Modulation of UVB-induced Carcinogenesis by Activation of Alternative DNA Repair Pathways

The molecular basis for ultraviolet (UV) light-induced nonmelanoma and melanoma skin cancers centers on cumulative genomic instability caused by inefficient DNA repair of dipyrimidine photoproducts. Inefficient DNA repair and subsequent translesion replication past these DNA lesions generate distinct molecular signatures of tandem CC to TT and C to T transitions at dipyrimidine sites. Since previous efforts to develop experimental strategies to enhance the repair capacity of basal keratinocytes have been limited, we have engineered the N-terminally truncated form (Δ228) UV endonuclease (UVDE) from Schizosaccharomyces pombe to include a TAT cell-penetrating peptide sequence with or without a nuclear localization signal (NLS): UVDE-TAT and UVDE-NLS-TAT. Further, a NLS was engineered onto a pyrimidine dimer glycosylase from Paramecium bursaria chlorella virus-1 (cv-pdg-NLS). Purified enzymes were encapsulated into liposomes and topically delivered to the dorsal surface of SKH1 hairless mice in a UVB-induced carcinogenesis study. Total tumor burden was significantly reduced in mice receiving either UVDE-TAT or UVDE-NLS-TAT versus control empty liposomes and time to death was significantly reduced with the UVDE-NLS-TAT. These data suggest that efficient delivery of exogenous enzymes for the initiation of repair of UVB-induced DNA damage may protect from UVB induction of squamous and basal cell carcinomas.

UVB-induced Carcinogenesis. A total of 40 SKH1 hairless mice were assigned to 4 treatment groups in the following experimental design: 1) empty liposome control, 2) UVDE-TAT, 3) UVDE-NLS-TAT, and 4) cv-pdg-NLS. An additional 10 mice were not treated with liposomes or UVB to assess frequencies of spontaneous skin lesions; however, these mice never developed any spontaneous skin tumors and are not included in further analyses. Beginning at 8 weeks of age, 0.2 mL of each liposomal formulation was uniformly applied to the dorsal skin of each mouse using a pre-moistened cotton swab. This amount was absorbed to apparent dryness Scientific RePORtS | (2018) 8:705 | DOI: 10.1038/s41598-017-17940-8 within 3-5 min after application. After 1 hr, mice were exposed to increasing doses of UVB irradiation in an 8-compartment chamber that was covered with a quartz plate to allow full UVB penetrance. This chamber maximized exposures to the backs of the mice.
To prevent significant skin irritation and burning, a UVB dose escalation strategy was used for the initial irradiations. As anticipated, following 9 UVB exposures (a total exposure of 2.0 kJ/m 2 ) over the course of 3 weeks, the dorsal skin showed evidence of mild redness, slight loss in elasticity, and thickening, without any evidence of sunburn. All mice continued to receive average weekly doses of ~22 kJ/m 2 and all mice were monitored three times per week for skin lesions. The earliest time for the formation of confirmed squamous cell carcinomas in the control empty liposome group was at 15 weeks, with a total cumulative exposure of ~217 kJ/m 2 . Analyses of the length of time to form the first tumor (≥2 mm diameter) in each irradiated mouse revealed that although there were trends for a delay in time to first tumor in all 3 active enzyme groups (shown in Fig. 2a,b and quantified by the median in Table 1), these delays were not statistically distinguishable from the control group, with nonsignificant p-values of 0.404 for UVDE-TAT, 0.788 for UVDE-NLS-TAT, and 0.668 for cv-pdg-NLS.
Following 23 weeks of UVB exposures, totaling 311 kJ/m 2 , all mice were photographed for comparative analyses. Representative images of mice treated with control empty liposome, UVDE-TAT, and UVDE-NLS-TAT are shown in Fig. 3, Panels a-c. Visual evaluation of the mice treated with control empty liposomes versus any of the enzyme-containing liposomes revealed a greater involvement of the dorsal surface area with various tumors, with 4 of the 10 mice having tumors that were >3 mm. In contrast, mice treated with liposomes containing an enzyme showed less severe damage, and no tumors >3 mm observed in either form of UVDE, and only one tumor >3 mm in the cv-pdg-NLS group. These observations indicated a suppressive effect on tumor formation with the treatment of either of the repair enzymes.
The total tumor burden (cumulative size of all tumors) for each mouse was analyzed at various time points, with the data for week 23 summarized in group-specific box plots for UVDE-TAT and UVDE-NLS-TAT (Fig. 4a) and for cv-pdg-NLS (Fig. 4b). At 23 weeks, the mean total tumor size per mouse for the empty, UVDE-TAT, UVDE-NLS-TAT, and cv-pdg-NLS liposome-treated groups were 12.2 mm, 2.9 mm, 2.1 mm, and 4.2 mm, respectively. The mean differences in total tumor size for UVDE-TAT, UVDE-NLS-TAT, and cv-pdg-NLS relative to the control group were not statistically significant (p = 0.092, 0.058, and 0.179, respectively). The empty-liposome treated group had a total of 28 tumors with an aggregate total tumor size of 122 mm at 23 weeks. In contrast, the UVDE-TAT, UVDE-NLS-TAT, and cv-pdg-NLS groups had 18, 15, and 19 tumors, respectively, and aggregated tumor sizes of 29 mm, 21 mm, and 42 mm, respectively. The 23-week time point was chosen for analyses because only 1 mouse (in the control empty liposome treated group) had developed a tumor that required the mouse to be euthanized. In this, and subsequent analyses of total tumor burden, mice that were euthanized prior to a specified time point were assigned tumor size values equal to their last measurements while alive. However, we acknowledge that using the final observed values of euthanized mice for these analyses is an underestimate of total tumor burden as it does not account for the anticipated tumor growth that would likely have continued to occur in these animals. Imputation of these final measurements at post-death time points was important since exclusion of euthanized mice would bias the total tumor size comparisons, since mice with the largest tumor burdens (>8 mm for a single tumor) would be unavailable for the between-group analyses.
Comparable analyses were also made at 33 weeks with these data summarized in Fig. 4c,d. The mean total tumor size per mouse was ~25 mm in the control group, ~12 mm in both the UVDE treatment groups, and 18.3 mm in the cv-pdg-NLS group. These tumor burden averages, when compared to empty liposome controls, were statistically significant for UVDE-TAT (p = 0.041) and UVDE-NLS-TAT (p = 0.045), but not for cv-pdg-NLS (p = 0.541). These findings demonstrate that, relative to the control empty liposomal treatment, application of liposomes containing UVDE-TAT or UVDE-NLS-TAT reduces the UVB-induced total tumor burden.
In addition, analyses were performed concerning time-to-death for each of the groups. To compare the risk of tumor-induced euthanasia for the three active enzyme groups relative to the control group, hazard ratios (HR) were estimated from a Cox regression model. Assuming proportional risks of euthanasia across all time points, estimated hazard (risk) ratios involving the control group were 0.35 (95% CI: 0.11, 1.05) for UVDE-TAT, 0.25 (95% CI: 0.07, 0.81) for UVDE-NLS-TAT, and 0.56 (95% CI: 0.21, 1.55) for cv-pdg-NLS. Thus, UVDE-NLS-TAT mice had 75% less (95% CI: 93% to 19% less) risk of being euthanized than control mice during the treatment period. Without having to assume a constant risk ratio across time, analyses of Kaplan-Meier survival curves of control vs UVDE-TAT and UVDE-NLS-TAT mice revealed survival advantages in these active treatment groups, with log-rank test p values of 0.135 and 0.037, respectively (Fig. 5a). A comparable analysis of another encapsulated enzyme compared to controls (Fig. 5b) revealed less of a survival advantages for cv-pdg-NLS (p = 0.598) than was observed for the two UVDE groups.
At the time when mice were euthanized, skin strips were harvested and processed for H&E staining. Even though the total tumor burden at 33 weeks and the time-to-death analyses revealed significant differences,   histologic analyses of the tumors that formed were indistinguishable among the 4 groups. All examined tumors were asymmetric and poorly circumscribed epithelial neoplasms, characterized by irregular aggregates extending into the superficial to deep dermis or, in some cases, into the subcutis and underlying muscle (Fig. 6). Atypical keratinocytes had enlarged, hyperchromatic and pleomorphic nuclei, were undergoing mitoses or apoptosis, and exhibited variable levels of eosinophilic cytoplasms. These features are highly characteristic of SCCs.

Discussion
The temporal hierarchy of NER of UV-induced dipyrimidine photoproducts in mammalian cells is characterized by not only rapid recognition and excision of 6-4 PPs (which are preferentially formed in open chromatin regions), but also preferential repair of CPDs in actively transcribed genes [22][23][24] . The remaining dipyrimidine photoproducts are repaired at greatly reduced rates that can extend over several days following a UV dose equivalent to a minimal sunburn. If cytosine bases are at either the 5′ or 3′ position or both, these damages can undergo spontaneous deamination to uracil that is highly mutagenic if replicated. CPDs remaining in the genome also function as one of the main causes of UV-induced immune suppression 25,26 . Thus, the potential deleterious effects of unrepaired DNA damage have the capacity to increase genomic instability and reduce immune surveillance, both of which increase the risk of cellular transformation. A strategy to mitigate these issues is to enhance DNA repair capacity in damaged cells by activating an alternative DNA repair pathway. Since less complex organisms possess alternative pathways to repair pyrimidine dimers, such as photoreactivation, BER, and NIR, the methodological challenge is to efficiently deliver sufficient levels of a repair enzyme to activate analogous alternative pathways in mammals. In the case of photoreactivation, there is not only the issue of delivering CPD-and 6-4 PP-specific photolyases, but also establishing conditions required to irradiate with sufficient amounts of appropriately tuned wavelengths of visible light. To avoid the problems associated with photoreactivation, the delivery of CPD-specific DNA glycosylase/AP lyases to initiate BER, or the introduction of UV endonucleases to activate NIR, offer feasible alternatives.
Differences in substrate specificities allow each enzyme to provide specific advantages, with pdgs recognizing only CPDs and ring-fragmented purines, and UVDEs removing both 6-4 PPs and CPDs, but not the ring-fragmented purines. We anticipated that initiation of repair of both dipyrimidine photoproducts may have important biological implications. Since NEIL1-and OGG1-initiated repair of the ring-fragmented purines is likely to be sufficient to minimize the effects of these specific UVB-induced DNA lesions 27 , it is hypothesized that the repair of both types of dipyrimidine photoproducts leads to suppression of UVB-induced carcinogenesis. Although not explicitly tested in this investigation, there are several potential mechanisms through which the topical delivery of UVDE could reduce the severity of UVB-induced carcinogenesis. These include: 1) increased, high-fidelity cellular DNA repair for CPDs and 6-4 PPs such that both dipyrimidine adducts are removed at an accelerated rate, thus reducing UVB-induced mutagenesis and carcinogenesis; 2) increased cell death in severely damaged cells via rapid initiation of single-strand breaks at CPDs and 6-4 PPs, thus killing potentially carcinogenic cells instead of undergoing error-prone replication; and 3) minimization of UV-induced immune suppression. Based on prior literature, we favor the mechanism of UVDE-initiated rapid repair of CPDs and 6-4 PPs in which the repair patch is anticipated to be synthesized with high fidelity. This conclusion is based on the following literature precedents. UVDE initiates repair at both dipyrimidine photoproducts by incision immediately 5′ to the damage 18,28,29 , followed by long-patch BER in conjunction with Rad27/Fen1, XRCC1, PARP1, and DNA polymerase [30][31][32][33] . Further, in UV-irradiated XPA cells expressing the Neurospora crassa UVDE, single-strand breaks were introduced immediately following irradiation and were efficiently repaired, resulting in enhanced survival approaching that of wild-type cells 31,32 . It was also demonstrated via expression of UVDE in repair-proficient and deficient S. pombe that introduction of the enzyme resulted in the initiation of repair of CPDs in both the nucleus and mitochondria, but that nuclear-localized repair was responsible for conferring enhanced survival in NER-deficient cells 34 . These mechanisms are inferred to occur via high fidelity reactions, since UVDE recognition of CPDs involves a quadruple flipping mechanism in which both the damaged dipyrimidine and the two complementary purines are extrahelical 35,36 .
Collectively, these data present consistent evidence that following UVB irradiation, UVDE-initiated repair results in rapid, long-patch processes, improving survival in DNA repair-deficient cells. The final product of long-patch BER is anticipated to be error-free since the undamaged complementary strand is available for repair-patch synthesis. This rapid repair of dipyrimidine photoproducts is also expected to greatly minimize UV-induced immune suppression since one of the primary signals for initiating this suppression is the duration of unrepaired CPDs in genomic DNAs 25,[37][38][39] . This conclusion is also consistent with data from the XP clinical trial using T4-pdg in which suppression of new tumors could at least partially be explained by enhanced immune surveillance.
Germane to our study using cv-pdg-NLS, previous investigations led by Dr. Daniel Yarosh and colleagues at Applied Genetics, Inc. used a similar long-term carcinogenesis protocol in mice. They demonstrated that administration of liposomal-encapsulated T4-pdg, immediately following UVB irradiation, increased the mean time to first 1 mm tumor from 19.5 weeks for heat-inactivated enzyme to 21.8 weeks for the active enzyme 40 . Further, by 30 weeks of treatment, the number of tumors per mouse was ~30% less in the T4-pdg treated mice versus the control. This suppression in carcinogenesis was also enzyme concentration dependent. There are many distinctions between the Yarosh investigations and our current study including 1) differences between T4-pdg and UVDE in terms of catalytic efficiencies, stabilities, and substrate specificities; 2) T4-pdg being a native enzyme while the cv-pdg enzyme was engineered to contain a NLS; 3) concentration differences of the encapsulated enzymes, with Yarosh et al. administering T4-pdg at 1.0 μg/mL, while the current study used enzyme concentrations of 10 μg/mL; 4) the timing of UVB irradiation, which occurred prior to T4-pdg application for the Yarosh study but 1 hr after topical enzyme delivery in this study; and 5) the number of mice used in each treatment group (25 in the Yarosh design vs. 10 in the current study). Despite these considerable differences in the study design and reagents, the trends of suppressing UVB-induced carcinogenesis by topical application of DNA repair enzymes were consistent.
Overall, these data provide compelling evidence for the benefit of topical delivery of DNA repair proteins that activate the alternative repair pathways for the delay or reduction of UVB-induced carcinogenesis. Future applications of these findings in human clinical trials are anticipated to first be carried out in XP patients and in organ-transplant patients who have higher rates of NMSC.

Cloning, Expression, Purification, and Encapsulation of UVDE Constructs. Engineering UVDE-
TAT and UVDE-NLS-TAT. Bacterial expression plasmids for UVDE-TAT and UVDE-NLS-TAT were created in a pET21b expression vector backbone to include a C-terminal hexahistidine tag. The UVDE construct was generated by subcloning the sequence for the S. pombe uve1 gene (GenBank accession number NP 596165.1) between the NdeI and HindIII sites into the multiple cloning sequence, leaving a short linker encoding for LAAALE between the last codon for the protein (K) and the hexahistidine tag (HHHHHH). The sequence encoding the AAALE was removed and substituted with sequences to encode the TAT peptide (YGRKKRRQRRR) with the intervening DNA sequence being 5′-TAT GGC CGC AAA AAG CGC CGT CAG CGC CGT CGC-3′, to generate the UVDE-TAT expression construct. Similarly, site directed mutagenesis was used to insert the NLS-TAT sequence (PKKRKRRLYGRKKRRQRRR), encoded by 5′-CCA AAG AAG AGG AAA AGG AGG CTA TAT GGC CGC AAA AAG CGC CGT CAG CGC CGT CGC-3′, yielding the UVDE-NLS-TAT expression construct. Cell paste was resuspended in lysis buffer (20 mM HEPES, pH 7.5, 20 mM imidazole, pH 7.5, 500 mM NaCl, 1 mM MnCl 2 , 10% glycerol, 0.5 mM PMSF, 14 mM β-mercaptoethanol) and lysed using a microfluidizer (Microfluidics, Inc., Model 110 L). The resulting lysate was cleared by centrifugation at 38,400 g for 60 min at 4 °C. The cleared lysate was applied to a 5 mL HisTrap HP column (GE Healthcare). The column was washed extensively with lysis buffer before the protein was eluted with a gradient of 20-300 mM imidazole and followed with 400 mM imidazole. Fractions containing UVDE-TAT or UVDE-NLS-TAT were pooled and loaded to a 5 mL HiTrap Heparin HP column (GE Healthcare). After extensive washes with Heparin Buffer A (25 mM HEPES pH 7.5, 500 mM NaCl, 10% glycerol, 1 mM MnCl 2 , 0.5 mM PMSF, 14 mM β-mercaptoethanol), the protein was eluted with a gradient of 500-1,000 mM NaCl. The heparin pool was concentrated and dialyzed into storage buffer (25 mM HEPES, pH 7.5, 500 mM NaCl, 1 mM MnCl 2 , 10% glycerol, 1 mM TCEP) or further purified by HiPrep 26/60 Sephacryl S-200 HR size exclusion column (GE Healthcare) in the storage buffer. The protein was concentrated using Amicon Ultra-15 centrifugal filter units (EMD Millipore), flash frozen in liquid nitrogen, and stored at −80 °C. The final yields of UVDE-TAT and UVDE-NLS-TAT were approximately 106 and 14 mg/L of E. coli culture, respectively. Photographs of the Coomassie-stained gels were not manipulated or spliced, but represent a single stained gel. Engineering cv-pdg-NLS. In order to create a cv-pdg-NLS expression vector (pcv-pdgNLS), the coding sequence for the gene encoding cv-pdg (from the Paramecium bursaria Chlorella virus 1 genome, PBCV-1: GenBank accession number JF411744.1) was engineered to contain a six-nucleotide spacer and a sequence encoding a nuclear localization signal (NLS) between the final codon and the stop translation codon (TGA). The NLS sequence (5′-CCC GGG CCA AAG AAA AAG AGG AAG AGG CTA-3′) encodes for the amino acid sequence PGPKKKRGRL. This expression construct was cloned into pET24a between the Nde1 and HindIII restriction sites. The sequences of all gene constructs were verified prior to expression studies. The plasmid was transformed into BL21 (DE3) cells and glycerol stocks were made for expression and purification.
Large-scale fermentation and purification of cv-pdg-NLS. Fed-batch culture was carried out in a stirred tank fermenter, with a base medium of minimal R/2 medium (2 g of (NH 4 ) 2 HPO 4 , 6.75 g of KH 2 PO 4 , 0.85 g of citric acid, and 0.7 g of MgSO 4 •7H 2 O per liter) supplemented with 20 g/L yeast extract and 0.5 ml/L trace metal solution (27 g of FeCl 3 •6H 2 O, 2 g of ZnCl 2 •4H 2 O, 1 g of CuCl 2 , 2 g of CoCl 2 •6H 2 O, 0.5 g of H 3 BO 3 , 1 g of CaCl 2 •2H 2 O, and 2 g of Na 2 MoO 4 •6H 2 O per liter of 1.2 N HCl), 8 g/L glucose, 10 g/L glycerol, 25 mg/L kanamycin, and 0.1 ml/L antifoam. An overnight seed culture was prepared from a glycerol stock of BL21 (DE3) with pET24a-cv-pdg-NLS in the supplemented R/2 medium without glycerol and antifoam. The batch phase was run at 37 °C. The pH was controlled at 7.0 by additions of 4 M NH 4 OH or 4 M H 3 PO 4 . The dissolved-oxygen concentration was controlled at 30% of air saturation (dO 2 controller was set to cascade between agitation speed and O 2 supplementation). When the OD 600 reached approximately 10, a glucose feed (50% glucose, 7 g/L MgSO 4 ) was initiated to maintain glucose level above 2 g/L. When the OD 600 reached 42, the feed was switched to glycerol (50% glycerol, 7 g/L MgSO 4 ) and the temperature was lowered to 19 °C over the course of about 30 min. Protein expression was induced by adding galactose to 2 g/L in one bolus. Biomass was harvested 5 hr post induction.
Cell paste was resuspended in lysis buffer (25 mM Tris-HCl, pH 8.0, 150 mM NaCl) and processed using a microfluidizer (Model M110L, Microfluidics, Inc.). The resulting lysate was cleared by centrifugation at 38,400 g for 60 min at 4 °C. The cleared lysate was applied to a Q Sepharose FF column (GE Healthcare). The Q flow-through was collected and loaded to a SP Sepharose HP column (GE Healthcare). The column was washed extensively with IEX Buffer A (25 mM Tris-HCl, pH 8.0, 150 mM NaCl) before step-elution with 0. Liposomes were composed of 2:2:5:1 molar ratio of DOPE:DOPC:CHEMS:Oleic Acid encapsulating cv-pdg-NLS, UVDE-TAT, and UVDE-NLS-TAT at 10 µg/mL each. Lipids were dissolved in chloroform and the solvent was evaporated using a Büchi (RE-121) rotary evaporator (Flawil, Switzerland) under vacuum for 4 hr. A thin lipid film was formed. The lipid film was hydrated with the repair enzymes in PBS buffer for 2 hr at 37 °C. The milky solution of liposomes was extruded consecutively 20 times through 2 μm, 1 μm, 400 nm and 200 nm polycarbonate membrane filter using a Lipex extruder connected to high pressure argon cylinder. The sizes of the liposomes were measured using Malvern Zetasizer nano ZS90 (Malvern, United Kingdom).
A total of 10 g of Carbomer 940 was added to 1000 mL of PBS to form a 1% hydrogel solution. The solution was mixed using a Gowe ® Electrical Compact Laboratory mixer. The pH of the mixture was adjusted to 7.4 by slow addition of NaOH. At pH above 6 the mixture forms a gel structure. The hydrogel was mixed thoroughly for 3 hr at room temperature. Liposomes were added to 1% hydrogel solution to make the final hydrogel concentration of 0.75%. The solutions were mixed thoroughly for 1 hr at room temperature and packaged in dermal syringes.

Animal Carcinogenesis and Tissue Preparation.
Female SKH1 hairless mice (6 weeks old) were obtained from Jackson Laboratories and group-housed at 5 mice per box. All animal protocols were pre-approved through the OHSU Institutional Animal Care & Use Committee and monitored by the Department of Comparative Medicine. All methods and procedures were performed in strict accordance with the policies and specified experimental design that was previously approved by the OHSU Institutional Animal Care & Use Committee. A total of 50 SKH1 hairless mice were randomly divided into 4 treatment groups (10 mice each), while an additional 10 mice were held with no treatments for observation of spontaneous tumor formation. After mice were housed for 2 weeks, all mice in each treatment group: 1) control empty liposome, 2) liposomes containing UVDE-TAT, 3) liposomes containing UVDE-NLS-TAT or 4) liposomes containing cv-pdg-NLS were treated as follows. One hr prior to UVB irradiation, 0.2 mL of the liposomal formulations were uniformly applied to the dorsal surface of each mouse using a pre-moistened cotton swab. This volume of liposome was absorbed within 3-5 min following application. Mice were irradiated 1 hr after application of the liposome in the morning on Monday, Wednesday, and Friday with UVB in a ventilated 8-chambered plexiglass box that was covered with a UVB-transmissible quartz plate. During the first 3 weeks with 9 total exposures, all irradiated mice were exposed to 225 J/m 2 per exposure to allow photoaging without sunburn or blistering. Following this 3-week period of UVB acclimation, mice received ∼22 kJ/m 2 /week and all mice were examined at least three times per week for an additional 30 weeks. Although mice were monitored at least three times per week for tumor formation, the recording of tumor sizes was performed at 15,18,19,20,21,22,23,24,26,28,30,32, and 33 weeks. At 23 weeks of UVB exposures, all mice were photographed for a visual record of the condition of the dorsal skin. Mice that developed either tumors >8 mm in diameter or ulcerated tumors were euthanized. Immediately following euthanasia, epidermal and dermal tissues were collected from representative tumor and non-tumor sections of the backs of the mice. Additionally, at the termination of the study (after 33 weeks), mice from the untreated group (no liposomes and no UVB treatment) were also euthanized and representative skin samples harvested. Tissues were fixed in aqueous-buffered zinc formalin (4% (w/v) formaldehyde and 600 ppm zinc) and after 2 days transferred into 70% (v/v) ethanol. Tissues were paraffin embedded and sections cut for hematoxylin and eosin (H&E) staining. Photographs were taken on a Zeiss ApoTome 2, using the Zeiss AxioCam 506 CCD color camera with no internal magnification. Scale bars were added using Image J software. Biostatistical Methods. The two time-to-event outcomes (death and first 2 mm tumor) were depicted graphically with nonparametric Kaplan-Meier curves for each treatment group and compared using the log-rank test. Reported log-rank p-values for pairwise comparisons between treatment groups were adjusted for multiplicity using the Tukey method 41 . Cox regression was applied to the time-to-death outcome to estimate hazard ratios for the risk of tumor-induced euthanasia when comparing each active treatment group to the control group. The Cox model assumption that the risks of death are proportional over the study period for the groups being compared was checked and found to be justifiable.
There were no missing values for total tumor size comparisons at either of the two time points of interest (23 weeks and 33 weeks) since a mouse's last measured value was carried over if it had been euthanized prior to the time point. Although the sample distributions of total tumor size were often positively skewed, this outcome was nonetheless analyzed using a one-way ANOVA as this method is known to be robust to deviations from the normality assumption. Moreover, discrepancies in the variance of total tumor size between treatment groups discouraged the use of a nonparametric method that assumes the distributions have equal spread. At each of the two time points of interest, the significance of total tumor burden differences between treatment groups was quantified with Tukey-adjusted pairwise p-values from the corresponding ANOVA model.
Statistical significance was ascribed to all effects with multiplicity-adjusted p-values < 0.05. Statistical analysis was done in SAS 9.4 and R 3.4.