AACR 2018 Proceedings: Abstracts 3028-5930

Proceedings of the AACR

Volume 59 | April 2018

Part A: Abstracts 3028-5930

TABLE OF CONTENTS

TUMOR BIOLOGY:

Cancer Imaging: Immunology and Systems Analysis in Vivo

Cancer Stem Cell Characterization

Carcinogenesis 1

Harnessing the Power of Cell Lines for Cancer Research

Immune Cells in the Microenvironment

Metastasis, Invasion, and Migration 1

Pediatrics 2: Preclinical Therapies, Resistance, and Stem Cells

Radiation Studies Using in Vitro and Computational Models

EPIDEMIOLOGY:

Biomarkers of Exogenous and Endogenous Risk Factors in Cancer Epidemiology

PREVENTION RESEARCH:

Prevention, Interception, and Early Detection Research

BIOINFORMATICS AND SYSTEMS BIOLOGY:

Sequence Analysis and Unique Database Resources

Systems and Computational Biology

MOLECULAR AND CELLULAR BIOLOGY / GENETICS:

Epigenetic Changes as Molecular Markers of Cancer

Exploring Oncogenic Transcription Factors

Genomic Instability

Genomic Profiling of Tumors 1

Genomic Profiling of Tumors 2

Kinases 2

Metabolomics

Oncogene Growth Factors and Their Receptors

Therapeutic Approaches

Ubiquitylation, Vesicles, and Membranes

CLINICAL RESEARCH:

Adoptive Cell Therapy 3

Biomarker Discovery 4

Immune Checkpoints 3

Liquid Biopsy 3

Molecular Classification of Tumors 1: Epigenetic Therapy, Functional and Molecular Imaging, and Tumor Heterogeneity

CANCER CHEMISTRY:

Drug Delivery

ENDOCRINOLOGY:

Steroid Receptors and Preclinical Studies of Endocrine-Related Cancers

IMMUNOLOGY:

Immunomodulatory Agents and Interventions 1

Innate Immune Responses in Cancer

Therapeutic Antibodies, Including Engineered Antibodies 3

EXPERIMENTAL AND MOLECULAR THERAPEUTICS:

Novel and Canonical Targets

Novel Assay Technology and Delivery Systems

Pharmacogenetics and Pharmacogenomics

Receptor Targeting and the Tumor Microenvironment

Resistance and Biology

Targeting Oncogenes, Tumor Suppressors, or Gene Products

Therapeutic Targeting

TUMOR BIOLOGY:

Carcinogenesis 2

Cell Adhesion and Extracellular Matrix

Determining How the Immune System Drives Tumor Progression

Mechanisms and Models of Gastrointestinal Malignancies

Molecular Imaging: Novel Probes and Preclinical Studies

Pediatrics 3: Signaling, Transcription, and Metastasis

Radiation Studies Using in Vivo and Clinical Models

Therapeutic Approaches to Metastasis

EPIDEMIOLOGY:

Biomarkers of Prognosis and Pharmacoepidemiology

Cancer in Minority Populations, Health Disparities, and Survivorship Research

BIOINFORMATICS AND SYSTEMS BIOLOGY:

Statistical Methods, Mathematical Modeling, and Molecular Modeling

MOLECULAR AND CELLULAR BIOLOGY / GENETICS:

Checkpoints and Cell Cycle Progression

Epigenomics

Genomic Profiling of Tumors 3

GTPases and Their Regulators and Effectors

Immunologic and Other Cancer Cell Death

MicroRNA Therapeutics

Noncoding RNAs: From Biology to Therapy

Post-transcriptional and Translational Control of Cell Fate

Tumor Suppressor Genes 1

Tumor-Stroma and Cell-Cell interactions

CLINICAL RESEARCH:

Diagnostic Biomarkers

Immune Checkpoints 4

Immune Mechanisms Invoked by Therapies 2

Liquid Biopsy 4

Molecular Classification of Tumors 2: Molecular Predictors of Response, Tumor Staging, and Correlation of Clinical and Molecular Markers

Novel Preclinical Therapies in Pediatric Solid Tumors

CANCER CHEMISTRY:

Drug Discovery Tools

IMMUNOLOGY:

Adaptive Immunity in Tumors

Immunomodulatory Agents and Interventions 2

New Immunosuppressive Mechanisms in Cancer

REGULATORY SCIENCE AND POLICY:

Regulatory Science and Science Health Policy

EXPERIMENTAL AND MOLECULAR THERAPEUTICS:

Canonical Targets 1

Combination Chemotherapy 1

DNA Damage and Cell Cycle Regulation Experimental Therapeutics

Novel Targets and Inhibitors

Microenvironmental and Cell Nonautonomous Factors in Mediating Therapeutic Resistance

Pharmacokinetics and Pharmacodynamics

CLINICAL RESEARCH:

Liquid Biopsy: Poster Discussion

Emerging Immunotherapy Targets and Combination Strategies to Overcome Treatment Resistance

EPIDEMIOLOGY:

Endogenous and Exogenous Factors in Cancer Risk and Mortality

EXPERIMENTAL AND MOLECULAR THERAPEUTICS:

A Therapeutic About Face: Reversing Drug Resistance

Early Novel Drug Development

IMMUNOLOGY:

Epigenetic and Metabolic Regulation of Cancer Immunity

MOLECULAR AND CELLULAR BIOLOGY / GENETICS:

Metabolism: Emerging Concepts and Therapy

Understanding the Genomic Dark Matter

PREVENTION RESEARCH:

Preclinical Studies of Cancer Prevention

TUMOR BIOLOGY:

Stemness and Cancer

Molecular Mechanisms Driving Metastasis

Expanding the Definition of the Tumor Microenvironment

TUMOR BIOLOGY:

Applications of 3D Models for Tumor Biology and Therapeutic Studies

Carcinogenesis 3

Carcinoma-Associated Fibroblasts in Tumor Progression

Dissecting Tumorigenesis in Vivo Using Genetic Approaches and Spontaneous Tumor Models

Metabolism and the Microbiome: Defining the Greater Microenvironment

Metastasis, Invasion, and Migration 2

Molecular Profiles, Circulating Cancer Cells, and Metastasis

The Systemic Microenvironment in Tumorigenesis

ENDOCRINOLOGY:

Clinical Endocrinology

EPIDEMIOLOGY:

Diet, Alcohol, Tobacco, and Other Lifestyle Risk Factors

PREVENTION RESEARCH:

Population and Behavioral Studies in Cancer

BIOINFORMATICS AND SYSTEMS BIOLOGY:

New Algorithms

MOLECULAR AND CELLULAR BIOLOGY / GENETICS:

DNA Methylation

Genomic Profiling of Tumors 4

Genomic Profiling of Tumors 5

MicroRNAs as Biomarkers

Signaling and Hormonal Inputs to Transcription Factor Regulation

Signaling and Therapy

Targets Affecting Metabolism

Tumor Suppressor Genes 2

CLINICAL RESEARCH:

Diagnostic and Prognostic Biomarkers in Clinical Trials

Immunomodulatory Agents and Interventions 3

Liquid Biopsy 5

Liquid Biopsy 6

Therapeutic Antibodies, Including Engineered Antibodies 4

Vaccines 2

CANCER CHEMISTRY:

Emerging Proteomic Technologies for Cancer Research

IMMUNOLOGY:

Emerging Tools and Models in Immuno-oncology Research

Immune Monitoring / Clinical Correlates

Neoantigens in Cancer

Oncogenes, Inflammation, and Cancer

EXPERIMENTAL AND MOLECULAR THERAPEUTICS:

Antibodies, Fusion Proteins, and Related Biologics

Canonical Targets 2

Combination Chemotherapy 2

Epigenetic and Metabolic Pathways in Mediating Therapeutic Resistance

Novel Targets and Therapeutics

Regulation of Gene Expression in Drug Resistance

Therapeutic Approaches Based on Gene Delivery and Vector System

Tuesday, April 17, 2018

TUMOR BIOLOGY:

Cancer Imaging: Immunology and Systems Analysis in Vivo

#3028

PET imaging with the bispecific ⁸⁹Zr-antibody ERY974 targeting CD3 and glypican 3 in tumor-bearing mouse models.

Stijn J. Waaijer,¹ Danique Giesen,¹ Takahiro Ishiguro,² Yuji Sano,² Norihisa Ohishi,² Athos Gianella-Borradori,³ Carolien P. Schröder,¹ Elisabeth G. de Vries,¹ Marjolijn N. Lub-de Hooge¹. ¹University Medical Center Groningen, Groningen, Netherlands; ²Chugai Pharmaceutical Co., Ltd., Japan; ³Chugai Pharma USA, NJ.

BACKGROUND ERY974, a modified monoclonal IgG4 bispecific antibody directed against human CD3 on T cells and glypican 3 (GPC3) on tumor cell, is currently in phase I clinical trial. The oncofetal protein GPC3 is overexpressed in several tumor types. Radiolabeling ERY974 with positron emission tomography (PET) isotope zirconium-89 (⁸⁹Zr) enables non-invasive molecular imaging of tumor targeting and whole-body distribution. We aimed to evaluate ⁸⁹Zr-ERY974 tumor targeting and effect of T cells on tumor uptake in mouse models, including a humanized mouse model.

METHODS ERY974 and two control molecules namely bispecific CD3xkeyhole limpet hemocyanin (KLH) and KLHxKLH antibodies were radiolabeled with ⁸⁹Zr. Studies were performed in immunodeficient NOD/Shi-SCID/IL-2Rgnull (NOG) as well as human CD34+ hematopoietic stem cell engrafted NOG mice (huNOG), all subcutaneously inoculated with GPC3 overexpressing human hepatocellular carcinoma HepG2 cells. Mice received 10 µg ⁸⁹Zr-ERY974, ⁸⁹Zr-CD3xKLH or ⁸⁹Zr-KLHxKLH intravenously, with subsequent µPET scanning at 24, 72, 120 and 168 h followed by ex vivo biodistribution. Organs of interest were quantified on µPET scans as mean standardized uptake value (SUVmean) and with ex vivo biodistribution as % injected dose/gram of tissue (%ID/g). Tumor, spleen and lymph nodes were analyzed with autoradiography and immunohistochemical CD3 staining.

RESULTS µPET imaging revealed increased tumor-to-blood ratio (TBR) of ⁸⁹Zr-ERY974 in NOG over time with maximal TBR of 2.2±0.3 at 168 h post tracer injection (pi). At 168 h, tumor uptake was specific as ⁸⁹Zr-CD3xKLH and ⁸⁹Zr-KLHxKLH showed a TBR of only 0.6±0.2 and 0.8±0.3, respectively. In huNOG mice human CD3+ T cells were present in tumor, spleen and lymph nodes. In huNOG mice tumor uptake of ⁸⁹Zr-ERY974 was higher than in NOG mice as measured on µPET scans (SUVmean at 168 h pi 6.9±2.6 vs 2.9±0.2; P<0.01) and with ex vivo biodistribution (60.9±26.2 %ID/g vs 16.7±2.3 %ID/g; P<0.001), whereas ⁸⁹Zr-CD3xKLH tumor uptake in both mouse models was lower (P<0.05) but were similar in these mouse models. Autoradiography 168 h following ⁸⁹Zr-ERY974 administration to huNOG mice showed ⁸⁹Zr in extensive T cell infiltrate areas in the tumors of huNOG mice, whereas T cell infiltrate was lower in tumors of ⁸⁹Zr-CD3xKLH and ⁸⁹Zr-KLHxKLH injected huNOG mice. Spleens of huNOG mice showed CD3+ specific uptake as ⁸⁹Zr-ERY974 and ⁸⁹Zr-CD3xKLH uptake were higher than ⁸⁹Zr-KLHxKLH uptake(P<0.05), whereas spleen uptake in NOG mice of the 3 tracers was similar. Moreover, in huNOG CD3+ mesenteric lymph nodes ⁸⁹Zr-ERY974 uptake was higher than ⁸⁹Zr-KLHxKLH uptake (P<0.05)

CONCLUSION ⁸⁹Zr-ERY974 demonstrates specific tumor uptake in NOG and huNOG mice, while in huNOG mice tumor uptake colocalized with T cell rich infiltrate and also uptake in in spleen and lymph nodes was observed.

#3029

FAP-mediated tumor accumulation of a T-cell agonistic FAP/4-1BB DARPin drug candidate analyzed by SPECT/CT and quantitative biodistribution.

Christian Reichen,¹ Ralph Bessey,¹ Christine DePasquale,² Stefan Imobersteg,² Martin Behe,² Alain Blanc,² Roger Schibli,² Alexander Link,¹ Laurent Juglair,¹ Joanna Taylor,¹ Patricia Schildknecht,¹ Julia Hepp,¹ Elmar vom Baur,¹ Hong Ji,¹ Christof Zitt,¹ Victor Levitsky,¹ Keith M. Dawson,¹ Michael T. Stumpp,¹ Dan Snell¹. ¹Molecular Partners AG, Zurich, Switzerland; ²Paul Scherrer Institute, Villigen, Switzerland.

Immunomodulating agents have revolutionized anti-cancer therapy. However, monotherapy is often not sufficient, and development of combination treatments is hampered by cumulative toxicity. In an attempt to overcome this challenge, a tumor-restricted agonistic 4-1BB/FAP DARPin drug candidate, which induces T-cell co-stimulation only when clustered by binding to fibroblast activation protein alpha (FAP) expressing cells, has been developed. FAP is a type II membrane-bound glycoprotein abundantly expressed in the stroma of many solid tumors by cancer-associated fibroblasts. As shown previously using in vitro and in vivo models (HT-29), co-stimulation induced by a FAP-targeted 4-1BB agonistic DARPin molecule leads to enhanced activation and expansion of CD8+ T-cells. To support clinical development of the drug candidate, tumor localization and accumulation were studied by whole-body SPECT/CT imaging and quantitative biodistribution using Indium-111 labeled DARPin molecules in a human colorectal adenocarcinoma (HT-29) xenograft model in CD1 nude mice. Immunohistochemical staining of tumor stroma confirmed local expression of FAP. Labeled 4-1BB/FAP DARPin molecules specifically accumulated in FAP-expressing tumor in vivo. SPECT/CT imaging and biodistribution revealed a maximum tumor accumulation of around 15% of the injected dose per gram of tissue around 72 h post injection. High tumor/blood ratios were observed one week post injection because the activity in the blood decreased according to the expected serum half-life of 26 h, determined in separate pharmacokinetic studies in BALB/c mice following single dose intravenous bolus injections. Based on the decrease of radioactivity in the tumor, a tumor residence half-life of approximately 4 days was calculated, indicating an extended tumor retention potentially due to FAP binding. No accumulation was observed in the muscle tissue that was choosen as a rather weakly-perfused control tissue. Taken together, FAP-targeting of a 4-1BB agonist DARPin molecule resulted in expected high tumor accumulation and retention compared to an untargeted version of the molecule, both relevant observations for further preclinical and clinical studies. These findings suggest that tumor-targeting via FAP has the potential to induce T-cell activation restricted to the tumor site, and thereby reducing toxicities caused by systemic 4-1BB activation. In conclusion, immunostimulatory drugs with tumor-targeted activity may have the potential to circumvent current limitations of immunotherapy and allow safe and effective use, in particular in combination therapy.

#3030

Immuno PET imaging of PD-L1 expression in syngeneic and human xenograft tumor mouse models using a site-specific ⁸⁹Zr labeled PD-L1 antibody.

Camilla Christensen,¹ Lotte K Kristensen,² Hanna Toftevall,³ Maria Nordgren,³ Brian J. Agnew,⁴ Carsten H. Nielsen,² Andreas Kjaer¹. ¹Dept. of Clinical Physiology, Nuclear Medicine & PET and Cluster for Molecular Imaging, Dept. of Biomedical Sciences, Rigshospitalet and University of Copenhagen, Copenhagen Ø, Denmark; ²Minerva Imaging, Copenhagen, Denmark; ³Genovis AB, Lund, Sweden; ⁴Thermo Fisher Scientific, Eugene, OR.

Objectives: One of the major checkpoints probed for therapy is programmed cell death protein 1 (PD-1) and its ligand PD-L1. Despite remarkable clinical efficacy, not all patients benefit from immune checkpoint therapy with PD-1/PD-L1 antibodies, and a cohort of non-responding patients exists. Tumor PD-L1 expression measured by immunohistochemistry has been applied as a biomarker for selecting patients to increase the response rate of PD-1/PD-L1 blockage, but the methods are invasive and prone to sampling error. The objective of the current work was therefore to develop an image-based biomarker for non-invasive Positron-Emission Tomography (PET) imaging of PD-L1 expression

Methods: Anti-PD-L1 (clone 6E11, Genentech) was site-specifically conjugated to dibenzocyclooctyne-Deferoxamine (DIBO-DFO) using GlyclNATOR® (EndoS2) and SiteClick™ technology. Site-specific conjugated 6E11 was subsequently radiolabeled with ⁸⁹Zr.

The immunoreactivity and KD value were determined by cell binding assays. In vitro stability was assessed by high-performance liquid chromatography (HPLC) and the degree of labeling (DOL) was determined by LC/MS. A dose escalating study was performed with or without co-injection with unlabeled 6E11 to determine the optimal mass dose for PET imaging. Longitudinal PET/CT imaging was performed at various time-points after tracer injection in HCC827 tumor (lung adenocarcinoma) bearing animals and ex vivo biodistribution was performed after the last imaging time-point. Additionally, PET/CT imaging studies were carried out in different human xenograft and syngeneic tumor models with varying degree of PD-L1 expression. The syngeneic tumors received either fractionated external radiation therapy (XRT) or mouse Interferon gamma (IFNγ) treatment 3 days prior to ⁸⁹Zr-DFO-6E11 PET/CT imaging in order to evaluate treatment induced up regulation of PD-L1 expression.

Results: ⁸⁹Zr-DFO-6E11 was successfully labeled with a radiochemical purity >99% and the KD value was determined to 0.23 nM. The HCC827 tumors were identified by ⁸⁹Zr-DFO-6E11 PET imaging (3.7 ± 0.2 %ID/g, mean ± SD), and co-injection of unlabeled 6E11 increased the relative tumor uptake. Ex vivo biodistribution confirmed the in vivo results (5.4 ± 1.7 %ID/g) at 144 hours post injection. Non-invasive PET /CT imaging with ⁸⁹Zr-DFO-6E11 was able to detect a treatment induced up regulation of PD-L1 expression following treatment with XRT or IFNγ.

Conclusions: Site-specific labeling of antibodies with ⁸⁹Zr is feasible for immuno-PET imaging and tumor PD-L1 imaging by ⁸⁹Zr-DFO-6E11. PET/CT is an attractive approach for non-invasive whole body visualization of PD-L1 expression and could potentially aid in therapy guidance. Site-specific labeled antibodies create the opportunity to utilize imaging biomarker-driven approaches to achieve best clinical outcomes in immunotherapy.

#3031

Positron emission tomography imaging of activated T cells by targeting OX40 reveals spatiotemporal immune dynamics and predicts response to in situ tumor vaccination.

Aaron T. Mayer, Israt S. Alam, Idit Sagiv-Barfi, Kezheng Wang, Ophir Vermesh, Debra K. Czerwinski, Emily M. Johnson, Michelle L. James, Ronald Levy, Sanjiv S. Gambhir. Stanford University, Stanford, CA.

Clinical success of cancer immunotherapies has renewed interest in imaging the behavior of immune cells. Due to the spatiotemporally varying signatures of immune response, it has been difficult to monitor and predict patient outcomes using traditional clinical tests. ImmunoPET, defined herein as positron emission tomography utilizing radiolabeled antibodies, has the potential to enable noninvasive, sensitive and longitudinal interrogation of immune cell subset and state. Cell states including activation, anergy, and exhaustion may be more prognostic of disease outcome than the presence of tumor-infiltrating immune cells alone. In particular, T cell activation is thought to be critical to treatment success across many classes of cancer immunotherapy. In this work, we present the first radionuclide imaging of OX40, a novel and specific biomarker of activated antigen-specific T cells. Activation dependent and T cell restricted expression of OX40 was validated in vitro via flow cytometric analysis. Cell uptake studies with radiolabeled ⁶⁴Cu-DOTA-AbOX40 demonstrated ~11 fold [p < .0001] higher uptake in dyna-bead activated T cells compared to resting. The tracer showed negligible nonspecific uptake in OX40 blocked or OX40-/- T cells and low background levels across a panel of 5 cancer cell lines tested. In vivo, ImmunoPET imaging revealed new insights into response following in situ tumor vaccination with CpG, an adjuvant immunotherapy currently in clinical trials. Balb-C mice bearing dual A20 lymphoma tumors were administered low dose CPG directly in the left tumor (n=7-10), while vehicle control mice received PBS (n=7-10). Early after vaccination, imaging revealed increased OX40 radiotracer uptake in the CPG treated tumor (TT) [~37%; p<0.05] confirmed by immunofluorescent staining. ViSNE, a visualization technique for high-dimensional cytometry data, classified OX40+ single cells in a cluster associated with a nonregulatory, activated CD4 T cell phenotype. CPG treatment led to local expansion of this unique OX40 cell population [~63%; p<0.05]. By late time points, a full systemic response emerged as evidenced by increased Luminex cytokine measurements in the plasma of CPG-treated mice. Unsupervised hierarchical clustering based on radiotracer or cytokine measurements correctly assigned mice into CPG-treated or vehicle cohorts, with few exceptions. More remarkably, a generalized linear regression model indicated early PET signal (mean %ID/g) in the local tumor environment to be highly predictive of response outcomes at late timepoints [r²=0.746]. OX40 ImmunoPET provides a readily translatable approach for monitoring activated T cells with high sensitivity and specificity. In this instance, integration of molecular imaging and computational immunology enabled systems-level interrogation of vaccine response.

#3032

Quantitative three-dimensional imaging cytometry of tumor immune microenvironment.

Steve Seung-Young Lee, Stephen J. Kron, Vytautas P. Bindokas. The University of Chicago, Chicago, IL.

The tumor microenvironment is a three-dimensional (3D) system of diverse cellular and non-cellular components whose heterogeneous structure is typically defined by haphazard growth of cancer cells and a disordered microvasculature. Tumor-immune cell interactions occur within this context, providing a challenge for analysis of the infiltrate and anti-tumor immune responses by two-dimensional (2D) methods such as immunohistochemistry (IHC). Toward overcoming the limitations of 2D methods, we introduce Transparent Tissue Tomography (T3) as a tool for quantitative 3D imaging cytometry of the tumor immune microenvironment. For T3 imaging cytometry, tumors were sliced into 400 μm macrosections to facilitate immunofluorescence staining, optical clearing, and confocal microscopic imaging. Macrosections were stained overnight with panels of four to six directly labeled fluorescent primary antibodies then cleared by D-fructose. The full volume of each macrosection was scanned in each channel using a confocal microscope and tumor images were tomographically reconstructed from the macrosection images. The tumor images were segmented to discriminate cell types, map biomarkers, and perform spatial analysis. As an application of T3 imaging cytometry, we examined the distribution of programmed death-ligand 1 (PD-L1) expression in spontaneous Her2+ mammary tumors formed in BALB-NeuT mice. T3 analysis of whole tumors determined PD-L1 expression by tumor cells at the periphery and CD31+ vascular endothelium in the core. For the first time, 3D tomographic projection disclosed vascular PD-L1 expression localized between the endothelium and inner layer of smooth muscle cells. In turn, T3 revealed a strong spatial correlation between CD45+ immune cell distribution and PD-L1 expression. Toward translation, T3 was adapted to analyze whole core needle biopsies. Using T3 to map CD3+CD8+ cytotoxic T cells (CTLs) in BALB-NeuT tumors revealed marked inhomogeneity across single 18 gauge needle cores, confirmed by serial section IHC assays after T3 imaging. Modeling on-treatment biopsy analysis, T3 confirmed a compound effect of radiation and anti-PD-L1 therapy on infiltration of effector cytotoxic lymphocytes such as granzymeB+ NK cells and T cells. Applying T3 imaging cytometry to analyze needle cores from EGFR+ human head and neck cancer tissue, we mapped the distribution of CD3+CD8+ CTLs with respect to the CD31+ microvasculature and tumor cells. By assessing multiple tumor parameters simultaneously at cellular resolution, T3 provides a unique window into the heterogeneity of the tumor immune microenvironment. In particular, spatial, multiparameter T3 analysis may serve as a tool to improve diagnostic, prognostic and predictive testing of patient biopsies as part of evaluation for immune checkpoint blockade therapy.

#3033

Immuno-PET detection of LAG-3 expressing intratumoral lymphocytes using the zirconium-89 radiolabeled fully human anti-LAG-3 antibody REGN3767.

Marcus P. Kelly, Richard Tavare, Jason T. Giurleo, Sosina Makonnen, Carlos Hickey, Makenzie A. Danton, T Cody Arnold, Dangshe Ma, Jie Dai, Jerry Pei, Jessica R. Kirshner, William C. Olson, Gavin. Thurston. Regeneron Pharmaceuticals, Tarrytown, NY.

Persistent antigen exposure and inflammatory signals in tumors induce expression of various co-inhibitory or immune checkpoint receptors on T cells, including programmed death protein 1 (PD-1) and Lymphocyte-Activation Gene 3 (LAG-3). Therapeutic antibodies blocking such co-inhibitory receptors have produced durable antitumor responses as single agents and in combinations. In order to monitor LAG-3 expression and potential changes in expression due to therapeutic intervention, we have developed a radionuclide-conjugated antibody to LAG-3 for immuno-PET. The fully human anti-LAG3 antibody REGN3767 was radiolabeled with the positron-emitting radionuclide Zirconium-89 (⁸⁹Zr) using the bifunctional chelator p-SCN-Bn-Deferoxamine (DFO). ⁸⁹Zr-REGN3767 demonstrated high radiochemical purity and immunoreactivity in cell binding assays. The ability of ⁸⁹Zr-REGN3767 to successfully identify LAG-3 expression in vivo was initially assessed using MC38 mouse tumors expressing human LAG-3 (MC38/hLAG-3) implanted into immune-deficient mice. ⁸⁹Zr-REGN3767 demonstrated higher uptake in MC38/hLAG-3 tumors compared to an ⁸⁹Zr-isotype control antibody using immuno-PET, and specificity was confirmed by ex vivo biodistribution at day 6 post radiotracer injection (~35 and ~5 %ID/g for ⁸⁹Zr-REGN3767 and ⁸⁹Zr-isotype, respectively). Furthermore, a dose titration study of ⁸⁹Zr-REGN3767 in immune deficient mice co-implanted subcutaneously with Raji lymphoma cells and human peripheral blood mononuclear cells (hPBMCs) demonstrated the ability of ⁸⁹Zr-REGN3767 to target LAG-3-expressing intratumoral T-cells. ⁸⁹Zr-REGN3767 immuno-PET and ex vivo biodistribution demonstrated specific localization to Raji/hPBMC co-implanted tumors; this uptake was significantly higher at antibody doses of 0.03 - 0.3 mg/kg than at 5 mg/kg. Doses of 0.03-0.3 mg/kg ⁸⁹Zr-REGN3767 were also able to detect LAG-3 positive T cells in the spleen. This study shows the ability of ⁸⁹Zr-REGN3767 to successfully image LAG-3 expressed on intratumoral and splenic T lymphocytes. This work supports the clinical translation of anti-LAG-3 immuno-PET for the assessment of LAG-3 expression, with the goal to investigate its utility for predicting and monitoring response to checkpoint blockade therapy.

#3034

¹⁸F-FDG-positron emission tomography (PET)/CT enables the identification of checkpoint inhibitor immunotherapy (CIT) responders by determination of CIT-induced metabolic changes in secondary lymphatic organs.

Johannes Schwenck, Barbara Schörg, Francesco Fiz, Kilian Wistuba-Hamprecht, Andrea Forschner, Thomas Eigentler, Benjamin Weide, Claus Garbe, Martin Röcken, Christina Pfannenberg, Bernd J. Pichler, Christian la Fougere, Manfred Kneilling. University of Tuebingen, Tuebingen, Germany.

Although the majority of patients with metastatic melanoma achieve a prolongation of overall survival by using checkpoint inhibitor based immunotherapy (CIT), there is still a larger number of patients who do not benefit from this therapy. As a CIT-induced systemic immune response is required to promote the anti-tumor effect we analyzed the glucose metabolism in secondary lymphoid organs such as the spleen by ¹⁸F-FDG-Positron Emission Tomography (PET). In preclinical studies, we were able to distinguish responders from non responders by focusing on the spleen ¹⁸F-FDG-uptake of mice with CIT in experimental tumor models. Thus, we aimed to gain deeper insights into impact of CIT on the metabolism in secondary lymphatic organs to identify responders and to stratify patients for differential treatment strategies.

We retrospectively analyzed ¹⁸F-FDG-PET/CT scans (baseline and post-therapy) of 38 patients with metastatic melanoma with CTLA-4 or PD-1 Ab treatment as third line therapy (21 responder: 5x nivolumab; 7x pembrolizumab; 9x ipilimumab; 17 non-responder: 2x nivolumab; 11x pembrolizumab; 4x ipilimumab). Spleen regions of interest (ROI) were defined in the CT data, copied to the coregistered PET and analyzed semiquantitatively. Total lesion glycolysis (TLG) was calculated by multiplication of the spleen volume and the SUVmean.

We determined in the baseline ¹⁸F-FDG-PET/CT-scans (prior to CIT), no significant differences in spleen volume (221±18 cm³ vs. 209 ±22 cm³) and in the spleen ¹⁸F-FDG-uptake (SUVmean: 1,74±0,06vs.1,72±0,05; TLG: 384±37 vs. 359±36) between responders and non-responders. In the follow up ¹⁸F-FDG-PET/CT-scans 110±68 days after onset of CIT we measured a similar increase in spleen volume in responders (+8±6%) and non-responders (+7±5%). 15 out of 21 responders revealed an enhanced spleen ¹⁸F-FDG uptake when compared to the baseline ¹⁸F-FDG-PET/CT-scans. The mean standard uptake values in the spleen of responders increased by +10±9% SUVmean. In sharp contrast, we determined hardly any change in the spleen ¹⁸F-FDG uptake of non-responders (SUVmean -1,3±2,6%). Additionally, the total lesion glycolysis (TLG) in the CIT-responders increased stronger (+25±22%) than in non-responders (+6±6%).

Our results suggest that CIT-induced metabolic changes in secondary lymphatic organs are associated with therapy responds. Thus, non invasive ¹⁸F-FDG-PET/CT investigations might represent a powerful tool to monitor CIT-induced systemic immune responses in patients. Consequently, preclinical research is prerequisite to uncover the exact mode of action of CIT-induced systemic immune response in secondary lymphatic organs. Moreover, prospective clinical studies are essential to evaluate the prognostic value of our method.

#3035

⁸⁹Zr-labeled anti-PD-L1 Probody therapeutic CX-072 biodistribution in mice bearing human xenograft or murine syngeneic tumors.

Danique Giesen,¹ Linda N. Broer,¹ Marjolijn N. Lub-de Hooge,¹ Irina Popova,² Bruce Howng,² Olga Vasiljeva,² Elisabeth G. de Vries,¹ Martin Pool¹. ¹University Medical Center Groningen, Groningen, Netherlands; ²CytomX Therapeutics Inc., San Francisco, CA.

BACKGROUND Immune checkpoint inhibiting antibodies have antitumor activity across several tumor types, but are not effective in all patients and can elicit side effects. CX-072, a fully human Probody™ therapeutic currently in a phase 1/2 clinical trial, is reactive to the murine and human programmed cell death-ligand 1 (PD-L1) immune checkpoint. Probody therapeutics are engineered antibodies with target-binding region blocking masking peptides, which can be preferentially cleaved by tumor-associated proteases, yielding fully active antibodies. CX-072 may thus preserve anti-tumor efficacy, while limiting side effects. We radiolabeled CX-072 with the positron emission tomography (PET) isotope zirconium-89 (⁸⁹Zr) to reveal its tumor targeting properties and whole body distribution using non-invasive PET imaging.

METHODS CX-072 and a non-specific Probody therapeutic control (PbCtrl) were radiolabeled with ⁸⁹Zr. For in vivo studies, PD-L1 expressing MDA-MB-231 human breast cancer cells were subcutaneously (sc) engrafted in Balb/c nude mice. To assess tracer protein dose dependency of the tumor uptake, mice received 10 μg ⁸⁹Zr-CX-072 or ⁸⁹Zr-PbCtrl (~5 MBq) supplemented with 0, 40 or 240 µg of unlabeled CX-072 or PbCtrl. To evaluate ⁸⁹Zr-CX-072 biodistribution in an immune-competent setting, C57BL6 mice were implanted sc with low PD-L1 expressing MC38 syngeneic murine colon adenocarcinoma cells. All mice underwent serial in vivo PET imaging 1, 3 and 6 days post injection (pi), quantified by mean standardized uptake value (SUVmean) and followed by ex vivo biodistribution. Activated Probody species in tissues were detected by Western capillary electrophoresis.

RESULTS PET imaging revealed increasing ⁸⁹Zr-CX-072 tumor accumulation between 1-6 days pi, with the highest SUVmean of 1.5 (± 0.2) observed for 10 µg at 6 days pi. Ex vivo biodistribution analysis showed 8.7 % injected dose per gram (%ID/g) tumor uptake for 10 µg ⁸⁹Zr-CX-072 versus 3.8 %ID/g for 10 µg ⁸⁹Zr-PbCtrl (P<0.01) in MDA-MB-231 xenografted mice. In the syngeneic MC38 model biodistribution analysis showed modest tumor uptake for 10 μg ⁸⁹Zr-CX-072 and ⁸⁹Zr-PbCtrl (6.5 vs 5.5 %ID/g, P=0.24; tumor-to-blood ratio of 0.61 vs 0.45, P<0.05). ⁸⁹Zr-CX-072 uptake in lymphoid tissues (spleen, lymph nodes) was similar to ⁸⁹Zr-PbCtrl. Activated Probody species were predominantly detected in tumor with lesser amounts present in lymphoid tissues.

CONCLUSION ⁸⁹Zr-CX-072 accumulates more in PD-L1-expressing tumor tissues than in lymphoid tissues. A sub-study of an ongoing clinical study (PROCLAIM-CX-072) is designed to validate study drug distribution in patients using a good manufacturing practice (GMP) quality ⁸⁹Zr-CX-072 tracer.

#3036

PEGylated hyaluronidase increases tumor uptake of ⁸⁹Zr-DFO-HuMab-5B1 (MVT-2163) in a CA19-9 positive hyaluronan-accumulating pancreatic cancer model.

Jonah Rainey,¹ Paul Maffuid,¹ Wolfgang W. Scholz,¹ Jack Ostrowski,¹ H Toni Jun,¹ Paul Resnick,¹ Xiaoming Li,² Jesse D. Bahn,² Susan Zimmerman,² Kelly Chen,² Barbara Blouw,² Curtis B. Thompson,² Daniel C. Maneval,² David W. Kang². ¹MabVax Therapeutics Holdings Inc., San Diego, CA; ²Halozyme Therapeutics, Inc., San Diego, CA.

Hyaluronan (HA) accumulates in the extracellular matrix of many solid tumors, and is associated with poor prognosis in pancreatic ductal adenocarcinoma. In non-clinical models, enzymatic degradation of HA with PEGylated recombinant human hyaluronidase PH20 (PEGPH20) has been shown to remodel the tumor stroma, reduce tumor interstitial fluid pressure, and expand tumor blood vessels, resulting in enhanced delivery of therapeutic and imaging agents, such as monoclonal antibodies. ⁸⁹Zr-DFO-HuMab-5B1 (MVT-2163) is a targeted ImmuoPET imaging agent for CA19-9 positive malignancies currently in clinical evaluation. CA19-9 plays a role in tumor adhesion and metastasis, and is an independent prognostic indicator of cancer survival. CA19-9 is expressed in pancreatic and other cancers, including small cell lung and colon. In this study, we aimed to show the effects of PEGPH20 on the biodistribution of MVT-2163 in a CA19-9 positive HA-rich human pancreatic tumor xenograft model.

Nude mice were peritibially implanted on the right hind limb with 5x10⁶ human pancreatic tumor cells engineered to overexpress hyaluronan synthase 3 (BxPC3/HAS3). Tumor volumes were measured by MR imaging, and mice were staged when average tumor size reached 320 mm³. The control group received intravenous (IV) vehicle, and the test group received PEGPH20 (IV, 1 mg/kg). Both control and test groups also received MVT-2163 (IV, 3 mg/kg) 24 hours later. PET images were captured from 2 to 120 hours post MVT-2163, and region of interest (ROI) of tumor and liver were analyzed (N=6/group). After the final PET image, tumors and livers were harvested for ex vivo gamma counting.

Ex vivo gamma counting demonstrated improved MVT-2163 accumulation in animals treated with PEGPH20, with a 50.9% increase in counts in the excised tumors. ROI analysis of the tumors showed increased tumor SUV of 8.0, 9.4, and 24.1% at 72, 96, and 120 hours post-injection, respectively, with PEGPH20 compared to vehicle. ROI analysis of liver regions showed an 11.4 to 26.4% reduction in SUV in the PEGPH20 group compared to control. Ex vivo counts of liver tissue confirmed the PET signal reduction with PEGPH20. Analysis of tumor-to-liver ratios showed average increases of 34.1, 35.7, and 58.5% at 72, 96, and 120 hours, respectively, post MVT-2163 injection in PEGPH20 treated mice.

In summary, as measured by SUV, PEGPH20 increased both the tumor uptake and the tumor-to-liver ratios of MVT-2163 in a CA19-9 positive xenograft mouse model of HA accumulating pancreatic cancer. Ex vivo analysis confirmed in vivo results. Taken together, the increased tumor uptake and the decreased liver uptake support further investigation into the potential clinical utility for the combination of PEGPH20 and MVT-2163.

#3037

Comparison of multiplexed imaging mass cytometry in FFPE tissue to monoplex immunohistochemistry.

Navi Mehra,¹ Carsten Schnatwinkel,¹ Elliott Ergon,¹ Joseph Krueger,¹ Karl Calara-Nielsen,² Brad Foulk,² Kirti Sharma,² Denis Smirnov,² Chandra Rao,² Tatiana Perova,² Rengasamy Boominathan². ¹Flagship Biosciences, Westminster, CO; ²Janssen Pharmaceuticals, Spring House, PA.

Introduction: Multiplexed analysis of limited tissue samples can improve our understanding of tumor biology and tumor microenvironment. Chromogenic and fluorescent multiplexed immunohistochemistry (IHC) approaches are available and offer great insights while conserving limited tissue however these approaches have their limitations. Multiplexed chromogenic IHC methods can at best accommodate up to 2-3 distinct markers. Fluorescence-based approaches can support higher degree of multiplexing however spectral overlap issues and differences in labeling efficiency/photostability complicate experimental procedures and data interpretation. Imaging mass cytometry system (IMC) has recently emerged as a novel technology for tissue imaging that enables multiplexed analysis protein expression (up to 40 markers) in a single tissue sample while circumventing the limitations of chromogenic and fluorescent IHC techniques. Metal-conjugated antibodies are used to perform qualitative and quantitative analysis of expression of multiple proteins of interest on a single formalin fixed paraffin embedded (FFPE) tissue slide. Here, we compare the performance of the IMC method with conventional, established IHC techniques using a small panel of markers.

Methods: Serial sections of FFPE tonsil and non-small cell lung carcinoma tissues were assessed by monoplex IHC and multiplex IMC for CD3 (Cell Signaling Technology, D7A6E), CD8 (LS Bio, C8/144B), CD68 (Abcam, KP1), PD-L1 (Spring Bio, SP142) and Histone H3 (D1H2). Digital image analysis using Flagship’s image analysis software was used to compare performance characteristics of multiplex IMC platform with standard monoplex chromogenic IHC.

Results: The staining patterns of the corresponding biomarkers were similar between IMC and IHC on sequential sections. Digital image analysis demonstrated concordance in the percentage of biomarker positive cells within analyzed matched IHC and IMC areas. It was also demonstrated that cellular image segmentation can be performed on IMC images thus allowing for utilization of various software packages for high dimensional single cell analysis of IMC data.

Conclusion: Comparative digital image analysis indicates that on FFPE tissues multiplexed IMC platform generates data comparable to that obtained from the monoplex chromogenic IHC platform. We believe that an IMC platform is a new tool capable of dramatically enhancing our ability to study biology of cancer using highly multiplexed analysis of limited tissue samples.

#3038

Intraoperative breast cancer margin detection via polarization-sensitive optical coherence tomography.

Marina Marjanovic,¹ Jianfeng Wang,¹ Yang Xu,¹ Eric J. Chaney,¹ Darold R. Spillman,¹ Anna M. Higham,² Natasha N. Luckey,² Kimberly A. Cradock,² George Z. Liu,² Stephen A. Boppart¹. ¹University of Illinois at Urbana-Champaign, Urbana, IL; ²Carle Foundation Hospital, Urbana, IL.

Current standard-of-care for breast conserving surgery (BCS) procedures relies on post-operative histopathology to provide a microscopic view and assessment of surgical margins. To avoid high reoperation rates there has been a great interest in developing new imaging solutions to visualize tissue intraoperatively, and provide real-time feedback on the margin status. Optical coherence tomography (OCT) is a high-resolution label-free, imaging technology that is the optical analogue to ultrasound imaging, except images are based on backscattered near-infrared light. Fundamentally, the contrast mechanism of OCT relies on the light scattering properties associated with different tissue structures. Unlike normal tissue with well-organized tissue structure, the structure of cancer tissue is often disorganized and typically characterized by variable cell sizes, abnormal shapes, and enlarged nuclei, resulting in different optical scattering properties, and enabling OCT to reveal differences between normal and cancerous tissues with high spatial resolution. We have previously demonstrated OCT for in vivo label-free intraoperative assessment of the breast conserving surgery (BCS) with a hand-held OCT probe that is integrated with a portable OCT system. Despite the technological advance to achieve in vivo OCT imaging during BCS, there remains a critical need to improve the OCT-based detection, differentiation, and diagnosis of breast tumor tissue.

Polarization-sensitive optical coherence tomography (PS-OCT) is a functional extension of standard OCT that maps changes in the polarization state of light induced by anisotropic tissue properties. By exploiting the interaction between the polarization state of light and tissue, additional structural and functional information can be extracted. Form birefringence is the main optical property present in tissues responsible for altering the polarization state of light. In breast tissue, birefringent structures include the connective tissue or normal stroma, composed largely of collagen fibers. In the presence of breast cancer, the birefringent collagen fibers are degraded, displaced, reorganized, and/or disrupted in ways that dramatically alter the PS-OCT signal, serving as a fundamentally new optical imaging biomarker for the presence of cancer in the breast.

In this study, we report data from 23 cancer surgeries, compared with healthy breast tissue samples obtained from the breast reduction surgeries. Our results show strong correlation of observed tissue birefringence and collagen fiber density, and the lack of the tissue birefringence, due to low collagen fiber density, in the cancerous areas. Our PS-OCT imaging system provides faster and more informative intraoperative tumor margin assessment. It has the potential to determine margin status in real-time during the surgical procedure and reduce reoperation rates.

#3039

Role of IDH mutation status on molecular and spatial heterogeneity in glial tumors across progression and recurrence.

Michael E. Berens,¹ Jill S. Barnholtz-Sloan,² Miribella Rusu,³ John Graf,³ Anup Sood,³ Sanghee Cho,³ Maria Zavodszky,³ Sara Byron,¹ Rebecca Halperin,¹ Yi Fritz,² Seungchan Kim,⁴ Fiona Ginty³. ¹TGen (The Translational Genomics Research Institute), Phoenix, AZ; ²Case Western Reserve University Comprehensive Cancer Center, Cleveland, OH; ³GE Global Research, Niskayuna, NY; ⁴Prairie View A&M University, Prairie View, TX.

We developed and deployed a workflow for generating multi-scale, multiparametric imaging data, feature extraction and/or converting to higher scales which equips multiple analysis approaches to differentiate clinically variable phenotypes of glial tumors. The workflow quantifies spatial heterogeneity (concordance of adjoining cells) and molecular heterogeneity (varied cell states determined by protein abundance) of glial tumors at the genomic, tissue, and medical imaging scales including IDH mutation status and progression/recurrence status. A panel of 24 multiplexed immunofluorescence (MxIF) markers (addressing 9 hallmarks of cancer) was used to profile single cells (in the thousands) in tissue sections from each of 31 glial tumors (ranging from primary grade II to IV, and recurrent grade IV). Pre-resection multi-parameter MR images were feature extracted from discreet habitats (necrosis, enhancing, and edema); whole exome and transcriptome sequencing from bulk viable tumor were analyzed. By MxIF, the various states of individual cells from treatment-naive patient specimens resolved unsupervised into 7 clusters, for which Cluster 2 (including cells from 9 patients) and Cluster 6 (including cells from 8 patients) contained the two larger bundles of patient cases. When separated into IDHmt and IDHwt cases, cells from IDHmt cases frequently contained cell populations dominated by a single cluster (low molecular heterogeneity); cells from cases with IDHwt represented multiple different clusters (high molecular heterogeneity). In grade III astrocytomas, and grade IV recurrent glioblastomas, spatial heterogeneity of the hallmark inducing angiogenesis was elevated in the IDHmt tumors compared to IDHwt, while between the same groups, molecular heterogeneity was lower in the IDHmt cases than wild type. Edema from T1w post contrast MR imaging was found to be elevated in IDHwt gliomas relative to IDHmt, while enhancement was reduced in IDHwt compared to IDHmt tumors. The findings demonstrate that IDHmt gliomas, irrespective of grade, show less edema, greater enhancement, and greater spatial heterogeneity of the inducing angiogenesis hallmark but lower molecular heterogeneity than IDHwt tumors. Molecular heterogeneity of cancer invasion also differed between IDHmt and IDHwt cases. Longer survival duration following diagnosis for patients with IDHmt gliomas may reflect generalized altered molecular and spatial heterogeneity, which is a phenotype evident on medical imaging. [Clinically-annotated specimens originated from the Ohio Brain Tumor Study and the Ivy GBM Clinical Trials Consortium]

#3040

Radiomics discriminates pseudo-progression from true progression in glioblastoma patients: A large-scale multi-institutional study.

Srishti Abrol,¹ Aikaterini Kotrotsou,¹ Ahmed Hassan,¹ Nabil Elshafeey,¹ Tagwa Idris,¹ Naveen Manohar,¹ Anand Agarwal,¹ Islam Hassan,¹ Kamel Salek,¹ Nikdokht Farid,² Carrie McDonald,² Shiao-Pei Weathers,¹ Naeim Bahrami,² Samuel Bergamaschi,³ Ahmed Elakkad,¹ Kristin Alfaro-Munoz,¹ Fanny Moron,⁴ Jason Huse,¹ Jeffrey Weinberg,¹ Sherise Ferguson,¹ Evangelos Kogias,⁵ Amy Heimberger,¹ Raymond Sawaya,¹ Ashok Kumar,¹ John de Groot,¹ Meng Law,³ Pascal Zinn,⁴ Rivka R. Colen¹. ¹UT MD Anderson Cancer Center, Houston, TX; ²University of California San Diego, San Diego, CA; ³University of South California, Los Angeles, CA; ⁴Baylor College of Medicine, Houston, TX; ⁵University Medical Center Freiburg, Germany.

BACKGROUND: Treatment-related imaging changes are often difficult to distinguish from true tumor progression. Treatment-related changes or pseudoprogression (PsP) subsequently subside or stabilize without any further treatment, whereas progressive tumor requires a more aggressive approach in patient management. Pseudoprogression can mimic true progression radiographically and may potentially alter the physician’s judgment about the recurrent disease. Hence, it can predispose a patient to overtreatment or be categorized as a non-responder and exclude him from the clinical trials. This study aims at assessing the potential of radiomics to discriminate PsP from progressive disease (PD) in glioblastoma (GBM) patients.

METHODS: We retrospectively evaluated 304 GBM patients with new or increased enhancement on conventional MRI after treatment, of which it was uncertain for PsP versus PD. 149 patients had the histopathological evidence of PD and 27 of PsP. Remaining 128 patients were categorized into PD and PsP based on RANO criteria performed by a board-certified radiologist. Volumetrics using 3D slicer 4.3.1 and radiomics texture analysis were performed of the enhancing lesion(s) in question.

RESULTS: Using the MRMR feature selection method, we identified 100 significant features that were used to build a SVM model. Five texture features (E, CS, SA, MP, CP) were found to be most predictive of pseudoprogression. On Leave One Out Cross-Validation (LOOCV), sensitivity, specificity and accuracy were 97%, 72%, and 90%, respectively. Using 70% of the patient data for training and 30% for validation, an AUC of 94% was achieved, with the sensitivity of 97% and specificity of 75%.

CONCLUSION: 3D radiomic texture features of conventional MRI successfully discriminated pseudoprogression from true progression in a large cohort of GBM patients.

#3041

Molecular, histopathologic, MRI and PET findings in syngeneic oncologic mouse models.

Julia Schueler,¹ Artem Shatillo,² Jussi Rytkönen,² Nina Zanella,¹ Antti Nurmi,² Tuulia Huhtala². ¹Charles River, Freiburg, Germany; ²Charles River, Kuopio, Finland.

Experimental tumors raised in rodents represent an important preclinical tool to develop innovative anticancer compounds before clinical testing. Amongst others such models include solid tumors raised in syngeneic fully immunocompetent hosts and tumors spontaneously growing in genetically engineered mice (GEM) and derivate thereof. These model platforms have gained additional value since the manipulation of the immune system to fight cancer has led to tangible benefits for cancer patients. Imaging has become important part of the basic and translational research. It enables e.g. metabolic activity, volumetric and biodistribution monitoring within same individual over time. Magnetic resonance imaging (MRI) and positron emission tomography (PET) are widely used for clinical diagnosis and disease follow up. Choosing the most suitable imaging application depends of the prioritization of features. MRI offers unprecedented soft tissue contrast, high spatial resolution and non-invasive nature renders MRI in rodents a perfect tool for preclinical work in oncological applications. Similarly to clinical setting, MRI in rodents is used for detection of the tumors, evaluation of therapeutic response and induced changes, with longitudinal follow-up for early detection of possible tumor recurrence. PET is an excellent tool to study tumor metabolism, location as well biodistribution of novel antibodies. The purpose of this work was to study molecular as well volumetric, metabolic and functional changes in syngeneic oncological mouse model using MRI, MRS and PET imaging. Our panel consists of 32 fully characterized mouse tumor models covering 12 major cancer types. We characterized those models by molecular profiling (whole exome sequencing and RNAseq), histopathological examination (tissue micro array as well as whole slide analysis of Ki67, CD31, SMA, TIL infiltration) and sensitivity towards checkpoint inhibitors. Amongst other findings we saw an enhanced activity of anti-CTLA-4 treatment in combination with anti-PD1 in 4T1 orthotopically implanted into the mammary fat pad in comparison to single agent therapy. In a follow-up study we implanted 14 mice with breast cancer cell line 4T1 orthotopically into the mammary fat pad. When tumors were palpable treatment with the checkpoint inhibitor combination started. During the course of the experiment we determined tumor volume of the primary lesion twice weekly. In addition to molecular markers, MRI, MRS and PET were applied to study changes in tumor over cancer progression. As a conclusion, several syngeneic oncological mouse models have been characterized using molecular profiling, histological techniques and in vivo imaging. These readouts provide a powerful and translational research tool together with oncological disease animal models allowing comprehensive evaluation of disease progression and treatment interventions for in vivo studies.

#3042

Nuclear morphology predicts prostate cancer metastasis at diagnosis.

Fangjin Huang,¹ Nathan Ing,¹ Eric N. Miller,² Hootan Salemi,¹ Michael S. Lewis,² Isla P. Garraway,² Arkadiusz Gertych,¹ Beatrice S. Knudsen¹. ¹Cedars-Sinai Medical Center, Los Angeles, CA; ²David Geffen School of Medicine at UCLA, Los Angeles, CA.

Background: Prostate needle biopsies (PNBX) that are obtained during the initial diagnostic workup are often the only source of tumor tissue that is available to determine the severity of prostate cancer (PC). Interestingly, tumor grades and pathologic features in PNBXs of men with lethal metastatic disease (M1 stage) are indistinguishable from those of high-grade nonmetastatic tumors of patients who do not progress after treatment (M0 stage). We hypothesize that the morphology of tumor nuclei can be used as a source of biomarker development to distinguish M1 tumors from high-grade localized M0 tumors.

Methods: Our study consists of a cohort of 85 high-grade M0 and 78 M1 cases, within a biorepository of 2150 patients at the Greater Los Angeles Veterans Affairs hospital. PNBX slides were digitized at 40X and pathologists annotated all cancer foci. These annotated regions were divided into smaller image tiles (dimensions) and fed into our digital image analysis pipeline for nuclear segmentation and feature extraction. We applied two distinct feature extraction methods to capture morphologic information within tumor nuclei. The 64 Handcrafted (HC) features describe nuclear properties such as shape, area, and chromatin conformation. The 62 Autoencoder (AE) features are abstract descriptors generated by a deep learning algorithm, which learns to redraw the nuclei. We built 2 machine learning models using AE or/and HC features for classification of M1 versus M0 cases. We divided our patients into 3 groups: training, testing and validation. A 7-fold cross-validation was performed in the training and testing sets and the area under the curve of the receiver operating characteristic (ROC) was calculated in the validation set.

Results: Model 1 utilizes processed features derived from patient-level distributions and generalized linear model. The estimated AUC for predicting M1 stage is 0.79 and 0.74 for HC and AE features, respectively. Model 2 uses the average of each features obtained from dominant group of nuclei within individual tiles and neural network models. Nuclei groups were assigned by unsupervised clustering method. For this model, the AUC of predicting M1 stage is 0.77 and 0.75 for HC and AE features, respectively.

Conclusion: By applying two distinct feature extraction methods and two approaches to summarize features, we obtain similar prediction accuracies. These results demonstrate that quantitative nuclear features contain information to classify M1 and M0 cases, which cannot be classified based on tumor grade or pathologic features.

#3043

Quantitative MRI during neoadjuvant therapy for predicting breast cancer response in the community setting.

Anna Sorace,¹ Jack Virostko,¹ Chengyue Wu,¹ Angela M. Jarrett,¹ Stephanie L. Barnes,¹ Debra Patt,² Boone Goodgame,³ Sarah Avery,⁴ Thomas E. Yankeelov¹. ¹University of Texas at Austin, Austin, TX; ²Texas Oncology, Austin, TX; ³Seton Healthcare Family, Austin, TX; ⁴Austin Radiological Association, Austin, TX.

Purpose: Quantification of accurate and early response to neoadjuvant therapy (NAT) provides the opportunity to replace an ineffective treatment with an alternative regimen, thereby potentially avoiding ineffective systemic therapy. Quantitative magnetic resonance imaging (MRI) has been shown to predict breast cancer response to treatment early during the course of NAT within academic medical centers. Integrating quantitative imaging techniques into community-based medical practices has the potential to reach a larger percentage of breast cancer patients, and allow more community center participation in clinical trials that require quantitative imaging. This study evaluated the reproducibility and accuracy of quantitative dynamic contrast-enhanced MRI (DCE-MRI) and diffusion-weighted MRI (DW-MRI) in the community setting and the ability to implement these techniques to predict the eventual response of breast tumors to NAT.

Experiment Procedure: MRI was performed at two community imaging centers and one academic research facility using a 3T Siemens Skyra scanners equipped with a breast coil. To assess reproducibility across sites, normal subjects (N=3) were scanned at three imaging centers. Quantitative T1 (required for pharmacokinetic modeling of DCE-MRI) and apparent diffusion coefficient (ADC) maps were calculated of normal fibroglandular breast tissue. Accuracy was tested through evaluation of T1 and DW-MRI in phantoms. Women undergoing NAT for breast cancer (N=10) were scanned with DCE-MRI and DW-MRI at baseline (prior to beginning therapy) and three longitudinal time points during the course of NAT to evaluate early prediction of response to therapy. Quantitative measures of ADC (evaluating cellularity from DW-MRI) and Ktrans (evaluating vascular perfusion and permeability from DCE-MRI) were calculated for the segmented tumor volume. Imaging was compared to pathology reports at the conclusion of NAT.

Results: Reproducibility scans of normal breast fibroglandular tissue yielded an average difference of 8.4% and 7.0% in T1 and ADC measurements, respectively, across sites. Phantom studies revealed accurate measurements of T1 mapping and ADC, with reproducibility measurements showing a difference of 2.8% and 1.6%, respectively. Patients achieving a pathological complete response (pCR) revealed a 13.8% ± 19.0% increase in the mean ADC values of the tumor from t1 (baseline, prior to beginning NAT) and t2 (following one round of NAT) and a 15.4% ± 40.9% decrease in mean Ktrans. Conversely, patients that did not achieve non-pCR had little change in ADC (-0.9% ± 12.6% change between t1 and t2) and a 15.3% ± 42.8% increase in Ktrans.

Conclusions: Quantitative MRI has been shown to be accurate and reproducible across community medical centers. Furthermore, the preliminary results discussed above parallel those previously reported in the academic research setting.

#3044

Label-free detection and identification of cancer-associated vesicles in human breast cancer.

Sixian You,¹ Yi Sun,¹ Haohua Tu,¹ Ronit Barkalifa,¹ Eric J. Chaney,¹ Marina Marjanovic,¹ Anna M. Higham,² Natasha N. Luckey,² Kimberly A. Cradock,² Z. George Liu,² Stephen A. Boppart¹. ¹University of Illinois at Urbana-Champaign, Urbana, IL; ²Carle Foundation Hospital, Urbana, IL.

Detection and identification of cancer-associated vesicles in their native tissue environment is essential to understand vesicle-mediated communication in cancer progression and potentially facilitate the treatment and management of cancer. Here we develop a label-free optical imaging method that can reliably detect cancer-associated vesicles in vitro, in vivo and in untreated human tissue ex vivo. Preliminary analysis conducted on fresh human breast tissue obtained from healthy and cancer subjects showed that a significant portion of vesicles from the cancer subjects have unique optical signatures in comparison with those from healthy subjects. Additionally, the spatial distribution of these vesicles with unique optical signatures in the tumor micro- and macro-environment appeared correlated with the staging of human breast cancer. These results suggest the potential capability of this methodology to identify and characterize the tumor associated vesicles in the authentic tumor microenvironment, which could open new windows in the explorations of their diagnostic values and physiological roles in cancer progression.

#3045

Pilot study of kernel-based dynamic (KBD) FDG-PET in patients with borderline resectable pancreatic cancer.

John P. Schwerkoske, Guobao Wang, Kit W. Tam, Jasmine C. Huynh, Heather H. Hunt, Michael L. Rusnak, Cameron C. Foster, Michael T. Corwin, Karen E. Matsukuma, Dorina Gui, May T. Cho, Richard J. Bold, Ramsey D. Badawi, Edward J. Kim. UC Davis, Sacramento, CA.

Introduction: Patients (pts) with borderline resectable (BR) pancreatic cancer (PC) derive a clear survival benefit when margin-negative resection can be achieved after neoadjuvant treatment (tx). As such, reliable imaging is desperately needed to identify those patients with the best chance of surgical benefit. Current imaging with CT/MRI and standard ¹⁸F-fludeoxyglucose (FDG) PET lacks sensitivity and specificity to resolve residual tumor contact with the large blood vessels of the vascular groove. Dynamic PET with kernel-based image reconstruction developed by our group has potential to overcome limits of standard static PET by robust parametric imaging of radiotracer kinetics. KBD FDG-PET provides the FDG kinetic parameter K1 for tumor perfusion and Ki for tumor glucose utilization. A lower Ki/K1 ratio indicates a better perfused, less metabolically active tumor and is derived by our method without need for separate perfusion CT/MRI. Methods: Pts with PC staged as BR as defined by consensus guidelines were eligible. Pts were assessed by CA 19-9 and CT/MRI along with KBD FDG-PET both pre-tx and post-tx. KBD FDG PET was performed by IV bolus of 10 mCi of FDG followed by 60-min PET data acquisition. A 3-compartment model with 5 micro kinetic parameters K1, k2, k3, k4, and fv was used where K1 (mL/min/mL) denotes the rate constant of FDG transport from plasma to tissue, k2 (1/min) the transport rate from tissue to plasma, k3 (1/min) the rate of FDG conversion to FDG 6-phosphate, k4 (1/min) the rate of dephosphorylation, and fv the fractional blood volume. Ki was calculated from micro parameters by the formula K1*k3/(k2+k3). Parametric images of these kinetic parameters were obtained by voxel-wise implementation of kinetic modeling. Results: 4 pts were enrolled, 3 pts had pre- and post-tx KBD FDG-PET, and 2 pts were resected. In the 3 evaluable pts, an overall 62% decrease in tumor glucose utilization was observed (mean pre-tx Ki=.03, [range .013-.045]; mean post-tx Ki=.011 [range .004-.016]. Mean pre-tx K1=.37 (range .26-.50) and mean post-tx K1=.38 (range .31-.46). Delta Ki was -17%, -47%, and -91% and delta Ki/K1 was -9%, -28%, and -95% in pts 1, 2, 3 respectively. Pt 3 had the highest pre-tx Ki and the lowest pre-tx K1, but had a >90% decrease in Ki and Ki/K1 post-tx, indicating a significant decrease in metabolic activity and concomitant increase in perfusion. Despite persistent superior mesenteric vein (SMV) abutment on standard post-tx imaging, margin-negative resection was achieved. In contrast, pt 1 had the lowest pre-tx Ki and highest pre-tx K1, but had a <20% decrease in Ki and Ki/K1 post-tx. Like Pt 3, Pt 1 also had SMV contact on standard post-tx imaging, but in contrast, experienced margin-positive resection. Conclusions: KBD FDG-PET is a promising modification of standard static PET that provides useful kinetic parameters for evaluation of PC resectability.

#3046

Evaluation of pancreatic ductal adenocarcinoma with acidoCEST MRI.

Rachel A. High,¹ Edward A. Randtke,¹ Kyle M. Jones,¹ Mark D. Pagel². ¹University of Arizona, Tucson, AZ; ²UT MD Anderson Cancer Center, Houston, TX.

Introduction: Our goal is to evaluate the role of using extracelluar pH (pHe) as a predictor of pancreatic ductal adenocarcinoma (PDAC) development using a non-invasive, in vivo imaging technique called acidoCEST MRI, to improve early detection of pancreatic tumors.

Methods: Spontaneous PDAC development was initiated by administering 14 caerulein injections over a 62 hour period to a KrasLSL.G12D/+; PdxCre mouse model. Caerulein induces pancreatitis which drives tumor development in the context of a Kras mutation. Animals were imaged with a Bruker BioSpin 7T MRI scanner with an acidoCEST MRI protocol at pre-injection, 1 hour and 48 hours post final injection of caerulein, and weekly until tumors reached a size of 200mm³. During all MR imaging, mice were anesthetized with 2.0% isofluorane, maintained at a respiration rate of 35-40 bpm, and maintained at a body temperature of 37°C. For acidoCEST MR imaging, mice were injected with 370 mg/mL iopamidol (200 μL IV bolus and 400 μL/hr IV infusion) and scanned with a 6 sec saturation time at 3.5 μT, with a 300 ms acquisition time. Pixel-wise parametric maps of pancreatic pHe values were generated via fitting the CEST spectrum with the Bloch-McConnell equations in MATLAB R2016a. Average pancreatic pHe was recorded.

Results: AcidoCEST MRI was successful in acquiring in vivo pHe of normal and tumor pancreatic tissue, both with sufficient uptake of the iopamidol contrast agent. The pHe values for healthy pancreatic tissue ranged from 6.83 to 7.33 with an average of 6.96 (n=14). 1 hour after the last caerulein injection, during pancreatic inflammation, pHe ranged from 6.78 to 7.00, with an average of 6.92 (n=6). Tumors began to reach a size of 200mm³ at 5 weeks post caerulein injections. In tumor tissue, pHe was 6.59 (n=1), demonstrating tumor acidosis.

Conclusions: We have demonstrated that acidoCEST MRI can be used to quantify pHe in healthy pancreatic tissue, pancreatitis, and pancreatic ductal adenocarcinoma. These pHe values will be used to determine if acidity in pancreatitis correlates with future development of PDAC and/or a more aggressive phenotype. Further pre-clinical studies using acidoCEST MRI may be performed to monitor early therapeutic response to pancreatic tumor treatment.

#3047

Development and initial application of a porcine-specific MRI and MRE protocol for liver imaging in a large animal cancer model.

Faith M. Thomas,¹ Bradley P. Sutton,¹ Tracey M. Wszalek,¹ Megan J. Dailey,¹ Kyle M. Schachtschneider,² Lawrence B. Schook,² Ron C. Gaba². ¹University of Illinois at Urbana-Champaign, Urbana, IL; ²University of Illinois at Chicago, Chicago, IL.

While global cancer mortality is generally decreasing, the incidence of hepatocellular carcinoma (HCC) is projected to increase given the growing prevalence of chronic liver diseases that increase the risk for carcinogenesis. The current diagnosis of liver cirrhosis is invasive tissue biopsy, and those at-risk undergo an HCC surveillance program of ultrasound imaging to survey tumor development. Such screening programs center on static radiologic imaging snapshots prescribed at arbitrary intervals based on empiric histologic cirrhosis staging schemes and observational tumor developmental data that neither reflect the transition between normal and diseased states nor biologically relevant disease thresholds. In contrast, magnetic resonance imaging (MRI) and magnetic resonance elastography (MRE) may provide a more foundational portrayal of early stages of liver disease and liver oncogenesis that could result in early diagnosis when curative therapies could be employed. In addition, MRI/MRE-correlated systemic molecular biomarkers indicative of specific disease stages also have potential to serve as important non-invasive diagnostic tools. In this study, we developed a clinically translatable MRI/MRE liver imaging protocol in a clinically relevant large animal platform, the Oncopig Cancer Model (OCM). The OCM is a novel transgenic swine model that recapitulates human HCC through development of site and cell specific tumors after Cre recombinase induced expression of heterozygous KRASG12D and TP53R167H transgenes. A clinical imaging workflow consisting of respiratory gated acquisitions was developed with the Siemens 3 T Prisma MRI scanner to include the following: T1 Flash in-phase and out-of-phase, T2-HASTE/BLADE acquisitions to provide motion-robust imaging, VIBE imaging, diffusion-weighted imaging with IntraVoxel Incoherent Motion (IVIM) for estimating blood flow, and MRE for quantifying liver stiffness. Additional MRI imaging for tumor characterization was also performed, spanning multiphase (arterial, venous, delayed) contrast-enhanced T1-weighted imaging. This novel imaging protocol was successfully applied in normal control Oncopig subjects (n = 2), resulting in high quality, high resolution MRI/MRE images for clinical interpretation. In the future, integration of MRI/MRE, histology, and molecular measures in Oncopigs presenting with liver cirrhosis and HCC may be employed to investigate biomarkers that may define normal versus diseased liver, and may serve as a new approach to identify optimal time points for initiation of disease screening and treatment administration for clinical translation.

#3048

Deep learning to predict survival prognosis for patients with non-small cell lung cancer using images and clinical data.

Edward H. Lee, Mu Zhou, Noah Gamboa, Kevin Brennan, Haruka Itakura, Viswam Nair, Sandy Napel, Simon Wong, Olivier Gevaert. Stanford University, Stanford, CA.

Goal

We present a deep learning framework to predict survival prognosis of patients with non-small cell lung cancer using CT images and clinical data developed and validated in multi-institutional cohorts.

Methods

We developed 3D convolutional neural networks (CNNs) using CT images, tumor nodule segmentation, clinical data (i.e. age, sex, histology, and stage) and survival times as input to predict survival of non-small cell lung cancer. Both the CT images and the segmentations were used as inputs for the CNN, the latter to localize the volumes-of-interests (VOIs) in the CT scans. We used three cohorts to evaluate our framework: Moffitt (n=186), Maastro (n=311), and Stanford (n=130). We investigated pre-training with the LIDC-IDRI dataset (n=1010) and compared deep learning performance with 1674 manually-curated radiomics features and clinical data. We used two ways to evaluate the models. First, we split the three cohorts into training and test sets by selecting two of the cohorts for training and the remaining for testing. Secondly, we combined all cohorts and performed a 10-fold stratified cross-validation. We used the Concordance Index (CI) to estimate model performance. Next, we evaluated priming whereby we use 10% of the validation cohort during training and test on the remaining patients in the test cohort. To ensure that the test set was identical for both priming and non-priming experiments, we discarded these samples for the non-priming tests.

Results

Table 1 shows the results of the two evaluation strategies. We observed consistently high CIs for all three cohorts. LIDC pretraining and priming yielded the highest CIs as compared to radiomic and clinical-only pipelines. For the 10-fold cross-validation experiment, our pipeline achieved an average CI=0.64.

Conclusion

Our study highlights the promising results of our pipeline to predict survival prognosis across three cohorts without the need to define radiomics features.

#3049

Dexamethasone pretreatment impairs the TS inhibition-mediated flare in thymidine salvage pathway activity.

Sharyn I. Katz, Xiao Chen, Yizeng Yang. Univ. of Pennsylvania, Philadelphia, PA.

Introduction: Inhibition of thymidylate synthase (TS) by pemetrexed, an inhibitor of thymidylate synthase, results in an early transient burst or flare in thymidine salvage pathway activity at 2 hrs of therapy, which, we have demonstrated in prior publications, can be measurable with FLT ([¹⁸F]thymidine)-positron emission tomography (PET) in non-small cell lung cancer (NSCLC). Administration of dexamethasone with pemetrexed-based therapy, as typically practiced in the clinic, could potentially confound this imaging approach since dexamethasone is known to inhibit expression of thymidine kinase 1, a key enzyme in the thymidine salvage pathway. Here we examine the potential impact of dexamethasone on the TS inhibition-mediated thymidine salvage pathway flare in NSCLC.

Materials and Methods: In order to determine NSCLC cell line sensitivity to dexamethasone and pemetrexed, IC50 studies were performed on NSCLC cell lines H23, H1975, H460, H1299. TS inhibition-mediated flare in thymidine salvage pathway activity was then measured at 2hrs of exposure to pemetrexed and cisplatin in NSCLC cells lines following using 3H-thymidine incorporation assays