cycle checkpoint response commonly seen in p53 mutant cells (Fig. 3A, B). While the subtle differences in baseline cell cycle distribution between B3-KO and C cells are interesting and a focus of future studies, they do not appear to result in increased radiation sensitivity of B3-KO cells.
Reduced capacity to repair double strand DNA breaks induced by ionizing radiation results in decreased clonogenic survival. Thus, in order to determine if DNA damage repair capacity contributes to radiosensitivity of B3-KO cells, we quanti ed gamma-H2AX ( H2AX) foci in C and B3-KO cells treated with sham or 2Gy RT at 30 minutes and 24 hours post treatment as a surrogate for repair of DNA-damage. As expected, increased H2AX foci were observed at the 30 minute time point in all cell lines (foci/nucleus increased 4-5 fold, Fig. 3E). Foci resolved to sham levels by 24 hours in all cell backgrounds, suggesting that the DNA-damage repair machinery is functioning similarly in both C and B3-KO cells. Interestingly, the mean number of foci per nucleus was moderately but reproducibly higher in HT3-B3-KO cells compared to HT3-C cells at 30 minutes post-2Gy, potentially indicating higher direct DNA damage in these cells, or perhaps more likely, different kinetics of response to DNA-damage (Fig. 3E). Representative uorescent images are shown in Fig. 3F. We also evaluated phosphorylation of ataxia-telangiectasia mutated (ATM) by Western blot as an indicator of response to DNA-damage; levels of phosphorylated-ATM were increased 30 minutes after 2Gy radiation, and decreased by 24 hours in all cell backgrounds (Fig. 3G, H). There was no apparent difference in 30 minute ATM phosphorylation between HT3-C and -B3-KO cells.

Cell death in B3-KO cells following RT is primarily necrotic
While ionizing radiation is known to induce cell death in tumour cells, the precise mechanism of this demise is unclear and could potentially involve multiple cell death pathways. Thus, we took a multifaceted approach to determine which cell death mechanism(s) occur in response to radiation in SERPINB3-C and B3-KO cells over time. First, we evaluated cells with transmission electron microscopy (TEM) to determine if cell death morphology induced by ionizing radiation was primarily apoptotic or necrotic. Untreated C and B3-KO cells showed cells with large nuclear:cytoplasmic ratio as expected, without apparent differences in morphology ( Fig. 4A, B, sham). In contrast, when treated with IR, phenotypic differences in C and B3-KO cells were striking. Dead B3-KO cells displayed necrotic morphology in both HT3 and SW756 ( Fig. 4A, B, 10Gy), with swollen and disintegrating nuclei (note size of B3-KO 10Gy nucleus is ~ 20-30µM in diameter compared to 10-15µM in sham-treated and C 10Gy conditions), and highly vacuolated cytoplasm. High magni cation images show breaks in the plasma membrane, cytoplasmic clearing, and largely normal mitochondrial morphology (Fig. 4A, B, inset). Irradiated HT3-C and SW756-C cells displayed enlarged cells with more frequent mitochondria, intact nuclear envelop. Blinded scoring of randomly sampled TEM images by two reviewers shows a signi cantly higher proportion of cells with necrotic morphology in irradiated B3-KO cells (~ 25%) compared to irradiated C cells (< 5%) (Fig. 4C). We also noted that the vacuoles seen in the irradiated B3-KO cells were membrane-bound, consistent with lysosomes, and in many cases the membranes were ruptured, releasing their contents into the cytoplasm (Fig. 4D, E).
Second, cells were treated with increasing doses of RT and analyzed by Western blot and uorescent microscopy at 24 hour intervals up to 96 hours. Western blots were probed for markers of multiple regulated cell death pathways including cleavage of end-effector caspase-3 and caspase-7 (apoptosis), and Poly (ADP-ribose) polymerase (PARP) cleavage products (apoptosis), gasdermin D (GSDMD) cleavage to p30 pore-forming product (pyroptosis), and phosphorylation of mixed lineage kinase domain like pseudokinase (MLKL) (necroptosis), and receptor interacting serine/threonine kinase 3 (RIPK3) (necroptosis  5A). There was qualitatively more caspase-3/7 cleavage in HT3-B3-KO cells compared to C cells, though low compared to the amount of cell death seen at those time points and compared to the apoptosis positive control (Fig. 5A, Supplemental Fig. 2C). In SW756 cells, no caspase-3 or caspase-7 cleavage was detected even 96 hours after 30Gy (Fig. 5B). We found no evidence of GSDMD cleavage, or p-MLKL, p-RIPK3/RIPK1 (markers of pyroptosis and necroptosis, respectively) in either HT3 or SW756 cells (Fig. 5A, B). Additional controls to demonstrate detection of markers by Western blot are shown in Supplemental Fig. 2C-E, including treatment with inducers and inhibitors of apoptosis (Supplemental Fig. 2C), necroptosis (Supplemental Fig. 2D), and pyroptosis (Supplemental Fig. 2E).
As radiation therapy-induced senescence can also contribute to decreased clonogenic survival, we evaluated if markers of senescence were differentially induced by RT in C and B3-KO cells. Sham-treated SW756-B3-KO cells had slightly higher percent positive senescence associated beta-galactosidase (SAβgal) cells compared to SW756-C cells (~ 15% versus ~ 10%) 24 hours after plating (sham, 0h, Fig. 5C, D); however, there was no difference 96 hours after sham-treatment, and SA-βgal positive cells were not signi cantly increased in either SW756-C or -B3-KO cells after 30Gy RT (Fig. 5C, D). HT3 cells did not have any SA-βgal positive cells in any elds of view either with sham or 10Gy RT (Supplemental Fig. 3). Similarly, Western blot analysis for BCL-2 and BAX, which are commonly increased and decreased, respectively, in senescent cells, were not signi cantly different between C and B3-KO cells, and did not change in either direction following treatment with ionizing radiation (Fig. 5E).
Since caspase-3 and caspase-7 cleavage was detected in vitro, we investigated hallmarks of apoptotic cell death in tumors in vivo following radiation. Flank xenografts were established in nude athymic mice from the HT3 cell lines as described above, randomized to sham or 10Gy radiation, and harvested 96 ultimately organismal death when exposed to various cytotoxic stressors 17 . This cell death mechanism has been further characterized and termed lysoptosis (citation for Good et al, Comms Biology). Given the conserved molecular function of SERPINB3 as an intracellular lysosomal cysteine protease inhibitor, we hypothesized that cervical tumor cells lacking SERPINB3 would be susceptible to LMP and lysosomal rupture following an insult such as ionizing radiation. TEM of irradiated B3-KO cells showed evidence of vesicular membrane rupture reminiscent of LMP in both HT3 and SW756 cells (Fig. 4D, E). To determine if lysosomal membrane integrity was lost prior to cell death or occurred as a post-mortem event, we performed live-cell time-lapse confocal microscopy to observe cells in the process of dying. Lysosomes were marked with the acidophilic dye LysoTracker™-deep red (green), and cell membrane permeability as a measure of cell death was determined by propidium iodide (red). C or B3-KO cells were treated with 10Gy and imaged during a period of expected high percent death, beginning approximately 72 or 96 hours after treatment. Dying B3-KO cells treated with radiation lost lysosomal integrity prior to loss of cell membrane integrity (Fig. 6A, C). HT3-C cells, in contrast, showed a delayed loss of Lysotracker staining after becoming PI positive (Fig. 6B, D).
Treatment of cells with the cysteine protease inhibitor E64d inhibited radiation-induced cell death in HT3-C and to a greater degree in HT3-B3-KO cells (Fig. 6E), providing further supporting evidence that lysosomal protease activity is required for cell death, particularly in the absence of SERPINB3.
The reactive site loop of SERPINB3 is required to protect cervical cancer cells against radiation-induced death and to promote tumor growth To determine if SERPINB3 is su cient to increase radioresistance in cervical cancer cells, SiHa (HPV16+/p53 wild type) and C33A (HPV-/p53 mutant) cell lines with no detectable SERPINB3 protein were used to generate stable clones expressing the wild-type SERPINB3 (B3) or an empty vector control (VC), with bicistronic expression of green uorescent protein (GFP) (Fig. 7A). Clonogenic survival was signi cantly higher in SiHa-B3 cells compared to SiHa-VC cells (DMF 1.25, Fig. 7B). C33A cells grow in a manner that is partially adherent with a population of viable cells that easily detach from the tissue culture dishes used for clonogenic survival assays, leading to high variability in the assay. Use of soft agar suspension resulted in disaggregated colonies with similar challenges. Therefore, the CellTiter-Glo® ATP-dependent reagent was employed to estimate cell viability in C33A cells. Indeed, we nd that C33A-B3 cells have higher cell viability following radiation treatment compared to C33A-VC cells, with similar ndings in the SiHa background (Fig. 7C).
To determine if the protease-inhibitory function of SERPINB3 is required for radioprotection in these cells, we generated isogenic stable clones expressing SERPINB3 with a single alanine to arginine amino acid substitution at the P14 residue (Schechter and Berger numbering scheme 23 ) of the C-terminal reactive site loop (RSL), corresponding to amino acid 341 in the hinge region (termed B3-A341R). The resultant protein blocks loop insertion after cleavage by target proteases and thus does not inhibit protease activity 7 . For functional analyses, a clone with similar levels of B3-A341R protein expression compared to the B3expressing clone was selected (Fig. 7A), and clonogenic survival and cell viability was compared to B3and VC-containing cell lines (Fig. 7B, C). Expression of B3-A341R did not protect cells from radiation and survival was similar to VC-expressing cells (DMF 0.93 compared to VC). Flank xenograft tumors of the C33A cell lines were established to determine effect of B3 and B3-A341R on tumor growth and in vivo radiation resistance. Sham treated C33A-B3 tumors overall grew faster than VC or B3-A341R tumors (Fig. 7D). After tumor establishment, half of the mice were randomized to tumor-directed radiation, and B3 tumors continued to grow after a single dose of 10Gy compared to VC and B3-A341R tumors (Fig. 7E). Tumor doubling time was not signi cantly different in sham-treated versus 10Gy irradiated B3-expressing tumors, whereas time to tumor doubling was signi cantly longer in irradiated VC-and A341R-containing tumors (Fig. 7F). Images of dissected tumors to portray relative tumor size at the time of sacri ce for sham irradiated and 10Gy treated tumors are shown in Fig. 7G. Taken together, these data suggest that SERPINB3 protects cells from ionizing radiation by inhibiting target lysosomal cysteine proteases.

Discussion
Since its isolation from human cervical tumors, SCCA has been shown to serve as a strong and consistent indicator of poor outcomes in cervical cancer and other types of cancers. While others have shown that SCCA1/SERPINB3 promotes tumor growth and resistance to cytotoxic agents, these data are the rst to show that SERPINB3 is directly responsible for radiation resistance of cervical cancer cells by inhibiting a lysosome mediated necrotic cell death pathway. Moreover, we show that the effect of SERPINB3 loss on radiation sensitivity in cervical tumor cells is similar if not greater than cisplatin, currently used as the standard of care to radiosensitize cervical cancer. These ndings suggest that SERPINB3 may serve as a therapeutic target for radiosensitization of resistant cervical cancers. Additionally, the data also implicate a lysosome-dependent regulated cell death pathway in radiationinduced cell death and resistance.
Ionizing radiation has a host of direct and indirect cellular effects on tumor cells ultimately leading to recovery and survival, or cell death. Direct damage to DNA and other macromolecules is known to lead to the induction of apoptosis primarily in hematologic cells, but apoptosis is not a predominant mode of cell death in most solid tumors. Following IR, the inability of the cell to undergo mitosis secondary to dysfunctional cell cycle checkpoints and accumulated unrepaired DNA-damage, entering a state known as mitotic catastrophe, thought to ultimately lead to cellular demise by secondary death pathways, or to enter a senescent state 24 . However, the dynamics of radiation-induced cell death, and the exact molecular mechanisms contributing to the cell fate decision are complex. The most recent recommendations from the Nomenclature Committee on Cell death describes myriad cell death subroutines, including cellintrinsic and -extrinsic modes 25 . Importantly, the committee stresses that there is often a great deal of interconnectivity between signal transduction cascades leading to one cell death mode versus another. Therefore, in order to truly understand the relative contributions of various lethal regulated cell death programs to radiation-induced tumor cell kill, it is critical to evaluate not only the presence of individual markers of cell death pathways at single time points and under single conditions, but multiple markers of suspected death modes at varying time points and dose levels. As such, in this study we undertook a broad evaluation to determine the intrinsic cell death routines contributing to radiation-induced death in SERPINB3-KO cells compared to control cells. The goal was to identify vulnerabilities exposed by the loss of SERPINB3 that can be exploited for therapeutic radiosensitization.
The rst nding was the prevalence of necrotic morphology in SERPINB3-KO cells following ionizing radiation. Although some end-effector caspase cleavage was detected by Western blot analysis, the amount was qualitatively low compared to the percent cell death observed. Pyroptosis is a form of necrotic regulated cell death typically induced by microbial pathogens, whereby cleavage of gasdermin D (GSDMD) by caspase-1, or in some cases gasdermin E (GSDME) cleavage by caspase-3, mediates its localization to the cell membrane. GSDMD-N or GSDME-N oligomerization forms pores resulting in rapid permeabilization of the cell membrane 26 . Despite some caspase-3 cleavage, we found no evidence of GSDMD or GSDME cleavage following radiation in either B3-C or B3-KO cells. Ferroptosis is another necrotic RCD which occurs independently of caspase activation and relies heavily on generation of ROS, a hallmark of ionizing radiation 27 . Nevertheless, ferrostatin-1 did not inhibit radiation-induced cell death in either C or B3-KO cells, and TEM images of necrotic cells revealed largely intact mitochondria suggesting ferroptosis is not a predominant form of radiation-induced cell death in B3-KO cells.
In C. elegans, the orthologue of SERPINB3, srp-6, serves as an intracellular cysteine protease inhibitor and was previously shown to protect animals against diverse toxic stressors by inhibiting lysosomal proteases and subsequent animal death 17 . In addition to hypo-osmotic conditions and heat shock, we also demonstrated that srp-6 null animals were more sensitive to oxidative stress and hypoxia. Additionally, several groups have shown that DNA-damage induced by topoisomerase inhibitors camptothecin and etoposide causes lysosomal membrane permeability 28, 29 . Therefore, we hypothesized that radiation, which causes both DNA damage and accumulation of oxidative species, might induce LMP in cervical cancer cells, particularly in cells lacking SERPINB3. Indeed, we observed TEM evidence of lysosome-like vesicle membrane rupture in B3-KO tumor cells treated with ionizing radiation, and loss of lysosomotropic staining by live cell microscopy in dying cells, similar to that seen in srp-6 null C. elegans. Further supporting the idea that SERPINB3 lysosomal protease inhibitor activity is responsible for its ability to protect cells against radiation, we found that the pharmacologic cysteine protease inhibitor E64d abrogated cell death in B3-KO cells. Additionally, expression of wild type but not the RSL mutant B3-A341R protected SERPINB3-low cervical cancer cells from radiation further supporting the importance of an intact RSL to bait-and-trap target lysosomal proteases released from the lysosome after exposure to ionizing radiation.
While the DNA-damage repair capacity appears intact in B3-KO cells, suggesting that inability to repair DNA-damage is not contributing to the radiosensitivity observed in these cells, we did observe that the initial quantity of H2AX foci, which is a surrogate for number of double strand DNA breaks, was higher in B3-KO cells speci cally in the HT3 background. The mechanism of this is not clear and is under current scrutiny; however, the myeloid and erythroid nuclear termination protein (MENT), a chicken Clade B or intreacellular SERPIN, is associated with heterchromatin 30 , thus there is precedence for nuclear function of SERPIN intracellular family proteins. Moreover, cysteine cathepsin proteases have been reported to localize to the nucleus 31,32 . It is conceivable that nuclear cathepsins may have a role in vulnerability to DNA-damage that is modi able by the presence or absence of SERPINB3. Perhaps more likely is a slightly different kinetics of foci formation in the C and B3-KO cells that will be discerned in future detailed analysis.
Importantly, we found similar effects of SERPINB3 on radiation-induced cell death and resistance in cervical tumor cell lines that are HPV positive and negative, and p53 wild type and mutant. These ndings suggest that radiation-induced lysosome mediated necrosis proceeds in a manner that is not dependent on functional p53.
The major limitation of the current study is the experimental di culty posed by the long time course of radiation-induced death. Instead of minutes to hours required for classic cell death inducers to result in cellular demise, most non-hematologic cells treated with ionizing radiation do not die for at least 48 hours after treatment, and in many cases several days after a single treatment. While we made every effort to capture an accurate view of cell death modes during each phase of this long time course, and in response to various doses of radiation, it is possible that less prevalent cell death events were not captured by Western blot, live cell imaging and TEM assays. Additionally, the time required to observe a single cell   11 . In brief, a single-vector lentiviral system was used, driving expression of Cas9 and the gRNA sequence. KO was con rmed by sequencing of the SERPINB3 gene, Western blot showing no protein product, and sequencing of most-likely off-target genes to con rm speci city. For the current paper, SiHa and C33A cell lines were engineered to stably express pULTRA (Addgene 24129) mammalian vector driving expression of the wild type SERPINB3 gene, or SERPINB3 mutant encoding a single alanine to arginine substitution at amino acid 341 (B3-P14m). Clonogenic cell survival assay 500-1000 cells per well were seeded in 6-well plates 24 hours prior to treatment with increasing doses of radiation (2, 4, 6 Gy x 1) and incubated for 1-3 weeks until control plates formed visible colonies (≥ 50 cells). IC50 concentration of cisplatin was added 1 hour prior to radiation, where indicated. Plates were xed and stained with 0.5% Crystal Violet, 30% Methanol, 10% Acetic Acid, 60% ddH2O for 30 minutes, rinsed in tap water and air dried at room temperature. Surviving fraction was calculated as the number of colonies ÷ (500 * plating e ciency) and plotted on a log 10 scale as per convention. The linear quadratic equation was t to each dataset using GraphPad Prism 8©. The dose modifying factor (DMF) was determined as the ratio of the dose resulting in 10% surviving fraction compared to control, indicated as the reference condition.
Cell Titer-Glo survival assay Cells were seeded to 50-70% con uence in 96-well dishes and cultured overnight prior to treatment with increasing doses of ionizing radiation. 24  In vivo tumor growth and radiation response

Tumor tissue analysis by histology and immunohistochemistry
At tumor harvest, mice were sacri ced using a CO 2 asphyxiation chamber and tumors immediately harvested and divided into three components for ash freezing, into 4% paraformaldehyde (PFA), and xation buffer for TEM as described above. PFA-xed tumor was sent and histology was performed by HistoWiz Inc. (histowiz.com) using a Standard Operating Procedure and fully automated work ow.
Samples were processed, embedded in para n, and sectioned at 4µm. Immunohistochemistry was performed on a Bond Rx autostainer (Leica Biosystems) using standard protocols. Antigen retrieval method was heat induced epitope retrieval (HIER) at pH = 6.0 for 20 minutes (CC3) and enzyme digestion for 10 minutes (TUNEL). Antibodies used were anti-cleaved caspase-3 (Asp175) (Cell Signaling, 9661S,

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.