Glypican-3 targeted delivery of 89Zr and 90Y as a theranostic radionuclide platform for hepatocellular carcinoma

Glypican-3 (GPC3) is a tumor associated antigen expressed by hepatocellular carcinoma (HCC) cells. This preclinical study evaluated the efficacy of a theranostic platform using a GPC3-targeting antibody αGPC3 conjugated to zirconium-89 (89Zr) and yttrium-90 (90Y) to identify, treat, and assess treatment response in a murine model of HCC. A murine orthotopic xenograft model of HCC was generated. Animals were injected with 89Zr-labeled αGPC3 and imaged with a small-animal positron emission/computerized tomography (PET/CT) imaging system (immuno-PET) before and 30 days after radioimmunotherapy (RIT) with 90Y-labeled αGPC3. Serum alpha fetoprotein (AFP), a marker of tumor burden, was measured. Gross tumor volume (GTV) and SUVmax by immuno-PET was measured using fixed intensity threshold and manual segmentation methods. Immuno-PET GTV measurements reliably quantified tumor burden prior to RIT, strongly correlating with serum AFP (R2 = 0.90). Serum AFP was significantly lower 30 days after RIT in 90Y-αGPC3 treated animals compared to those untreated (p = 0.01) or treated with non-radiolabeled αGPC3 (p = 0.02). Immuno-PET GTV measurements strongly correlated with tumor burden after RIT (R2 = 0.87), and GTV of animals treated with 90Y-αGPC3 was lower than in animals who did not receive treatment or were treated with non-radiolabeled αGPC3, although this only trended toward statistical significance. A theranostic platform utilizing GPC3 targeted 89Zr and 90Y effectively imaged, treated, and assessed response after radioimmunotherapy in a GPC3-expressing HCC xenograft model.

Hepatocellular carcinoma (HCC) is the most common histologic subtype of liver cancer with forecasts projecting an increase in incidence in coming decades [1][2][3] . HCC most commonly presents in the setting of chronic liver disease, often in an advanced stage involving multiple diseased liver segments 4,5 . In cases of advanced HCC, overall survival is poor with the standard-of-care systemic therapy (e.g. sorafenib, lenvatinib) and checkpoint inhibition immunotherapy (e.g. nivolumab, pembrolizumab) only providing marginal survival benefit [6][7][8][9][10] . Novel therapies that combine immune checkpoint inhibitors with anti-angiogenic agents (e.g. atezolizumab and bevacizumab) or tyrosine kinase inhibitor (e.g. pembrolizumab and lenvatitinib) have recently demonstrated promise in advanced HCC, although there are limiting side effects [11][12][13][14] . Locoregional therapies, such as transarterial chemoembolization and radioembolization, are important tools in management of locoregional and advanced HCC. They have been shown to prolong survival in unresectable HCC and can facilitate downstaging to transplant or surgical resection [15][16][17] . Surgical resection is a mainstay of therapy in patients with resectable disease and preserved liver function, however recurrence in the liver remnant is common due to de novo carcinogenesis, intrahepatic metastasis, and missed lesions during initial staging 18 . Therefore, improved diagnostic and therapeutic adjuncts are needed to better stage and treat patients with HCC.
Targeted radionuclide theranostics is an emerging field that promises new and personalized approaches to cancer therapy. To date, numerous targeted radionuclide diagnostic and therapeutic technologies are in www.nature.com/scientificreports/ development, including positron-emitting diagnostic agents for targeted PET imaging and beta-and alphaemitting agents as radionuclide therapies, or radioimmunotherapy (RIT). Somatostatin receptor (SSTR) targeted Galium-68 ( 68 Ga) DOTATATE radiopeptide imaging and Lutetium-177 ( 177 Lu) DOTATATE peptide receptor radionuclide therapy (PRRT) compose the clinical theranostic platform available for the management of advanced midgut neuroendocrine tumors [19][20][21][22][23] . Currently, there are no clinically approved radionuclide technologies for the management of HCC; however, several antibody-targeting agents are in preclinical development [24][25][26][27] . Glypican-3 (GPC3) is a cell surface antigen highly expressed on HCC that has been used for a variety of targeted therapies including chimeric antigen receptor T-cell (CAR-T) therapy, vaccine and radionuclide therapy 28,29 . GPC3 is expressed on over 75% of HCC tumors and is not found on normal tissues 30,31 . GPC3 is variably expressed in HCC, and although highest cell surface expression is associated with de-differentiated tumors, it is also present on well-differentiated tumors and dysplastic nodules indicating it may be effective in identifying early neoplastic lesions 32 . It is therefore an attractive target for developing a theranostic platform which combines detection and treatment of both early and advanced disease.
Delivery of radioimmunoconjugates for a theranostic approach, including initial tumor identification and staging, subsequent targeted radioimmunotherapy, and assessment of treatment response has not been reported in HCC. To this end, our group developed GPC3-targeted radioimmunoconjugates using zirconium-89 ( 89 Zr-αGPC3) for immuno-PET imaging and ytrrium-90 ( 90 Y-αGPC3) for cytotoxic radioimmunotherapy to serve as functional agents for a theranostic approach 24,33 . In this proof-of-concept study, we test the utility of sequentially delivering radioimmunoconjugates to identify, treat and assess treatment response in a murine orthotopic xenograft model of HCC.

Results
Conjugated αGPC3 maintains binding affinity for GPC3. Binding to cell surface GPC3 by unconjugated, DFO-and DOTA-conjugated αGPC3 was confirmed by flow cytometry (Fig. 1). Normalized geometric mean fluorescence for αGPC3-DFO and αGPC3-DOTA were significantly elevated compared to control. These validated conjugates were subsequently radiolabeled with 89 Zr for immuno-PET studies and 90 Y for RIT. 89 Zr-αGPC3 immuno-PET reliably identifies small tumors and measures tumor volume. Orthotopic tumor establishment was confirmed with serum AFP measurement and IVIS bioluminescence imaging six weeks after subcapsular HepG2 cell injection. To stage tumors prior to RIT, 29 animals underwent 89 Zr-DFO-αGPC3 immuno-PET imaging and GTV was measured along with serum AFP sampling. All animals demonstrated discrete localizations of increased PET intensity at the subcapsular injection site with minimal background signal (Fig. 2). Immuno-PET was sensitive in detecting both small and large tumors, with GTVs ranging from 0.009cm 3 to 0.35cm 3 . Individual animal immuno-PET GTV measurements strongly correlated with serum AFP concentrations (R 2 of 0.90) (Fig. 3a), demonstrating its reliability in assessing tumor burden and measuring tumor volume. Maximum standardized uptake (SUV max ) were calculated from PET images prior to RIT and ranged from 6 to 53.5 and did not correlate with tumor size (data not shown). 90 Y-αGPC3 RIT decreases tumor burden. Two days after 89 Zr-DFO-αGPC3 immuno-PET imaging, the animals received either 11.1 MBq 90 Y-DOTA-αGPC3, αGPC3-DOTA alone by tail vein injection or no treatment. Tumor burden was not statistically different between groups prior to RIT, with mean serum AFP concentrations of 185,986 ± 127,154 ng/mL in the no treatment control group, 166,370 ± 119,541 ng/mL in the αGPC3-DOTA control group, and 180,613 ± 155,557 ng/mL in the 90 Y-DOTA-αGPC3 group (p = 0.97). Pre-RIT GTV was 0.16 ± 0.10cm 3 in the no treatment, 0.15 ± 0.13cm 3 in the αGPC3-DOTA control groups, and 0.14 ± 0.12cm 3 in the 90 Y-DOTA-αGPC3 group (p = 0.96) (Fig. 4a,b). www.nature.com/scientificreports/ Thirty days after RIT injection, mean serum AFP was measured and animals underwent a second round of 89 Zr-DFO-αGPC3 immuno-PET imaging. Mean serum AFP concentration was lower in the 90 Y-DOTA-αGPC3 treatment group compared to both no treatment and αGPC3-DOTA control groups (Fig. 5a). Compared to pretreatment levels, mean serum AFP concentration decreased 57% to 77,834 ± 90,970 ng/mL, in the 90 Y-DOTA-αGPC3 treatment group, and increased 150% to 428,221 ± 444,127 ng/mL, and 143% to 403,835 ± 446,544 ng/ mL, in the no treatment and αGPC3-DOTA control groups, respectively. A significant reduction in serum AFP 30 days after 90 Y-DOTA-αGPC3 RIT was observed in all but one animal. One untreated animal required euthanization prior to completion of the study due to large tumor size. Zr-DFO-αGPC3 immuno-PET in assessing treatment response, 29 animals imaged prior to RIT underwent a second round of immuno-PET imaging 30 days after RIT. Again, we observed discrete localizations of increased PET intensity in the left lobe of the liver in all imaged animals, indicating that serial exposure to our antibody does not limit binding of additional radiotracer doses, likely through replenishment of surface antigen (Fig. 5a). Similar to pre-treatment immuno-PET imaging, GTVs correlated with serum AFP concentration after therapy (R 2 = 0.87, Figs. 3b, 5b), confirming its accuracy in measuring tumor volume after serial antibody exposure. No significant correlation between pre-treatment SUV max and response to 90 Y RIT was observed. Lower intensity  Immuno-PET imaging with 89 Zr-DFO-αGPC3 demonstrated an increase in GTV of 226% to 0.51 ± 0.70cm 3 , and 115% to 0.31 ± 0.35cm 3 , in untreated and non-radiolabeled αGPC-DOTA animals, respectively, and a decrease in GTV in 90 Y treatment animals by 62% to 0.10 ± 0.10cm 3 , compared to pretreatment volumes, in a trend toward statistically significant differences between groups (p = 0.13) (Fig. 4b). Of note, due to limitations in imaging logistics during 2nd round, several randomly selected animals were not imaged by immuno-PET. Three animals in the no treatment control group were not imaged, two of them having large tumors as measured by serum AFP (570,000 and 1,114,000 ng/mL). Two animals in the non-radiolabeled αGPC-DOTA control group were not imaged, both of them found to have large tumors by serum AFP (322,000 and 1,114,000 ng/mL). One animal in the 90 Y-DOTA-αGPC3 group was not imaged, with moderate sized tumor by serum AFP (95,000 ng/ mL).

Discussion
This preclinical study demonstrates the utility in combining 89 Zr-DFO-αGPC3 and 90 Y-DOTA-αGPC3 for detection and therapy in a specific theranostic approach to stage, treat, and assess response to treatment in an orthotopic xenograft model of HCC, highlighting the potential direct clinical relevance for patients with HCC.
To test theranostic feasibility, we delivered our radiolabeled αGPC3 sequentially over time to assess efficacy in the setting of serial antibody exposure. Using 89 Zr-DFO-αGPC3 immuno-PET, we reliably measured volume of sub-centimeter implanted tumors, which correlated well with markers of tumor burden indicating utility in initial clinical staging. 90 Y-DOTA-αGPC3 RIT effectively reduced tumor burden with a decrease in serum AFP and 89 Zr-DFO-αGPC3 immuno-PET tumor volume compared to control. Correlations between serum AFP and 89 Zr-DFO-αGPC3 immuno-PET tumor volume remained strong after RIT suggesting 89 Zr-DFO-αGPC3 immuno-PET may be useful in assessing response to therapy. These findings confirm the effectiveness of each radioimmunoconjugate in the presence of serial antibody exposure. If effective in patients with HCC, this targetspecific diagnostic and therapeutic approach would represent a major advance over current technologies used in HCC that would specifically benefit patients with GPC3 expressing tumors. This study validates the 89 Zr-DFO-αGPC3 immuno-PET agent as a highly sensitive and specific molecular imaging agent for imaging GPC3-positive tumors. All animals with tumors identified by bioluminescent imaging and serum AFP measurement had detectable tumors by immuno-PET before and after RIT. 89 Zr-DFO-αGPC3 immuno-PET detected sub-centimeter tumors, with tumor sizes ranging from 0.009 to 1.9cm 3 , and reproducibly measured volume using standardized segmentation technology 34,35 . Although there are several limitations associated with PET GTV quantification 36 , we were reassured to see a strong correlation between measured GTV and serum AFP before and after RIT. Further studies to validate the accuracy of this tumor volume quantification are currently ongoing. www.nature.com/scientificreports/ We observed a reduction in tumor size after treatment with 90 Y-DOTA-αGPC3. This effect was observed in animals exposed to antibody for PET imaging just a week prior, indicating that the antibody binding sites were not saturated or that bound sites were quickly recycled. The antitumor effect may be explained by antibody-dependent cellular cytotoxicity which has been observed with other αGPC3 antibodies targeting different epitopes, including GC33, which has been evaluated in clinical trials 37 . However, in our study, we did not observe a measurable antitumor effect in the non-radiolabeled antibody control group, indicating that this antitumor effect was due to the presence of targeted radiation. This targeted RIT is delivered systemically and may be an important new therapy for patients with advanced GPC3-expressing HCC.
This study is the first to report a combined PET/RIT approach to the treatment of HCC and represents an advance in the field towards a more personalized and targeted radionuclide approach. This contrasts with the current 90 Y radionuclide therapy used in HCC which is delivered in microspheres via a transarterial approach. The antibody directed approach offers a systemically administered alternative and would benefit from GPC3 targeting. Although only applicable to GPC3 + tumors, these patients would be selected for therapy with the 89 Zr-DFO-αGPC3 immuno-PET, increasing likelihood of treatment success. This theranostic platform for HCC complements the growing field of radionuclide theranostics for management of solid tumors. Since the 1980s, several theranostic agents were developed including those for colorectal cancer 38 , breast cancer 39 , renal cell carcinomas 40 , non-Hodgkin's lymphoma 41 and prostate cancer 42 . The most widely utilized agent is used in the management of neuroendocrine tumors with 68 Ga-DOTATATE PET for identification and 177 Lu DOTATATE as PRRT 19,43 . These studies support the development of more theranostic platforms to personalize patient selection with targeted molecular imaging followed by targeted therapy.
There are several limitations to our preclinical study. Due to PET scanner availability, not all animals were imaged after RIT, including several control animals with high tumor burden. Although shown to be generally widely expressed on HCC, the extent of GPC3 expression varies and may limit the sensitivity of the tumor volume quantification and efficacy of RIT for all HCCs observed in clinical practice. We did not observe an association between SUV max and response to radioimmunotherapy or tumor size in this study. In fact, a weak correlation was observed between higher SUV max and smaller tumor size.This may be explained by reduced radiotracer uptake from insufficient perfusion or intratumoral necrosis in our model, which warrants investigation in future studies. Utilizing a cell line with high GPC3 expression in an immunocompromised xenograft does not adequately represent the complex human HCC tumor microenvironment and favors the function of our GPC3-targeting agents; however, this proof-of-concept study was important to demonstrate the potential of our platform. Additional experiments demonstrating the limit of GPC3 expression needed for detection and effective therapy with our theranostic platform are needed. Furthermore, although PET image segmentation is limited by PET technique and its partial volume effect, we still observed very strong correlations between tumor volume by PET and validated measures of tumor volume, particularly for smaller tumors. More investigations will be needed to assess the accuracy of PET outside this range.

Conclusion
In this study, we report GPC3 targeted radionuclide delivery of 89 Zr/ 90 Y as a promising theranostic approach to detect, treat, and monitor treatment in an orthotopic xenograft model of HCC. By utilizing the 89 Zr-DFO-αGPC3 immuno-PET imaging platform to assess the efficacy of 90 Y-DOTA-αGPC3 RIT, we confirm the continued utility of our radioimmunoconjugates with serial antibody exposure, demonstrating feasibility of this theranostic platform as a future approach to improve diagnosis and treatment of patients with HCC.

Materials and methods
Cell lines and tissue culture. GPC3-positive HepG2-Red-FLuc HCC cells expressing Luciferase were purchased from PerkinElmer (RRID:CVCL5I98, Bioware, cat. no. BW134280) and were maintained as previously described 33 . Cells were cultured in a monolayer at 37 °C in Dulbecco's modified Eagle Medium (DMEM; Gibco) supplemented with 10% fetal bovine serum (FBS; Gibco) in a humidified chamber with 5% CO 2 . Cells were grown until 70-80% confluent, detached with 0.25% trypsin and passaged into new cell culture flask per manufacturer's instructions (PerkinElmer). Production of αGPC3 IgG1. As previously described, αGPC3 IgG1-producing hybridomas were generated through the Fred Hutchinson Cancer Research Center antibody core facility 44 . 1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acid (DOTA). To demetallate the αGPC3 antibody for labeling with radiometals, it was concentrated to 6 mg/mL and then dialyzed against metal-free saline (150 mM NaCl, and 1 mM EDTA adjusted to pH 7 and passed over a Chelex 100 column) for 3 days at 4 °C with a minimum of three buffer changes a day. The day before the conjugation reaction, the antibody was dialyzed for an additional two buffer changes to replace the saline with demetallated HEPES buffer (50 mM HEPES (N-(2-hydroxyethyl)piperazine-N′-ethanesulfonic acid, 150 mM NaCl, and 1 mM EDTA adjusted to pH 8.5 and passed over a Chelex 100 column). In acid washed microcentrifuge tubes, the demetallated αGPC3 antibody was conjugated with 10 equivalents of either p-SCN-Bn-DFO or p-SCN-Bn-DOTA (both from Macrocyclics) as a solution at 10 mg/mL in DMSO. These reactions were allowed to run overnight at room temperature with gentle mixing. The reaction mixtures were then dialyzed against a metal-free citrate buffer (50 mM sodium citrate and 150 mM NaCl with pH 5.5) over 3 days at 4 °C followed by dialysis against 150 mM saline (pH 7.0) for another 3 days. Each buffer change contained Chelex resin to scavenge metals. The final conjugates were then stored in acid washed tubes at 4 °C.

Scientific Reports
| (2021) 11:3731 | https://doi.org/10.1038/s41598-021-82172-w www.nature.com/scientificreports/ 89 Zr-Labeling of αGPC3-DFO. Demetallated 2 M sodium carbonate was added to 89 Zr to adjust the pH to 7.0-7.5. To this solution was added HEPES buffer at pH 7.0, followed by αGPC3-DFO, prepared as above. The mixture was then incubated for 2 h at room temperature. The labeled antibody was separated from unreacted 89 Zr via a PD-10 column (GE Healthcare) eluted with PBS prior to analysis by HPLC) to verify radiochemical purity. Acid-washed vials and pipette tips were used for all steps. The radiochemical yield was 78% and the radiochemical purity was > 95%. 90 Y-Labeling of αGPC3-DOTA. αGPC3-DOTA was radiolabeled with 90 Y as previously described 33 . Acidwashed vials and pipette tips were used for all steps. The radiochemical yield was 90% and the radiochemical purity was > 98%.
Flow cytometry. Flow cytometry was used to evaluate the in vitro binding of the αGPC3-DFO and αGPC3-DOTA conjugates. HepG2-Red-FLuc cells were resuspended in cold phosphate-buffered saline (PBS) at a concentration of 1 × 10 6 cells/mL. One microgram of unconjugated αGPC3, αGPC3-DFO or αGPC3-DOTA was added to the cell suspension and incubated for 45 min at 4 °C. After incubation, samples were washed in cold PBS, and incubated with 1 μg of PE-labeled goat-α-mouse IgG1 secondary antibody (BD Biosciences, San Jose, CA) on ice for 30 min in the dark. Cells were then washed in cold PBS and were analyzed with a LSRII (BD Biosciences, San Jose, CA) using the FACS Diva software. A minimum of 10,000 cells were analyzed for each sample in triplicate. Data analysis was performed on the FlowJo software, version 8.8.6 (TreeStar, Ashland, OR).
Orthotopic xenograft model. This study was performed in accordance with the University of Washington Office of Animal Welfare guidelines for the humane use of animals, and all procedures were reviewed and approved by the Institutional Animal Care and Use Committee (Protocol #4304-02) and was carried out in compliance with the ARRIVE guidelines 45 . The orthotopic xenograft model was generated as previously 33 . After a week of acclimatizing, approximately 2 × 10 6 Luciferase-expressing GPC3-positive HepG2 cells in 25μL of Dulbecco's modified Eagle medium containing 50% Matrigel (BD Biosciences) were injected into the subcapsular space of the left hepatic lobe. Six weeks after hepatic subcapsular cell injection, a 75 mg/kg intraperitoneal injection of VivoGlo luciferin (Promega) was administered and bioluminescent imaging was performed using an IVIS Lumina II system (PerkinElmer) to verify tumor establishment.

Measurement of serum AFP.
To monitor orthotopic tumor growth, whole blood was obtained from animals by submandibular bleed and collected in EDTA-coated Eppendorf tubes 33,46 . Serum was extracted from the fresh whole blood, then frozen and allowed to decay 10 half-lives (~ 27 days for 90 Y, ~ 33 days for 89 Zr) if found to be radioactive in accordance with the University of Washington Environmental Health and Safety policy. Frozen serum specimens were thawed on ice and the serum concentration of AFP was determined on the UniCel Dxl 800 Access Immunoassay System (Beckman Coulter) using an Access AFP Alpha-fetoprotein pack (Quest Diagnostics) and reported in nanograms/milliliter.
In vivo 90 Y radioimmunotherapy. HepG2 tumor bearing animals were assigned to one of three experimental groups based on serum AFP measurements to ensure comparable cohorts and were grouped in same cage (n = 12 in each group). Animals either received no injection, 70 μg αGPC3-DOTA without radioisotope or 70 μg DOTA-αGPC3 radiolabeled with 11.1 MBq (300uCi) of 90 Y (Perkin Elmer), administered via tail vein injection. Thirty days after RIT injection, randomly selected animals were imaged by PET/CT and then all animals were euthanized for serum AFP collection to evaluate for early anti-tumor effect of RIT. Primary outcome measure was serum AFP and gross tumor volume assessed on immuno-PET. Randomly selected livers were harvested and placed in 10% (w/v) neutral-buffered formalin. Animals were euthanized if they did not meet standard body-conditioning scores set by our Office of Animal Welfare.
Small-animal positron emission tomography. 89 Zr-DFO-αGPC3 imaging studies were performed using the Inveon PET/CT scanner (Siemens Medical Solutions USA, INC. Molecular Imaging, Knoxville, TN), which was calibrated for 89 Zr. Whole-body PET and CT imaging was performed on animals anesthetized with 1-2% isoflurane anesthesia in 100% oxygen at 1L/min in a temperature-controlled bed with respiratory monitoring. Tumor-bearing animals were injected with 11.1 MBq (300uCi) of 89 Zr-DFO-αGPC3 (~ 70 μg antibody) via the tail vein 1 week before (n = 29) and 4 weeks (n = 27) after RIT injection. Pre-RIT PET/CT imaging was performed on days 4 and 5 after 89 Zr-DFO-αGPC3 injection and post-RIT imaging was performed on days 6 and 7 after injection due to equipment availability. Animals first had a 60-min PET scan followed by a CT scan, which enabled scatter and attenuation correction. Due to scanner malfunctions during post-RIT imaging, 10 animal scans were PET-only and therefore not corrected for scatter or attenuation. Due to limited imaging time on PET scanner, not all animals were imaged. Animals were randomly selected and were imaged before and after RIT.
PET tumor volume and intensity measurements. PET images were reconstructed using ordered subset expectation maximization/shifted Poisson maximum a posteriori (OSEM3D/SP-MAP; 2 iterations 18 subsets) with a 256 × 256 matrix, target resolution of 1.5 mm, zoom factor 1.3, and corrections for scatter and attenuation. CT images were reconstructed using forward back projection with a Shepp-Logan filter, slight noise reduction, and appropriate beam hardening corrections. Maximum Standardized Uptake Values (SUV max ) were determined using MIM software (MIM Software Inc., Cleveland, OH USA). Dose information for each mouse was entered as 8 www.nature.com/scientificreports/ time. Because two animals were present in each image, the process was repeated for the second mouse. Gross tumor volume (GTV) measurement and image analysis were performed with Horos software (Nimble Co., Annapolis, MD USA). Localizations of increased intensity in the midline upper abdomen as anticipated based on anatomy and surgical procedure were classified as tumors. GTV was measured using a combination of fixed thresholding and manual segmentation methods as previously described 35 . A fixed, relative threshold of 40% of the maximum signal intensity was set as the lower threshold for each growing region. For larger tumors with diffuse, heterogeneous uptake, manual visual segmentation was performed. The total and mean signal intensity of each volume of interest was obtained and was decay and dose-corrected. Since 10 animals did not have attenuation or scatter correction, GTV measurement was performed with images that were not attenuated or corrected for scatter to maintain consistency of measurements across cohort. All segmentation was performed by a user blinded to experimental conditions of each animal.
Immunohistochemistry. Formalin-fixed, paraffin-embedded tissues were sectioned at 4 microns onto positively charged slides. Slides were de-paraffinized in xylene and rehydrated through graded ethanol. Heat mediated antigen retrieval was performed in 10 mM sodium citrate buffer (pH 6.0). Incubations were performed with cleaved caspase-3 (CC3) primary antibody (

Ethics approval. This study was approved by the Institutional Review Board and the Institutional Animal
Care and Use Committee at the University of Washington.