Regulation of Anti-Tumor Immunity and Immunotherapy Response in Colorectal Cancer

Regulation of Anti-tumor Immunity and Immunotherapy Response in Colorectal Cancer

A DISSERTATION SUBMITTED TO THE FACULTY OF THE UNIVERSITY OF MINNESOTA BY

Xianda Zhao

IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

Advisor: Subbaya Subramanian, M.S., Ph.D.

JUNE 2020

Acknowledgment My graduate work and this thesis are the results of a collective effort put forth by my family, friends, mentors, and the MICaB graduate program. If I have to use one word to summarize the past years, I would say gratefulness.

I am often told that good mentors are essential for laying the groundwork for success. The mentors that I gained over the last ten years substantially helped and guided me. I would never have the opportunity to study in the U.S. without the fundamental training that my mentors have provided to me at the medical school of Wuhan University, China. They offered me to participate in their scientific and clinical researches with a flexible schedule. They treated me like a formal member, allowed me to explore my ideas, and, most importantly, encouraged me to strive for the best. I will be forever grateful to those invaluable resources that they brought to me, and those holidays that they voluntarily spend on guiding my research.

My graduate study at the University of Minnesota is the happiest and an unforgettable time in my life. I always firmly believe the excellent training environment in the U.S. is the best I have ever experienced. I am extremely grateful to my mentor Dr. Subbaya Subramanian, who provides me a friendly, highly creative, and encouraging training environment. He is always supportive of testing new concepts, applying for research funds, and performing cutting-edge experiments that are critical for academic development. His thoughtful discussion and unlimited supports substantially accelerated my research progression. I am also so fortunate to have my thesis committee members Dr. Tim Starr, Dr. Subbaya Subramanian, Dr. Emil Lou, and Dr. Amy Skubitz. They ask thoughtful questions about my research and provide insightful feedback. Most impressively, they spend so much time to comment on my manuscript and correct it word by word. Their enthusiasm in science, thoughtfulness in experiment design, and meticulousness in manuscript preparation set up a model for my future career.

I am also thankful to our current and previous lab members, Dr. Lihua Li, Dechen Wangmo, Ce Yuan, Beminet Kassaye, Duha Alshareef, and Nile Liu. They have been working incredibly hard to push my projects forward and finish up our experiments. Of course, my work cannot be done without the immeasurable supports from the MICaB graduate program (DGSs Dr. Stephen Jameson and Dr. Wade Bresnahan, Ms. Megan Ruf, and Ms. Louise Shand), who escorted every step of my study. The generous research grant from the Minnesota Colorectal

Cancer Research Foundation and the Doctoral Dissertation Fellowship from the University of Minnesota also supported my work.

I owe my deepest gratitude to my parents and grandparents. Their unlimited love and unconditional support helped me differentiating from all intensive competitions from elementary school to college and achieving my best. Last but not least, I owe my pure gratefulness to my wife, Yuyu He, who redefined happiness to me and substantially helped my work with her intelligence in data science.

Dedication This thesis is dedicated to everyone who provided me with intellectual and emotional support over the past years.

iii

Abstract Immune checkpoint blockade therapy (ICBT) has revolutionized the treatment and management of numerous cancers, yet a substantial proportion of colorectal cancer (CRC) patients are resistant. Most importantly, the mechanisms that cause ICBT resistance in CRC patients are mostly unclear. Both clinical and laboratory studies implied that both tumor cell- intrinsic factors and traditional cancer therapies (e.g., chemotherapy and oncogenic pathway- targeted therapy) have regulatory effects on anti-tumor immunity and ICBT efficacy. In the present thesis, we first characterized the pathological and immunological features of different pre- clinical CRC models. We reported the feasibility of using small animal endoscopy to establish mouse orthotopic colon tumors. We found that tumors grown orthotopically in the colon microenvironment develop better immune infiltration than tumors with the same genetic background growing in a subcutaneous microenvironment. These data indicated that the tissue microenvironment impacts anti-tumor immunity. Meanwhile, we observed that the endoscopy- guided cancer cell line-originated orthotopic CRC model is much more sensitive to ICBT over the subcutaneous model, making it not suitable for experiments that require ICBT-resistant tumors. This observation made us decide to use the subcutaneous tumor models, which are ICBT- resistant, for understanding cancer immunotherapy resistance.

In the second section, we evaluated the impact of tumor-draining lymph nodes (TdLNs) and chemotherapy on ICBT efficacy. Specifically, we demonstrated TdLNs are critical for tumor antigen-specific T-cell response in early-stage tumors. However, TdLNs shift from an immunoreactive to an immunotolerant environment during tumor development. In mice with advanced primary tumors, TdLNs are not the major reservoir of tumor antigen-specific T cells. To evaluate the impacts of those immunotolerant TdLNs on ICBT response, we established a surgical model to mimic tumor recurrence in situ. We surgically removed the established primary tumors with or without concurrent TdLNs resection. Then, we inoculated secondary tumors, which are in the same lymphatic drainage area as the primary tumors, to mimic tumor recurrency. Notably, removing those immunotolerant TdLNs concurrently with established primary tumors did not affect the ICBT response on localized secondary tumors.

In another set of experiments, we evaluated the impacts of chemotherapy (5-FU) on ICBT efficacy. We revealed that using 5-FU as induction treatment for ICBT increased tumor visibility to immune cells, decreased immunosuppressive cells in the tumor microenvironment, and limited chemotherapy-induced T-cell depletion. However, sustained chemotherapy impaired iv the efficacy of ICBT by suppressing the host immune system and depleting tumor-infiltrating T cells. Therefore, the sequential combination of chemotherapy with ICBT may result in a better response than the sustained chemotherapy and ICBT combination.

Finally, we investigated how tumor cells regulate T-cell activation via intercellular communication based on extracellular vesicles (EVs). Specifically, we revealed that tumor cells- secreted EVs (TEVs) containing microRNA miR-424 suppressed the CD28-CD80/86 costimulatory pathway in tumor-infiltrating T cells and dendritic cells. Modified TEVs with miR- 424 knocked down enhanced T-cell mediated antitumor immune response in CRC tumor models and increased the response to ICBT. Intravenous injection of modified TEVs induced tumor antigen-specific immune responses. Moreover, injections of modified TEVs boosted the ICBT efficacy in CRC models that mimic treatment-refractory late-stage disease.

Collectively, the present study improves our understanding of CRC anti-tumor immune regulation and proposed novel treatment for ICBT resistant human CRC.

Table of Contents Acknowledgment ...... i Dedication ...... iii Abstract ...... iv Table of Contents ...... vi List of Tables ...... ix List of Figures ...... x 1. Chapter 1 Introduction ...... 1 1.1 Mechanisms of immune checkpoint blockades therapy for cancers ...... 2 1.2 Tumor response and intrinsic resistance to immune checkpoint blockades therapy ...... 10 1.2.1 Tumor cells mediated immune checkpoint blockades therapy resistance ...... 10 1.2.2 Host mediated immune checkpoint blockade therapy resistance ...... 14 1.2.3 Overcoming the immune checkpoint blockades therapy resistance by combining with traditional therapies ...... 16 1.3 Tumor cells derived extracellular vesicles in anti-tumor immunity ...... 18 1.4 Challenges of immune checkpoint blockade therapy in colorectal cancer ...... 20 1.5 Conclusions and future directions ...... 21 1.6 Publications ...... 23 2. Chapter 2 Characterizing The Immunological and Pathological Features of Different CRC Mouse Models ...... 24 2.1 Introduction ...... 25 2.2 Results ...... 26 2.2.1 Establishment of orthotopic colorectal tumors in mouse with endoscopy-guide microinjection ...... 26 2.2.2 Fewer cells are required to establish subcutaneous models compared to orthotopic models 28 2.2.3 Orthotopic tumors have more adaptive immune-cell infiltration than subcutaneous tumors 28 2.2.4 NK cells increased, and myeloid-derived suppressive cells are decreased in orthotopic tumors 31 2.2.5 Overall antitumor immune response increased in orthotopic tumors ...... 31 2.2.6 Immune checkpoint expression and efficacy of ICBT differed between orthotopic and subcutaneous models ...... 35 2.2.7 Surgical based mouse cecum tumor model mimics advanced stage human CRC tumors 35 2.3 Discussion ...... 38 2.4 Methods and Material ...... 41 2.4.1 Cell culture ...... 41 2.4.2 Mice and treatment ...... 41 2.4.3 Tumor implantation ...... 41 2.4.4 In vivo imaging ...... 42 2.4.5 Flow cytometry ...... 42 vi

2.4.6 Histology and immunostaining ...... 43 2.4.7 RT-qPCR ...... 43 2.4.8 ELISA ...... 44 2.4.9 Statistical analysis...... 44 2.5 Publication ...... 45 3. Chapter 3 Impact of Tumor-draining Lymph Nodes and Chemotherapy on Cancer Immunotherapy Response ...... 46 3.1 Introduction ...... 47 3.2 Results ...... 48 3.2.1 TdLNs are essential for antitumor immune activation and immunotherapy response in early-stage disease ...... 48 3.2.2 TdLNs are not necessary for immunotherapy response in advanced disease tumor models 50 3.2.3 TdLNs shift from an immunoreactive to an immunotolerant environment and tumor- antigen specific T cells disseminate during tumor development ...... 57 3.2.4 Sequential treatment of 5-FU and anti-4-1BB or anti-PD-1 leads to better responses than concurrent treatment ...... 60 3.2.5 Sequential treatment of 5-FU and anti-4-1BB or anti-PD-1 stimulates a strong antitumor immune response ...... 63 3.3 Discussion ...... 70 3.4 Methods and Material ...... 73 3.4.1 Cell cultures ...... 73 3.4.2 Mice ...... 73 3.4.3 Subcutaneous tumor induction ...... 74 3.4.4 Identification of major tumor-draining lymph nodes ...... 74 3.4.5 Subcutaneous tumor and TdLNs resection ...... 74 3.4.6 RT-qPCR ...... 75 3.4.7 Flow cytometry ...... 75 3.4.8 Mass cytometry...... 76 3.4.9 Mouse IFN-γ enzyme-linked immunosorbent assays ...... 76 3.4.10 Histology ...... 77 3.4.11 Mouse treatments and T-cell depletion ...... 77 3.4.12 5-FU toxicity evaluation ...... 77 3.4.13 Statistical analysis...... 78 3.5 Publication ...... 82 4. Chapter 4 Tumor Secreted Extracellular Vesicles Regulate T-cell Costimulation and Can Be Manipulated to Induce Tumor-Specific T-cell Responses in Colorectal Cancer ...... 83 4.1 Introduction ...... 84 4.2 Results ...... 85 4.2.1 Costimulatory molecules CD28, CD80, and CD86 are downregulated on human CRC infiltrating immune cells ...... 85 4.2.2 CD28-CD80/86 costimulatory pathway is necessary for ICBT response ...... 87 4.2.3 MicroRNA-424 is overexpressed in human CRC and inhibits CD28 and CD80 expression ...... 89

vii

4.2.4 TEVs transfer miR-424 from human CRC cells to tumor-infiltrating T cells and DCs 92 4.2.5 Blocking tumor cell-derived miR-424 inhibits tumor development in an immune- dependent manner ...... 98 4.2.6 Blocking tumor cell-derived miR-424 enhanced adaptive antitumor immunity and sensitizes advanced tumors to ICBT ...... 102 4.2.7 Depleting functional miR-424 induces antitumor immunogenicity of TEVs ...... 105 4.2.8 TEVs with miR-424 knockdown enhance ICBT efficacy in an advanced CRC pre- clinical tumor model ...... 108 4.3 Discussion ...... 110 4.4 Methods and Material ...... 114 4.4.1 Human samples...... 114 4.4.2 Tumor cell lines ...... 115 4.4.3 Mice ...... 115 4.4.4 Human CRC organoids culture ...... 116 4.4.5 Isolation of extracellular vesicles (EVs) ...... 116 4.4.6 Characterization of isolated EVs ...... 117 4.4.7 Immunofluorescence and histology ...... 117 4.4.8 Mouse subcutaneous tumor model ...... 118 4.4.9 Mouse cecum orthotopic tumor model ...... 118 4.4.10 Mouse treatment ...... 119 4.4.11 Mouse in vivo imaging ...... 119 4.4.12 T cells and antigen presentation cells in vitro assays ...... 119 4.4.13 Enzyme-linked immunosorbent assays (ELISAs) for IL-2 and D-dimer ...... 120 4.4.14 Cytokine array ...... 121 4.4.15 Tissue digestion and flow/mass cytometry ...... 121 4.4.16 Locked nucleic acids and in situ hybridization (LNA-ISH) ...... 122 4.4.17 Dual-luciferase assay ...... 122 4.4.18 DNA, mRNA and miRNA analyses ...... 122 4.4.19 Western blotting ...... 124 4.4.20 Statistical analysis...... 124 5. Chapter 5 Conclusions and Future Directions ...... 134 Bibliography ...... 142 Appendix ...... 158

viii

List of Tables Table 1.1 Summary of immune checkpoints ...... 8 Table 3.1. gp70 mRNA expression in different tissues...... 54 Table 3.2. Key material table for chapter 3...... 79 Table 4.1. Key resources table for chapter 4...... 125

List of Figures Figure 1.1. Overview of anti-tumor immune response and cancer immune checkpoint blockades therapy...... 3 Figure 1.2 Stimulatory and inhibitory immune checkpoints in the regulation of tumor-specific cytotoxic T lymphocytes (CTLs)...... 4 Figure 1.3 Interactions between T-cell receptor (TCR), CD28, CTLA-4, and PD-1...... 6 Figure 1.4 MHC-I and IFN-γ dysfunction in tumor cells cause ICBT resistance...... 11 Figure 1.5 Immunoregulation induced by oncogenic pathways in the tumor immune microenvironment...... 12 Figure 1.6 Overview of tumor cells derived EVs on tumor development...... 18 Figure 2.1. Tumorigenesis in orthotopic and subcutaneous models was distinct...... 27 Figure 2.2. Adaptive immune cell infiltration was higher in orthotopic tumors than in subcutaneous tumors...... 30 Figure 2.3. Innate immune cell infiltration in orthotopic and subcutaneous models was distinct. 32 Figure 2.4. Macrophage subtypes in the subcutaneous and orthotopic tumors...... 33 Figure 2.5. Expression of IL2, IL6, granzyme B, and IFNγ varied in orthotopic and subcutaneous tumors...... 34 Figure 2.6. Immune checkpoint profile and response to ICBT differed in orthotopic and subcutaneous tumors...... 36 Figure 2.7. The response of subcutaneous tumors to ICBT...... 37 Figure 2.8. CD31 staining in the subcutaneous and orthotopic tumors...... 37 Figure 2.9. The features of surgically implanted mouse cecum tumors...... 38 Figure 3.1. Identification of tumor-draining lymph nodes (TdLNs) in mice...... 49 Figure 3.2. Physical changes and histology of TdLNs, NdLNs, and spleen of tumor-bearing mice...... 51 Figure 3.3. Impact of TdLNs on tumor initiation and immunotherapy response in early-stage tumor models...... 52 Figure 3.4. Gating of tumor antigen-specific CD8+ T cells...... 53 Figure 3.5. Impact of TdLNs on tumor recurrence and immunotherapy response in advanced stage tumor models...... 55 Figure 3.6. Immune features in secondary tumors with or without TdLNs...... 56 Figure 3.7. Functional status of TdLNs and tumor antigen-specific T-cell distribution in tumor- bearing mice with advanced disease...... 59 Figure 3.8. 5-FU and anti-4-1BB sequential treatment elicits strong antitumor activity...... 61 Figure 3.9. T-cell depleting effects of 5-FU and anti-CD3 treatment...... 62 Figure 3.10. 5-FU and anti-4-1BB sequential treatment on secondary tumors that mimic tumor recurrence...... 63 Figure 3.11. Side effects of different treatments...... 64 Figure 3.12. Gating of the tumor-infiltrating immune cells...... 66 Figure 3.13 Tumor immunological response to 5-FU and anti-4-1BB treatments in CT26 tumors...... 67 Figure 3.14. Tumor immunological response to 5-FU and anti-4-1BB treatments in MC38 tumors...... 68

Figure 3.15. Tumor immunological response to 5-FU and anti-PD-1 treatments in CT26 tumors...... 69 Figure 4.1. Expression of costimulatory molecules CD28, CD80, and CD86 on tumor-infiltrating immune cells in human CRC ...... 86 Figure 4.2 Supplementary data to Figure 4.1...... 88 Figure 4.3 miR-424 is highly expressed in CRC and negatively regulate CD28 and CD80 expression...... 91 Figure 4.4 Supplementary data to Figure 4.3...... 92 Figure 4.5 Tumor cells derived EVs transfer miR-424 to immune cells and downregulate CD28 and CD80...... 95 Figure 4.6 Supplementary data to Figure 4.5...... 97 Figure 4.7 Tumor EVs with functional miR-424 accelerated tumor growth by suppressing antitumor immunity...... 99 Figure 4.8 Supplementary data to Figure 4.7...... 101 Figure 4.9 The depletion of functional miR-424 enhanced the costimulatory molecules CD28 and CD80 expression, antitumor immunity, and response to ICBT...... 103 Figure 4.10 Supplementary data to Figure 4.9...... 104 Figure 4.11 TEVs without functional miR-424 are efficient in stimulating antitumor immune response...... 106 Figure 4.12 Supplementary data to Figure 4.11...... 107 Figure 4.13 TEVs without miR-424 enhanced the efficacy of ICBT in established orthoptic tumors...... 109 Figure 4.14 Supplementary data to Figure 4.13...... 110 Figure 5.1. Graphic summary of Chapter 3...... 137 Figure 5.2. Graphic summary of Chapter 4...... 140

1. Chapter 1

Introduction

1.1 Mechanisms of immune checkpoint blockades therapy for cancers The use of chemotherapy to treat cancer patients began at the start of the 20th century1. Yet radiotherapy and surgery dominated the field until it became clear that combination therapy could significantly improve survival outcomes in patients with various advanced cancers, such as acute childhood leukemia and advanced Hodgkin's disease1. Chemotherapy then became part of standard clinical practice, representing the first generation of treatment strategies in medical oncology. The elucidation and understanding of the molecular mechanisms that regulate cancer growth and the malignant transformation of normal cells ushered in the second generation of treatment strategies: targeted therapy. Using inhibitors that specifically target abnormal cell proliferation and tumor growth pathways, clinicians have seen significant therapeutic effects in patients with certain cancers, such as breast cancer and prostate cancer. Immunotherapy, the third generation of treatment strategies, kills tumor cells by improving the host immune response of cytotoxic T cells. It has shown promise, with an antitumor pattern different from that of both chemotherapy and targeted therapy (Figure 1.1) and has profoundly changed the paradigm of cancer treatment.

Tumors are communities of not only malignant cells but also surrounding stromal cells, including fibroblasts, epithelial cells, and, even more critical, infiltrating immune cells2. Understanding how the immune system affects cancer initiation and progression has been one of the most daunting challenges in oncology. Over the past two decades, studies have made it clear that the communication between cancer cells and the immune system is a dynamic process. In fact, the immune system plays a dual role. It can suppress tumors by destroying cancer cells and inhibiting outgrowth; or, it can promote tumor progression by selecting aggressive immune evasive subclones and modulating the tumor microenvironment to facilitate tumor outgrowth3-6. According to the commonly accepted concept of cancer immunoediting, the tumor immune response has three sequential phases: elimination, equilibrium, and escape3,6,7. In the elimination phase, forces of both innate and adaptive immunity engage cooperatively to destroy developing tumors before they are clinically apparent3,6,7. However, if rare cancer variant cells escape during the elimination phase, they may enter the equilibrium phase, in which their outgrowth is prevented and they are maintained in a dormant state3,6,7. Editing of tumor immunogenicity occurs during the equilibrium phase. The immune selected cancer cells are then no longer recognized by the forces of adaptive immunity, become insensitive to immune effectors, and induce an immunosuppressive state in the tumor microenvironment3,6,7. Next, these immune insensitive

2 tumor cells enter the escape phase to form a fully thriving tumor whose outgrowth cannot be controlled3,6,7. These findings indicated that immunotherapies must be able to re-stimulate the anti-tumor immunity to eliminate the immune escaped tumor cells. The major steps of the antitumor immune response are outlined in Figure 1.1.

Figure 1.1. Overview of anti-tumor immune response and cancer immune checkpoint blockades therapy.

(A) Basis of antitumor immunity. Antitumor immunity can be broadly divided into three major stages. The process involving seven key steps begins with the release of tumor antigens and ends with the elimination of cancer cells. The primary cell types involved in each of these seven steps are illustrated. (B) and (C) Schematic representation of cancer treatment responses observed with various therapeutic strategies. Cancer immunotherapy demonstrates a unique feature of long-term tumor control. It is expected that patients will have a better clinical outcome with personalized immunotherapy that overcomes the intrinsic and acquired resistance, in comparison with

3 traditional therapies. This figure is reprinted with permission from Zhao et al., Cancer Research, 2017. Multiple tumor immunosuppressive mechanisms causing tumor cell immune escape have been demonstrated, but the immune checkpoint mechanism has garnered extensive attention, given the clinical success of targeting inhibitory checkpoint molecules8,9. Immune checkpoints are molecules in the immune system that either turn up or turn down costimulatory signals (Figure 1.2). By inhibiting the immune checkpoints that suppress T-cell activation (such as cytotoxic T-lymphocyte–associated antigen 4 (CTLA-4) and programmed death 1 (PD- 1)/programed death ligand 1 (PD-L1)), immune checkpoint blockade therapy (ICBT) has resulted in a response rate as high as 40% in melanoma patients, with some durable effects10-12. However, in large-scale use in patients with solid tumors, ICBT remains ineffective, for two main reasons: intrinsic resistance (i.e., insensitivity of the tumors to ICBT) and adaptive resistance (which occurs promptly after ICBT)13,14. Recent studies have revealed some of the fundamental mechanisms of CTLA-4 and PD-1/PD-L1 mediated immune regulation15.

Figure 1.2 Stimulatory and inhibitory immune checkpoints in the regulation of tumor- specific cytotoxic T lymphocytes (CTLs). 4

In the tumor-draining lymph node, dendritic cells (DCs) present tumor antigens to naïve T cells, and induce T-cell activation. The major histocompatibility complex (MHC) and T-cell receptor (TCR) signaling pathway provide the primary signal for T-cell activation, while the immune checkpoints expressed on T cells and their ligands on DCs, costimulate or limit T-cell activation. The function of tumor-specific CTLs in the tumor immune microenvironment is under the control of regulatory cells (such as macrophages, immature and suppressive monocytes, regulatory B cells, and regulatory T cells) as well as immune checkpoint molecules. Costimulatory and inhibitory immune checkpoints are presented in tumor-specific CTLs. Ligands are expressed on tumor cells and other stromal cells, such as cancer-associated fibroblasts (CAFs) and myeloid- derived suppressor cells (MDSCs). Costimulatory immune checkpoints are represented by green labels and the inhibitory immune checkpoints by red labels. This figure is reprinted with permission from Zhao et al., Pharmacology & Therapeutics, 2018.

PD-1 is a cell surface receptor that belongs to the immunoglobulin superfamily and is expressed on T cells (Figure 1.3), which are activated by the T-cell receptor (TCR) engagement and pro-B cells16. PD-1 expression requires transcriptional activation and takes about 12 hours. Once stimulated, PD-1 directly inhibits TCR-mediated downstream signaling and increases T-cell migration within tissues. The continually moving T cells have only a limited amount of time to survey the surface of interacting cells for the presence of the cognate peptide-MHC [major histocompatibility complex] complex, so T cells can “pass over” target cells that have a lower peptide-MHC complex level17. Recent studies also illustrated the interactions between PD-1 and CD28, which is the major positive co-stimulatory pathway of T-cell activation18-20. Upon activation, the PD-1 signaling recruits SHP2 phosphatase to preferentially dephosphorylate the CD28 signaling18. Without the presence of CD28 signaling, PD-1 blockade fails to rescue the exhausted T cells in the tumor microenvironment, leading to anti-PD-1 resistance19.

PD-1 binds to 2 ligands: PD-L1 (B7-H1, CD274) and PD-L2 (B7-DC, CD273); those ligands share 37% of their sequence homology and are arose via gene duplication21,22. The major stimulator of PD-L1 expression is inflammatory cytokine interferon-gamma (IFN-γ) produced by activated T cells and by natural killer (NK) cells; PD-L1 is primarily found on activated hematopoietic cells and epithelial cells, including cancer cells23,24. In cancer cells, the oncogenic pathways, such as EGFR, KRAS, and MYC, also controls tumor cell-secreted PD-L1 level in the tumor microenvironment (discussed later in detail)25. In contrast to the wide expression range of PD-L1, PD-L2 expression is more selective. Induced by interleukin-4 (IL-4) rather than by INF-γ, PD-L2 is expressed on activated dendritic cells (DCs) and macrophages26.

Figure 1.3 Interactions between T-cell receptor (TCR), CD28, CTLA-4, and PD-1.

The schematic shows the molecular interactions and downstream events induced by TCR, CD28, CTLA4, and PD-1. TCR and CD28 provide positive signaling for T-cell activation. CTLA-4 inhibits T-cell activation by competing ligands with CD28 and dephosphorylating the CD28’s downstream molecules. PD-1 impedes T-cell activation by dephosphorylating the downstream events of CD28 and TCR. However, more detailed mechanisms of those events are unclear. Unlike PD-1, which dampens T-cell activation, CD28 is expressed in T cells to enhance TCR signaling when TCR is engaged by the peptide–MHC complex (Figure 1.3). CD28 is the receptor of CD80 (B7.1), CD86 (B7.2), and inducible costimulatory (ICOS)-L proteins, all of which are expressed in the antigen-presenting cells (APCs), such as DCs and monocytes27. As a regulatory feedback inhibitory mechanism, CTLA-4 is expressed in activated T cells that already express CD28; it shares both CD80 and CD86 as its ligands, along with CD28, and shows a higher affinity for both ligands than for CD2828. Therefore, CTLA-4 expression in activated T cells dampens CD28 co-stimulation by competing or depleting ligands of the CD28 signaling pathway. In addition, biochemical evidence has suggested that CTLA-4 recruits phosphatases to attenuate the TCR signal29. Activation of phosphatases PP2A and SHP2 is critical in counteracting both tyrosine and serine/threonine kinase signals induced by TCR and CD2830. Moreover, CTLA-4 induces T-cell motility and overrules the TCR-induced stop signal required for stable conjugate formation between T cells and APCs. This mechanism leads to fewer contact periods between T cells and APCs, resulting in decreased cytokine production and in T-cell proliferation30.

The functions of CTLA-4 might be more than the competition with CD80 and CD86 for the CD28 co-stimulation. Regulatory T cells (Treg) are essential for immunosuppression in the tumor microenvironment. Recent studies confirmed that CTLA-4 promotes Foxp3 induction and Treg differentiation31. In mouse tumors, anti-CTLA-4 depleted the CTLA-4 expressing Tregs, thus alleviating the immunosuppression in tumor microenvironment32. CTLA-4 blockade also induced the expansion of ICOS+ Th1-like CD4 T cells in tumor microenvironment33. However, these effects are not validated in human tumors34, indicating the complicated effects of CTLA- 4/anti-CTLA-4 on tumor immune responses.

Beside the PD-1 and CLTA-4 systems, some cancers have a signature of T-cell suppressive mechanisms that include other inhibitory pathways, such as indoleamine 2,3- dioxygenase (IDO), OX40 (CD134), 4-1BB (CD137), LAG3, and P-selectin glycoprotein ligand- 1 (PSGL-1) (summarized in table 1.1). However, current investigations regarding these newly identified immune checkpoints are primarily limited to molecular biology level, pre-clinical stages, or early phase clinical trials.

Here, we outline the current understanding of intrinsic resistant mechanisms of solid tumors to ICBT, the challenges of ICBT in treating human colorectal cancer patients and discuss the potential strategies to overcome those challenges.

Table 1.1 Summary of immune checkpoints

Stimulatory checkpoint Modulation methods Current Status molecules CD27 Anti-CD27 monoclonal antibody Phase II clinical trial CD28 Pre-activation by CD28signaling T cells Phase II clinical or CD28 expressing T cells adoptive trial transfer CD40 Anti-CD40 monoclonal antibody or Phase II clinical sugar-engineered antibody, CD40 ligand trial CD122 (interleukin-2 CD122-biased agonist Phase I clinical receptor beta subunit) trial CD137/41BB Anti-CD137 monoclonal antibody, 41BB Phase II clinical expressing T cells adoptive transfer trial OX40/CD134 Anti-OX40 monoclonal antibody Phase I clinical trial Glucocorticoid-induced Anti-GITR monoclonal antibody Phase I clinical TNFR family-related gene trial (GITR) Inducible T-cell Anti-ICOS monoclonal antibody Phase I clinical costimulatory trial (ICOS/CD278)

Inhibitory checkpoint molecules Adenosine A2A receptor Blockade of the action of adenosine that Phase I clinical (A2AR) is produced by tumors trial B7-H3/CD276 Anti-B7-H3 monoclonal antibody Phase I clinical trial B7-H4/VTCN1 NA Basic study35 B-and T-lymphocyte NA Basic study36 attenuator (BTLA/CD272) Cytotoxic T-lymphocyte- Anti-CTLA-4 monoclonal antibody Approved for associated protein 4(CTLA- clinic 4/CD152) Indoleamine 2,3- IDO peptide vaccination and inhibitor Phase II clinical dioxygenase (IDO) trial

Killer-cell Anti-KIR monoclonal antibody Phase I clinical immunoglobulin-like trial receptor (KIR) Lymphocyte activation Anti-LAG3 monoclonal antibody Phase I clinical gene-3 (LAG3) trial Programmed death 1(PD-1) Anti-PD-1 monoclonal antibody and Approved for anti-PD-L1 monoclonal antibody clinic T-cell immunoglobulin Anti-TIM-3 monoclonal antibody Phase I clinical domain and mucin domain trial 3 (TIM-3) V-domain Ig suppressor of Anti-VISTA monoclonal antibody Phase II clinical T-cell activation (VISTA) trial P-selectin glycoprotein Anti-PSGL-1 antibody Basic study37 ligand-1 (PSGL-1) NA = not applicable This table is reprinted with permission from Zhao et al., Pharmacology & Therapeutics, 2018.

1.2 Tumor response and intrinsic resistance to immune checkpoint blockades therapy ICBT has become a cornerstone of cancer immunotherapy and has elicited durable clinical responses in drug-sensitive patients. But resistance to ICBT is a significant barrier, limiting the number of patients who can benefit. In patients with untreated melanoma, the response rate to ICBT as a monotherapy has been only 15% to 40%10. For human CRC, ICBT is only promising in 30-60% of the microsatellite instable-High (MSI-H) subtype tumors38. Unfortunately, most CRC patients (>85%) have microsatellite stable (MSS) tumors39,40, that do not respond to ICBT. In this section, we will review the currently validated mechanisms causing ICBT resistance in solid tumors.

1.2.1 Tumor cells mediated immune checkpoint blockades therapy resistance By its very nature, cancer involves the accumulation of mutations in the genome of transformed cells, resulting in abnormal phenotypes in those cells. During the process of T-cell maturation, T cells that recognize autoantigen are cleared away; therefore, the tumor antigens encoded by mutated genes (mutations that are not seen in healthy individuals) trigger a robust antitumor immune response. The individual features of cancer cells determine the intensity of the initial step of antitumor immunity: tumor antigen release. The mutation burden varies widely among different types of tumors, ranging from as few as ten mutations to thousands41. Initial studies analyzing ICBT response found that tumors that have a more substantial mutation burden would be more likely to be seen by the immune system and would respond better to ICBT42-44. In animal tumor models, artificially generated MSI phenotype significantly increased tumor neoantigen number and elicited a better ICBT response45. Further clinical investigations revealed that the MSI phenotype induced high mutational load in tumors is correlated with a higher ICBT response rate, leading to the approval of ICBT in multiple MSI tumors9.

Accumulation of neoantigens does not necessarily elicit a potent anti-tumor immune response. Tumor cells' visibility to T cells and sensitivity to T-cell effectors are decisive for the fate of tumor cells. Tumor cells are notorious for the dysfunction of interferon-γ (IFNγ) signaling pathway and loss of major histocompatibility complex (MHC) expression (Figure 1.4). A productive T-cell response against tumor cells relies on the expression of IFNγ receptors in the tumor cells, which activate the Janus kinase (JAK)-signal transducer and activator of transcription (STAT) signaling to stimulate PD-L1 expression. Mutations in the IFNγ signaling proteins are commonly detected in tumors resistant to ICBT46. Dysfunction of the IFNγ signaling pathway

10 impairs adaptive PD-L1 expression, the predominant ligand of PD-1, leading to insensitivity to anti-PD-1/PD-L1 treatment. The IFNγ signaling pathway is also critical for enhancing the expression of the MHC-I molecules on tumor cells. Tumor cells lacking functional interferon signaling may have little or no MHC-I expression, permitting immune escape. MHC-I antigen presentation also requires coordinated expression of several genes, including TAP1, TAP2, B2M, and the immunoproteasome genes. However, mutations of these genes are noticed in tumor cells47,48. Taken together, mutations in the immune pathway-related genes in tumor cells result in decreased tumor cell visibility and sensitivity to T cells, which finally lead to ICBT resistance.

Figure 1.4 MHC-I and IFN-γ dysfunction in tumor cells cause ICBT resistance.

Mutations in the β2M gene lead to loss of functional MHC-I expression on the tumor cell surface. The IFN-γ pathway is deficient in tumor cells due to the JAK1/2 gene mutation. Defects in the MHC-I expression and IFN-γ signaling transduction ultimately cause tumor cells to resist immune cells-mediated tumor death. This figure is reprinted with permission from Zhao et al., Cancers, 2020. Besides the mutations that produce neoantigens and impair immune response pathways in tumor cells, the driver mutations in oncogenes and in tumor suppressor genes (TSG) also influence the tumor cells-mediated immune regulation (Figure 1.5). Among them, the influence of PTEN mutation in anti-PD-1 efficacy is well described. Inactivation of PTEN, a TSG in multiple cancer types, correlated with decreased T-cell infiltration in tumor tissue and with inferior outcomes after anti-PD-1 therapy49-51. Mechanistically, how PTEN mutations cause immunosuppression in the tumor microenvironment is mostly unknown. Loss of PTEN in tumors

11 might increase the expression of immunosuppressive factors, such as PD-L152. However, this mechanism is not widely validated in different cancer types yet49.

Figure 1.5 Immunoregulation induced by oncogenic pathways in the tumor immune microenvironment.

Activation of oncogenic pathways induces immunotolerance mainly through several mechanisms: (1) Secreting inhibitory immune checkpoints, such as PD-L1. On abnormal activation of KRAS, EGFR, BRAF, and MYC, and formation of the fusion gene EML4-ALK, expression of the inhibitory immune checkpoint PD-L1 increases. The key downstream pathways that mediate PD- L1 expression, such as MAPK and STAT3, are shared by different mechanisms. BRAF mutations increase interleukin (IL)-1 secretion, thereby inducing PD-L1 and PD-L2 expression in cancer associate fibroblasts (CAFs). CDK5 and KIT are identified as key mediators of PD-L1 expression stimulated by interferon-gamma (IFN-γ). (2) Recruiting immunosuppressive cells by secreting cytokines and chemokines. BRAF and KRAS mutations upregulate secretion of IL-10, IL-6, TGF- β, VEGF, and CCL2, which are critical for recruiting immunosuppressive cells. (3) Enhancing the function of immunosuppressive cells. Activation of the PI3K pathway supports the myeloid cells with immunosuppressive phenotypes. (4) Inducing resistance to T-cell killing. BRAF mutations downregulate MDA and MHC-I expression, resulting in impaired recognition of tumor cells by T 12 cells. Loss of PTEN function induces resistance to T-cell killing via enhancing autophagy in tumor cells. MYC upregulates CD47, the “do not eat me” signal, which contributes to the resistance of phagocyte-dependent tumor clearance. This figure is reprinted with permission from Zhao et al., Pharmacology & Therapeutics, 2018. RAS/RAF/mitogen-activated protein kinase (MAPK) pathway is another well- characterized oncogenic pathway regulating anti-tumor immunity. Notable effects of KRAS mutation have been recognized as induction and maintenance of potent immunosuppression in tumors. Two primary mechanisms by which KRAS mutants induce immunosuppression have been elucidated. First, tumor cells with KRAS mutation express significantly more PD-L1 via the MAPK and STAT3 dependent mechanisms53,54. Second, tumor cells carrying KRAS mutants aggregate immunosuppressive Tregs in the tumor immune55. KRAS mutants activate the MEK- MAPK-AP1 signaling in tumor cells, thereby stimulating the secretion of interleukin (IL)-10 and transforming growth factor-beta (TGFβ) required for Treg induction55. The silencing of KRAS mutants reversed the immunosuppressive effect produced by Tregs.

Mutations in the BRAF gene also constitutively activate MAPK. BRAFV600 mutations, such as BRAFV600E, have shown immunosuppressive effects with multiple mechanisms, including upregulating immunosuppressive cytokines (IL-10, IL-6, CCL2, and vascular endothelial growth factor (VEGF)), internalizing MHC-I molecules expressed on tumor cells, and stimulating PD-L1/PD-L1 expression on cancer-associated fibroblasts. Other recently identified immunoregulating oncogenic mutations involve phosphoinositide 3-kinase (PI3K), epidermal growth factor receptor (EGFR), echinoderm microtubule-associated protein-like 4 (EML4), cyclin-dependent kinase 5 (CDK5), and MYC25. Current evidence supports that those oncogenic mutations primarily induce PD-L1 production from tumor cells25, thus facilitating PD-1/PD-L1 mediated T-cell exhaustion in the tumor microenvironment.

Copy number variations also affect the immunotherapy response. Focal copy number loss of β2-microglobulin, a part of the MHC-I, was observed in patients resistant to CTLA-4 blockade56. Whole-exome sequencing analysis indicated that arm and whole-chromosome levels of copy number alterations also influence ICBT response by changing the expression of multiple genes in immune-related pathways56,57. The effect of mutational load and burden of copy number loss on ICBT response might be non-redundant, suggesting the potential utility of a combinatorial biomarker to optimize ICBT response prediction56.

To summarize, recent studies demonstrated that neoantigens in tumor cells are the driving force of anti-tumor immune response. Deficiency in the immune-related pathways in tumor cells nullifies T-cell mediated tumor killing. Activation of certain oncogenes and inactivation of tumor suppressor genes induced immunotolerance in the tumor immune microenvironment. Theses data established our initial understanding of how tumor-intrinsic mechanisms affect anti-tumor immunity in tumor progression and ICBT response. However, most studies are descriptive and are predominantly focused on PD-L1 expression regulation, Treg induction, IFNγ signaling, and MHC-I expression. Whether or not the tumor-derived factors affect other T-cell activation mechanisms, such as CD28-CD80/86 co-stimulation, remain entirely unknown.

1.2.2 Host mediated immune checkpoint blockade therapy resistance The host factors, such as tumor-draining lymph nodes (TdLNs) and tumor microenvironment components, control necessary steps in anti-tumor immunity (Figure 1.1). It is now well known that ligands of inhibitory immune checkpoints are highly expressed in the tumor microenvironment and that they can suppress the activity of tumor-specific T cells. Extensive analyses of clinical tumor tissues that are responsive or resistant to ICBT have revealed several critical mechanisms by which the host factors affect T cells priming, tumor microenvironment infiltration, and their killing functions.

One of the incipient steps of anti-tumor immune response is priming tumor-specific T cells in the TdLNs. The neoantigens in tumor cells provide the immunogenic peptide for T-cell recognition. However, the host factors, especially the APCs in the tumor microenvironment, determine the intensity of antigen presentation and T cells stimulation (Figure 1). Current evidence support that a few specific DCs subsets, including the CD103+ (in mouse, CD141+ in human) conventional type-1 DCs, are decisive for CD8+ T-cell priming in TdLNs58,59. These CD103+ DCs are very efficient in processing tumor-specific antigens and in transporting them to TdLNs through the CCR7/CCL19/21 chemotaxis. This process results in both direct CD8+ T cells stimulation and antigen hand-off to resident APCs58. Expanding the CD103+ DCs increased tumor response to anti-PD-L1 in melanoma mouse model59. Targeting the immune checkpoints expressed on CD103+ DCs, such as T-cell immunoglobulin and mucin-domain containing-3 (TIM-3), also enhanced anti-tumor immunity60. The conventional type-2 DCs are another subset of DCs involved in the T-cell priming mechanisms. These DCs traffic from tumor to TdLNs, to present tumor antigens to conventional CD4+ T cells61. They serve as one of the major effectors of anti-CTLA-4 treatment and are responsible for anti-CTLA-4 efficacy61. Even though the

14 critical subsets of DCs that carry out the T-cell priming process are identified, the roles of those DCs in modulating intrinsic ICBT resistance remain unclear. Few studies have noted whether or not those DCs were dysfunctional in cancer patients who did not respond to ICBT. Therefore, understanding whether or not the dysfunction of intratumoral DCs is a mechanism causing intrinsic ICBT resistance is urgent.

After they are primed in the TdLNs, those cancer-specific T cells move through the vascular system to the tumor site. Successful T-cell infiltration not only relies on appropriate chemokine attraction but also depends on the tumor vasculature function. By comparing T-cell- rich and T-cell-poor tumors, researchers have found that the apoptosis-inducing Fas ligand (FasL), which is expressed in tumor vasculature, negatively controls tumor-infiltrating CD8+ T cells62. Tumors with high levels of endothelial FasL tend to have few CD8+ T cells but an abundant number of Tregs, which are protected from FasL-mediated cell death62. Multiple factors other than FasL, such as endothelin B receptor (ETBR), are also involved in vasculature- controlled T-cell infiltration63. In pre-clinical pharmacologic models, inhibition of ETBR substantially increased T-cell adhesion to endothelial cells, resulting in enhanced T-cell function63. Recent findings also identified that the production of CXCL9/10 in the tumor microenvironment is necessary for sufficient T-cell infiltration64. Lack of CXCL9/10-producing CD103+ DCs in the tumor microenvironment restricts activated tumor-specific T-cell recruitment from the peripheral blood64.

Increased numbers of infiltrated tumor-specific T cells through the tumor capillary bed will be of little importance if T cells never have a chance to reach the vicinity of cancer cells. The impact on immune regulation of cancer-associated fibroblasts (CAF) has become more evident, especially the role of fibroblast activation protein-a (FAP)-positive CAFs in the T-cell distribution in tumor tissues65,66. CAFs restrict T cells from tumor cells in two ways. First, the extracellular matrix that they produce forms a physical barrier to prevent T-cell migration to tumor regions67. Second, experimental studies have shown that FAP+ CAFs block T cells from reaching tumor cells by secreting CXCL12, which coats cancer cells with a biochemical barrier and by recruiting MDSCs into tumor microenvironment65,68. The detailed mechanisms are not yet clear, but, at least in pre-clinical cancer models, targeting the FAP+ CAFs shows promise in reversing resistance to ICBT65,69.

The tumor-infiltrating T cells’ cytotoxic activity is also regulated by the tumor microenvironment components, such as the immune checkpoint molecules expressed on stromal cells and tumor-infiltrating myeloid-derived cells70. Mechanistically, Kaneda et al.71 reported that PI3Kγ regulates the pro-/anti-inflammatory behavior of tumor-infiltrating myeloid-derived cells. In tumor-bearing mice, PI3Kγ inhibition enhanced expression of MHC-II and pro-inflammatory cytokines but inhibited expression of immunosuppressive factors in tumor-associated macrophages. More tumor-infiltrating CD8+ T cells were found in tumors with Pik3cg−/− but not in wild-type macrophages. Strikingly, in in vitro experiments, PI3Kγ inhibition did not activate T cells directly. Thus, PI3Kγ regulates the anti-tumor immune response by controlling the switch between immune-stimulatory and suppressive macrophages, but not by directly inactivating T- cell function. At the molecular level, PI3Kγ/AKT signaling inhibited NFκB while stimulating C/EBPβ activation. Inactivated NFκB and activated C/EBPβ induce transcriptional programs that promote the immunosuppressive phenotype of macrophages. In contrast, selective inactivation of macrophages by PI3Kγ stimulated NFκB activation and inhibited C/EBPβ, thereby switching the immunosuppressive phenotype of tumor-associated macrophages. De Henau et al.72 extended the immunoregulatory effects of PI3Kγ to suppressive myeloid cells. Intrinsic resistance to ICBT was associated with the infiltration of suppressive myeloid cells in the tumor immune microenvironment. In ICBT-resistant tumors with extensive infiltration of myeloid cells, selective targeting of PI3Kγ reduced tumor growth and metastasis and induced a higher level of antitumor T-cell activity.

To summarize, current data suggest that the host factors induce immunotolerance to tumors. Dysfunction of the host elements impairs T-cell priming, infiltration, and cytotoxic function, resulting in intrinsic ICBT resistance. However, how tumor cells orchestrate with host factors to induce intrinsic ICBT resistance remain mostly unknown.

1.2.3 Overcoming the immune checkpoint blockades therapy resistance by combining with traditional therapies With more robust evidence becoming available on mechanisms of resistance to ICBT, efforts are being made to derive actionable combination strategies to combat such resistance. For reversing intrinsic resistance, the fundamental focus is on transforming immunologically “cold” tumors to “hot” tumors, i.e., to increase immune infiltration and function in the tumor microenvironment. The traditional cancer treatments, including chemotherapy and radiotherapy, have a complementary tumor-eliminating pattern with the ICBT. Conventional cancer therapies

16 offer a fast and robust response; however, the duration of response is short, because of the rapid development of adaptive drug resistance. Although immunotherapies like ICBT have a slower onset of action, the survival benefit is durable. Moreover, recent studies have pointed out that chemotherapeutic drugs and radiation affect the immune compartment in the tumor microenvironment73,74. Therefore, combining traditional cancer treatments to overcome ICBT resistance is being extensively studied.

Chemotherapy is primarily designed to target cancer cells, which proliferate at a higher rate than most normal cells. However, chemotherapy also has significant effects on immune regulation74. Traditionally, conventional chemotherapy has been considered immunosuppressive due to its lymphogenic toxicity. However, an increasing number of experimental studies have suggested that chemotherapy with DNA-damaging agents can promote antitumor immunity. In a mouse lung cancer model, oxaliplatin combined with cyclophosphamide significantly induced immunogenicity of tumor cells75. Mechanistically, tumor expression and secretion of DAMPs— such as cytosolic DNA, high-mobility group box 1 (HMGB1), calreticulin, hyaluronan, and heat shock protein—are enhanced by DNA-damaging agents75-77. The result is that cancer cells and stromal cells increase secretion of type I IFN and other chemokines, to facilitate the function of APCs. DNA-damaging agents can cause mutations in tumor cells. It was expected that these drug-induced mutations could elicit anti-tumor immune responses. However, clinical studies revealed that chemotherapy only weakly increases neoantigen number, which is not enough to induce an immune response in ovarian cancers78. Other potential mechanisms of chemotherapy- induced anti-tumor immunity involve depleting immunosuppressive cells, such as Tregs and myeloid-derived suppressive cells in the tumor microenvironment79.

Radiotherapy, another routinely performed traditional anti-cancer treatment, also enhances MHC class I surface expression and secretion of DAMPs. It activates DCs and increases the cross-presentation of tumor antigens to naïve T cells for T-cell priming80-82. Fas surface expression on tumor cells is also upregulated by radiotherapy, making cells sensitive to programmed cell death induced by the engagement of Fas with FasL on T cells. Other immunologic targets of radiation include Treg populations and immune checkpoints73. Taken together, recent data supported the potential of combining traditional cancer treatments that have a favorable immune effect in treating tumors that are resistant to ICBT monotherapy. However, we are still unclear about the detailed mechanisms of theses combinatory therapies and their optimal schedules. Both pre-clinical and clinical studies are therefore required to investigate the 17 impact of traditional cancer treatments on the anti-tumor immune response for an optimal outcome.

1.3 Tumor cells derived extracellular vesicles in anti-tumor immunity The discovery of extracellular vesicles (EVs) opens a new angle to understand the intracellular interactions in multicellular organisms, including the tumor tissues (Figure 1.6). Biologically, EVs are groups of highly heterogeneous vesicles secreted from their parental cells with different mechanisms. Although the classification of EVs is continuously evolving, they generally fall into two categories: ectosomes and exosomes83. Ectosomes are generated directly from the budding of the cell plasma membrane, whereas, exosomes are assembled in the cytoplasm and released from the cell plasma membrane. The ectosomes are in the size range of ~50nm to 1μm in diameter. By contrast, the exosomes are smaller, with the size range of ~40nm to 160nm. Because of the overlapping of the size range, EVs samples isolated by the current methods, such as ultracentrifugation, contain both exosomes and a small proportion of ectosomes. Here, we will use the term EVs rather than exosomes throughout the dissertation for an accurate definition.

Figure 1.6 Overview of tumor cells derived EVs on tumor development. 18

The EVs secreted by wild-type tumor cells (wild-type EVs) accelerate tumor progression through multiple mechanisms. However, the modified EVs secreted from irradiated tumor cells are immunogenic, therefore suppressing tumor development. The biological functions of EVs are determined by their receptors anchored on the bilayer lipid membrane and the cargos inside the vesicles. EVs characterization has found both the cytoplasmic and nuclear components of the parental cells in the EVs. The membrane components consist of lipids and proteins, like a typical cell membrane. Because of their different origins, the exosomal and ectosomal membranes are not precisely the same. The exosomal membrane contains large amounts of cholesterol, sphingolipids, and phosphatidylserines, which are usually found on the inner side of cell membranes. The tetraspanins, such as CD9, CD63, CD81, etc., are also enriched in the exosomal membrane. MHC molecules that are proteins related to antigen presentation and the cell adhesion molecule integrin were found in the EVs membranes84,85. The encapsulated molecules in EVs include proteins and nucleic acids. The protein composition contains cytoskeletal, heat-shock, and nuclear proteins and is also known to change depending on the type of cell producing it and the surrounding microenvironment84. Almost all kinds of nucleic acids encapsulated in the cells, including RNA and DNA, have been confirmed in the EVs84. Once captured by recipient cells, the EVs encapsulated nucleic acids can be translated to proteins or exert their regulatory effects on recipient cells. However, similar to the protein cargos, the nucleic acid cargos are consistently changing with the parental cell types and statuses.

Tumor cells-derived EVs may induce or promote neoplasia. EVs from breast and prostate cancer cells induce neoplasia through sharing their mRNA and miRNA cargos to surrounding cells86,87. miR-125b, miR-130, miR-155, miR-200, as well as Hras and Kras mRNAs in aggressive tumor cells secreted EVs participated in the neoplastic reprogramming and tumor formation of surrounding stem cells and precancerous cells87,88. Besides, recent studies reported the functions of tumor cells secreted EVs on preparing metastatic niches. EVs derived from the metastatic tumor tissues target on the surrounding blood vessels, changing the structure of the vascular endothelial cells, and finally facilitating breast tumor cells to penetrate the capillary walls89,90. In the metastatic ovarian cancer niche, mesothelial cells of the peritoneum can take up tumor cells derived EVs containing MMP-1 mRNA. The MMP-1 mRNA was translated in the mesothelial cells and induced cell apoptosis. Dysfunction in the mesothelial cells led to the formation of a hole in the peritoneum, thus favoring the peritoneal dissemination of the cancer91.

Tumor cells-derived EVs also regulate anti-tumor immunity and ICBT response. PD-L1, the ligand of PD-1, was identified in tumor cells derived EVs92-96. The EVs incorporated PD-L1 showed similar immunosuppressive effects on the CD8+ T cells, like the PD-L1 expressed on the cell membrane. Notably, the EVs derived PD-L1 could induce immunosuppression at distant sites, such as the TdLNs, when EVs are transported to those organs. More importantly, PD-L1 in EVs appears to be resistant to anti-PD-L1 antibody blockade95, suggesting the correlation between tumor cells-derived EVs and ICBT resistance. The role of EVs in anti-tumor immunity was also shown to be relevant to their miRNA cargos that have regulatory effects on the immune pathways. An analysis of the human melanoma EVs demonstrated that the melanoma-derived EVs were taken up by CD8+ T cells to downregulate T-cell responses through diminishing T-cell receptor (TCR) signaling and cytokine secretions. Further experiments validated that miRNAs, such as miR-3187-3p, miR-498, miR-122, miR149, and miR-181a/b, are responsible for the decreased TCR signaling97.

Despite such tumor-promoting mechanisms, the tumor cells derived EVs have shown the potential to fight cancers. Analyzing the tumor cells derived EVs provide information on the primary/metastatic tumors. Novel cancer detection methods are being developed by examining the tumor cells-derived EVs98. The tumor cells-derived EVs may also suppress tumor development by delivering tumor specific-antigens to initiate DCs antigen presentation and T-cell priming. DCs pulsed by tumor cells derived EVs in vitro showed anti-tumor immunity by eliciting T-cell response99. Moreover, recent studies demonstrated the potential of modifying tumor cells derived EVs to stimulate anti-tumor immunity100,101. Irradiated tumor cells secreted EVs have shown immunostimulatory effects rather than immunosuppressive effects. Treating the tumor-bearing mice with irradiated tumor cells secreted EVs reduced tumor burden and prolonged survival time100,101. Taken together, these data supported that wild type tumor cells secreted EVs are tumor-promoting, and modifying the cargos in EVs will enhance their immunogenicity to stimulate anti-tumor immunity.

1.4 Challenges of immune checkpoint blockade therapy in colorectal cancer In both men and women, CRC is the third most common type of cancer, with more than 53,000 estimated deaths102. Although the clinical activity of ICBT is consistent in some forms of cancer, there is an apparent dichotomy of clinical benefit in patients with metastatic CRC103. The sentinel trial using PD-1 blockade was a phase 1 trial that enrolled 296 patients with heavily pretreated cancer; 19 of them had chemorefractory CRC. Objective responses were seen in 20 patients with melanoma, non-small cell lung cancer, and renal carcinoma, but no objective responses were observed in any of these 19 patients with CRC104. A separate phase 1 study reported an objective response in 1 of 14 patients with heavily pretreated CRC105. Based on this sentinel response (1 of 33 total CRC patients), investigators hypothesized that tumors in the exceptional responder harbored MSI. A phase 2 trial was planned to assess the efficacy of the PD-1 inhibitor pembrolizumab in 41 patients with heavily pretreated carcinomas, 32 of whom had cancers of the colon or rectum. Of these 32 patients, 11 harbored MSI, and 21 were MSS. The DNA mismatch repair (MMR)-deficient cohorts included patients with inherited germline MMR deficiency (Lynch syndrome) as well as patients with sporadic MMR-deficient tumors. When the response was assessed, 40% of the MSI patients showed a partial response, and none of the 21 MSS tumors showed objective response8. Based primarily on these findings, in May 2017, the US Food and Drug Administration (FDA) approved the use of pembrolizumab in patients with heavily pretreated forms of MSI-metastatic carcinoma. In the case of CRC explicitly, the approval specifies that patients must have been treated and become refractory to standard-of-care chemotherapeutic drugs 5-fluorouracil, oxaliplatin, and irinotecan, or the equivalent.

Although the MSI/MSS phenotype is considered as a biomarker for ICBT response9,45, this stratification alone cannot adequately explain the observed difference in treatment response, since immune cell infiltration is also observed in a large proportion of MSS tumors40,106. Meanwhile, even for the MSI-CRC tumors with sufficient immune infiltration, not all of them respond to ICBT38. The current clinical evidence suggested that the ICBT responsive rate for MSI-CRC tumors ranged from 30%-60% in different cohorts38. These fundamental observations lead us to hypothesize that the MSI/MSS phenotype, which is correlated with neoantigen loads, only partially covered the ICBT intrinsic resistance mechanisms. Notably, why ICBT cannot functionally rescue the existing tumor-infiltrating immune cells in the majority of CRC tumors is mostly unknown.

1.5 Conclusions and future directions The past decades have seen substantial advances in our understanding of immune checkpoint activity and in its clinical success. Still, because of intrinsic resistance, only a small proportion of patients have benefited so far from ICBT. Recent studies have revealed multiple mechanisms of the intrinsic ICBT resistance and potential ways to overcome it. However, key challenges remain before we can transform the ICBT resistant tumors, such as the MSS-CRC

21 tumors and part of the MSI-CRC tumors, to sensitive ones. To further elucidate the mechanisms by which CRC tumors resist to ICBT, we primarily focus on the following three aspects:

1. Characterizing the immunological and pathological features of different CRC mouse models to determine the suitable models for specific study aims.

2. Measuring the effects of traditional cancer therapies (primary tumor and TdLNs resection and chemotherapy) on immunotherapy response in recurrent diseases.

3. Understanding how the tumor cells affect anti-tumor immunity and ICBT response in CRC tumors.

1.6 Publications This chapter has been modified (with permission) from the published articles:

Zhao X, Subramanian S. Intrinsic Resistance of Solid Tumors to Immune Checkpoint Blockade Therapy. Cancer Res. 2017 Feb 15;77(4):817-822. (https://www.ncbi.nlm.nih.gov/pubmed/28159861)

Zhao X, Subramanian S. Oncogenic pathways that affect antitumor immune response and immune checkpoint blockade therapy. Pharmacol Ther. 2018 Jan;181:76-84.

(https://www.ncbi.nlm.nih.gov/pubmed/28720430)

Zhao X, May A, Lou E, Subramanian S. Genotypic and phenotypic signatures to predict immune checkpoint blockade therapy response in patients with colorectal cancer. Transl Res. 2018 Jun;196:62-70.

(https://www.ncbi.nlm.nih.gov/pubmed/29518351)

Zhao X, Yuan C, Rieth JM, Wangmo D, Subramanian S. Novel Methods to Overcome Acquired Resistance to Immunotherapy. Current Applications for Overcoming Resistance to Targeted Therapies, 97-129.

(https://link.springer.com/chapter/10.1007/978-3-030-21477-7_4)

Zhao X, Wangmo D, Subramanian S. Acquired resistance to immune checkpoint blockade therapies. Cancers. Cancers, 2020 May 5;12(5):E1161.

(https://pubmed.ncbi.nlm.nih.gov/32380703/?from_single_result=32380703&expanded_search_q uery=32380703)

2. Chapter 2

Characterizing The Immunological and Pathological Features of Different CRC Mouse Models

2.1 Introduction Tumor development in humans is regulated by tumor-specific adaptive immune responses107. Recently developed therapies that enhance the immune response, such as chimeric antigen receptor (CAR) T cells and immune checkpoint blockade therapies (ICBTs), have resulted in remarkable outcomes in certain cancers108,109. For example, checkpoint blockade antibodies targeting PD-1, PD-L1, or CTLA-4 have resulted in significantly improved survival in patients with advanced drug-resistant melanoma, lung cancer, and renal cancer12,110,111. However, in patients with colorectal cancer (CRC), which ranks among the most common malignancies in the United States, the efficacy of these therapies was much less remarkable44,112.

Preclinical CRC immunotherapy studies have primarily depended on syngeneic subcutaneous tumor models113,114. Recent research has shown that tumor location determines tissue-specific recruitment of tumor-associated macrophages in melanoma model115. Therefore, the classic subcutaneous models may not mimic the immune tumor microenvironment of actual human CRC and may be a major barrier in efforts to translate findings on immunotherapy from the laboratory to the clinic116. Moreover, two immune-infiltrated subtypes of CRC have been seen in human patients: the well-infiltrated subtype and the poorly infiltrated subtype117,118. However, the current subcutaneous CRC models are poorly infiltrated119. Meanwhile, the subcutaneous models lack metastatic tumor formation, limiting their values in studying the late-stage disease. Therefore, it is critical to establish preclinical models that have the same location with and better mimics the immune well-infiltrated human CRC features, especially their baseline immune response. Such models would be complementary to the classic subcutaneous models and would facilitate investigation of the mechanisms of ICBT resistance in different clinical settings.

Orthotopic tumor models have several advantages over chemically- or genetically induced tumor models. In general, orthotopic transplants are easier and faster to establish, and they are located within a microenvironment comparable to that of human diseases120. In CRC, orthotopic tumor models have usually required a laparotomy, which can cause a strong inflammation response, which is a potential confounding factor of experimental outcomes, making it unsuitable for immunotherapy studies, which started on the early-stage of tumors121. Given advances in small-animal endoscopic technologies, we established the novel orthotopic CRC model that uses endoscopy-guided microinjection to establish orthotopic tumors in the colon wall in mice122. Our minimally invasive model does not provoke an inflammatory response and is particularly suitable for immunotherapy studies. In this study, we investigated the key 25 characteristics of our model, including immune infiltration and responses to ICBT. We compare endoscopy-based orthotopic tumors with subcutaneous tumors established from the same cell line and demonstrate a significant difference in the immune response. Meanwhile, we will validate the metastatic potential of the previous reported surgical based mouse cecum models, which could be used to mimic advanced stage human CRC tumors. These findings highlight that tumor location influences immune responses in CRC animal models and the importance of model selection in preclinical immunotherapy studies.

2.2 Results 2.2.1 Establishment of orthotopic colorectal tumors in mouse with endoscopy- guide microinjection To establish a standard procedure of tumor cell implantation in the mouse colon wall, we tested different anesthesia options. We found that colon spasms and colonic secretion, in response to endoscopic examination or needle puncture, were the most common issues leading to implantation failure. The frequency and degree of colon spasms and colonic secretion were higher in mice anesthetized with Avertin; those issues were controlled when we administered a combination of ketamine, xylazine, and atropine. During the process of tumor cell implantation, a positive lifting of the colonic mucosa at the implantation sites indicated successful cell inoculation. We have achieved an 80% success rate in implanting orthotopic mouse CRC tumors with continuous practicing.

When we used endoscopy to monitor tumor growth, we saw abnormal protrusions in the colon lumen around 1 week after injection in some mice (Figure 2.1A, i-ii); subsequently, rapid tumor growth induced colonic obstruction (Figure 2.1A, iii). In other mice, as tumors invaded the submucosal layer and expanded toward the serosa, we could not detect tumor growth by bright light endoscopy; however, we did see stiffness, brittleness, and heavy bleeding in the mucosa (Figure 2.1A, iv-vi). Histologic analysis indicated that orthotopic tumors were growing in the submucosa layer and invading the muscularis layer (Figure 2.1A, vii). We detected orthotopic tumors around 3 weeks after implantation of 105 CT26-Luc cells (Fig. 1B) by IVIS. A direct correlation between tumor volume and the number of injected tumor cells can be found in our orthotopic model (Figure 2.1C).

Figure 2.1. Tumorigenesis in orthotopic and subcutaneous models was distinct.

(A, i-iii) Orthotopic tumors were established by endoscopy-guide microinjection in the colon wall of BABL/c mice. In some mice, tumors and tumor-caused colon obstruction can be seen directly in the colon lumen (arrows indicate tumors). (A, iv-vi) In other mice, abnormal movement, stiffness, and bleeding can be seen in the colon. (A, vii) H&E staining showed tumor tissue in the submucosa layer. The invasion margin can be seen in the muscularis externa. (B) An in vivo imaging system can be used for monitoring in our orthotopic model. (C) Injection of different 27 numbers of CT26 cells (105 and 106) led to a significant difference in tumor volume at autopsy (arrows indicate tumors). (D) At 4 weeks after injection of 103 HT29 tumor cells in either the colon wall or the subcutaneous connective tissue of athymic nude mice (n = 4 in each model, iii, the arrows indicate tumors), tumorigenesis occurred only in the subcutaneous model. Arrows indicate the representative injection sites in the colon wall, but no tumors formed (i-ii). (E) At 4 weeks after injection of 104 CT26 cells in either the colon wall (i; n = 3) or the subcutaneous connective tissue (E, ii; n = 4) of BALB/c mice, subcutaneous tumors formed (i); however, no orthotopic tumors formed (E, ii). An autopsy confirmed the results of imaging (data not shown). (F) Injection of 105 MC38 did not induce tumor formation in the colon wall (n = 8) of C57/B6 mice, whereas it did cause subcutaneous tumor formation (n = 8). H&E, hematoxylin and eosin. For all panels, data are plotted as the mean ± SD. t-test was performed between indicated two groups, **P < 0.01. This figure is reprinted with permission from Zhao et al., Oncotarget, 2017. 2.2.2 Fewer cells are required to establish subcutaneous models compared to orthotopic models In athymic nude mice, orthotopic injection of 103 HT29 cells could induce tumor formation at 4 weeks in the subcutaneous model, but not in the orthotopic model (Figure 2.1D). However, orthotopic injection of 105 HT29 cells was sufficient to induce tumor formation and invasion at 4 weeks in the orthotopic model (Data not shown). Similar results were seen in our two syngeneic orthotopic models. For example, injection of 104 CT26 cells induced tumor formation at 4 weeks in subcutaneous tissue, but not in the colon wall in BALB/C mice (Figure 2.1E; results confirmed by autopsy, data not shown) and injection of 105 MC38 cells induced tumor formation at 4 weeks in subcutaneous tissue, but not in the colon wall in C57BL/6 mice (Figure 2.1F). Taken together, our data indicate that more cells are required to initiate tumorigenesis in the orthotopic model compared to the subcutaneous model.

2.2.3 Orthotopic tumors have more adaptive immune-cell infiltration than subcutaneous tumors Immune cell infiltration in CRC is highly variable between patients, and increased infiltration of immune cells was associated with better outcomes40,106. Immunostaining with T-cell markers (CD3 and CD8) in human CRC samples, we also observed two CRC subtypes, based on T-cell infiltration: well-infiltrated tumors (Figure 2.2A, i-iii) and poorly-infiltrated tumors (Figure 2.2A, iv). In samples of the normal colon from BALB/c mice, the microenvironment that orthotopic tumors grew in, we found very few T cells except in the Peyer patches (Figure 2.2B). This observation may be reflective of the lack of antigen exposure in these mice and their controlled housing environment123. In the orthotopic tumor tissue, we found T cells in both the margins (Figure 2.2C, i) and the central parts (Figure 2.2C, ii) of tumors; we saw both CD4+ T cells (Figure 2.2C, iii) and CD8+ T cells (Figure 2.2C, iv). In contrast, in subcutaneous tumors established by the same cell line, we found a very small number of T cells (Figure 2.2D). Flow 28 cytometry analysis showed the same trends of T-cell infiltration in both models. We then determined the proportion of CD8+ T cells in these two models and found no differences. Since B-cells are important in antigen presentation and adaptive immunity. We checked B-cell numbers in both these models and found that the number of tumor-infiltrating B cells was higher in orthotopic tumors than in subcutaneous tumors (Figure 2.2E). The RT-qPCR analysis showed expression of chemokines related to T-cell migration was higher in orthotopic tumors than in subcutaneous tumors (Figure 2.2F).

Figure 2.2. Adaptive immune cell infiltration was higher in orthotopic tumors than in subcutaneous tumors.

(A) In human CRC samples, both well-infiltrated i-iii) and poorly-infiltrated (iv) tumors were found. (B) In normal BALB/c mouse colon, CD3+ cells were usually seen in the mucosa lymph tissues. (C) In orthotopic tumors, CD3+ cells were seen in the tumor margins (i) and in the central 30 parts of the tumor (ii). Both CD4+ and CD8+ T cells were seen in orthotopic tumors (iii-iv). (D) In subcutaneous tumors, CD3+ cells were also detected, but not as many as in orthotopic tumors. (E) Flow cytometry analysis indicated more B cells and more T cells in orthotopic tumors than in subcutaneous tumors. The proportion of CD8+ T cells was the same in the two models. (F) In orthotopic tumors, mRNA expression of CXCR3, CXCL9, CXCL10, and CXCL11 was higher. The expression level was presented as fold change, refers to the lowest expression. N, normal; T, tumor. % total refers to the total number of cells in tumor tissue. For all panels, data are plotted as the mean ± SD. t-test was performed between indicated two groups, ***P < 0.001; ****P < 0.0001. This figure is reprinted with permission from Zhao et al., Oncotarget, 2017. 2.2.4 NK cells increased, and myeloid-derived suppressive cells are decreased in orthotopic tumors Innate immune cells are also critical in regulating anti-tumor immune responses. Therefore, we compared the innate immune profiles between these two models. We observed high levels of NKp46+ cells in human normal colon tissue (Figure 2.3A, i). However, in CRC patient samples, we observed differential levels of NKp46+ cells. (Figure 2.3A, ii, iii). In mice tissue samples, we found more NKp46+ cells in orthotopic tumors than in subcutaneous tumors (Figure 2.3B, C). RT-qPCR analysis of transcript levels of genes that encode cytokines or chemokines related to natural killer (NK) cell functions, we found higher expression of IL12, IL15, and IL18 in orthotopic tumors than subcutaneous tumors (Figure 2.3D). Flow cytometry analyses showed no difference in the number of dendritic cells (DCs) and macrophages between the two models (Figure 2.3E). We have carried out staining for CD80+ (M1 subtype) and CD206+ (M2 subtype) in both of our CRC models. Our data shows there was no significant difference in the proportion of the M1 and M2 subtypes (Figure 2.4). However, we found more CD11b+, CD11c-, and Ly6C+ or Ly6G+ myeloid-derived suppressive cells (MDSCs in subcutaneous tumors than orthotopic tumors (Figure 2.3F).

2.2.5 Overall antitumor immune response increased in orthotopic tumors Our analysis indicates a more robust immune cell presence in the orthotopic model. Further, we determined the overall inflammatory and antitumor immune response intensity by measuring the concentration of IL6, IL2, IFNγ and granzyme B. In orthotopic tumors (compared with subcutaneous tumors, normal colon tissue, and cell line), we found that CT26 cells and normal colon tissues did not express IL6, IL2, IFNγ, and granzyme B. However, compared with subcutaneous tumors, orthotopic tumors that grew in the colon microenvironment showed higher expression of all the cytokines tested (Figure 2.5). These findings support our observation that orthotopic tumors tend to have more antitumor immune cells and fewer immunosuppressive cells, suggesting that orthotropic tumors have a stronger overall antitumor immune response than subcutaneous tumors. 31

Figure 2.3. Innate immune cell infiltration in orthotopic and subcutaneous models was distinct.

(A) In human colon samples, NKp46+ cells were found in normal colon and in some CRC cases. (B-C) In tumor models, NKp46+ cells were more frequent in orthotopic tumors than in subcutaneous tumors. mRNA expression of IL-12, IL-15, IL-18, and CCL5 (all related to NK cells) was higher in orthotopic tumors. (D) The expression level was presented as fold change, refers to the lowest expression. (E) Flow cytometry showed no difference in the number of DCs and macrophages in the two models. (F) More MDSCs, important immunosuppressive cells, were found in subcutaneous tumors. A higher proportion of CD11b+ CD11c- cells were found in 32 subcutaneous tumors. The proportion of Ly6C+ and Ly6G+ cells (in the CD11b+ CD11c- population) was the same in the two models (data not shown). IL, interleukin; N, normal; T, tumor. % total refers to the total number of cells in tumor tissue. For all panels, data are plotted as the mean ± SD. t-test was performed between indicated two groups, *P < 0.05. This figure is reprinted with permission from Zhao et al., Oncotarget, 2017.

Figure 2.4. Macrophage subtypes in the subcutaneous and orthotopic tumors.

(A) F4/80 and CD206 were co-stained in subcutaneous and orthotopic tumors for M2 phenotype macrophages. (B) F4/80 and CD80 were co-stained for M1 phenotype macrophages. Most macrophages observed were the M2 phenotype. Representative data from each group (n=5) was shown. This figure is reprinted with permission from Zhao et al., Oncotarget, 2017.

Figure 2.5. Expression of IL2, IL6, granzyme B, and IFNγ varied in orthotopic and subcutaneous tumors.

Using the enzyme-linked immunosorbent assay (ELISA) method, we measured the expression of some proinflammatory and cytotoxic cytokines, such as IL2 (A), IL6 (B), granzyme B (C), and IFNγ (D). In CT26 tumor cell culture and in normal colon tissue in BALB/c mice, those cytokines could not be detected. But in both orthotopic and subcutaneous tumors in mice, we detected the expression of those cytokines: expression was higher in orthotopic tumors than in subcutaneous tumors. IFN-γ, interferon-gamma; IL, interleukin. For all panels, data are plotted as the mean ± SD. t-test was performed between indicated two groups, *P < 0.05; ** P < 0.01. This figure is reprinted with permission from Zhao et al., Oncotarget, 2017.

2.2.6 Immune checkpoint expression and efficacy of ICBT differed between orthotopic and subcutaneous models Besides the number of immune cells, the activation status of cytotoxic immune cells is also critical in regulating the anti-tumor immune response. The functional status of immune cells in tumors was determined by measuring the expression levels of immune checkpoints and T-cell activation markers. We generated orthotopic and subcutaneous CRC models by injecting the same number (105) of cells. In subcutaneous tumors, we observed higher levels of immune checkpoints such as CTLA-4 and PD-1 on the T cells compared to orthotopic tumors (Figure 2.6A). Further, PD-L1 expression was also higher in subcutaneous tumors than in orthotopic tumors (Figure 2.6A). However, we observed no significant differences in the expression of T- cell activation markers such as CD62L, CD44, in tumor-infiltrating T cells in both tumor models (Data not shown). We further investigated whether these two tumor models with different immune profiles show varying responses to immune checkpoint blockade therapy (ICBT). Towards this, mice were either treated with IgG or anti-PD1 plus anti-CTLA4 in six doses for three weeks (Figure 2.6B). This treatment regime induced response in both orthotopic and subcutaneous tumors (Figure 2.6C-E). After six doses of treatment, we did not observe any tumor growth in the orthotopic model (Figure 2.6C). However, in the subcutaneous model, although the tumors responded to ICBT (Figure 2.6D, E and Figure 2.7), residual tumors were still observed. To determine tumor vascularization that may have a potential confounding effect on ICBT treatment, we stained CD31 in two of our CRC models. Our staining patterns suggest that no obvious difference in capillary density in these two models (Figure 2.8).

2.2.7 Surgical based mouse cecum tumor model mimics advanced stage human CRC tumors Although the subcutaneous model and our novel endoscopy-guided orthotopic CRC model have distinct immunological features, none of them can represent the late-stage human CRC tumors, which have metastatic lesions. Previous studies have shown that tumor cells/tissues implanted on the mouse cecum developed extensive metastasis124,125. By surgically implanting CT26 tumor cells/tissues to the mouse cecum, we established the mouse cecum tumors (Figure 2.9). We observed an aggressive primary tumor formation on the cecum wall. With the development of tumors, the mice showed signs of intestinal obstruction. All mice developed lymphatic metastasis and peritoneal cavity seeding. In around 20% of mice, we observed metastatic tumors in the liver and spleen (Figure 2.9). These data validated that the mouse cecum

35 tumor models are suitable for mimicking the late-stage human CRC tumors with metastatic lesions.

Figure 2.6. Immune checkpoint profile and response to ICBT differed in orthotopic and subcutaneous tumors.

(A) Expression of inhibitory immune checkpoints PD-1 and CLTA-4 was higher on tumor- infiltrating T cells in subcutaneous tumors. The expression of PDL1 was higher in subcutaneous tumors. (B) To compare the drug sensitivity of the 2 models, we administered anti-CLTA-4 and anti-PD1 with moderate intensity. (C) At the endpoint, orthotopic tumors showed a better response; they were thoroughly blocked by ICBT, whereas, (D-E) the subcutaneous tumors can only be partially controlled by ICBT. ICBT, immune checkpoint blockade therapy; IFN-γ, interferon-gamma; T, tumor. For all panels, data are plotted as the mean ± SD. t-test was performed between indicated two groups, *P < 0.05; **P < 0.01. This figure is reprinted with permission from Zhao et al., Oncotarget, 2017.

Figure 2.7. The response of subcutaneous tumors to ICBT.

(A) After the ICBT, the number of tumor-infiltrating T cells dramatically increased. (B-C) Expression of CD62L on T cells decreased, and expression of IFNγ increased, indicating a more activated phenotype of tumor-infiltrating T-cell after ICBT. ICBT, immune checkpoint blockade therapy; IFN-γ, interferon-gamma; T, tumor. For all panels, data are plotted as the mean ± SD. t- test was performed between indicated two groups, *P < 0.05; ****P < 0.0001. This figure is reprinted with permission from Zhao et al., Oncotarget, 2017.

Figure 2.8. CD31 staining in the subcutaneous and orthotopic tumors.

(A) Representative pictures of CD31 staining. (B) No significant difference was observed in CD31 positive cells. This figure is reprinted with permission from Zhao et al., Oncotarget, 2017.

Figure 2.9. The features of surgically implanted mouse cecum tumors.

(A) The in vivo imaging showed the progression of surgically implanted cecum tumors. (B) The primary orthotopic cecum tumor and liver, metastatic spleen lesions were shown. (C) The histology of the primary tumor was shown. 1: tumor cells in submucosa; 2: tumor cells in muscularis externa; 3: tumor cells outside the colon wall; 4: tumor cells in the mucosa; 5: liver tissue; 6: tumor cells spreading to the surface of the liver; 7: tumor metastasis in the liver.

2.3 Discussion Studies in mice are frequently used to provide the rationale for testing anti-tumor therapies in phase 1 clinical trials126. The traditional approach, using human xenografts in immunocompromised mice, is not amenable to testing immunomodulatory anti-tumor agents due to the absence of a physiological immune system. The mouse models of tumors, including human xenograft tumors and syngeneic tumors, were suitable for preclinical studies that tested cytotoxic drugs126. However, immunotherapy has a different mechanism for eliminating tumors: it kills tumor cells by enhancing or rebuilding the antitumor immune response in patients (rather than by directly killing tumor cells). The efficacy of immunotherapy, especially ICBT, largely depends on the immune microenvironment of the tumors127,128.

To maximize opportunities to translate novel immunotherapy strategies from preclinical studies to clinical application, CRC mouse models are needed that can mimic physiologically relevant microenvironment116. In this study, we analyzed two immune-competent models using murine CRC cell lines (CT26 and MC38) in their matched immune-competent hosts (BALB/C and C56BL/6). We demonstrate that these models can be used to test the efficacy of checkpoint blockade therapy, anti-CTLA-4 and anti-PD-1. Furthermore, we demonstrate that the immune 38 profile and response is different when comparing an orthotopic tumor microenvironment to a subcutaneous microenvironment in the skin.

In line with previous studies, we observed human CRC subtypes with immune cells well and poorly-infiltrated tumors117,129. The poorly immune-infiltrated tumors are insensitive to ICBT130, while the well-infiltrated tumors are potentially more sensitive to ICBT131. Our investigation of the adaptive immune profiles of orthotopic and subcutaneous tumors in mice represents a key step in mimicking different clinical settings in mouse models. The two CRC tumor models produced diverse immune phenotypes in terms of T-cell infiltration. Orthotopic tumors were better infiltrated with T-cell than subcutaneous tumors and observed the same trend with B-cells.

Innate immune cells, depending on their differentiation and functional status, either suppressed or promoted tumor formation132. NK-cell is the primary cell type in innate immunity has antitumor functions133. Enhancing the antitumor effects of NK cells via heterodimeric bispecific single-chain variable fragments (scFv) killer engagers has been very promising in preclinical models134. In human CRC tissue samples, we observed a subset of samples with infiltrated NK cells. In mice, orthotopic tumors had more NK-cell infiltration than subcutaneous tumors. Notably, subcutaneous tumors had more immunosuppressive cells, such as MDSCs, compared to orthotopic tumors. On the other hand, we found the same proportion of antigen- presenting cells, such as DCs and macrophages in these two tumor models. In terms of expression of IL-2, IL-6, IFN-γ and granzyme B, the orthotopic tumors showed relatively higher levels than subcutaneous tumors.

By comparing the two mouse models, we found that inhibitory immune checkpoint proteins, such as CTLA4, PD1, and PDL1, were expressed at lower levels on orthotopic tumors compared to subcutaneous tumors. Such checkpoints are significant factors in regulating the activation of adaptive immune cells, especially T cells135,136. These findings suggested that T cells in subcutaneous tumors were activated but under strong suppression of immunosuppressive factors, including immune checkpoints. These data were in line with our results that subcutaneous tumors have a weak immune response, further confirmed the distinction of immune profiles in these two models.

Response to ICBT in these two models was also investigated. Multiple ICBT protocols with significant differences in treatment intensity have been reported in preclinical studies75,137,138. 39

It also has been demonstrated that dual blockade of the PD-1 and CTLA-4 pathways increased the antitumor effects via enhancing immune effector cell/regulatory T-cell ratio in an animal model139. Considering that the CT26 tumors were relatively sensitive to ICBT140, we administrated immune checkpoint blockades with moderate intensity. This treatment plan would rule out false positive or negative rustles due to inappropriate therapeutic dose. We found that tumors in both models responded to ICBT. But orthotopic tumors were more sensitive: they were totally blocked by ICBT. Subcutaneous tumors responded to ICBT, based on reduced growth; yet at the end of our study period, small subcutaneous tumors were still present. Taking all data together, the subcutaneous model mimics a weak immune infiltrated and heavily immunosuppressive phenotype, whereas the orthotopic model can mimic relatively well-immune infiltrated CRC in a physiologically relevant microenvironment.

As a prerequisite for translating innovative immunotherapy strategies from bench to bedside, appropriate experimental CRC models are needed to mimic different clinical settings accurately. Our study investigated the key characteristics of a novel orthotopic CRC model and showed significant differences between the orthotopic model and the subcutaneous model in immune profiles and ICBT. Our study indicates that there remains a role for the subcutaneous model in immunotherapeutic studies because they are easy to establish, are very stable, and they mimic a relatively sparse immune microenvironment. Our orthotopic model, on the other hand, provides another useful option that better mimics CRC tumors with higher levels of immune infiltration in human patients. Finally, we validated the metastatic potential of surgically implanted mouse cecum tumors, which has been reported previously124,141,142. This data indicated the feasibility of using cecum tumor models to mimic human late-stage CRC tumors with extensive abdominal metastasis.

Taken together, our study identifies the influence of tumor location on immune response in CRC mouse models. Technically, we contributed a detailed protocol for establishing endoscopy-guided mouse CRC orthotopic models. Moreover, we highlight the significance of model selection in immunotherapy studies and demonstrate a role for our novel endoscopy guided orthotopic CRC model that could supplement current subcutaneous models to increase the translational potential of preclinical CRC immunotherapeutic studies.

2.4 Methods and Material 2.4.1 Cell culture CT26, a CRC cell line generated from BABL/C mice, was purchased from the American Type Culture Collection (ATCC, Manassas, VA) and was cultured in RPMI-1640 medium with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin (pen-strep). MC38, a CRC cell line generated from C57BL/6 mice, was a gift from Dr. Nicholas Haining (Harvard University) and was cultured in Dulbecco’s modified Eagle’s medium (DMEM) with 10% FBS and 1% pen- strep. HT29, a human CRC line, was purchased from ATCC and cultured in McCoy’s 5A medium with 10% FBS and 1% pen-strep. For in vivo imaging experiments, we generated a stably transfected CT26 cell line with firefly luciferase (CT26-Luc). Cell lines obtained from ATCC resource were authenticated by the vendor.

2.4.2 Mice and treatment BALB/c mice (6-8 weeks old, Jackson Laboratory, Bar Harbor, ME) were used for grafts using CT26 and CT26-Luc cells. C57/B6 mice (6-8 weeks old, Charles River Laboratories, Wilmington, MA) were used for MC38 grafts. Athymic nude mice (6-8 weeks old, Charles River Laboratories) were used for HT29 grafts. All mice were kept in a specific pathogen-free facility and had unrestricted access to water and food and a controlled 12-hr day-night cycle. Animal studies were approved by the institutional animal care and use committee of the University of Minnesota.

After tumor cell implantation, mice were randomly separated into a treatment arm and a control arm. In the treatment arm, mice were injected with anti-mouse PD1 (Clone: RMP1-14, 10 mg/kg, twice per week) and anti-mouse CLTA4 (Clone: UC10-4F10-11, 5 mg/kg, twice per week) (both from BioCell Technology LLC, Newport Beach, CA). In the control arm, mice were injected with an IgG isotype (15 mg/kg, twice per week) (BioLegend, San Diego, CA).

2.4.3 Tumor implantation To perform in vivo high-resolution colonoscopies, we used the Mainz Coloview mini- endoscopic system (Karl Storz Endoskope, Tuttlingen, Germany). For orthotopic tumor cell implantation in the colon wall, we anesthetized mice with ketamine (100 mg/kg) combined with xylazine (10 mg/kg) via intraperitoneal injection. To minimize colon movement, contraction, and secretion, we administered atropine (0.04 mg/kg, intraperitoneally). After tumor cell implantation, mice were put on a heating pad until fully recovered. When monitoring tumors, we anesthetized mice with Avertin (250 to 500 mg/kg, intraperitoneally) 5 minutes before the

41 procedure to minimize the duration of anesthesia. For subcutaneous tumor cell implantation, we suspended tumor cells in a Matrigel matrix, then inoculated the suspension in the flank of legs.

The mouse cecum CRC model was established by surgery. 2-4 hours before surgery, Buprenorphine (SR) 2mg/kg is administered subcutaneously for pre-emptive analgesia. Then, mice are anesthetized with ketamine (100mg/kg)/xylazine(10mg/kg). We then position the mouse on its back on fix the forelegs in a V shape with tape. The abdomen of the mouse was shaved, and the skin was prepared by wiping with Povidone-iodine prep pads and then an alcohol prep pad. CT26 cells were resuspended in matrigel. The surgical area was isolated by placing the sterile drape around the incision area on top of the abdomen. A 15mm vertical midline incision was made on the skin. We then incised the linea alba to separate the rectus abdominis muscles and opened the abdomen. The cecum was located and carefully placed horizontally on top of a histo- cassette. A matrigel drop containing the CT26 cells (2*105) was inoculated under the serosa layer. The inoculation site was carefully wiped by an alcohol pad to remove any potential leaking. The cecum was placed back into the abdomen using cotton swabs drenched in PBS. The abdominal wall and skin opening were then closed. Mice were monitored for at least 7 days; more analgesics were given when needed.

2.4.4 In vivo imaging To monitor orthotopic CRC tumors established by CT26-Luc cells, we used the IVIS Spectrum in vivo imaging system (PerkinElmer, Waltham, MA); 10 minutes before imaging, we injected mice with D-luciferin, GoldBio (150mg/kg, intraperitoneally) and then anesthetized them with isoflurane. For all mice, we set the exposure time of imaging as 60 sec.

2.4.5 Flow cytometry Flow cytometry was used to measure immune cell infiltration and activation markers. Harvested tumor tissue was digested in a solution of collagenase IV (5 mg/ml) and deoxyribonuclease (DNase, 50 units/ml) at 37° C for 1 hr and filtered through a 40-μm cell strainer. Cells were centrifuged and resuspended in red blood cell lysis buffer for 10 minutes, followed by another round of centrifugation.

The following antibodies were purchased from BioLegend or BD Biosciences (San Jose, CA): CD3 (17A2), CD19 (6D5), CD4 (GK1.5), CD8 (53-6.7), CD11b (M1/70), CD11c (N418), CD28 (37.51), PD1 (29F.1A12), CTLA4 (UC10-4B9), PDL1 (10F.9G2), CD44 (IM7), CD62L (MEL-14), interferon-gamma (IFNγ, XMG1.2), Ly6C (HK1.4), Ly6G (1A8), F4/80 (BM8), and I-Ad (AMS-32.1). Cells were stained with surface marker antibodies first, and then fixed and 42 permeabilized for staining with intracellular markers. Data were analyzed using FlowJo software (Tree Star, Inc., Ashland, OR).

2.4.6 Histology and immunostaining Immediately after mice were sacrificed, tumor tissue was fixed in 10% formalin before paraffin embedding. Standard procedures were used for hematoxylin and eosin (H&E) staining. For immunostaining, formalin-fixed, paraffin-embedded tissue was treated with xylene, rehydrated with ethanol, and heated in a microwave with the citric buffer to retrieve antigens. For blocking purposes, the tissues were incubated for 30 minutes, with 5% bovine serum albumin buffer. Followed by overnight incubation at 4° C, with primary antibodies: anti-CD3 antibody, anti-CD4 antibody, and anti-NKp46 antibody (Abcam, Cambridge, United Kingdom) at 1:100 dilutions, anti-CD8 antibody (Novus Biologicals, Minneapolis, MN) at 1:20 dilution, anti-CD31 (Novus Biologicals, Minneapolis, MN) at 1:100, anti-F4/80 (R&D system Minneapolis, MN) at 1:100, anti-CD206 (R&D system Minneapolis, MN) at 1:100, as well as anti-CD80 (Novus Biologicals, Minneapolis, MN) at 1:100. After washing, tissues were incubated with fluorescence-conjugated secondary antibodies at room temperature for 1 hr. Slides were prepared with antifade mountant with 4’,6-diamidino-2-phenylindole (DAPI).

2.4.7 RT-qPCR The mirVana microRNA (miRNA) Isolation Kit was used to extract RNA. We used 500 ng of total RNA for real-time quantitative reverse-transcriptase polymerase chain reaction (RT- qPCR) analysis with the QuantiTect Reverse Transcription Kit (Qiagen, Hilden, Germany). To measure cDNA samples, we used the LightCycler 480 Instrument (Roche Life Science, Indianapolis, IN) normalized to 18S ribosomal RNA (rRNA) expression. Primer sequences are: 18srRNA (Forward: GTTGGTTTTCGGAACTGAGG, Reverse: AGTCGGCATCGTTTATGGTC), NCR1 (Forward: TTGCCAACTGAAGACTGCCA, Reverse: TCCCTCTGTGAGCCCTAGTC), CD19 (Forward: GTCATTGCAAGGTCAGCAGTG, Reverse: GGGGTCAGTCATTCGCTTCC), CCL5 (Forward: ATATGGCTCGGACACCACTC, Reverse: ACTTGGCGGTTCCTTCGAG), IL-2 (Forward: ATGAACTTGGACCTCTGCGG, Reverse: GTCCACCACAGTTGCTGACT), IL-18 (Forward: TCAGACAACTTTGGCCGACT, Reverse: CAGTCTGGTCTGGGGTTCAC), IL-15 (Forward: TCCCAGTTGCAAAGTTACTGC, Reverse: TTCTCCTCCAGCTCCTCACA), IL-12 (Forward: CCGGTCCAGCATGTGTCAAT, Reverse: GGACTGGCTAAGACACCTGG), CXCL11 (Forward: CTTATGTTCAAACAGGGGCG, Reverse: TGCATTATGAGGCGAGCTT),

CXCL10 (Forward: GCTGCAACTGCATCCATATC, Reverse: AGGAGCCCTTTTAGACCTTT), CXCL9 (Forward: GAAGTCCGCTGTTCTTTTCC, Reverse: TTGACTTCCGTTCTTCAGTGT), CXCR3 (Forward: CCTGCATAGTTGTATGGGGT, Reverse: ATATGGGGCATAGCAGTAGGC).

2.4.8 ELISA ELISAs were performed to measure granzyme B, IFNγ, IL6, and IL2 using ELISA kits from Affymetrix (Santa Clara, CA) according to the manufacturer's protocol. All samples were normalized based on protein concentrations measured using a BCA protein assay (Pierce Chemical Company, Dallas, TX).

2.4.9 Statistical analysis For all statistical analyses, we used GraphPad Prism 6.0 (GraphPad Software, San Diego, CA). To compare the treatment arm and the control arm, we used the Student t-test. For multiple group data, we used the one-way analysis of variance (ANOVA) method; for multiple pairwise comparisons, we performed a Bonferroni post hoc adjustment. All data are plotted as the mean ± standard deviation (SD). Two-sided P values < 0.05 were considered statistically significant.

2.5 Publication This chapter has been modified (with permission) from the published articles:

Zhao X, Li L, Starr TK, Subramanian S. Tumor location impacts immune response in mouse models of colon cancer. Oncotarget. 2017 Jun 9;8(33):54775-54787.

(https://www.ncbi.nlm.nih.gov/pubmed/28903381)

3. Chapter 3

Impact of Tumor-draining Lymph Nodes and Chemotherapy on Cancer Immunotherapy Response

3.1 Introduction Immune checkpoint blockade therapies (ICBT) such as anti-CTLA-4 and anti-PD-1/PD- L1, have transformed the therapeutic landscape of cancers, including melanoma and tumors with microsatellite instability9,104,143. Nonetheless, as with more traditional forms of systemic chemotherapy options, many patients manifest either intrinsic or acquired resistance leading to treatment failure13,144,145. Multiple mechanisms that influence tumor response to ICBTs have been identified— the mutational load in tumor cells, the degree of T-cell exhaustion, tumor microenvironmental functions, and intestinal microbiota13,144,145. In most cases, ICBT is used for treating patients with heavily pretreated tumors. The interactions between first-line therapy may influence tumor response to subsequently administered ICBTs due to tumor evolution and heterogeneity. In most patients with solid tumors, common interventions before ICBT include resection of primary tumors with concurrent resection of draining lymph nodes followed by administration of chemotherapies and/or targeted therapies9,146,147. However, minimal information is known about whether these interventions will impact tumor response to ICBT.

Tumor-draining lymph nodes (TdLNs), which are usually resected concurrently with the primary tumors, have shown dual impacts on tumor development and treatment. On the one hand, TdLNs are critical peripheral lymphatic organs where tumor antigens are presented by dendritic cells to naïve T cells to elicit antitumor immunity148-150. Thus, loss of TdLNs weakens immunosurveillance mechanisms and increases the likelihood of tumor initiation and progression148-151. On the other hand, TdLNs are affected by immunosuppressive factors released by tumor cells during tumor progression. These immunosuppressive factors can suppress the function of TdLNs, making them immune‐privileged sites152-156. Based on these facts, we hypothesize that TdLN resection is an essential factor that influences long-term tumor immunity and response to ICBT. In this study, we used tumor models representing different disease stages to elucidate the impacts of TdLN resection on ICBT efficacy and understand the underlying mechanisms of those effects.

The immunoregulatory effects of chemotherapies have been investigated in multiple cancer models with different chemotherapy drugs. Chemotherapy drugs such as oxaliplatin, paclitaxel, and 5-fluorouracil (5-FU) have shown positive effects in antitumor immunity either by eliciting a tumor-specific T-cell response or by reducing immunosuppressive factors in the tumor microenvironment75,157,158. Bone marrow suppression, which is a common side effect of chemotherapies, causes leukopenia that affects antitumor immunity. Because chemotherapies 47 have dual effects on regulating antitumor immunity, we hypothesize that combining chemotherapy with ICBT has diverse effects on the antitumor immune response. Consequently, an appropriate combinatory strategy will be critical in determining tumor response. In this study, we used 5-FU, which blocks DNA replication, as a representative chemotherapeutic drug to study the factors that influence the effects of chemotherapy on ICBT.

Mouse models are critical for pre-clinical cancer studies; most published studies have been performed on primary tumor models. To better represent the clinical conditions in which most immunotherapies are administered, we established a mouse tumor model that allows evaluation of the immunotherapeutic response in secondary tumors after primary tumor resection with or without concurrent TdLN removal. We also included anti-PD-1 (antagonist to inhibitory immune checkpoints) and anti-4-1BB (agonist to stimulatory immune checkpoints) to better represent ICBT with different mechanisms159,160.

3.2 Results 3.2.1 TdLNs are essential for antitumor immune activation and immunotherapy response in early-stage disease We first needed to identify TdLNs in the subcutaneous tumor model. We injected Evans blue and Alexa Fluor 488 into tumors established in the right flank of the mice to trace lymphatic drainage (Figure 3.1A). Evans blue staining was detected in the right inguinal and axillary lymph nodes 10 min after injection (Figure 3.1B). To develop a more sensitive method for detection, we used flow cytometry to trace the Alexa Fluor 488 drainage in lymphatic organs for up to 48 hr. Again, the right inguinal and right axillary lymph nodes showed the highest fluorescence intensity (Figure 3.1C). Other lymph nodes, such as right brachial and right popliteal lymph nodes, also showed increased fluorescence signal after injection, but the signal intensity was significantly lower than in the right inguinal and right axillary lymph nodes (Figure 3.1C). Also, increased weight was observed in the spleen, and right inguinal and axillary lymph nodes during tumor development, suggesting an immune response occurred in these lymphatic organs (Figure 3.2A- J). Collectively, these results indicated that the right inguinal and right axillary lymph nodes are the sentinel TdLNs in our tumor model.

Figure 3.1. Identification of tumor-draining lymph nodes (TdLNs) in mice.

(A) Evan blue dye or Alexa Fluor® 488 dye was injected into the tumor in the right hinge flank to trace TdLNs. (B) 10 min post Evan blue dye injection in the right hinge flank tumor, the right inguinal (RI), and right axillary (RA) lymph nodes (LNs) were stained. The deeper color was seen at 30 min and 60 min post-injection. The representative data from three independent experiments were shown. (C) Flow cytometry was used for detecting the Alexa Fluor® 488 dye distribution in lymphatic organs (drainage from the right hinge flank tumor injection site). The RI LN and RA LN showed the highest FITC signal and were identified as the major TdLNs. The other LNs were identified as the non-draining lymph nodes (NdLNs) (n=3 in each group, data displayed as means ± SEMs, RI: right inguinal, LI: left inguinal, RP: right popliteal, LP: left popliteal, RB: right brachial, LB: left brachial, RA: right axillary, LA: left axillary, M: mesenteric). This figure is reprinted with permission from Zhao et al., iScience, 2020. Next, we evaluated the impact of TdLNs on tumor initiation and antitumor immune response stimulation. Resection of TdLNs, but not non-draining lymph nodes (NdLNs), before tumor cell inoculation significantly accelerated tumor development in both CT26 and MC38 tumor models (Figure 3.3A). We then analyzed the stimulation of antitumor immunity with and without TdLNs. We used the frequency of tumor antigen-specific CD8+ T cells (10 days after tumor cells inoculation) as an indicator of antitumor immune response stimulation (Figure 3.4). More tumor antigen-specific CD8+ T cells were detected in the right and left brachial lymph nodes and spleen of tumor-bearing mice with intact TdLNs (Figure 3.3B). 4-1BB (CD137) provides essential costimulatory signaling for T cells, and its agonist has shown tumor- eliminating effects in mice159. To test the effects of TdLNs on the immunotherapeutic response, we administered two injections of anti-4-1BB shortly after tumor cell inoculation to simulate

49 patients with minimal disease burden. The prophylactic anti-4-1BB treatments successfully prevented tumor development in mice with intact TdLNs, and anti-tumor immunity memory was established, as evidenced by their rejection of secondary tumors. These effects were not seen in mice with resected TdLNs (Figure 3.3C). These data demonstrated that TdLNs are critical for anti-tumor immunity activation, and loss of TdLNs leads to rapid early-stage tumor growth even with the potent T-cell co-stimulatory agonist.

3.2.2 TdLNs are not necessary for immunotherapy response in advanced disease tumor models Since recurrence after primary tumor resection is one of the major causes for treatment failure, we evaluated the impact of TdLNs on tumor recurrence and the response to immunotherapy in mouse models after advanced primary tumor resection. We allowed the primary tumor to grow to a relatively large volume and then resected the primary tumor with and without concurrent TdLN resection. Secondary tumors were then inoculated to mimic localized tumor recurrence (Figure 3.5A). We confirmed a clean primary tumor resection margin in our models (Table 3.1), allowing all secondary tumors to start with a comparable baseline. TdLNs were also subjected to histological analysis to confirm that no metastasis developed in TdLNs (Figure 3.2K). Notably, in our well-controlled model, the secondary tumor growth rate was similar in mice with and without TdLNs (Figure 3.5A, B). In another group of TdLNs resected mice, we depleted T cells to study the impact of systemic immunity on subsequent tumor development. As predicted, the secondary tumor developed rapidly in mice with impaired systemic immunity (Figure 3.5A, B). Together, these results indicate that tumor recurrence is accelerated by impaired systemic immunity but not by impaired regional immunity (TdLNs resection).

Figure 3.2. Physical changes and histology of TdLNs, NdLNs, and spleen of tumor-bearing mice.

(A-J) The weight of spleen and major superficial LNs (both TdLNs and NdLNs) were measured at different time points of tumor development. During tumor development, significant splenomegaly was observed. Evident lymphadenopathy was observed in the TdLNs rather than in the NdLNs during tumor development (n=4 in each group, t-test was performed between indicated groups, data were displayed as means ± SEMs, *p<0.05, **p<0.01, ***p<0.001, statistical analyses without significance were not shown). (K) At the late-stage of tumor development, the histology of TdLNs and NdLNs was evaluated. The TdLNs were larger than NdLNs and naïve LNs (taken from tumor-free mice). No metastasis was observed in TdLNs. Representative data from three independent experiments were shown (Scale bars: 250μm in 40X images and 50μm in 200X images). This figure is reprinted with permission from Zhao et al., iScience, 2020.

Figure 3.3. Impact of TdLNs on tumor initiation and immunotherapy response in early- stage tumor models.

(A) Experimental schedule and tumor growth curves in mice with or without TdLNs. Both CT26 (BALB/c mouse as the host) and MC38 (C57BL/6 mouse as the host) subcutaneous models were enrolled in the experiment. Mice were pre-conditioned by TdLNs resection (right inguinal and axillary LNs), NdLNs resection (left inguinal and axillary LNs), or sham surgery prior to tumor inoculation. Accelerated tumor growth was observed in mice without TdLNs (n=5 in each group, One-way ANOVA test between all groups, data represent each individual mouse, *p<0.05, **p<0.01, ***p<0.001). (B) Distribution of tumor antigen (gp70) specific CD8+ T cells in tumor- bearing mice with or without TdLNs. Less tumor antigen-specific CD8+ T cells were detected in the right and left brachial lymph nodes, and spleen of TdLNs resected tumor-bearing mice (n=4 in each group, Tukey's multiple comparisons test between two groups, data were displayed as means ± SEMs, n.s.: no significance, **p<0.01). (C) Experimental schedule and early-stage tumor response to anti-4-1BB treatment. Two injections of anti-4-1BB were given shortly after

52 tumor inoculation. The treatment prevented tumor development in tumor-bearing mice with intact TdLNs. Rechallenge of the tumor cells did not form tumors in all anti-4-1BB cured mice (n=4 in each group, One-way ANOVA test between all groups, data represent each individual mouse, ****p<0.0001). This figure is reprinted with permission from Zhao et al., iScience, 2020.

Figure 3.4. Gating of tumor antigen-specific CD8+ T cells.

(A) Representative gating process of CD8+ gp70 (tumor) antigen-specific T cells in peripheral lymphatic organs. (B) Representative gating process of CD8+ gp70 (tumor) antigen-specific T cells in tumor tissues. The negative tetramer and lymphatic organs from naïve mice were used as controls for gating. This figure is reprinted with permission from Zhao et al., iScience, 2020. Table 3.1. gp70 mRNA expression in different tissues.

Model Tumor tissue Tumor marginal skin & Normal connective tissues post-surgery skin tissue

CT26 (Babl/c) Positive Negative Negative

MC38 (C57BL/6) Positive Negative Negative

Next, we asked whether TdLNs resection altered immune infiltration in secondary tumors. The major immune cell types were evaluated in secondary tumors (Figure 3.6). Total tumor-infiltrating T cells, PD-1 high expression T cells, and MDSCs were not altered in secondary tumors either with or without TdLNs. PD-L1 expression was similar. The frequency of CD103+ DCs and lymphatic endothelial cells were significantly higher in the secondary MC38 tumors with TdLNs (Figure 3.6). However, in the CT26 model, only the lymphatic endothelial cell frequency was statistically higher in secondary tumors with TdLNs than without TdLNs. The frequency of CD103+ DCs showed a similar trend but did not reach statistical significance (Figure 3.6).

Immunotherapies are typically prescribed to patients who have undergone advanced primary tumor resection. In another pre-clinical model, we administrated anti-4-1BB and anti- PD-1 to study whether TdLN resection will lead to immunotherapy resistance. To mimic clinical conditions, we resected the established primary tumor, both with and without concurrent TdLN resection. We then inoculated the secondary tumor to mimic localized tumor recurrence. A 6-day gap was allowed between the secondary tumor inoculation and any treatment (Figure 3.5C). This allows the tumor to connect with systemic circulation and to establish the tumor microenvironment. Then, the mice were treated with anti-4-1BB or anti-PD-1 (Figure 3.5C). Notably, both anti-4-1BB and anti-PD-1 treatments were efficient in controlling secondary tumor initiation. Secondary tumor control was maintained after TdLN resection (Figure 3.5C), suggesting that TdLN resection may not be a major influencing factor on the efficacy of ICBT when used as adjuvant therapy in late-stage disease.

Figure 3.5. Impact of TdLNs on tumor recurrence and immunotherapy response in advanced stage tumor models.

(A) The experimental schedule. Resection of TdLNs did not accelerate localized secondary tumor (mimicking recurrent tumor) development in both CT26 and MC38 subcutaneous tumor models. However, systemic deletion of T cells significantly accelerated secondary tumor development in both tumor models (n = 8-10 in each group, both individual and summarized curves were shown, t-tests were performed between the TdLN resected and T-cell depleted groups, data were 55 displayed as means ± SEMs, t-test was performed between the TdLN(-) and TdLN(-) T-cell depleted groups, *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001). (B) Systemic depletion of T cells, but not TdLN resection, led to a shorter survival time of mice due to secondary tumor development (n= 8-10 in each group, log-rank test between indicated groups, **p<0.01). (C) Response to anti-4-1BB and anti-PD-1 treatment was tested in localized secondary tumors with or without TdLNs. Anti-4-1BB and anti-PD-1 treatments suppressed secondary tumor growth in both TdLN intact and resected mice (n=5 in each group, data were displayed as means ± SEMs). This figure is reprinted with permission from Zhao et al., iScience, 2020.

Figure 3.6. Immune features in secondary tumors with or without TdLNs.

(A) The frequency of lymphatic endothelia cells was higher in CT26 secondary tumors (mimicking recurrent tumors) with TdLNs than secondary tumors without TdLNs (n=8 in each 56 group, t-test). The total tumor-infiltrating T-cell frequency, PD-1 high expression T cells frequency, CD103+ dendritic cells (DCs) frequency, myeloid-derived suppressive cells (MDSCs) frequency, and PD-L1 expression were similar in two groups (n=8 in each group, t-test was performed, data displayed as means ± SEMs, n.s.: no significance, *p<0.05). (B) The experiments were repeated in the MC38 tumor model. The frequency of lymphatic endothelial cells and CD103+ DCs were higher in secondary tumors with TdLNs than in secondary tumors without TdLNs (n=4 in each group, t-test was performed, data displayed as means ± SEMs, n.s.: no significance, *p<0.05). This figure is reprinted with permission from Zhao et al., iScience, 2020. 3.2.3 TdLNs shift from an immunoreactive to an immunotolerant environment and tumor-antigen specific T cells disseminate during tumor development Based on the above results, we then hypothesized that immunosuppression in TdLNs and systemic spreading of tumor antigen-specific T cells during tumor development make the TdLNs less important for late-stage tumors compared with early-stage tumors. We collected the TdLNs (right inguinal and axillary LNs) and NdLNs (left inguinal and axillary LNs) at different stages of tumor development for analysis and compared them with the naïve LNs. The frequency of CD62L-CD4+ T cells was significantly higher in TdLNs than in NdLNs when tumors were small. However, the differences disappeared once the tumors became large (Figure 3.7A). CD80, a crucial co-stimulatory molecule, was higher on APCs in TdLNs than in NdLNs and naïve LNs at early-stage disease (Figure 3.7B). However, with tumor development, the CD80 level on APCs in TdLNs dropped (Figure 3B). As the receptor of CD80, CD28 is highly expressed on CD4+ and CD8+ T cells in TdLNs of early-stage tumors but decreased dramatically during tumor development (Figure 3.7C). Previous studies showed that CD28 is downregulated in T cells, which are repetitively exposed to antigens161,162. Therefore, high numbers of T cells with lower CD28 levels may be the product of repeated activation in the TdLNs of late-stage tumors. However, recent studies indicated that sustained CD28 expression after T-cell priming is required for T-cell function and response to further stimulations, including immune checkpoint inhibitors19,163. IFN-γ is highly produced by functional T cells. However, decreased IFN-γ concentration was observed in TdLNs during tumor development (Figure 3.7D). These data suggested that immune cells in TdLNs of late-stage tumors may not function as properly as in the TdLNs of early-stage tumors, shifting the TdLNs from an immunoreactive to the immunotolerant environment during tumor development.

The amount and distribution of tumor antigen-specific T cells also influence antitumor immunity and immunotherapy response164. We measured the distribution of tumor antigen- specific T cells in mice with established tumors (volume 500-700mm3) (Figure 3.4). As expected, the frequency of gp70 specific CD8+ T cells was highest in the tumor-infiltrating CD8+ T cells

57 population. The gp70 specific CD8+ T cells were detected in all major peripheral lymphatic organs, including spleen, TdLNs, NdLNs, and blood (Figure 3.7E). The proportion of CD4+ (around 55%-65% of all T cells) and CD8+ (around 25%-35% of all T cells) T cells in TdLNs was comparable with that in the naïve mice LNs (data not shown). Considering that TdLNs are only a very small proportion of the lymphatic system, our data suggested that in advanced tumor conditions, TdLNs are not the primary reservoir of tumor antigen-specific T cells. The widely distributed tumor antigen-specific T cells in peripheral lymphatic organs could be the responders of immunotherapies for controlling localized residual tumor (minimal secondary tumors in our model) recurrence.

Figure 3.7. Functional status of TdLNs and tumor antigen-specific T-cell distribution in tumor-bearing mice with advanced disease.

(A) More activated (CD62L-) CD4+ T cells were observed in TdLNs than NdLNs on 7 days post tumor cells inoculation. However, at the late stage of tumor development, the proportion of activated CD4+ T cells was similar in TdLNs and NdLNs. The proportion of activated CD8+ T cells were close in TdLNs and NdLNs during tumor development (n=4 in each group, t-tests were performed, data were displayed as means ± SEMs, n.s.: no significance, **p<0.01). (B) CD80 expression level on antigen presentation cells (APCs) was higher in TdLNs than NdLNs on 7 days post tumor cells inoculation (n=4 in each group, t-tests were performed, data were displayed as means ± SEMs, n.s.: no significance, ***p<0.001, MFI: mean fluorescent intensity). (C) The proportion of CD28high T cells (both CD4+ and CD8+) in TdLNs was decreased during tumor development (n=4 in each group, t-tests were performed, data were displayed as means ± SEMs, n.s.: no significance, ***p<0.001). (D) The concentration of IFNγ in TdLNs was higher at the early stage of tumor development than the late-stage (n=4 in each group, t-tests were performed, data were displayed as means ± SEMs, n.s.: no significance, *p<0.05). (E) At the established tumor model (volume 500-700mm3), systemic distribution of tumor antigen (gp70) specific CD8+ 59

T cells were detected in multiple lymphatic organs and the tumor microenvironment. The tumor microenvironment has the highest frequency of gp70 specific CD8+ T cells than lymphatic organs (n=6 in each group, t-tests were performed between indicated groups, data were displayed as means ± SEMs, n.s.: no significance, *p<0.05, **p<0.01, ***p<0.001). This figure is reprinted with permission from Zhao et al., iScience, 2020. 3.2.4 Sequential treatment of 5-FU and anti-4-1BB or anti-PD-1 leads to better responses than concurrent treatment In addition to the primary tumor and TdLN resection, chemotherapy is a critical factor potentially affecting the efficacy of immunotherapies. Since our preceding data showed that TdLN resection might not affect the immunotherapeutic response, we then focused on the impacts of chemotherapy on immunotherapies. Several mechanisms by which chemotherapies regulate anti-tumor immunity have been identified75,165-167. However, no study has analyzed whether the schedule of combining chemotherapies with immunotherapies influences their synergetic effects. To investigate the impact of different combination therapy schedules on tumor response, we compared sequential versus concurrent 5-FU and anti-4-1BB or anti-PD-1 therapy in mouse models. The IgG and anti-4-1BB monotherapy in immunocompetent and T-cell depleted mice served as control groups (Figure 3.8A). In mice with established tumors, anti-4-1BB monotherapy delayed tumor growth and prolonged mice survival time (Figure 3.8B, C). Anti- CD3 impaired systemic immunity by suppressing T-cell populations (Figure 3.9). In an established tumor model, anti-CD3 preconditioning nullified the anti-tumor effects of anti-4-1BB (Figure 3.8B, C), indicating that intact systemic immunity was required for anti-4-1BB response. 5-FU also delayed tumor development in established tumor models (Figure 3.8B, C). We then combined anti-4-1BB with the 5-FU treatment and found no noticeable improvement in mice survival time (Figure 3.8B, C). In another cohort of mice, the 5-FU treatment was used as induction, and then later, anti-4-1BB was added as the maintenance treatment (Figure 3.8B, C). To determine an appropriate sequential treatment strategy, we tested the dynamics of 5-FU induced T-cell depletion (Figure 3.9). In the sequential treatment, anti-4-1BB was given when the T-cell population had almost recovered from the 5-FU treatment. Mice treated with sequential combination therapy had the longest survival time and the most effective tumor control of all cohorts (Figure 3.8B, C).

Next, we compared the 5-FU and anti-4-1BB sequential and concurrent treatments in a more clinically relevant model. In this model, we performed resection of the established primary tumor together with its TdLNs and induced localized secondary tumors for treatment (Figure 3.10A). Over 60 days of the experiment, the sequential treatment showed better tumor

60 suppression than concurrent treatment (Figure 3.10B). Anti-PD-1 is an FDA-approved class of cancer-directed immunotherapy with different mechanisms than anti-4-1BB. To test whether the conclusion from the anti-4-1BB treatment was generalizable to the anti-PD-1 treatment, we combined 5-FU and anti-PD-1 in concurrent and sequential schedules. Again, the 5-FU and anti- PD-1 given in sequence showed better tumor control than when administered concurrently (Figure 3.10C).

Figure 3.8. 5-FU and anti-4-1BB sequential treatment elicits strong antitumor activity.

(A) Tumor (500mm3 to 700mm3 in volume) bearing mice were randomly assigned to 6 treatment groups: IgG (one dose/3 days), 5-FU monotherapy (one dose/12 days), anti-4-1BB monotherapy (one dose/3 days), anti-CD3 therapy (one dose/3 days) and anti-4-1BB therapy (one dose/3 days, two days after anti-CD3), 5-FU (one dose) and anti-4-1BB (1 dose/3 day, starting at 9 days post 5-FU) sequential therapy, and 5-FU (one dose/12 days) and anti-4-1BB (1 dose/3 day, started at the same day of 5-FU) concurrent therapy. The treatment was continued until the endpoint of follow-up. (B) CT26 tumor response to different treatments. The 5-FU and anti-4-1BB sequential treatment significantly prolonged survival time of the tumor-bearing mice (n=4 in each group for 61 the tumor growth curve, data were displayed as means ± SEMs, n=7 in each group for the mouse survival curve, log-rank test between all survival curves, ****p<0.0001). (C) The same experiments of panel B were repeated in the MC38 tumor model (n=4 in each group for the tumor growth curve, data were displayed as means ± SEMs, n=7 in each group for the mouse survival curve, log-rank test between all survival curves, ****p<0.0001). This figure is reprinted with permission from Zhao et al., iScience, 2020.

Figure 3.9. T-cell depleting effects of 5-FU and anti-CD3 treatment.

(A) 5-FU treatment on naïve mice depleted T cells in lymphatic organs, blood circulation, and bone marrow. The T-cell population was recovered around 9 days after 5-FU treatment (n=3 in each group, data were displayed as means ± SEMs). (B) Single-dose of anti-CD3 treatment on naïve mice depleted T cells for around 3 days (n=3 in each group, data were displayed as means ± SEMs). (C-D) A combination of anti-4-1BB with 5-FU didn't rescue the T-cell depletion induced by 5-FU treatment (n=3 in each group, t-test was performed, data were displayed as means ± SEMs, n.s.: no significance). This figure is reprinted with permission from Zhao et al., iScience, 2020.

Figure 3.10. 5-FU and anti-4-1BB sequential treatment on secondary tumors that mimic tumor recurrence.

(A) The primary tumor and TdLNs were resected when tumors are at around 500mm3 in volume. Secondary tumors were induced and treated by different strategies at 300mm3-350mm3 in volume. Some mice rejected the secondary tumors and were not included in the therapeutic study. (B-C) The 5-FU and anti-4-1BB or anti-PD-1 sequential treatment were more efficient than the 5- FU and anti-4-1BB or anti-PD-1 concurrent treatment in controlling secondary tumors in CT26 and MC38 models (n=7 in each group, data were displayed as means ± SEMs, t-test was performed between the sequential and concurrent treatment groups, * p<0.05, ICBT: immune checkpoint blockade therapy (anti-4-1BB or anti-PD-1)). This figure is reprinted with permission from Zhao et al., iScience, 2020. Toxicity is a primary concern for cancer treatments, especially in combination therapy. We took this into account by evaluating the side effects of each treatment. 5-FU monotherapy and 5-FU and anti-4-1BB concurrent combination therapy caused severe body weight loss and diarrhea during the treatment (Figure 3.11). In contrast, the 5-FU and anti-4-1BB sequential combination therapy showed slight or no side effects for the duration of the experiment (Figure 3.11).

3.2.5 Sequential treatment of 5-FU and anti-4-1BB or anti-PD-1 stimulates a strong antitumor immune response Our pre-clinical models suggested that 5-FU and anti-4-1BB or anti-PD-1 sequential treatment has superior tumor controlling effects than the concurrent treatment schedule. We investigated the potential mechanisms of this result by performing mass cytometry to generate a comprehensive immune landscape characterization in tumor tissues (Figure 3.12). Notably, CD80 and CD86 expression were upregulated after 5-FU and anti-4-1BB sequential treatment in CT26 tumors (Figure 3.13A). High expression of these two critical co-stimulatory factors suggests 63 enhanced tumor visibility by T cells. The expression of PD-L1 on tumor tissue was not significantly changed among different groups (Figure 3.13A). Furthermore, tumor immune infiltration studies showed that anti-4-1BB monotherapy stimulated tumor-infiltrating T-cell proliferation and increased the CD8+ T-cell versus regulatory T-cell ratio (Figure 3.13B, C, D). In sum, these experiments showed that 5-FU and anti-4-1BB sequential treatment alone maintained the positive effects of anti-4-1BB on the T cell population (Figure 3.13B, C, D).

Figure 3.11. Side effects of different treatments.

(A-B) The mouse body weight was measured on day 12, 24, and 32 during treatment. On day 32, the mice treated with the 5-FU and anti-4-1BB sequential treatment have higher body weight than the mice treated with the 5-FU and anti-4-1BB concurrent treatment (n=7 in each group, t-test was performed between indicated groups at the last time point, the individual value was shown, ****p<0.0001). (C-D) Diarrhea assessment was performed at the endpoint of mice follow-up to evaluate the side effects on mouse intestine. The 5-FU and anti-4-1BB concurrent but not sequential treatment caused severe diarrhea (n=7 in each group, t-test was performed between indicated groups, data were displayed as means ± SEMs, ****p<0.0001). This figure is reprinted with permission from Zhao et al., iScience, 2020. Meanwhile, the tumors treated by the sequential therapy had the lowest MDSC frequency and highest NK cell frequency (Figure 3.13F, H). PD-1 expression on CD8+ T cells and macrophages frequency were similar among all groups (Figure 3.13E, G). CD103+ DC frequency trended higher in the anti-4-1BB monotherapy group but was not statistically significant (Figure 3.13I). We repeated the same experiment in MC38 tumors (Figure 3.14) and obtained similar results as in CT26 tumors with most parameters tested. However, CD80 and CD86 expression levels in

MC38 tumors were not increased significantly by 5-FU and anti-4-1BB sequential treatments. This difference between MC38 and CT26 tumors indicates the tumor-dependent effects of the treatment. In CT26 tumors, we also evaluated the immunoregulatory effects of 5-FU and anti-PD-1 combination (Figure 3.15). Tumor-wide expression of PD-L1, CD86, and CD80 was increased in 5-FU and anti-PD-1 sequential treatment group. In addition, the frequencies of total T cells, proliferating CD8+ T cells, and NK cells were highest in tumors treated by 5-FU and anti-PD-1 sequential therapy (Figure 3.15). Notably, the frequency of MDSCs was decreased by 5-FU monotherapy and combination treatment (Figure 3.15). These findings showed the immunological impacts of different treatment strategies and reinforced that using 5-FU as an induction treatment and then anti-4-1BB or anti-PD-1 as maintenance treatments produces the most prominent and synergic effect in reversing the immunosuppressive tumor microenvironment.

Figure 3.12. Gating of the tumor-infiltrating immune cells.

(A) The major tumor-infiltrating immune cell populations were showed in the tSNE plots. (B) Manual gating of the major tumor-infiltrating immune cells. The alive cell population was first identified, and the immune cells (CD45+) were then gated. The gating of T-cell populations (CD45+CD3+CD8+ for CD8+ T cells, CD45+CD3+CD4+CD25+Foxp3+ for Tregs), NK cells (CD45+CD3-CD11b-CD11c-CD49b+), macrophages (CD45+CD3-CD11b+F4/80+), myeloid- derived suppressive cells (MDSCs, CD45+CD3-CD11b+Gr-1+), and CD103+ DCs (CD45+CD3- CD11b-CD11c+I-A/I-E+CD103+) were showed here. The same markers were used throughout the study to identify the immune cells. This figure is reprinted with permission from Zhao et al., iScience, 2020. 66

Figure 3.13 Tumor immunological response to 5-FU and anti-4-1BB treatments in CT26 tumors.

(A) ViSNE plot showed the single-cell level expression of PD-L1, Ki-67, CD80, and CD86 in the tumor tissue. PD-L1, Ki-67, CD80, and CD86 expression were quantified in whole tumor tissue. The 5-FU and anti-4-1BB sequential treatment significantly upregulated CD86 and CD80 expression in tumor tissues (n=3 in each group, data were displayed as means ± SEMs, t-test was performed between the indicated groups, n.s.: no significance, *p<0.05, ***p<0.001, MSI: mean signal intensity). (B-I) The tumor-infiltrating T cells frequency, CD8/Treg ratio, Ki-67+CD8+ T cells frequency, expression of PD-1 on CD8+ T cells, myeloid-derived suppressive cells (MDSCs) frequency, macrophages frequency, NK cells frequency, and CD103+ dendritic cells (DCs) 67 frequency were measured in tumors treated by different strategies (n=3 in each group, data were displayed as means ± SEMs, t-test was performed between the indicated groups, n.s.: no significance, *p<0.05, **p<0.01, MSI: mean signal intensity, Anti-4: Anti-4-1BB, 5-FU + anti-4 S.: 5-FU and anti-4-1BB sequential combination treatment, 5-FU + anti-4 C.: 5-FU and anti-4- 1BB concurrent combination treatment). This figure is reprinted with permission from Zhao et al., iScience, 2020.

Figure 3.14. Tumor immunological response to 5-FU and anti-4-1BB treatments in MC38 tumors.

(A) ViSNE plot showed single-cell level expression of PD-L1, Ki-67, CD80, and CD86 in the MC38 tumor tissue. The 5-FU and anti-4-1BB sequential treatment significantly upregulated CD80 while decreased Ki-67 expression in tumor tissues (n=4 in each group, t-test was performed between indicated groups, data were displayed as means ± SEMs, MSI: mean signal intensity, n.s.: no significance, *p<0.05, **p<0.01). (B-I) The tumor-infiltrating T-cell frequency, CD8/Treg ratio, and Ki-67+CD8+ T-cell frequency were higher in the sequential treatment than the concurrent treatment group. The myeloid-derived suppressive cells (MDSCs) were depleted in 5-FU treated groups. The 5-FU and anti-4-1BB sequential treatment were compared with the anti- 4-1BB monotherapy, 5-FU monotherapy, and 5-FU and anti-4-1BB concurrent treatment (n=4 in each group, t-test was performed between indicated groups, data were displayed as means ± SEMs, *p<0.05, **p<0.01, *** p<0.001, ****p<0.0001, MSI: mean signal intensity, n.s.: no significance, Anti-4: Anti-4-1BB, 5-FU + anti-4 S.: 5-FU and anti-4-1BB sequential combination treatment, 5-FU + anti-4 C.: 5-FU and anti-4-1BB concurrent combination treatment). This figure is reprinted with permission from Zhao et al., iScience, 2020.

Figure 3.15. Tumor immunological response to 5-FU and anti-PD-1 treatments in CT26 tumors.

(A-D) The 5-FU and anti-PD-1 sequential treatment significantly increased PD-L1, CD80, and CD86 expression in CT26 tumors (n=5 in each group, t-test was performed between indicated groups, data were displayed as means ± SEMs, MSI: mean signal intensity, n.s.: no significance, *p<0.05, **p<0.01, *** p<0.001, ****p<0.0001). (E-H) The 5-FU and anti-PD-1 sequential treatment significantly increased tumor-infiltrating T-cell frequency, Ki-67+CD8+ T-cell frequency, and NK cell frequency in tumor tissues, compared with the concurrent treatment. The myeloid-derived suppressive cells (MDSCs) were depleted in 5-FU treated groups (n=5 in each group, t-test was performed between indicated groups, data were displayed as means ± SEMs, *p<0.05, **p<0.01, *** p<0.001, ****p<0.0001, MSI: mean signal intensity, n.s.: no 69 significance, 5-FU + anti-PD-1 S.: 5-FU and anti-PD-1 sequential combination treatment, 5-FU + anti-PD-1 C.: 5-FU and anti-PD-1 concurrent combination treatment). This figure is reprinted with permission from Zhao et al., iScience, 2020.

3.3 Discussion Immunotherapies are mostly used as second- or third-line treatments for treatment- refractory tumors. However, studies that investigate the impact of different clinical conditions and combination strategies on tumor immunotherapy are limited. Here, we comprehensively profiled the impacts of tumor-draining lymph nodes (TdLNs) resection and different chemotherapy combination schedules on ICBT responses.

Surgery has been a dominant strategy for several decades to prevent, diagnose, stage, and treat cancers. Radical surgery—a procedure that removes blood supply to the tumor, lymph nodes, and sometimes adjacent structures—is routinely performed in many cancers such as colorectal cancer, breast cancer, and lung cancer. Early-stage cancer patients have excellent disease control with surgery alone, yet advanced diseases require more comprehensive treatments, including chemotherapies, oncogenic pathway targeted therapies, and immunotherapies. Currently, most immunotherapies are used as adjuvant treatments (given after surgeries). TdLNs are the primary lymphatic organs where antitumor immune responses are initiated148,149,156,168,169. In mouse models with resected TdLNs before tumor cell inoculation, we observed that removal of TdLNs significantly accelerated tumor growth and compromised response to immunotherapy. These data uncovered a key role for TdLNs in preventing cancer cells from evading antitumor immunity at early stages. Mechanistically, TdLNs resection in early-stage disease led to inadequate antitumor immune simulation, featured by a low frequency of tumor antigen-specific T cells in lymphatic organs. Our observations were in line with previous studies, highlighting the significance of TdLNs in initiating antitumor immunity and regulating immunotherapy response in early-stage disease170.

Recently, Fransen et al. reported that TdLNs are determining factors of PD-1/PD-L1 immune checkpoint therapies in early-stage tumor models170. However, whether the TdLNs are critical for immunotherapy response in recurrent tumor models, which represent a major clinical issue, had not been addressed. In our study, we established a model to mimic tumor recurrence from residual tumor lesions after primary tumor and TdLN resection. We first thoroughly resected the primary tumors either with or without TdLN resection and confirmed a clean surgical

70 margin. We then inoculated tumor cells in situ to induce a secondary tumor. This method allows all secondary tumors to have a relatively similar baseline volume and growth dynamic before any treatment. We also allow the localized secondary tumors to connect with systemic circulation and establish a tumor microenvironment before treatment was initiated. Our well-designed model provided a platform for an unbiased evaluation of treatment efficacy in residual disease after primary tumor resection.

With our model, we found that resection of TdLNs in advanced tumors did not influence localized secondary tumor immunity and response to immunotherapies (anti-PD-1 and anti-4- 1BB). Furthermore, we investigated the factors that determine the significance of TdLNs in antitumor immunity and immunotherapeutic response. Previous findings indicated that the bidirectional crosstalk between tumor cells and TdLNs allowed remodeling of each other during tumor progression148,149,153,154,156. Immunosuppressive factors derived from tumors such as TGF- beta, can drain to TdLNs and induce an immunosuppressive microenvironment153,171. We tested the hypothesis that the antitumor function of TdLNs is impaired in advanced tumor models. We compared the immune responses in naïve LNs, TdLNs of early-stage and advanced tumors and demonstrated a trend between potent immunosuppression in TdLNs and tumor progression. Although the TdLNs eventually became immunotolerant, the distribution of tumor antigen- specific T cells is extensive in lymphatic tissues in advanced tumors. Resection of TdLNs did not significantly reduce the population of tumor-antigen specific T cells that respond to immunotherapies. Our data corroborate with previous reports showing strong immunosuppression development in TdLNs of human cancers155,172. This explains why the resection of TdLNs may not influence the antitumor immunity in late-stage tumor models. Finally, it is also important to understand that the resected TdLNs in our experimental models might have developed immunotolerance. However, since humans have more TdLNs than the mouse model, immunoreactive TdLNs do exist in certain circumstances and might influence immunotherapy response150,173. Therefore, it will be critical to evaluate the functional status of TdLNs in humans before extending our conclusions to human cancers.

Systemic therapies, such as chemotherapies, are used to treat primary tumors, eradicate micrometastatic disease, or stabilize the disease in widespread incurable conditions1. Chemotherapies have the advantages of being fast-acting and effective; thus, they are widely administered as the primary treatment for combinational strategies1. Combinations of chemotherapies with immunotherapies are extensively discussed and currently tested in 71 preclinical models and clinical trials75,167,174,175. Comprehensive studies have revealed the mechanisms by which chemotherapy can promote antitumor immunity by induction of immunogenic cell death and disruption of tumor microenvironment components that are used to evade the immune response165,176-179. However, cancer chemotherapies are also considered immunosuppressive due to their cytotoxic effects on immune cells. Therefore, we speculated that the same chemotherapy might have different impacts on anti-tumor immunity, either stimulatory or inhibitory, depending on the specific combination schedules. We used 5-FU, a common chemotherapeutic agent, as a representative agent, to study the influences of different chemotherapeutic and immunotherapeutic combination strategies on the anti-tumor immune response.

Through extensive study of 5-FU induced immune responses, we revealed both systemic immunosuppressive effects and immune-stimulating effects in the tumor microenvironment. 5-FU treatment upregulated CD80 expression and depleted MDSCs. CD80 is a protein found on antigen-presenting cells as well as tumor cells and belongs to the B7 family; it provides a costimulatory signal necessary for activating T cells and natural killer cells180-183. Thus, the upregulation of CD80 in tumor tissue induced by 5-FU treatment will potentially lead to increased tumor visibility by T cells. MDSCs are a heterogeneous population of cells that potently suppress T-cell responses184,185. By depleting MDSCs in tumor tissue, 5-FU treatment may potentiate antitumor immunity by eliminating the negative regulations. These findings are also supported by a previous report186. In addition to the immunogenic effects, we also observed that 5-FU treatment suppressed the T-cell population in the tumor microenvironment. Thus, avoiding the immunosuppressive effects and preserving the immunogenic effects of 5-FU treatment will determine the response of 5-FU and immunotherapy combinations.

In our study, the administration of anti-4-1BB or anti-PD-1 after 5-FU treatment significantly improved tumor responses. In this combination strategy, anti-4-1BB or anti-PD-1 selectively boosted response of T-cell and NK cells while the 5-FU treatment increased tumor visibility and suppressed MDSCs. However, when anti-4-1BB or anti-PD-1 was added to the repetitive 5-FU treatment, less synergistic effects were observed. Our data highlighted the importance of determining the best schedule for designing a successful chemo-immunotherapy combination. In addition to timing, dosing is another potential factor that affects the chemotherapy-induced immune response. Low dose chemotherapies have shown special

72 immunoregulatory effects in tumor models187,188. Further studies are needed to test different chemotherapy doses on the chemo-immunotherapy combination.

In conclusion, our research investigated how traditional cancer treatments will affect novel immunotherapies in clinically relevant tumor models. Our findings indicate that TdLN resection can have an adverse impact on anti-tumor immunity, but only in early-stage tumor models. In advanced tumor models, resection of immunotolerant TdLNs during primary tumor surgery does not significantly alter anti-tumor immunity or immunotherapy response in secondary tumors that mimic localized tumor recurrence. Meanwhile, minimizing the immunosuppression and strengthening the immunogenic effects of traditional cancer therapies are critical for immunotherapy induced durable cancer remission. Specifically, sequential cytotoxic chemotherapy followed by immunotherapy produced a significantly higher degree of anti-tumor response compared to concurrent combination therapy. These findings highlight the need to test immunotherapies in tumor models that more closely mimic different clinical conditions and establish references for designing clinical trials to determine the most effective cancer immunotherapy strategies.

3.4 Methods and Material 3.4.1 Cell cultures Murine CRC cell lines CT26 (purchased from American Type Culture Collection (ATCC)) and MC38 (gift from Dr. Nicholas Haining) were used for the study and were authenticated by STR profiling. CT26 cells were maintained in complete RPMI-1640 medium (GIBCO BRL), supplemented with 10% heat-inactivated FBS (Thermo Fisher Scientific), 100 IU/mL penicillin, and 100 μg/mL streptomycin (Invitrogen Life Technologies). MC38 cells were cultured in the complete DMEM medium (GIBCO BRL) with the same supplements as the RPMI 1640 medium. All cells were routinely authenticated and tested for mycoplasma.

3.4.2 Mice Wild type BALB/c mice (6-8 weeks old, Jackson Laboratory) and C57BL/6 mice (6-8 weeks old, Charles River Laboratories) were used for animal studies. All mice were kept in a specific pathogen-free facility with fully autoclaved cages to minimize non-tumor specific immune activation. Animal studies were approved by the institutional animal care and use committee (IACUC). All mice are female. We do not expect any influence of gender on our study aims. 73

3.4.3 Subcutaneous tumor induction For the subcutaneous syngeneic model, the cells were harvested at low passages, washed, and resuspended in Matrigel matrix (Corning Inc.) before injection. Mice were shaved right before injection. CT26 (2×105 cells/injection) or MC38 (5×105 cells/injection) cells were inoculated subcutaneously into the right hind-flank of 6 to 8-week-old female BALB/c or C57BL/6 mice. The same amount of tumor cells was used for the re-challenge experiment in Figure 1. Tumor length and width were measured every three to seven days, and the volume was calculated according to the formula (length × width2)/2. Mice were divided into different experimental groups at random when tumors reached a specific size.

3.4.4 Identification of major tumor-draining lymph nodes To identify the major tumor-draining lymph nodes (TdLNs), we injected 50μl 1% Evans blue (Sigma-Aldrich) or 50μl 1% Alexa Fluor® 488 dye (Thermo Fisher Scientific) into the subcutaneous tumor (~400-500mm3) at the right hind flank. The left and right inguinal LNs, axillary LNs, brachial LNs, popliteal LNs, and mesentery LNs were taken at 10 min, 30 min, and 60 min post Evan blue injection. For the fluorescence-labeled group, we collected LNs at 0.5h, 3h, 24h, and 48h post-injection. The intact LNs were visually examined for Evans blue staining. LNs, spleen, and tumor tissues were ground and meshed for single-cell suspension, which was measured by flow cytometry for Alexa Fluor® 488 dye signal. To evaluate the physical change of LNs and spleen during tumor development, we weighted LNs and spleen from naïve mice and mice with different sizes of the tumor (100-200mm3, 500-700mm3, or 1200-1500mm3).

3.4.5 Subcutaneous tumor and TdLNs resection Primary tumors were resected when they reached the indicated volume, as shown in the experimental schematics in each figure. Tumor-bearing mice were anesthetized with Ketamine (100 mg/kg) and Xylazine (10 mg/kg) by intraperitoneal injection. To minimize animal pain, we administrated Buprenorphine (slow-releasing, 2 mg/kg) subcutaneously 2 hours before anesthesia. Mice were prepared by removing hair from the skin region over the tumor. We prepared the skin by wiping with iodine prep pads and then alcohol prep pads. Resections were performed by elliptical incisions, 5mm left to the subcutaneous tumors. With iris scissors, we separated the capsule of subcutaneous tumors from the surrounding connective tissue to isolate and resect intact tumors. Once tumors were removed from the adjacent fascia, the incisions were sutured with 5/0 vicryl ties (polyglactin 910, Ethicon). For the TdLNs resection, the TdLNs were located based on the superficial anatomic landmark points. The mice were prepared, as mentioned above. A 5-10mm incision was made, and TdLNs were removed. Then the skin was sutured with 74

5/0 vicryl ties. For tumor rechallenge, 1 day after surgery, we inoculated the secondary tumor (CT26: 5×105 cells/injection, MC38: 1×106 cells/injection) to the surgical site to mimic tumor recurrence.

3.4.6 RT-qPCR We used the murine leukemia virus envelope gp70 as a biomarker of tumor burden. Biopsies were collected from normal mouse skin, tumor tissue, and surgical margin after tumor resection. The mirVana microRNA (miRNA) Isolation Kit (Thermo Fisher Scientific) was used to extract total RNA from these biopsies. 500 ng of total RNA was used for establishing the cDNA library with the QuantiTect Reverse Transcription Kit (Qiagen). We used the LightCycler 480 Instrument (Roche Life Science) to measure 18S ribosomal RNA (rRNA) and gp70 expression. Primers used: 18S rRNA forward primer: GTTGGTTTTCGGAACTGAGG, 18S rRNA reverse primer: AGTCGGCATCGTTTATGGTC, gp70 forward primer: AAAGTGACACATGCCCACAA, gp70 reverse primer: CCCCAAGAGGCACAATAGAA189.

3.4.7 Flow cytometry Flow cytometry was used to measure tumor tissue immune infiltration, tumor antigen- specific T cells, and immune cell functions. Harvested tumor tissues were chopped into small pieces (around 3mm x 3mm) and then digested in a solution of collagenase IV (1 mg/ml) and deoxyribonuclease (DNase, 50 units/ml) at 37°C for 1 hr with shaking. The digested tissue was then meshed and filtered through a 70 μm cell strainer. The cell suspension was centrifuged and resuspended in red blood cell lysis buffer for 15 minutes at room temperature for eliminating red blood cells. Another centrifugation was performed to get the cell pellet for staining. For the lymphatic organs, we directly meshed the tissue and filtered through a 40 μm cell strainer to get the single-cell suspension, followed by red blood cell elimination.

Following the tissue sample preparation, cells were stained with the fixable cell viability dye and then cell surface marker antibodies for a 15 min incubation at 4°C. Next, cells were fixed and permeabilized for intracellular staining for a 30 min incubation at RT. The cells were finally stained with intracellular markers (30 min at RT) and analyzed on a BD FACS-CANTO instrument (BD Biosciences). To analyze the tumor antigen-specific T cells, we performed H-2Ld MuLV gp70-SPSYVYHQF APC conjugated tetramer (MBL International) staining by following the manufacturer's instruction, before antibody staining. The influenza hemagglutinin- IYSTVASSL APC conjugated tetramer (MBL International), which should only stain a very minimal population of T cells in mice without influenza hemagglutinin stimulation, was used as a 75 negative control for ruling out false positive in the tetramer staining and setting up the gate for gp70 tetramer. Lymphatic tissues from naïve mice were also used as negative controls. According to the manufacture’s instruction and our preliminary experiment optimization, we used the anti- CD8 (clone KT15) antibody (MBL International) to further reduce the false-positive rate of the tetramer staining. All antibodies for flow cytometry were purchased from Biolegend and summarized in supplementary materials. Data were analyzed using FlowJo software (Tree Star, Inc.).

3.4.8 Mass cytometry Details on antibodies and reagents used are listed in supplementary table 2. We purchased the prelabeled antibodies from Fluidigm Corporation and unlabeled antibodies (MaxPar® Ready purified) from Biolegend. Conjugation of the purified antibodies with metal tags was performed by using the MaxPar X8 antibody labeling kit (Fluidigm Corporation) according to the manufacturer's instructions. The metal tagged antibodies were then validated and titrated in positive control and negative control samples.

Tumor samples were collected and digested using standard flow cytometry procedures. A total of 3 million single cells were used for each mass cytometry staining. In brief, the single-cell pellets were first incubated with Cell-ID Cisplatin with a final concentration of 5 µM for 5 min at RT to identify dead cells. Cells were then washed and blocked by Fc-receptor blocking solution. Cell membrane staining was then performed with metal-conjugated antibodies for 30 min at RT. After staining, cells were fixed and permeabilized. The intracellular staining antibodies were then added and incubated for 45 min at RT. Finally, cells were labeled with 1 ml 1,000× diluted 125 µM Cell-ID intercalator-Ir to stain all cells in MaxPar Fix and Perm Buffer overnight at 4 °C. EQ Four Element Calibration Beads with the reference EQ passport P13H2302 were added to each staining tube right before data acquisition by a CyTOF 2 mass cytometer. The mass cytometry data were then normalized and exported for gating on alive single cells, which were then imported to the Cytobank software. A t-SNE analysis was performed with default parameters (perplexity, 30; iterations, 1,000) on all cell types in tumor samples.

3.4.9 Mouse IFN-γ enzyme-linked immunosorbent assays Mouse naïve lymph nodes and TdLNs were collected, weighed, and ground in 100 μl RIPA lysis and extraction buffer. After the tissues were lysed, the total protein was used for enzyme-linked immunosorbent assay (ELISA, Affymetrix) to detect mouse IFNγ, by following the manufacturer’s protocol. 76

3.4.10 Histology Mouse naïve lymph nodes, TdLNs, and non-tumor-draining lymph nodes (NdLNs) were collected and fixed in 10% formalin for 24 hr. Tissues were embedded in paraffin and cut for hematoxylin and eosin (H&E) staining. The whole tissue sections were scanned and analyzed for potential metastatic tumor cells.

3.4.11 Mouse treatments and T-cell depletion We tested the effects of 5-FU treatment and 5-FU and anti-4-1BB combination treatment on T-cell depletion in vivo. Intraperitoneal administration of anti-CD3 treatment (clone: 17A2, BioXcell, 5 mg/kg every 3 days) was given to induce T-cell depleted mice. One dose of 5-FU (150 mg/kg) or 5-FU (150 mg/kg) and anti-4-1BB (5 mg/kg) combination treatment was given intraperitoneally in naïve mice. Mice were sampled on days 2, 4, 7, and 9 after treatment for quantifying T cells in lymph nodes, spleen, bone marrow, and blood circulation.

For treatment propose, mice were treated with IgG (5 mg/kg as an anti-4-1BB control, 10 mg/kg as an anti-PD-1 control), 5-FU (150 mg/kg), anti-4-1BB agonist (5 mg/kg，clone: 3H3), or anti-PD-1 (10 mg/kg, clone: RMP1-14) for treatment purpose. For the 5-FU monotherapy, one dose of 5-FU was given every 12 days to minimize the severe side effects. For anti-4-1BB and IgG monotherapy, mice were treated every 3 days. For the 5-FU and anti-4-1BB sequential treatment, anti-4-1BB treatment started 9 days after one dose 5-FU treatment and continued as 3 days per injection after that. For the 5-FU and anti-4-1BB concurrent treatment, we added the anti-4-1BB cycle to the 5-FU cycle. The anti-PD-1 was used as the same as the anti-4-1BB cycle. All treatments were given intraperitoneally and continued until the endpoint of study design. The treatment starting points and endpoints varied in different experiments for different purposes and were shown in the individual figure or figure legend.

3.4.12 5-FU toxicity evaluation We recorded animal body weight and diarrhea scores after treatments. Mice were weighed on day 12, 24, and 32 after treatment. The diarrhea score was assessed at the endpoint of each treatment by using a 4-point scoring system: 0=normal stool; 1=slight diarrhea (soft formed stool without perianal staining of the coat); 2=moderate diarrhea (unformed stool with moderate perianal staining of the coat); and 3=severe diarrhea (watery stool with severe perianal staining of the coat)190.

3.4.13 Statistical analysis All statistical analyses and graphing were performed using GraphPad Prism software (Version 6). Data were displayed as means ± SEMs. For comparison of two groups of quantitative data, paired or unpaired Student’s t-test was performed. When applicable, one-way analysis of variance (ANOVA) was utilized for multiple groups’ comparison, followed by post hoc (Tukey's) multiple comparisons test. Kaplan-Meier curves were plotted to visualize mouse survival, and log-rank tests were used to compare survival outcomes between subgroups. A two- tail P value of less than 0.05 was considered statistically significant.

Table 3.2. Key material table for chapter 3.

Reagent for immune assays Clone Vendor Identifier

Anti-mouse CD3-FITC 17A2 BioLegend 100204

Anti-mouse CD28-PE 37.51 BioLegend 102106

Anti-mouse PD-1-PerCP/Cy5.5 29F.1A12 BioLegend 135208

Anti-mouse CD62L-PE/Cy7 MEL-14 BioLegend 104418

Anti-mouse CD8a-APC/Cy7 53-6.7 BioLegend 100714

Anti-mouse LAG-3-BV421 C9B7W BioLegend 125221

Anti-mouse/human CD44-PE IM7 BioLegend 103008

Anti-mouse CD19-Pacific Blue 6D5 BioLegend 115523

Anti-mouse CD4-BV510 GK1.5 BioLegend 100449

Anti-mouse CD86-PE GL-1 BioLegend 105007

Anti-mouse F4/80-PE/Cy5 BM8 BioLegend 123111

Anti-mouse CD80-PE/Cy7 16-10A1 BioLegend 104734

Anti-mouse/human CD11b- M1/70 BioLegend 101212 APC

Anti-mouse I-A/I-E-APC/Cy7 M5/114.15.2 BioLegend 107628

Anti-mouse CD11c-BV510 N418 BioLegend 117338

Anti-mouse CD45-Pacific Blue 30-F11 BioLegend 103126

Anti-mouse CD45-FITC 30-F11 BioLegend 103108

Anti-mouse CD3- 17A2 BioLegend 100218 PerCP/Cyanine5.5

Anti-mouse CD3ε-PE/Cy7 145-2C11 BioLegend 100320

Anti-mouse CD103-Pacific Blue 2E7 BioLegend 121418

Anti-mouse Gr-1-PE/Cy7 RB6-8C5 BioLegend 108416

Anti-mouse CD45-BV510 30-F11 BioLegend 103138

Anti-mouse Podoplanin- 8.1.1 BioLegend 127418 APC/Cy7

Anti-mouse CD31-Pacific Blue 390 BioLegend 102422

Anti-mouse CD45-89Y 30-F11 Fluidigm 3089005B

Anti-mouse Ly-6G-41Pr 1A8 Fluidigm 3141008B

Anti-mouse CD11c-142Nd N418 Fluidigm 3142003B

Anti-mouse CD4-145Nd RM4-5 Fluidigm 3145002B

Anti-mouse F4/80-146Nd BM8 Fluidigm 3146008B

Anti-mouse Gr-1-147Sm RB6-8C5 BioLegend/Fluidigm 108449/201147B

Anti-mouse CD11b-148Nd M1/70 Fluidigm 3148003B

Anti-mouse CD19-149Sm 6D5 Fluidigm 3149002B

Anti-mouse CD25-150Nd 3C7 Fluidigm 3150002B

Anti-mouse CD28-151Eu 37.51 Fluidigm 3151005B

Anti-mouse CD3e-152Sm 145-2C11 Fluidigm 3152004B

Anti-mouse CD274-153Eu 10F.9G2 Fluidigm 3153016B

Anti-mouse CD152-154Sm UC10-4B9 Fluidigm 3154008B

Anti-mouse CD279-155Gd RMP1-30 BioLegend/Fluidigm 109113/201155A

Anti-mouse CD335-156Gd 29A1.4 BioLegend/Fluidigm 137625/201156B

Anti-mouse Foxp3-158Gd FJK-16s Fluidigm 3158003A

Anti-mouse RORgt-B2D-159Tb B2D Fluidigm 3159019B

Anti-mouse CD62L-160Gd MEL-14 Fluidigm 3160008B

Anti-mouse Ki-67-161Dy B56 Fluidigm 3161007B

Anti-mouse Ly-6C-162Dy HK1.4 Fluidigm 3162014B

Anti-mouse CD197-164Dy 4B12 Fluidigm 3164013A

Anti-mouse IFNg-165Ho XMG1.2 Fluidigm 3165003B

Anti-mouse IL-4-166Er 11B11 Fluidigm 3166003B

Anti-mouse CD103-167Er 2E7 BioLegend/Fluidigm 121402/ 201167B

Anti-mouse CD8a-168Er 53-6.7 Fluidigm 3168003B

Anti-mouse CD49b-170Er HMa2 Fluidigm 3170008B

Anti-mouse CD80-171Yb 16-10A1 Fluidigm 3171008B

Anti-mouse CD86-172Yb GL1 Fluidigm 3172016B

Anti-mouse Granzyme B-173Yb GB11 Fluidigm 3173006B

Anti-mouse CD127-174Yb A7R34 Fluidigm 3174013B

Anti-mouse CD44-176Yb IM7 BioLegend/Fluidigm 103051/201176B

Anti-mouse I-A/I-E-209Bi M5/114.15.2 Fluidigm 3209006B

Cell-ID™ Intercalator-Ir NA Fluidigm 201192B

Cell-ID™ Cisplatin NA Fluidigm 201064

H-2Ld MuLV gp70 Tetramer- NA MBL International TB-M521-2 APC

IFN gamma Mouse ELISA Kit NA Thermo Fisher BMS606 Scientific

Zombie Violet™ Fixable NA BioLegend 423113 Viability Kit

Zombie Aqua™ Fixable NA BioLegend 423101 Viability Kit

Zombie Green™ Fixable NA BioLegend 423111 Viability Kit

MuLV gp70 Tetramer-APC NA MBL International TB-M521-2

Influenza HA Tetramer-APC NA MBL International TS-M520-2

Anti-mouse CD8-FITC KT15 MBL International D271-4

Drugs or antibodies for Clone Vendor Identifier treatment

Anti-mouse 4-1BB (CD137) 3H3 BioXcell BE0239

Anti-mouse PD-1 RMP1-14 BioXcell BP0146

Anti-mouse CD3 17A2 BioXcell BE0002

5-Fluorouracil NA Intas DB00544 Pharmaceuticals

Mouse Age Vendor Identifier

BALB/cJ 6-8 weeks The Jackson 000651 Laboratory

C57BL/6 6-8 weeks Charles River 027

3.5 Publication This chapter has been modified (with permission) from the published articles:

Zhao X, Kassaye B, Wangmo D, Lou E, Subramanian S. Chemotherapy but not the tumor- draining lymph nodes determine the immunotherapy response in secondary tumors. iSCIENCE. 2020 May 22; 23(5): 101056

(https://www.ncbi.nlm.nih.gov/pubmed/32344378)

4. Chapter 4

Tumor Secreted Extracellular Vesicles Regulate T-cell Costimulation and Can Be Manipulated to Induce Tumor-Specific T-cell Responses in Colorectal Cancer

4.1 Introduction Therapies targeting immune checkpoint pathways, including the cytotoxic T lymphocyte- associated protein (CTLA-4) and the programmed cell death-1 (PD-1)/PD-1 ligand-1 (PD-L1), have revolutionized cancer treatment191-193. For colorectal cancer (CRC), immune checkpoint blockade therapy (ICBT) is effective in 30-60% of the microsatellite instable-High (MSI-H) subtype38. Unfortunately, most CRC patients (>85%) have microsatellite stable (MSS) tumors39,40, that do not respond to ICBT. The MSI-H tumors are generally associated with a higher mutational load and are considered immunogenic194,195. Currently, the MSI/MSS phenotype is considered as a biomarker for ICBT response9,45. However, this MSI/MSS based stratification alone cannot adequately explain the observed difference in treatment response since immune cell infiltration is also observed in a large proportion of MSS tumors40,106. A mechanistic understanding of why the existing tumor-infiltrating immune cells are not functionally rescued by ICBT is critical for improving outcomes in both MSS- and MSI- CRC patients.

Intercellular communications between tumor cells and immune cells in the tumor microenvironment (TME) affect the antitumor immune response and the tumor response to ICBT. One way tumor cells can affect surrounding cells is by secreting extracellular vesicles (EVs)98, and tumor cell-derived EVs (TEVs) can be either immunogenic or immunosuppressive depending on their cargo. For example, PD-L1 in TEVs have strong immunosuppressive effects93,95, while the tumor antigens and other immune-stimulating factors carried in EVs can induce immune responses101,196,197. Investigation of additional tumor intrinsic mechanisms and immunosuppressive components in EVs that incite ICBT failures are highly warranted.

It has been demonstrated that CD28-CD80/86 costimulatory signals are critical for ICBT response18,19. For example, PD-1 recruits the Shp2 phosphatase, which preferentially inhibits CD28 signaling18. This suggests that anti-PD-1 efficacy is, in part, due to rescuing CD28- CD80/86 signaling. Another study demonstrated that CD28-CD80/86 signaling in tumor- infiltrating CD8+ T cells is essential for anti-PD-1 response19. Markedly, CD80 expression on the antigen-presenting cells (APCs) was recognized as a novel mechanism that restricts PD-1 signaling on interacting T cells by forming cis-heterodimers with PD-L1198. This CD80-PD-L1 interaction inhibits CD80 signaling through CTLA-4, thus attenuating the immunosuppression mediated by CTLA-4199. These reports led us to ask if other mechanisms, particularly tumor- intrinsic ones, led to CD28-CD80/86 signal dysregulation. We found this signaling axis was disrupted in human CRC patients by TEVs. Additionally, we investigated if depleting the 84 potential CD28-CD80/86 suppressive factors in TEVs would enhance their immunogenic effects, thus improving antitumor immunity.

4.2 Results 4.2.1 Costimulatory molecules CD28, CD80, and CD86 are downregulated on human CRC infiltrating immune cells We determined the overall adaptive antitumor immune response in human CRC tissues (23 MSI subtype and 45 MSS subtype) by examining the T-cell infiltration in the TME. Tumors with well- and poorly- infiltrated T cells were observed in both microsatellite instable (MSI) and microsatellite stable (MSS) subtypes (Figure 4.1A). Overall, the MSI-CRC had significantly higher numbers of tumor-infiltrating T cells compared to MSS-CRC (Figure 4.1A). This observation corroborated our immune signature analysis of the TCGA-CRC dataset (Figure 4.2A) and previous reports in CRC40,44,200.

Next, we measured the immune checkpoint genes (ICG) expression in CRC. Genes expression analysis of a set of 50 ICGs, including CD28, CD80, and CD86, in the TCGA-CRC dataset, stratified tumors into ICG- high and -low expression groups. Both MSI- and MSS-CRC tumors were seen in each group (Figure 4.1B). At the protein level, we measured CD28-CD80/86 costimulatory molecules on human CRC infiltrating T cells and dendritic cells (DCs). Flow cytometry analysis of a cohort of 10 primary human CRC fresh tissues showed that CD4+ T cells generally had higher levels of CD28 than CD8+ T cells (Figures 4.1C & 4.2B). However, variable levels of CD28 expression were observed on both CD4+ and CD8+ T cells (Figure 4.1C). Likewise, tumor-infiltrating DCs had variable levels of CD80/86 (Figure 4.1C). These data, though, are composites of TILs and DCs throughout the entirety of the tumor and do not account for potential spatial differences. To specifically study the CD28-CD80/86 costimulatory molecules on T cells and DCs localized in the center of tumors, we collected 68 archived human CRC tissues (23 MSI subtype and 45 MSS subtype) for immunofluorescence analysis. Only T cells and DCs that intercalated within or between the epithelial component of the CRC/normal colon tissue were selected for scoring (Figures 4.1D & 4.2C-D). No/low expression of CD28 (score<0.5, Figure 4.2C) on T cells was observed in ~60% of all cases in both MSI- and MSS- CRC tumors (Figure 4.1D). Similarly, DCs located in the central tumor showed no/low expression of CD80/86 (score<0.5, Figure 4.2C) in ~50%/~33% of all cases (score<0.5, Figure 4.1D). Taken together, our data showed that the expression of CD28 and CD80/86 is highly variable and can be weak or absent on tumor- infiltrating T cells and DCs in both MSI- and MSS-CRC tumors.

Figure 4.1. Expression of costimulatory molecules CD28, CD80, and CD86 on tumor- infiltrating immune cells in human CRC

(A) Multiplex immunofluorescence staining of tumor-infiltrating T cells (CD3, CD8) and myeloid cells (CD11b) in human CRC tissue samples. (24 MSI cases, 47 MSS cases, Mean±SEM, t-test, ****p<0.0001) (B) Immune checkpoints expression profile in the TCGA CRC dataset. Hierarchical clustering was applied to CRC patients expressing different levels of immune checkpoint genes (list below). The top two clusters were selected to define patients with high or low immune checkpoint gene expression. (C) Flow cytometry analysis of CD28, CD80, and CD86 on immune cells in fresh human CRC tissues. (Mean±SEM, n=10, t-test, *p<0.05) (D) Multiplex immunofluorescence staining of CD28 on T cells and CD80/86 on DCs in a human CRC cohort (23 MSI cases, 45 MSS cases). Only T cells and DCs in the tumor center were scored. Magnification 400X. (E) Impact of CD28 and CD80/86 expression on ICBT (anti-CTLA- 4 + anti-PD-1) in subcutaneous tumors. ICBT restricted early-stage tumor development in WT mice. However, Cd28-/- mice and Cd80/86-/- mice were resistant to ICBT. Antibody blocking of CD80 and CD86 induced tumor resistance to ICBT. (Individual data for each case, n=5, t-test, **p<0.01, ****p<0.0001, Immune checkpoint gene list: CD274, PDCD1LG2, PDCD1, CD80, CD86, CD28, CTLA4, ICOS, ICOSLG, CD276, VTCN1, C10orf54, HHLA2, TMIGD2, Tnfrsf9, Tnfsf9, TNFRSF14, BTLA, CD160, TNFSF14, LAG3, TNFRSF4, TNFSF4, CD70, CD27,

CD40, CD40LG, TNFRSF18, TNFSF18, PVR, CD226, TIGIT, CD96, CD48, CD244, HAVCR2, LGALS9, ADORA1, ADORA2A, IDO1, TDO2, CEACAM1, CD47, SIRPA, BTN2A1, CD209, CD200R1, CD200R1, TNFSF15, TNFRSF25) See also Figure 4.2.

4.2.2 CD28-CD80/86 costimulatory pathway is necessary for ICBT response Having observed weak/no expression of CD28-CD80/86 on tumor-infiltrating immune cells in human CRC, we tested whether loss of CD28-CD80/86 expression affected tumor response to ICBT in a subcutaneous mouse model of CRC. Our data showed that a combination of anti-PD-1 and anti-CTLA-4 could significantly suppress the growth of early-stage (100- 200mm3) subcutaneous tumors (MC38 mouse CRC cell line with MSI phenotype) (Figure 4.1E). Single drug treatment with either anti-PD-1 or anti-CTLA-4 was found to be less effective than combination treatments (data not shown). However, in both Cd28-/- and Cd80/86-/- mice, the combination treatment did not suppress growth (Figure 4.1E & 4.2E). In human CRC, loss of CD28 and CD80/86 expression occurs in response to a tumor-dependent mechanism, rather than a deficient genetic mechanism. To model this, we administered anti-CD80 and anti-CD86 blocking antibodies to wild type mice bearing early-stage tumors to mimic acquired deficiency (depletion after tumor growth). Notably, blocking CD80 after the initial tumor development abolished the efficacy of ICBT (Figure 4.1E & 4.2E). Blocking CD86 showed a similar but weaker effect than the CD80 (Figure 4.1E & 4.2E). Reanalysis of published data201 indicated that tumors responding to immune checkpoint inhibitors have higher levels of CD28 and CD80/86 expression (Figure 4.2F). These results suggested that adequate CD28 and CD80/86 expression were necessary for effective ICBT, and these findings are consistent with previous studies18,19.

Figure 4.2 Supplementary data to Figure 4.1.

(A) The total immune infiltration and proportion of immune cells were estimated in CRC, kidney renal clear cell carcinoma (KIRC), lung adenocarcinoma (LUAD), lung squamous cell carcinoma 88

(LUSC), and skin cutaneous melanoma (SKCM), through TCGA database analysis (C: control, normal tissue, T: tumor tissue). (B) Representative gating strategy used to identify T-cells and dendritic cells (DCs) in human tissues. (C) Representative photos showed the scoring strategy to measure the expression of CD28 on CD3+ cells, and CD80 and CD86 on CD11c+ cells. The final score of each case is calculated as the mean score of five independent 400X fields. (D) Representative staining of CD28 on CD3+ cells, and CD80 and CD86 on CD11c+ cells in human normal colon tissues. (E) The individual tumor growth curves of figure 4.1E. (F) Co-stimulatory molecules expression profiling was analyzed in non–small cell lung cancer, head and neck squamous cell cancer, and melanoma patients after anti-PD-1 treatment. The heatmap showed a higher trend of CD28 and CD80/86 expression in patients with complete response (CR), partial response (PR), and stable disease (SD) than in patients with progressive disease (PD). 4.2.3 MicroRNA-424 is overexpressed in human CRC and inhibits CD28 and CD80 expression MicroRNAs (miRNAs) are small non-coding RNAs that post-transcriptionally regulate gene expression202. To determine if miRNAs regulate CD28-CD80/86 costimulatory molecules in human CRC, we first identified miRNAs that target CD28 and CD80/86 by target prediction (Figure 4.3A). TCGA-CRC data revealed that miR-424 is upregulated in CRC but downregulated in normal colon tissues (Figure 4.3A-B). By in situ hybridization (ISH), we validated that miR- 424 is expressed in cancer tissue, but not in normal colon tissue (Figure 4.3C and Figure 4.4A). In another cohort of 21 human CRC and patient-matched normal tissues, we quantified miR-424 expression by qRT-PCR (Figure 4.3D). Fourteen out of 21 CRC cases showed at least 2-fold upregulation of miR-424 in contrast to corresponding normal colon tissues (Figure 4.3D). Hence, miR-424, which potentially inhibits CD28 and CD80 expression, appears upregulated in 2/3 of human CRCs.

We next sought to confirm the inhibitory effects of miR-424 on CD28 and CD80. In dual-luciferase reporter vectors, we inserted the 3'UTR sequences of CD28 or CD80 (Figures 4.3E and 4.4B). Reporter assays showed that miR-424 significantly reduced the luciferase signal (Figures 4.3E and 4.4B). In the reciprocal experiments, site-directed mutations on the 3'UTRs nullified the effects of miR-424 (Figure 4.3E and 4.4B). We then isolated primary human T cells and DCs that endogenously expressed CD28 or CD80 from healthy donors' blood and transfected them with miR-424 or scrambled miRs (Figure 4.3F). Again, miR-424 showed a strong inhibitory effect on CD28 and CD80 protein expression (Figure 4.3F). These data confirm that miR-424 binds to 3'UTRs of CD28 and CD80 and suppress their expression on T cells and DCs, respectively.

Subsequently, we extended our investigations to mouse CRC models established by subcutaneous injections of CT26 (MSS phenotype) and MC38 (MSI phenotype) mouse CRC cell 89 lines. We assessed the immune cell landscape and miR-424 expression changes during different stages of tumor growth (100-200mm3, 500-700mm3, and 1000-1200mm3). Notably, we observed a gradient decrease of tumor-infiltrating immune cell frequencies and CD28 and CD80/86 expression during tumor development (Figures 4.3G and 4.4C-E). In contrast, miR-424 expression significantly increased with tumor growth in both CT26 and MC38 tumors (Figures 4.3G and 4.4C-E). These findings indicate that increased miR-424 expression is correlated with low CD28 and CD80 protein expression and an immunosuppressive phenotype during tumor growth.

Figure 4.3 miR-424 is highly expressed in CRC and negatively regulate CD28 and CD80 expression.

(A) Schematic of the process to identify miR-424 as a mechanism to downregulate the CD28 costimulatory pathway in human CRC. (B) miR-424 expression in human CRC and normal colon cases (TCGA dataset). (Kruskal-Wallis test, ****p<0.0001) (C) In situ hybridization (ISH) of miR-424 expression in human tissues. Representative staining of the normal colon (left panel) and CRC tissue (right panel). (n=10) (D) qRT-PCR analysis of miR-424 expression in paired human CRC and normal colon tissues. Scale shows Fold change (FC). (E) Dual-luciferase reporter assay of miR-424 on the 3’UTR of human CD28 and CD80. (Mean±SEM, n=3, t-test, 91

**p<0.01, ****p<0.0001, n.s. no significance) (F) Primary human CD4+ and CD8+ T cells and DCs transfected with miR-424 mimic showed lower CD28 and CD80 protein expression, respectively. (Mean±SEM, n=3, t-test, *p<0.05, **p<0.01) (G) Heatmap representation of miR- 424 expression level and the immune landscapes during various phases of CT26 tumor growth. (n=3, t-test, p-values listed) See also Figure 4.4.

Figure 4.4 Supplementary data to Figure 4.3.

(A) The positive (U6) and negative (scramble) controls used for the ISH experiments. (B) Dual- luciferase assay confirmed the binding of miR-424 on the three prime untranslated regions (3’UTR) of mouse CD28 and CD80. (Mean±SEM, n=3, t-test, **p<0.01, ***p<0.001, n.s. no significance). (C) miR-424 expression level and the immune landscapes during the MC38 tumor progression, from early-stage tumors to advanced-stage tumors, were shown in the heatmap. (n=3, t-test, p-value was listed). (D) Representative gating strategy used to identify tumor- infiltrating T-cells (FSC-Alow, SSC-Alow, CD45+, CD3+, CD4+/CD8+) in mouse tumor tissues. (E) Representative gating strategy used to identify tumor-infiltrating CD103+ DCs (CD45+, MHCII+, CD11c+, CD11bdim, CD103+) in mouse tumor tissues. 4.2.4 TEVs transfer miR-424 from human CRC cells to tumor-infiltrating T cells and DCs As our data indicated that miR-424 upregulation in human CRC was a potent inhibitor of CD28 and CD80 expression, we sought to identify the mechanism that led to the accumulation of 92 miR-424 in human CRC and tumor-infiltrating T cells and DCs. Strikingly, the level of miR-424 was significantly higher in human and mouse CRC cell lines and primary human CRC tumor organoids, than to the species matched naïve T cells and DCs (Figure 4.5A-B). Extracellular vesicles (EVs) are critical in mediating the transfer of microRNAs and intercellular communication in the TME98,203,204. Therefore, we investigated whether tumor cell-expressed miR-424 could be transferred to T cells and DCs, leading to strong functional effects in the recipient cells. TEVs enriched for exosomes (average size ~130nm) from human and mouse CRC cell lines were isolated and characterized (Figure 4.6A-C). In these TEVs, we observed high levels of miR-424 by absolute qRT-PCR (Figure 4.5A). Using confocal microscopy, we confirmed the uptake of fluorescence-labeled TEVs by T cells and DCs cells in vitro (Figure 4.5C). We then examined TEVs mediated miR-424 transfer, in a transwell coculture system that contained tumor cells and T cells or DCs. Notably, the presence of tumor cells significantly increased the miR-424 level in T cells and DCs (Figure 4.5D). Treatment with the sphingomyelinase inhibitor GW4869 could block TEVs production without affecting tumor cell viability (Figure 4.6D). When TEVs production is blocked by GW4869, tumor cells failed to cause an increase in miR-424 levels in cocultured T cells and DCs (Figure 4.5D).

We extended our in vitro work to mouse models to validate that TEVs are responsible for transferring miR-424 from tumor cells to T cells and DCs in a physiologically relevant condition. WT CT26 cells or CT26 cells expressing CD63-GFP (a marker to trace EVs) were used to establish in vivo tumors. The GFP signal was detected in the tumor-infiltrating (CT26-CD63- GFP) T cells and DCs (Figure 4.5E). However, T cells and DCs in the tumor-draining lymph nodes (TdLNs) and spleen showed no/weak GFP signal (Figures 4.5E and 4.6F). T cells and DCs sorted from mice implanted with WT tumors by flow cytometry showed higher levels of miR-424 than T cells and DCs isolated from mice implanted with miR-424 knockout (miR-424KO) tumors and naïve mice spleen (Figure 4.5F). Further, we determined that the activation of T cells and DCs did not regulate their endogenous miR-424 expression (Figure 4.6G). Thus, the accumulation of miR-424 in tumor-infiltrating T cells and DCs is due to the uptake of miR-424 containing TEVs.

Next, we determined the functionality of TEVs with/without functional miR-424 on the CD28-CD80/86 costimulatory pathway. We first characterized TEVs isolated from WT tumor cells, miR-424 inhibitor (miR-424 antisense oligonucleotides) transfected tumor cells and miR- 424KO tumor cells. We confirmed that tumor cells expressing the miR-424 inhibitor packaged 93 the inhibitor into secreted EVs (miR-424i-EV) (Figure 4.6H). Further, we validated that miR- 424i-EV and miR-424KO-EV (EVs derived from miR-424KO tumor cells) do not have functional miR-424 in them and cannot regulate CD28 and CD80 (Figure 4.6I). In a human in vitro model, we incubated Jurkat T cells with the Raji Burkitt's lymphoma cell line to mimic a T cell-APC cell interaction. The addition of TEVs derived from the human CRC cell line HT29, which have normal miR-424 inhibited CD28 and CD80 expression on Jurkat and Raji cells, respectively (Figure 4.5G). Transfecting HT29 with a miR-424 knock-down construct abolished this effect, while transfection with control did not (Figure 4.5G). Next, we tested the effect of TEVs generated from CRC patient-derived tumor organoids on T-cell activation, which is indicated by IL-2 production. Tumor organoids retain the heterogeneity of tumors and are considered to be advanced in vitro models205,206. In this experiment, we used a transwell assay with Jurkat/Raji cells in the upper chamber and CRC organoids in the lower chamber (Figure 4.5H). The presence of tumor organoids reduced IL-2 production from the Jurkat-Raji conjugation system. Noticeably, blocking EVs secretion by GW4869 rescued IL-2 production (Figure 4.5H).

Finally, we investigated how miR-424 expression is upregulated and regulated in human CRC cells. Hypoxia is a key feature that differentiates tumors from normal tissues and is commonly observed in human CRC207. Induction of hypoxia stabilized the transcription factor HIF-1α expression in HT29 cells and led to the upregulation of miR-424 and increased EV production (Figure 4.6J). Our findings align with previous reports in other experimental systems208-210. Collectively, these data reveal that hypoxia in tumor cells causes upregulation of miR-424 that is delivered to tumor-infiltrating T cells and DCs by TEVs where it suppressed CD28-CD80/86 costimulatory signaling.

Figure 4.5 Tumor cells derived EVs transfer miR-424 to immune cells and downregulate CD28 and CD80.

(A) Relative levels of miR-424 expression in human and mouse CRC cell lines, primary human and mouse T cells and DCs, and primary human CRC organoids. (B) Representative photomicrograph of primary human CRC organoids. (C) Confocal microscopy confirmed EVs uptake by T cells and DCs. CT26 cells derived EVs were stained with Dio dye and incubated with mouse naïve T cells and DCs. (D) CT26 cells cocultured with naïve T cells and DCs in a transwell system with and without EV inhibitor GW4869. miR-424 levels in T cells and DCs in each experimental condition. (Mean±SEM, n=3, t-test, *p<0.05, **p<0.01) (E) Flow cytometry 95 analysis showed GFP signal in a proportion of tumor-infiltrating T cells and DCs of the CT26- CD63-GFP tumors. Wild type (upper row) and CD63-GFP transfected CT26 cells (lower row) were used to establish subcutaneous tumors. T cells and DCs in the CT26-CD63-GFP tumor- draining lymph nodes (TdLNs) showed a very weak GFP signal. (n=3 in each condition) (F) miR-424 levels in T cells and DCs isolated from the spleen of naïve mice and subcutaneous tumors (established by wild type (WT) tumor cells or miR-424 knockout (KO) tumor cells) by flow sorting. (Mean±SEM, n=3, t-test, *p<0.05) (G) Expression levels of CD28 on Jurkat cells and CD80 on Raji cells in conjugation system: WTEVs, miRi-Ctrl-EV, and miR-424i-EV derived from human HT29 CRC cells were used for each set. (Mean±SEM, n=3, t-test, *p<0.05, **p<0.01, ***p<0.001) (H) Human CRC derived organoids were cocultured with the Jurkat and Raji cells conjugation system in a transwell system with or without GW4869. IL-2 production in the coculture system was measured. (Mean±SEM, n=3, t-test, *p<0.05) See also Figure 4.6.

Figure 4.6 Supplementary data to Figure 4.5.

(A) Transmission electron microscopy was used to capture the image of TEVs (CT26 tumor cell line as a representative). (B) Nanosight system was used to measure the size of TEVs (CT26 tumor cell line as a representative). (C) CD9 and CD63 protein expression were detected on the surface of TEVs (CT26 tumor cell line as a representative). (D) GW4869 reduced tumor cells EVs production but no viability. CT26 tumor cell line as a representative. (Means±SEM, n=3, t- test, ***p<0.001) (E) A repeat of experiments in Figure 4.5D in MC38 tumor cell line. (Means± SEM, n=3, t-test, *p<0.05) (F) T-cells and DCs in the spleen (SPL) of CT26-CD63-GFP tumor-

97 bearing mice showed no GFP signal. Related to Figure 4.5E. (n=3 in each condition) (G) Stimulation of naïve mouse T-cells (by CD3/CD28 activation beads) and DCs (by lipopolysaccharide) didn’t increase miR-424 expression. (Mean±SEM, n=3, t-test, n.s. no significance) (H) The complementary sequence of miR-424 inhibitor was cloned into the dual- luciferase vector (PsiCHECK-2). The vector was used to confirm the presence of functional miR- 424 inhibitors in miR-424i-EVs. (Mean±SEM, n=3, t-test, *p<0.05) (I) Dual-luciferase assay confirmed functional miR-424 in CT26 TEVs could bind with the 3’UTRs of CD28 and CD80. No functional miR-424 was detected in miR-424i-EV and miR-424KO-EV. (Mean±SEM, n=3, t- test, *p<0.05, **p<0.01, ***p<0.001) (J) HIF-1α accumulation was induced by hypoxia in HT29 cells. HIF-1α accumulation led to increased miR-424 expression and EVs production in HT29 cells. (Mean±SEM, n=3, t-test, *p<0.05, **p<0.01, n.s. no significance)

4.2.5 Blocking tumor cell-derived miR-424 inhibits tumor development in an immune-dependent manner Because TEVs with miR-424 suppressed the CD28-CD80/86 costimulatory pathway, we hypothesized that tumor-derived miR-424 impacts mouse CRC tumor development in vivo. Stably blocking the expression of Rab27a (a protein regulating exosome secretion) using shRAB27a reduced TEV production by MC38 tumor cells (Figure 4.7A). Suppressing TEVs’ secretion significantly reduced MC38 tumor growth in vivo (Figure 4.7B). In another experiment, mice were subcutaneously injected with tumor cells that cannot produce endogenous miR-424 containing TEVs. The resultant tumors were then injected with TEVs containing functional miR- 424, which significantly increased tumor growth (Figure 4.7C). These data demonstrate that miR- 424 in TEVs accelerates tumor development in immune-competent mice.

Next, we tested the impact of tumor cell-derived miR-424 on tumor formation and growth. Notably, when functional miR-424 was depleted in tumor cells and TEVs, tumor formation and growth rates decreased significantly (Figures 4.7D and 4.8A). In the host mice that rejected miR-424i tumors (lacking functional miR-424), we examined antitumor specific immune response (Figures 4.7E and 4.8B). In TdLNs and spleen of these mice, we found an expansion of gp70 tumor antigen-specific CD8+ T cells (Figure 4.7E). gp70 is a known tumor antigen present in CT26 cells. We got a similar result in the MC38 tumor model with p15E tumor antigen- specific tetramer staining (Figure 4.8B). These data indicated that a robust antitumor immune response could contribute to the rejection of tumor formation in those mice. miRNAs have multiple autonomous targets in cell function. Therefore, to confirm that the antitumor immunity was the predominant reason for the slower development of tumors lacking functional miR-424, we repeated the experiments in severe combined immunodeficient (SCID) mice, Cd28-/- mice, and Cd80/86-/-mice (Figures 4.7F and 4.8A). Tumor growth rate, proliferation (Ki-67+), and apoptotic cell (cleaved-caspase-3+) frequency were similar among tumors with/without functional miR-424 98 in SCID mice (Figure 4.7G). Blocking miR-424 did not alter tumor growth in Cd28-/- and Cd80/86-/- mice (Figure 4.8A). Transcriptome analyses of tumor cells revealed that blocking functional miR-424 did not extensively regulate cell-autonomous gene expression in tumor cells (Figures 4.7H and 4.8C-E). Taken together, these results indicated that blocking tumor cell- derived miR-424 effectively inhibits tumor development in a CD28-CD80/86 dependent manner.

Figure 4.7 Tumor EVs with functional miR-424 accelerated tumor growth by suppressing anti-tumor immunity.

(A) The knockdown of Rab27a significantly reduced EVs production in MC38 cells. (Mean±SEM, n=3, t-test, *p<0.05) (B) Growth curve of MC38 subcutaneous tumors established by MC38-Rab27aKD and MC38-Scramble-Ctrl cells. (Mean±SEM, n=5, t-test, **p<0.01, ***p<0.001, ****p<0.0001) (C) CT26 EVs with functional miR-424 (CRISPR-Ctrl EV) or without miR-424 (miR-424KO EV) were injected into the CT26-miR-424KO tumors in mice. Exogenous compensation of TEVs with miR-424 enhanced the growth of tumors without endogenous miR-424. (Mean±SEM, n=5, t-test, ***p<0.001, ****p<0.0001) (D) CT26 cells with functional miR-424 (WT, miRi-Ctrl, CRISPR-Ctrl, miR-424OE) and CT26 tumor cells without

99 functional miR-424 (miR-4242i, miR-424KO) were used to induce tumor formation in immune- competent BALB/c mice. (n=10-20, Kaplan-Meier method for survival curve, log-rank test for survival analysis, ****p<0.0001) (E) Antitumor specific immune response was detected in tumor-free mice that rejected the miR-424i CT26 cells. (Mean±SEM, n=6, t-test, *p<0.05, ***p<0.001) (F) The depletion of the functional miR-424 did not affect CT26 tumor growth pattern in SCID mice (n=4, the figure showed individual curves) (G) Proliferation (ki-67) and apoptosis (cleaved-caspase-3) of tumor cells was not influenced by functional miR-424 depletion. (Mean±SEM, n=4, t-test, n.s. no significance) (H) Upper panels: Differentially expressed genes in CT26 cells with and without functional miR-424. Lower panel: Gene Ontology analysis of biological pathways. See also Figure 4.8.

100

Figure 4.8 Supplementary data to Figure 4.7.

(A) A repeat of experiments in Figure 4.7D with MC38 tumor cell line in WT C57BL/6 mice, Cd28-/- mice, and Cd80/86-/- mice. (n=10-20, Kaplan-Meier method for survival curve, log-rank test for survival analysis, ***p<0.001, ****p<0.0001) (B) Anti-tumor specific immune response was detected in the tumor-free mice that were injected with miR-424i MC38 tumor cells. (Means±SEM, n=4, t-test, *p<0.05) (C) mRNA-Seq data showed the proportion of differentially expressed genes in MC38 cells with and without functional miR-424. (D) The Gene Ontology analysis showed biological pathways that were up/down-regulated in MC38 cells with and without functional miR-424. (E) miRNA-Seq data showed the proportion of differentially expressed miRNAs in CT26 and MC38 cells with and without functional miR-424.

101

4.2.6 Blocking tumor cell-derived miR-424 enhanced adaptive antitumor immunity and sensitizes advanced tumors to ICBT Since blocking tumor cell-derived miR-424 suppressed tumor growth by an immunity dependent manner, we then analyzed the impacts of tumor cell-derived miR-4242 on anti-tumor immune effectors. Immunohistological analyses on the CT26 and MC38 tumor sections showed that decreasing tumor cell-derived miR-424 led to a higher T-cell infiltration (Figures 4.9A and 4.10A). T cells were observed in both peri- and intra-tumor areas (Figures 4.9A and 4.10A). Analyses of antitumor immune landscapes were performed in tumor tissues, TdLNs, and the spleen of tumor-bearing mice (Figures 4.9B-C and 4.10B-D). Blocking tumor cell-derived miR- 424 predominantly altered the antitumor immune landscape in both CT26 and MC38 tumor microenvironments (Figures 4.9B and 4.10B) but not in the TdLNs and spleen of tumor-bearing mice (Figures 4.9C and 4.10C-D). We found that T cells in tumors that lack miR-424 had elevated CD28 expression compared to T cells in control tumors. Similarly, the CD80 expression on DCs in miR-424 blocked tumors was higher than the control tumors (Figures 4.9B and 4.10B). A higher concentration of IL-2 was observed in miR-424 blocked tumors as well (Figures 4.9B and 4.10B). Other parameters, such as total immune infiltration, T-cell infiltration, and proliferation, were enhanced by blocking tumor cell-derived miR-424 (Figures 4.9B and 4.10B). Remarkably, the major immunosuppressive mechanisms such as tumor tissue PD-L1 expression and PD-1 expression on CD4+ T cells were higher in the miR-424 blocked tumors (Figures 4.9B and 4.10B). The frequencies of Treg and myeloid-derived suppressor cells (MDSCs) showed a trend of increase (Figures 4.9B and 4.10B). These mechanisms indicate a potential negative feedback response.

Our observations point out that the expression of CD28 on T cells and CD80 on DCs was downregulated in advanced tumors with elevated miR-424 (Figures 4.3G and Figure 4.4C). Because CD28 and CD80 are directly involved in the efficacy of anti-PD-1/PD-L1 treatments18,19,198,199, we reasoned that blocking tumor cell-derived functional miR-424 would enhance tumor response to ICBT. We tested anti-PD-1/CTLA-4 treatments in early-stage tumors and established tumors with/without functional miR-424 expression (Figures 4.9D and 4.10E). Notably, anti-PD-1/CTLA-4 treatments in established WT tumors were less effective than in early-stage tumors (Figures 4.9D and 4.10E). In contrast, blocking tumor cell-derived miR-424 in combination with ICBT caused tumor rejection or reduction in established tumor growth in WT mice. However, the treatment is ineffective in Cd28-/- mice and Cd80/86-/- mice, even with blocked miR-424 in tumor cells (Figures 4.9D and 4.10E). Systemic assessments by mass 102 cytometry demonstrated distinctive antitumor immune landscapes in anti-PD-1/CTLA-4 treated established tumors with/without miR-424 (Figure 4.9E). A favorable phenotype with high tumor- infiltrating T-cell frequency and function was only seen in treated tumors that lacked functional miR-424 (Figure 4.9E-F). These data support the significance of tumor cell-derived miR-424 in determining tumor response to ICBT.

Figure 4.9 The depletion of functional miR-424 enhanced the costimulatory molecules CD28 and CD80 expression, antitumor immunity, and response to ICBT.

(A) The overall T-cell infiltration in CT26 tumors with (WT, miRi-Ctrl) and without (miR-4242i) functional miR-424. T cells were observed in both peripheral and central regions of tumors. (Mean±SEM, n=6, t-test, **p<0.01 n.s. no significance) (B) Heatmap representation of antitumor 103 immune profiles in CT26 tumors without functional miR-424 and control tumors. Right panels: Levels of CD28 and CD80 expression on tumor-infiltrating CD4+ and CD8+ T cells and CD103+ DCs. Intratumoral IL-2 concentration was measured in all three groups. (Mean±SEM, n=6, t-test, *p<0.05, **p<0.01 n.s. no significance) (C) Expression of CD28 on T cells and CD80 on antigen-presenting cells (APCs) in tumor-draining lymph nodes (TdLNs) and spleen (SPL) of MC38 tumor-bearing mice. (Mean±SEM, n=6, t-test, n.s. no significance) (D) In the MC38 tumor model, ICBT is effective for treating the early-stage tumors (volume ~100mm3) but not for controlling the large tumors (volume 500-700mm3). Blocking functional miR-424 in tumor cells enhanced the response of established tumors to ICBT in WT C57BL/6 mice, but not in Cd28-/- mice and Cd80/86-/- mice. (E) Mass cytometry analysis of immune landscape in established tumors treated with ICBT. (n= 4 cases/condition). A representative t-SNE plot is showing immune cell infiltration for each condition. (F) Tumor-infiltrating T-cell proliferation induced by ICBT in established tumors without functional miR-424. (Mean±SEM, n=4, t-test, *p<0.05) See also Figure 4.10.

Figure 4.10 Supplementary data to Figure 4.9.

(A) The overall T-cell infiltration in MC38 tumors with (WT, miRi-Ctrl) and without (miR- 4242i) functional miR-424. T-cells were observed in both peripheral and central parts of tumors. (Mean±SEM, n=6, t-test, **p<0.01 n.s. no significance) (B) The anti-tumor immune profiles in MC38 tumors without functional miR-424 and control tumors were shown in the heatmap. Quantitate analysis was shown between miR-424i and miRi-Ctrl groups at the right of heatmap. Right panels: Quantitate analyses of the CD28 expression on tumor-infiltrating CD4+ and CD8+ T-cells, CD80 expression on tumor-infiltrating CD103+DCs, and the intratumoral IL-2 104 concentration were shown in all three groups. (Mean±SEM, n=6, t-test, *p<0.05, **p<0.01 n.s. no significance) (C) The immune landscape of TdLNs and SPL in the CT26 (WT, miRi-Ctrl, miR-424i) tumor-bearing mice. Quantitate analyses were shown at the right of the heatmap. (D) The immune landscape of TdLNs and SPL in the MC38 (WT, miRi-Ctrl, miR-424i) tumor- bearing mice. Quantitate analyses were shown at the right of the heatmap. (E) In the CT26 tumor model, ICBT is effective for treating the early-stage tumors (volume at around 100mm3) but not for controlling the established tumors (volume at around 500-700mm3). Depleting the tumor cells derived functional miR-424 enhanced the response of established tumor response to ICBT. 4.2.7 Depleting functional miR-424 induces antitumor immunogenicity of TEVs Wild type TEVs are generally considered immunosuppressive211,212. On the other hand, TEVs do contain tumor antigens that can induce an antitumor immune response95,101,196,197. Therefore, we speculated that depleting immunosuppressive factors, such as miR-424, would potentially enhance the immunogenicity of TEVs. To address this, we systemically administered TEVs with/without functional miR-424 to naïve mice. We first validated that the systemically administered TEVs could be taken in by immune cells in the peripheral lymphatic organs (Figure 4.11A). We then found that the administration of TEVs lacking miR-424 (termed modified TEVs) could stimulate the expansion of tumor antigen-specific CD8+ T cells in peripheral lymphatic organs of naïve mice without inducing extensive non-specific T-cell activation (Figure 4.11B). Moreover, to investigate whether a modified TEVs induced tumor-specific immune response can reject tumors in vivo, we induced tumors in mice preconditioned with TEVs (Figure 4.11C). All mice treated with saline and TEVs with functional miR-424 developed CT26/MC38 tumors (Figure 4.11C). Remarkably, mice treated with modified TEVs were immunized to the tumor cells or only developed minimal CT26/MC38 tumors (Figure 4.11C). Repetition in Tlr4-/- mice showed similar results, indicating that the antitumor specific immune response, rather than potential non-specific immune responses, account for the protective effects of modified TEVs (Figure 4.11C). In SCID mice, Cd28-/- mice, and Cd80/86-/- mice, the treatment failed to prevent tumor formation (Figure 4.11C), indicating that the effect is highly dependent on the CD28- CD80/86 costimulatory pathway.

We monitored body weight, blood thrombotic status, and serum cytokine levels of recipient mice (Figure 4.12A-C) to evaluate treatment safety. No differences were observed between untreated and treated groups (Figure 4.12A-C). Histological analysis of the major organs was performed and confirmed that the short-term modified TEV treatment (2 doses in one week, i.v., 10μg protein per dose) did not cause severe damage to the heart, liver, lung, kidney, small intestine, or spleen (Figure 4.12D). Cumulatively, our findings highlighted that modified TEVs are immunogenic and can stimulate antitumor immunity in a safe manner. 105

Figure 4.11 TEVs without functional miR-424 are efficient in stimulating antitumor immune response.

(A) Representative flow cytometry dot plots showing intravenously injected TEVs (CD63-GFP labeled) were taken by T cells and DCs in the peripheral lymphatic organs (n=3/group). (B) Tail vein injection of TEVs (10µg) with different status of miR-424 function in naïve mice. TEVs without functional miR-424 (miR-424i-EV) stimulate tumor antigen-specific (gp70) T-cell expansion in the peripheral lymphatic organs. Non-specific T-cell activation (CD62L- frequency) was also evaluated. (Mean±SEM, n=4, t-test, *p<0.05 n.s. no significance) (C) TEVs (10µg) with different miR-424 functional status was injected to naïve mice intravenously. Two weeks after injection, the mice were challenged with syngeneic tumor cells subcutaneously. Naïve WT mice preconditioned by TEVs without functional miR-424 (miR-424i-EV and miR-424KOEV) showed a lower tumor burden than mice preconditioned by TEVs with functional miR-424 or saline. The experiments were repeated in Tlr4-/- mice, SCID mice, Cd28-/- mice, and Cd80/86-/- mice. The tumor burden was measured two weeks after tumor cells inoculation. Left half: CT26 model; Right half: MC38 model. (Mean±SEM, n=4-12, t-test, *p<0.05, ***p<0.001, ****p<0.0001, n.s. no significance) See also Figure 4.12.

106

Figure 4.12 Supplementary data to Figure 4.11.

(A) The influence of short-term TEVs injection on the naïve Babl/c mice growth curve was shown (2 doses, 10ug/dose). (B) Thrombosis level (indicated by the serum D-dimer level) in naïve Babl/c mice treated by short-term TEVs injection (2 doses, 10ug/dose). (Mean±SEM, n=3, t-test, n.s. no significance) (C) Serum cytokine levels in naïve Babl/c mice treated by short-term TEVs injection (2 doses, 10ug/dose). The upper panels showed representative data of the 107 cytokine array (n=3). The lower panel showed the quantitative results of C5a, ICAM-1, CXCL12. (Mean±SEM, n=3, t-test, n.s. no significance) (D) Histological analysis of the major organs (spleen, lung, heart, liver, kidney, and small intestine). Tissue samples were taken from mice treated by short-term TEVs injection (2 doses, 10ug/dose). (n=6, one representative picture from each group) 4.2.8 TEVs with miR-424 knockdown enhance ICBT efficacy in an advanced CRC pre-clinical tumor model Given the potent antitumor immunogenic features of the modified TEVs, we further examined their therapeutic value in advanced CRC pre-clinical tumors. As a preliminary test, we administered the modified TEVs as a treatment for early (100-200mm3) and established (500- 700mm3) subcutaneous tumors and found that the modified TEVs treatment alone was effective in minimal tumors, but not advanced tumors (Figure 4.14A). Since we have shown that the advanced tumors have high levels of immunosuppressive factors (Figure 4.3G), we reasoned that the antitumor immune response initiated by modified TEVs was not sustained in the advanced tumors. Thus, we proposed to combine the modified TEVs with ICBT to treat immunosuppressive tumors.

To mimic human late-stage CRC, we developed a cecum based orthotopic transplantation preclinical model213-215. Both established primary tumors in the cecum and metastatic lesions in the liver and peritoneal cavity were observed in this model (Figure 2.9), indicating the advantages of using it to study treatment for stage IV CRC. For established tumors, two doses of modified TEVs were given during the 1st week of treatment for induction and anti-PD-1/CLTA-4 were given twice a week until the endpoint for sustention (Figure 4.13A). Notably, the combination of modified TEVs with ICBT significantly reduced metastatic tumor burden and extended overall survival to more than 50% compared with ICBT only group (Figure 4.13B-C). Tumors in each group were subjected to immune cell analyses, which showed that the combination treatment with modified TEVs and ICBT stimulated the most potent antitumor immunity with the highest T-cell infiltration frequency and activity (Figures 4.13D-E and 4.14D). Thus, combination treatment using modified TEVs that can stimulate tumor-specific T-cell expansion in peripheral lymphatic organs along with ICBT that depletes immunosuppressive factors in the TME, highlights a potentially novel therapeutic approach to delay progression in late-stage CRC.

108

Figure 4.13 TEVs without miR-424 enhanced the efficacy of ICBT in established orthoptic tumors.

(A) The schematic of treatment design. CT26-Luc cells were implanted into the cecum of naïve mice to establish advanced stage orthotopic tumors. Tumor development was monitored by periodic in vivo imaging. Mice with established tumors were assigned to different treatments (see methods for dose details, representative images from each treatment group are shown). Mice were either sacrificed after two weeks of treatment for evaluating disease burden and immune infiltration or followed up for survival analysis. (B) Metastatic tumor burden evaluated after two weeks of treatment. (Mean±SEM, n=5, t-test for two-group comparisons, one-way ANOVA for multiple groups comparison, *p<0.05, n.s. no significance) (C) Kaplan-Meier overall survival analysis of mice in each group. The mean survival time listed in the table. (n=12, log-rank test for survival analysis, **p<0.01, n.s. no significance) (D) Heatmap representation of immune profiles of tumors treated with different therapies. (E) Quantitative analysis of panel D. Amount of total CD8+ T-cell, IFN-γhigh CD4+ T-cell, and granzyme Bhigh CD8+ T-cell was significantly higher in tumors treated by TEVs without miR-424 and ICBT than in tumors treated by ICBT alone. (Mean±SEM, n=5, t-test for indicated two groups, **p<0.01, ***p<0.001, n.s. no significance) See also Figure 4.14.

109

Figure 4.14 Supplementary data to Figure 4.13.

(A) Left: Response of early-stage tumors (100-200mm3) to short-term different TEVs injection (2 doses, 10ug/dose). Right: Response of established tumors (500-700mm3) to short-term different TEVs injection (2 doses, 10ug/dose). (Mean±SEM, n=5, one-way ANOVA for multiple groups comparison, *p<0.05) (B) The supplementary quantitative analysis of figure 4.13D. 4.3 Discussion In this study, we demonstrate a critical immunosuppressive role of tumor cell-derived miR-424 as a negative regulator of CD28-CD80/86 costimulatory signaling in tumor-infiltrating T cells and DCs. Blocking miR-424 in TEVs alleviated the immunosuppressive effects and enhanced the immune stimulatory effects of TEVs, making them a potential adjuvant treatment for ICBT used to treat CRC. It was noteworthy that the CT26 and MC38 cells used in this study are representative of CRC-MSS and -MSI phenotypes, respectively216. Thus, our data revealed a shared mechanism of T-cell costimulation in these CRC subtypes. In both men and women, CRC is the third most common type of cancer102.

110

CRC patients were among the first cohort of cancer patients tested for ICBT; however, the majority of late-stage CRC patients were unresponsive to ICBT in the early clinical trials. After several trials, it became evident that almost all responders had the MSI subtype8,105. Although the mutational loads in different CRC subtypes are considered as one of the major determining factors of ICBT response195, other mechanisms regulating immune response are likely to exist. Indeed, around 40-70% of the immune infiltrated late-stage MSI CRC with high mutational load failed in ICBT8,38. Furthermore, MSS CRC that is considered resistant to ICBT can still have a high frequency of tumor-infiltrating T cells40,106,217. Our observation in the TCGA CRC dataset and a cohort of 71 CRC patient samples validated that a proportion of the MSS tumors have comparable T-cell infiltration as with MSI tumors (Figure 4.1). This evidence supports the hypothesis that independent of the MSI/MSS classification, there are other potential mechanisms controlling the adaptive immune response in CRC, which needs further investigation.

ICBT functions through blocking the inhibitory immune checkpoint molecules expressed on tumor cells, T cells, and other TME components. However, to rescue exhausted T cells by blocking PD-1/PD-L1 signaling, positive costimulation from the CD28-CD80/86 pathway is required19. Furthermore, biochemical studies demonstrated that T-cell expressed CD28 is a primary target for PD-1 mediated inhibition18. In infectious conditions, CD28 expression is also required after T-cell priming for helper T-cell-mediated immune protection163, and the presence of CD28-mediated signaling prevents the induction of anergy in T cells218. All of these observations prompted us to investigate whether the CD28-CD80/86 costimulatory signaling is a prerequisite for ICBT response and whether this signaling is preserved in the CRC immune microenvironment. Remarkably, our analysis of human CRC indicated that some tumor- infiltrating T cells and DCs lack expression of CD28 and CD80/86, respectively (Figure 4.1). Moreover, deleting or blocking the CD28-CD80/86 costimulatory signaling nullified the efficacy of ICBT in mouse CRC models that have immune cell infiltrations (Figure 4.1). Our findings are consistent with early reports in non-small cell lung cancer and melanoma tumors19,219 and support the hypothesis that the lack of intact CD28-CD80/86 costimulatory signals on infiltrated T cells and DCs restricts ICBT response in human CRC.

In non-cancerous conditions, CD28 expression levels decrease during antigenic exposure but rapidly revert back to pre-stimulation levels. However, sustained T-cell stimulation causes CD28 downregulation and eventual loss. CD28 downregulation with T-cell activation 111 predominantly involves transcriptional repression and increased protein turnover as a T-cell intrinsic negative feedback mechanism220. Biosynthesis and expression of CD80/86 on DCs are stimulated by antigen presentation and are associated with DC maturation. In cancerous conditions, the effects of the tumor and stromal cells on T cells and DCs through direct contact and/or paracrine mechanisms cannot be ignored25,144,221. Our previous studies and other reports have established that abnormal expression of miRNAs is associated with CRC pathobiology222-224. Therefore, we sought to investigate tumor cell-secreted miRNAs as a tumor intrinsic mechanism of CD28-CD80/86 downregulation in immune cells. We identified that miR-424s negatively regulate CD28 and CD80 expression (Figure 4.2). miR-424 was upregulated in CRC compared to the normal colon, T cells, and DCs (Figures 4.2 and 4.3). In tumors, miR-424 was upregulated in tumor cells and was transported to surrounding T cells and DCs via TEVs, leading to the downregulation of CD28 and CD80 proteins and defective T-cell function (Figure 4.3). Increased miR-424 levels in tumor cells were induced by hypoxia and accompanied by immune exclusion (Figure 4.2 and 4.3). These data validated that tumor intrinsic mechanisms can be an initiating factor for immunosuppression. Of course, we cannot exclude other unknown mechanisms that upregulate miR-424. The immune editing during tumor development may also contribute to the selection and accumulation of miR-424 producing tumor cells.

In mouse tumors, depleting or blocking miR-424 in tumor cells significantly delayed tumor growth in an immune-dependent manner (Figure 4.4). Immune landscape analyses demonstrated that lack of tumor-derived miR-424 rescued CD28 expression on both tumor- infiltrating CD4+ and CD8+ T cells. CD80 expression on the tumor-infiltrating CD103+ DCs that are critical for tumor antigen presentation and effector T-cell trafficking58-60,64 were increased in tumors lacking miR-424. Notably, T-cell frequency and function were significantly higher in those tumors. As a consequence, lack of tumor-derived miR-424 and restoration of CD28- CD80/86 signaling reversed tumor intrinsic resistance to anti-PD-1/CTLA-4 treatment (Figure 4.5). Our data reinforce previous findings that intact CD28-CD80/86 costimulatory pathways could attenuate PD-1 mediated T-cell exhaustion and were associated with better response to anti- PD-1 treatment19,20,225. Recent mechanistic studies describing the immunostimulatory and therapeutic significances of CD80 expression on APCs198,199 also support our findings. In summary, our studies establish that TEVs delivered miR-424 is a crucial mechanism causing defective CD28-CD80/86 costimulatory signaling in human CRC that can induce ICBT resistance.

112

Blocking tumor cell-derived miR-424 restored the defective CD28-CD80/86 costimulatory signaling in tumors and resulted in a better response to ICBT (Figure 4.5). However, in clinical conditions, it is difficult to genetically alter tumor cells or specifically block miRNAs in the TME. TEVs carry tumor antigens and are being tested as potential cell-free cancer vaccines101,196. Therefore, to translate our mechanistic findings to potential therapeutic approaches, we considered enhancing the immunostimulatory aspects of TEVs by eliminating the miR-424 induced immunosuppression. Our modified TEVs (without immunosuppressive miR- 424) stimulated a tumor-specific immune response that was independent of the tumor's endogenous antigen presentation and T-cell priming. Systemic administration of modified TEVs in naïve mice induced tumor-specific immune responses and are protective of tumor cell challenge in the preconditioned mice (Figure 4.6). Careful monitoring of the health of the recipient mice indicated there were no serious adverse side effects of this treatment (Figure 4.12). A recent phage 2 stage clinical trial suggested that utilizing the anti-PD-1/anti-CTLA-4 combination treatment as neoadjuvant therapy in early-stage CRC might have a better response than being used as adjuvant therapy in late-stage tumors226. Therefore, we established an orthotopic CRC model that mimics human metastatic CRC tumors to test the therapeutic efficacy of modified TEVs226. In such an aggressively progressing CRC model, we successfully tested combining the modified TEVs with ICBT (Figure 4.7). This combination resulted in an increased immune response and mouse survival rate over the treatment with ICBT alone (Figure 4.7). This is most likely due to the modified TEVs ability to stimulate anti-tumor specific T-cell responses, and that could be sustained by ICBT. Similarly, a recent study using TEVs produced by irradiated mouse breast cancer cells demonstrated the enhanced costimulatory functions of DCs101. These findings, together with our results, underscore the potential of using modified TEVs with enhanced immunostimulatory effects in stimulating antitumor immunity and treating ICBT resistant tumors.

Our results have several important implications for clinical translation. First, we show that tumor cells may suppress positive costimulatory pathways in tumor-infiltrating immune cells. To revitalize the tumor-infiltrating T cells, protecting or restoring the positive costimulatory pathways may be equally important as suppressing the inhibitory pathways. This concept is backed by a recent report using CD28 stimulatory mechanisms to synergize with anti-PD-1 blocking antibody20. Second, our data showing that modified TEVs could stimulate anti-tumor T- cell responses and reverse ICBT resistance in advanced CRC models suggest potential

113 therapeutics for ICBT resistant tumors. WT TEVs are considered immunosuppressive in general227,228. However, recent findings identified the immunosuppressive factors in those TEVs and showed that blocking them can alleviate the immunosuppression93,95,96. Further, adding immunostimulatory factors into the TEVs can even enhance their ability to initiate immune responses101. Traditional TEVs related therapies used WT TEVs to pulse DCs in vitro, and TEVs loaded DCs were infused to treat tumors229-231. This method avoids the potential negative effects of TEVs on the immune system in vivo. However, this approach is time-consuming and did not consider the negative effects of WT TEVs that may impart functional defects in DCs, leading to treatment failure. Administration of the modified TEVs can minimize the negative effects of WT TEVs on immune response and, at the same time, efficiently stimulate tumor-specific immunity in vivo. Modified TEVs could be an alternative to the current TEVs based cancer treatments.

Taken together, our study reinforces previous reports regarding positive costimulatory pathway and TEVs in ICBT19,95. We deciphered a novel mechanism of the CRC immune evasion and an innovative concept of modified TEVs based treatment. It is noteworthy that miR-424 may have strong effects on other immune cell populations or immune processes, considering it has a large number of predicted immune gene targets. Our current results do not exclude the contribution of other potential immune evasive mechanisms of ICBT failure14. A comprehensive understanding of immune suppressive mechanisms and accurate blocking of these suppressive factors are critical and necessary for a successful treatment. Finally, an array of potent immunosuppressive factors have been identified as well92,95-97. Blocking a bulk of these negative regulators may further enhance the efficacy of modified TEVs on stimulating antitumor immunity but requires more investigations.

4.4 Methods and Material 4.4.1 Human samples Human colorectal cancer (CRC) biospecimens and cancer patients’ peripheral blood were consented and collected through the Biological Materials Procurement Network (BioNet) of the University of Minnesota and the Cooperative Human Tissue Network, under an IRB approved protocol. Formalin-Fixed Paraffin-Embedded (FFPE) samples were used for immunofluorescent imaging studies, and freshly excised samples were used for flow cytometry and qRT-PCR analyses. All freshly resected specimens were placed in ice-cold RMPI-1640 medium in a 50mL conical tube and immediately transported to the laboratory for processing. Health donors derived

114 peripheral blood was provided by the memorial blood centers with an approved IRB protocol. All human CRC tumors collected for the study were treatment naïve.

4.4.2 Tumor cell lines Human CRC cell lines HT116 (ATCC, CCL-247), HT29 (ATCC, HTB-38), DLD-1 (ATCC, CCL-221), and SW480 (ATCC, CCL-228) were purchased from the American Type Culture Collection (ATCC). Mouse CRC cell line CT26 (ATCC, CRL-2638) was purchased from the ATCC. Mouse CRC cell line MC38 was kindly provided by Dr. Nicholas Haining. Human Jurkat cells (ATCC, TIB-152) and Raji cells (ATCC, CCL-86) were purchased from ATCC. Jurkat cell, Raji cells, CT26 cells, and DLD-1 cells were maintained in complete RPMI-1640 medium (GIBCO BRL), supplemented with 10% heat-inactivated FBS (Thermo Fisher Scientific), 100 IU/mL penicillin, and 100 μg/mL streptomycin (Invitrogen Life Technologies). SW480 cells and MC38 cells were cultured in the complete DMEM medium (GIBCO BRL), and HCT116 and HT29 cells were cultured in McCoy's 5a Medium (Thermo Fisher Scientific) with the same supplements as the RPMI 1640 medium. All cells were routinely authenticated and tested for mycoplasma.

Derivatives of these parental cell lines were established for different experimental proposes. For in vivo tumor imaging, CT26-Luc was created through the transduction of CT26 with a firefly luciferase expression construct. The CT26/MC38-miR-424KO cell lines were constructed by the CRISPR/Cas9 technology and single clone selection. The guide RNA flanking the miR-424 was purchased from the Canopy Biosciences. The CT26/MC38-miR-424i and CT26/MC38-miR-424OE were established by transducing the wild type cells with constructs that either express the anti-miR-424 sequence (miR-424i) or the miR-424 sequence (miR-424OE). To knock down the Rab27a expression (Rab27aKD), we transduced the MC38 cell line with an shRNA library that would express five different Rab27a-targeting siRNAs. Vectors that express scramble sequences were used to establish the control (Ctrl) cell lines for the corresponding genetically edited cell lines. The lentiviral system (HEK-293T cells (ATCC, CRL-3216), Lipofectamine™ 3000 Transfection Reagent (Invitrogen), and pPACKH1-XL packaging vector (SBI) were used for making stably edited cell lines. The manufacture information of each construct was listed in the material table.

4.4.3 Mice All animal studies were approved by the institutional animal care and use committee (IACUC) of the University of Minnesota. All mice were kept in a specific pathogen-free 115 condition with fully autoclaved cages to minimize non-tumor specific immune activation. The following mice were purchased from the Jackson Laboratory: BALB/cJ, B6.129S2-Cd28tm1Mak/J, B6.129S4-Cd80tm1Shr Cd86tm2Shr/J, NOD.CB17-Prkdcscid/J, B6.CB17-Prkdcscid/SzJ, and B6(Cg)- Tlr4tm1.2Karp/J. The C57BL/6 mice were purchased from The Charles River Laboratories. Mice were bred in house and were used for experiments at the age of 6-8 weeks.

4.4.4 Human CRC organoids culture Surgically resected human CRC tissues were obtained from the BioNet, University of Minnesota. Fresh tumor tissue was processed as previously described232 with few modifications. Briefly, the tissue was cut into small pieces and washed with ice-cold PBS several times until the solution appears clear with no connective tissues and debris. The tissue was subsequently digested with digestion buffer (DMEM, 2.5% FBS, penicillin/streptomycin, 75U/mL collagenase IV) for 60 min at 37°C on a shaker. The digested tissue was passed through a 70uM cell strainer to remove any debris and large fragments. The cells were centrifuged at 300g for 3 mins at 4°C. The cell pellet was suspended in matrigel (growth factor reduced) at 3,000 cells/well. 33uL of the cell suspension was dispensed as a core in a 24 well culture plate. The matrigel was allowed to polymerize in the incubator at 37°C. The organoids were cultured in advanced ADMEM supplemented with penicillin/streptomycin, 10mM HEPES, 1X Glutamax, 1X B27, 1X N2, 1mM N-acetylcysteine, 100ug/mL Primocin, and 10nM Gastrin. The following niche factors were used: 10uM SB202190, 10uM Y-27632, 50ng/mL human EGF, 0.5mM A83-01, 10nM Prostaglandin E2, 10mM Nicotinamide and 50% L-WRN conditioned media (Wnt, R-spondin and Noggin). The organoids were passaged every week by incubating in TrypLE Express for 5 mins at 37°C and were plated in a new 24 well plate. Fresh complete medium was supplemented every two days.

4.4.5 Isolation of extracellular vesicles (EVs) Before EVs isolation, cells were cultured in media supplemented with 10% exosome depleted FBS (Gibco) for 24h. A standard differential centrifugation protocol was used to purify EVs fro the cell culture supernatant. In brief, culture supernatant was centrifuged at 300g for 10min to remove cells. The supernatant was then centrifuged at 3,000g for 10min to remove dead cells. Debris was pelleted after centrifugation at 10,000g for 30min. Supernatants were then centrifuged at 100,000g for 70min at 4°C (Beckman Coulter). The pelleted EVs were suspended in PBS and collected by another ultracentrifugation at 100,000g for 70min at 4°C to minimize protein. A similar method was used to isolate human CRC organoids derived EVs. To block the

116

EVs production from tumor cells and tumor organoids, 30μM GW4869 was added to the culture medium.

4.4.6 Characterization of isolated EVs We first imaged the purified EVs using electron microscopy. Briefly, purified EVs were resuspended in PBS and were dropped on formvar carbon-coated nickel grids. After staining with 2% uranyl acetate, grids were air-dried and visualized using a Tecnai G2 Spirit BioTWIN transmission electron microscope. The size and concentration of EVs isolated from cell culture supernatant were determined using a NanoSight LM-10 instrument (Malvern Instruments), which is equipped with a fast video capture camera and a particle-tracking software. Five independent microscopic fields were captured and analyzed for each sample. The individual data were merged into a single histogram plot.

We also tested the EVs-related protein markers on our purified EVs. The aldehyde/sulfate beads (10μl, Life Technologies) were added to 200μl PBS-resuspended EVs (10μg). The beads and EVs mixture were mixed using a benchtop rotator for 15 minutes at room temperature. 600μl PBS was then added to the solution, and mixing was continued overnight at 4°C. Then 400μl glycine (1M) was added, mixed, and incubated for 1h at room temperature. The solution was then spun down. The beads were resuspended in 100μl of 10% BSA in PBS and incubated for 1h at room temperature. The beads were pelleted down again and resuspended in 100μl of FACS staining buffer and stained for CD63, CD9, and isotype control for 30min at room temperature. The beads were finally washed three times with PBS and analyzed by a BD FACSCanto analyzer.

4.4.7 Immunofluorescence and histology Sections of FFPE tissues were deparaffinized with xylene, rehydrated with gradient ethanol, and subjected to antigen retrieval with heated antigen retrieval buffers (AR6 Buffer or AR9 Buffer, PerkinElmer). Sections were then blocked in 5% Bovine serum albumin (BSA) blocking buffer for 30min. Permeabilization was performed for intracellular staining by incubating the sections for 10 min with PBS containing 0.25% Triton X-100. Primary antibodies were applied and incubated overnight at 4℃. Then the sections were washed and incubated with secondary antibodies (1:1000 dilution) for 1h at room temperature. Slides were washed and mounted with ProLong Gold anti-fade mounting medium (Invitrogen) for imaging. For the multiplex immunofluorescent assays, we used the Opal 7-Color Automation IHC Kit (PerkinElmer). The tests were optimized and performed by following the instruction provided in the Kit. 117

The primary anti-human antibodies include CD3 (1:100, Clone CD3-12), CD8 (1:100, Clone 144B), E-cadherin (1:200, Clone M168), CD28 (1:50, Polyclonal), CD80 (1:100, Clone 2A2), CD86 (1:100, Clone EP1158Y), and CD11c (1:100, Clone EP1347Y and ITGAX/1242). Anti-mouse primary antibodies include Ki-67 (1:100, Polyclonal), cleaved-caspase-3 (1:50, Polyclonal), and CD3 (1:100, Clone CD3-12). Quantitative image analysis was performed by counting positive percentage and evaluating fluorescent signal intensity in at least five randomly selected fields of each section. The manufacture information of each antibody was included in the material table.

To analyze tissue histology, we conducted the hematoxylin-eosin (H&E) staining. The H&E staining kit was purchased from Abcam. We followed the manufacture’s instruction to optimize the staining protocol for each tissue type.

4.4.8 Mouse subcutaneous tumor model To establish the subcutaneous tumor model, we injected 2*105 CT26 tumor cells that were resuspended in 100μl matrigel (BD Biosciences) to the right flank of BALB/c mice. Similarly, 5*105 MC38 cells were used for the C57BL/6 mice. After tumor cell injection, tumors were typically measured 2-3 times per week using an electronic caliper. Tumor volume was calculated through the formula Volume = (Width2*Length)/2. Mice were considered as death when tumors exceeded the volume of 1000mm3 or 1500mm3, depending on the experimental setup.

4.4.9 Mouse cecum orthotopic tumor model Mouse cecum orthotopic CRC model was established for evaluating the impacts of modified tumor cells derived EVs (TEVs) on anti-PD-1/anti-CTLA-4 response. Buprenorphine (SR) 2mg/kg is administered subcutaneously 2 hours before surgery for pre-emptive analgesia. Then, mice are anesthetized with ketamine (100mg/kg)/xylazine(10mg/kg). We then position the mouse on its back to fix the forelegs in a V shape with tape. The abdomen of the mouse was shaved. The abdominal skin was prepared by wiping with povidone-iodine prep pads and then alcohol prep pads. The surgical area was isolated by placing the sterile drape around the incision area on top of the abdomen. A 15mm vertical midline incision was made on the skin. We then incised the linea alba to separate the rectus abdominis muscles and opened the abdomen. The cecum was located and carefully placed horizontally on top of a histo-cassette. A matrigel drop containing the CT26-Luc cells (2*105) was inoculated under the serosa layer. The inoculation site was carefully wiped by alcohol pads to remove any potential leaking. The cecum was placed back 118 into the abdomen using cotton swabs drenched in PBS. The abdominal wall and skin opening were then closed. Mice were monitored for at least seven days to recover, and more analgesics were given when needed.

4.4.10 Mouse treatment Mice were treated with IgG (10 mg/kg as an anti-CTLA-4 or anti-PD-1 control, once per 3 days), anti-PD-1 (10 mg/kg, clone: RMP1-14, once per 3 days), and anti-CTLA-4 (10 mg/kg, clone: UC10-4F10-11, once per 3 days) for treatment purpose. For blocking the CD80 and CD86 signaling, we administrated anti-CD80 (40 mg/kg, clone: 16-10A1) or anti-CD86 (40 mg/kg, clone: GL-1) to the mice one day before anti-CTLA-4/anti-PD-1 treatment starts. All antibodies were given intraperitoneally and continued until the endpoint of the study design. For EVs treatment, we injected 15μg EVs through the tail vein injections. The treatment starting points and endpoints varied in different experimental setups for different purposes (showed in the individual figure).

4.4.11 Mouse in vivo imaging We used the IVIS Spectrum in vivo imaging system (PerkinElmer) to monitor the orthotopic CRC tumor growth. 10 mins before image capturing, we injected the mice with D- Luciferin (150mg/kg, intraperitoneally, GoldBio) and anesthetized them with isoflurane. We set an exposure time of imaging to 60 sec and used the 540nm filter for signal collection.

4.4.12 T cells and antigen presentation cells in vitro assays Primary human T cells were purified from healthy human peripheral blood monocellular cells (PBMCs). Briefly, the human PBMCs were isolated by gradient centrifugation. Then the Dynabeads™ Untouched™ Human T Cells Kit (Invitrogen) was used to purify naïve T cells from PBMCs. We followed the manufactures’ protocol for detailed procedures. Recombinant IL-2 (30 U/mL) and DynabeadsTM Human T-Activator CD3/CD28 (25μl per 1×106 T-cells) were added to the cell culture for activating the naïve T cells. The activated human T-cells were cultured for up to seven days.

We induced primary human dendritic cells (DCs) from PBMCs by culturing them in complete RPMI-1640 medium supplemented with 250 IU/mL IL-4 and 800 IU/mL GM-CSF for two days. Then the cells were centrifuged and cultured in fresh complete RPMI-1640 medium supplemented with fresh 2,000 IU/mL IL-4 and 2,000 IU/mL GM-CSF. On day 6, cells were resuspended in fresh complete RPMI-1640 medium supplemented with 2,000 IU/mL IL-6, 400 IU/mL IL-1β, 2,000 IU/mL TNF-α, and 100 ng/mL LPS, and were cultured for another 24 hours. 119

On day 7, we used flow cytometry (CD14, CD1a, HLA-DR, CD80) to determine the viability, yield, and absolute cell count of induced DCs.

Mouse PBMCs were isolated from mouse spleen and inguinal, axillary, and brachial lymph nodes with a similar method as human PBMCs isolation. Mouse T-cells and DCs were purified from PBMCs with the Dynabeads™ Untouched™ Mouse T Cells Kit and Dynabeads™ Mouse DC Kit. We performed electroporation (Neon, Invitrogen) to transfect T-cells and DCs with 100nM synthesized miR-424. Parameters for T cells (107/ml): pulse voltage: 2,100V, pulse width: 20ms, pulse number: 1. Parameters for DCs (5*106/ml): pulse voltage: 1,500V, pulse width: 30ms, pulse number: 1.

To study the impacts of tumor cells derived EVs (TEVs) on the interaction between T- cells and APCs, we used the Jurkat and Raji cell co-culture system. Raji B cells were pre- incubated with 30ng/ml SEE superantigen (Toxin Technology) in RMPI-1640 medium for 30min at 37°C. Cells were then washed by PBS twice to remove free SEE. After antigen loading, 0.3 million antigen-loaded Raji B cells and 0.3 million Jurkat T cells were mixed in one well of a 12- well plate. At the same time, 2μg of TEVs were added. Then the plate was centrifuged at 300g for 1min to initiated cell-cell contact. Cells were then incubated at the regular cell culture condition for 24h before further analyses.

The same Jurka-Raji cells conjugation system was used for understanding the impacts of human tumor organoids derived EVs on co-stimulation. Briefly, 0.1 million antigen-loaded Raji B cells and 0.1 Jurkat T cells were mixed and loaded to one insert of a 12 well transwell co- culture system (Costar, 3.0μm). 105 primary human CRC organoids cells were seeded on the bottom of the transwell co-culture system to produce organoids derived EVs. To block the EVs production, 30μM GW4869 was added to the co-culture assay. The concentration of IL-2 was quantified in the co-culture media.

4.4.13 Enzyme-linked immunosorbent assays (ELISAs) for IL-2 and D-dimer Tests were performed to quantify IL2 in tumor tissue and cell culture media using ELISA kits (Invitrogen). To evaluate the intensity of thrombosis upon TEVs treatment, we collected mouse plasma after treatment and measured the D-dimer concentration using ELISA kits (LSBio). For ELISA tests, all samples were normalized based on protein concentrations acquired using a BCA protein assay (Pierce Chemical). We followed the manufacturer’s instructions for each step of the experiment.

120

4.4.14 Cytokine array We purchased the Proteome Profiler Mouse Cytokine Array Kit (Panel A) to detect cytokines in mouse serum. Serum was isolated from the whole blood drawn from the mice treated with either the wild type or modified TEVs. 100ul of mouse serum was added to each membrane (determined by preliminary experiments). We followed the manufacturer's standard instructions for incubation time and image acquisition.

4.4.15 Tissue digestion and flow/mass cytometry Mouse and human fresh tumor tissues were cut into small pieces (2mm*2mm) and were digested by collagenase I and IV (0.5 mg/ml for each) and deoxyribonuclease (50 units/ml) for 1h at 37°C. Digested tissues were loaded on a cell strainer with 70μm holes. Then the tissues were fully meshed and flushed through the cell strainer into a 50ml conical tube. Red blood cells lysis buffer was applied to remove RBCs, and the remaining cells were counted for further steps. For flow cytometry, the diluted Zombie fixable viability dye (Biolegend) was used to differentiate live and dead cells. Cells were then stained with antibody cocktails for cell surface markers. Cells were subjected to cytoplasm staining or intranuclear staining after cell surface staining when needed. The Cyto-Fast Fix-Perm Buffer Set (BioLegend) and True-Nuclear Transcription Factor Buffer Set (BioLegend) were used for intracellular and intranuclear staining. We followed the manufactures’ instructions for all steps. To avoid signal decay, we analyzed all samples in a BD FACSCanto (BD Biosciences) cytometer right after the staining steps. For lymphatic tissue staining, the enzyme digestion was omitted, and 40μm cell strainers were used to replace the 70μm cell strainers used for tumor tissues. All flow cytometry data were analyzed with the FlowJo Version 10 software (BD Biosciences).

We performed mass cytometry to evaluate the tumor immune landscape changes induced by immunotherapies. Details on antibodies and reagents used were listed in the key resources table. Metal pre-labeled antibodies were purchased from Fluidigm Corporation. All unlabeled antibodies were purchased (MaxPar® Ready purified) from Biolegend. We performed antibody- metal tag conjugation by using the MaxPar X8 antibody labeling kit (Fluidigm Corporation) according to the manufacturer's instructions. All manually labeled antibodies were validated and titrated in positive control and negative control samples.

Single cells were isolated from mouse tumor tissues and used for mass cytometry staining. We followed the public protocol from Fluidigm Corporation (Maxpar Cell Surface Staining) for detailed procedures. After antibody staining, EQ Four Element Calibration Beads 121 with the reference EQ passport P13H2302 were added to each staining tube right before data acquisition by a CyTOF 2 mass cytometer. The mass cytometry data were normalized across all cases before analysis. A t-SNE analysis was performed on all viable cells in tumor samples.

4.4.16 Locked nucleic acids and in situ hybridization (LNA-ISH) ISH for miRNAs was carried out on FFPE CRC, and normal colon sections with Double- DIG labeled LNATM probe for miR-424 (Qiagen) using miRNA ISH optimization Kit (BioChain). Double-DIG labeled probe for U6 (Qiagen) was used as a positive control. Double- DIG labeled scrambled microRNA (Qiagen) was used as a negative control. We followed the manufacturer's protocol for the ISH procedure. The temperature for probe incubation was optimized based on the probe’s Tm. Counterstaining was performed with Nuclear Fast Red™ (Sigma) and mounted with Eukitt quick-hardening mounting medium (Sigma).

4.4.17 Dual-luciferase assay The 3’UTR of human and mouse CD28 and CD80 genes were cloned with sticky ends (XhoI and NotI) on both sides. The dual-luciferase backbone plasmid (with its original multiple cloning site linker sequence) was purchased from Promega. Successful insertion of the cloned 3’UTR DNA into backbone plasmids was validated by enzyme digestion and sequencing. Site- directed mutations were performed to construct the mutated 3’UTR (3’UTR-mut) vector. Finally, the new constructions’ concentration was measured by NanoDrop 2000 (Thermo Scientific).

1μg of constructed dual-luciferase vectors and 100nM synthesized miR‐424 mimic (ThermoFisher Scientific) were co-transfected into HEK-293T cells using lipofectamine 3000 for 24h. Scrambled miR mimics with none target in the human/mouse genome served as a negative control for miRNA. Cells treated with transfection vehicles only were applied for the mock group. After transfection, protein in each cell pellet was harvested using the passive lysis buffer. We detected the luciferase activity by using the Dual‐Luciferase® Reporter Assay System (Promega) and a plate reader. The relative luciferase activity value was calculated as firefly luciferase signal intensity versus the renilla luciferase signal intensity. The result was shown as a fold change of the vehicle group.

4.4.18 DNA, mRNA and miRNA analyses For all DNA analyses, DNA was extracted with a DNA Extraction kit from Qiagen. The size of the DNA sequences was analyzed by electrophoresis. Total RNA extraction from tissues, cells, and EVs, was performed with the mirVana Total/miRNA RNA Isolation kit (Life Technologies/ Invitrogen). Extractions were carried out according to the manufacturer's 122 instructions. 500ng of total RNA was used for establishing the cDNA library with the miScript II RT Kit (Qiagen). qRT-PCR was performed with the SYBR Green I Master kit (Roche Applied Science) in a LightCycler 480. miRNA sequence-specific forward primer and universal reverse primers were used for mature miRNA analysis. U6 small nuclear RNA was used as an internal control for relative miRNA quantification. To compare the amount of miR-424 in EVs samples with cell line samples, we performed quantitative qRT-PCR analysis. The synthesized miR-424 sequence was used to establish a standard curve for absolute quantification.

High-throughput sequencing was carried out according to the previously described methods233,234. The total RNA was isolated by the mirVana Total RNA Isolation kit (Life Technologies). The Mayo Clinic Genome Analysis Core performed library creation and sequencing. For RNA-Seq, 1μg of total RNA was used to create each sequencing library using the Truseq RNA Sample Preparation Kit (Illumina, RS-122-2001). Libraries were then sequenced across two lanes on an HiSeq 4000 (Illumina) instrument with the 100 base pair paired-end mode. For miRNA-Seq, the sequencing libraries were prepared using 1μg of total RNA per the manufacturer’s instructions for the NEBNext multiplex small-RNA kit (New England Biolabs, Ipswich, MA). Sequencing was conducted using one lane on an HiSeq 4000 (Illumina) instrument with the 50 base pair paired-end mode.

The FASTQ files were analyzed using a customized pipeline (gopher-pipelines; https://bitbucket.org/jgarbe/gopher-pipelines/overview) developed and maintained by the Minnesota Supercomputing Institute. Briefly, sequencing quality was first checked using FastQC v0.11.5 (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/), followed by adapters and low quality reads trimming using Trimmomatic v0.33 (http://www.usadellab.org/cms/index.php?page=trimmomatic)235. Reads passing trimming were then aligned to GRCm38/mm10 reference genome using HISAT2 v2.0.2 (https://ccb.jhu.edu/software/hisat2/index.shtml)236. The transcripts abundances were counted using subread v1.4.6 (http://subread.sourceforge.net/)(Liao et al., 2014). Differential gene expression analysis was then performed in R v3.4.3 using the edgeR package (https://bioconductor.org/packages/release/bioc/html/edgeR.html)237.

Quality control (QC) of miRNA sequencing data was performed using the FastQC before and after the adaptor trimming with Trimmomatic. Then, the paired-end reads were assembled using the PANDAseq and aligned to the GRCm38/mm10 reference genome using the

123 bowtie2238,239. Finally, the total mature miRNA counts were generated with HTSeq240. MicroRNA differential expression was analyzed using the edgeR package in R (version 3.4.3).

RNA-Seq data in The Cancer Genome Atlas (TCGA) datasets were analyzed for immune cell subtype classifications. We performed the analysis by using the CIBERSORT (https://cibersort.stanford.edu/) following the recommended settings241.

4.4.19 Western blotting Cells were incubated in low oxygen (1%) condition for 24h to induce hypoxia. For western blotting, total cellular protein (30μg) from each sample was separated by SDS-PAGE, transferred to PVDF membranes, and subjected to primary antibody incubation. Antibodies for Western blot analysis include anti-HIF-1α (Abcam), anti-Rab27a (Abcam), and anti-Beta-Actin (Cell Signaling Technology). All primary antibodies were incubated at 4°C overnight. Peroxidase-linked anti-mouse IgG and peroxidase-linked anti-rabbit IgG were used as secondary antibodies. All secondary antibodies were incubated for 1h at room temperature. Pierce ECL western blotting substrate (Thermo Scientific) was used for image development.

4.4.20 Statistical analysis We used GraphPad Prism v.6.0 to perform statistical analyses. The normality of distribution was determined by the D’Agostino–Pearson omnibus normality test. Levene's test was used for assessing the equality of variances. For normally distributed data with homogeneity of variance, the significance of mean differences was determined using either two-tailed unpaired or paired Student’s t-tests. Non-parametric Mann-Whitney U-tests (for unpaired groups) and Wilcoxon matched-pairs tests (for paired groups) were conducted for non-normally distributed data. One-way ANOVA was used when more than two groups for comparisons were needed. Lifespan curves were plotted with the Kaplan–Meier method, and P values were obtained using a log-rank (Mantel-Cox) test. The Chi-square test was performed to analyze contingency tables. Error bars shown in graphical data represent Mean ± SEM. A two-tailed value of P < 0.05 was considered statistically significant.

124

Table 4.1. Key resources table for chapter 4.

REAGENT FOR IMMUNOASSAY SOURCE IDENTIFIER

Anti-human/mouse CD3 Abcam Cat #: ab11089

Anti-human CD28 Biorbyt Cat #: orb10295

Anti-human CD80 Abcam Cat #: ab86473

Anti-human CD86 Abcam Cat #: ab53004

Anti-human E-cadherin Abcam Cat #: ab76055

Anti-human CD11c Abcam Cat #: ab52632

Anti-human CD11c Abcam Cat #: ab212508

Anti-mouse Ki-67 Abcam Cat #: ab15580

Anti-mouse Cleaved-caspase 3 R&D systems Cat #: AF835

Anti-human CD8 Abcam Cat #: ab17147

Goat anti-Rabbit Secondary Antibody, Alexa Invitrogen Cat #: A11008 Fluor 488

Goat anti-Mouse Secondary Antibody, Alexa Invitrogen Cat #: A11001 Fluor 488

Goat anti-Rabbit Secondary Antibody, Alexa Invitrogen Cat #: A11011 Fluor 568

Goat anti-Rat Secondary Antibody, Alexa Fluor Invitrogen Cat #: A11077 568

Donkey anti-Rat Secondary Antibody, Alexa Invitrogen Cat #: A21208 Fluor 488

Goat anti-Mouse Secondary Antibody, Alexa Invitrogen Cat #: A11004 Fluor 568

Anti-human CD63-FITC (H5C6) BioLegend Cat #: 353006

Anti-human CD9-FITC (HI9a) BioLegend Cat #: 312104

Anti-mouse CD63-PE (NVG-2) BioLegend Cat #: 143904

Anti-mouse CD9-FITC (MZ3) BioLegend Cat #: 124808

Anti-human HLA-DR-PE (L243) BioLegend Cat #: 307606

Anti-human CD14-Pacific Blue (63D3) BioLegend Cat #: 367122

125

Anti-human CD1a-FITC (HI149) BioLegend Cat #: 300104

Anti-human CD80-BV 510 (2D10) BioLegend Cat #: 305234

Anti-human CD45-APC/Cy7 (HI30) BioLegend Cat #: 304014

Anti-human CD56-Pacific Blue (5.1H11) BioLegend Cat #: 362519

Anti-human CD45-Pacific Blue (2D1) BioLegend Cat #: 368539

Anti-human CD19-Pacific Blue (HIB19) BioLegend Cat #: 302224

Anti-human CD8a-APC/Cy7 (HIT8a) BioLegend Cat #: 300926

Anti-human CD3-PE/Cy7 (HIT3a) BioLegend Cat #: 300316

Anti-human CD4-PE (A161A1) BioLegend Cat #: 357404

Anti-human CD28-PerCP/Cy5.5 (CD28.2) BioLegend Cat #: 302922

Anti-human CD11b-PE/Cy5 (ICRF44) BioLegend Cat #: 301308

Anti-human CD3-Pacific Blue (UCHT1) BioLegend Cat #: 300417

Anti-human CD86-PE/Cy7 (IT2.2) BioLegend Cat #: 305422

Anti-human CD11c-APC (3.9) BioLegend Cat #: 301614

Anti-mouse/human FOXP3-PE (150D) BioLegend Cat #: 320008

Anti-mouse CD25-APC (PC61) BioLegend Cat #: 102012

Anti-mouse Ki-67-Pacific Blue (16A8) BioLegend Cat #: 652422

Anti-mouse CD3-FITC (17A2) BioLegend Cat #: 100204

Anti-mouse CD19-FITC (6D5) BioLegend Cat #: 115505

Anti-mouse/human CD11b-FITC (M1/70) BioLegend Cat #: 101205

Anti-mouse NK-1.1-FITC (PK136) BioLegend Cat #: 108705

Anti-mouse CD28-PE (37.51) BioLegend Cat #: 102106

Anti-mouse PD-1-PerCP/Cy5.5 (29F.1A12) BioLegend Cat #: 135208

Anti-mouse CD62L-PE/Cy7 (MEL-14) BioLegend Cat #: 104418

Anti-mouse CD8a-APC/Cy7 (53-6.7) BioLegend Cat #: 100714

Anti-mouse NK-1.1-Pacific Blue (PK136) BioLegend Cat #: 108721

Anti-mouse/human CD44-PE (IM7) BioLegend Cat #: 103008

Anti-mouse CD19-Pacific Blue (6D5) BioLegend Cat #: 115523

126

Anti-mouse CD4-BV510 (GK1.5) BioLegend Cat #: 100449

Anti-mouse CD86-PE (GL-1) BioLegend Cat #: 105007

Anti-mouse F4/80-PE/Cy5 (BM8) BioLegend Cat #: 123111

Anti-mouse CD80-PE/Cy7 (16-10A1) BioLegend Cat #: 104734

Anti-mouse/human CD11b-APC (M1/70) BioLegend Cat #: 101212

Anti-mouse I-A/I-E-APC/Cy7 (M5/114.15.2) BioLegend Cat #: 107628

Anti-mouse CD11c-BV510 (N418) BioLegend Cat #: 117338

Anti-mouse CD45-Pacific Blue (30-F11) BioLegend Cat #: 103126

Anti-mouse CD45-FITC (30-F11) BioLegend Cat #: 103108

Anti-mouse CD3-PerCP/Cy5.5 (17A2) BioLegend Cat #: 100218

Anti-mouse CD3ε-PE/Cy7 (145-2C11) BioLegend Cat #: 100320

Anti-mouse CD103-Pacific Blue (2E7) BioLegend Cat #: 121418

Anti-mouse Gr-1-PE/Cy7 (RB6-8C5) BioLegend Cat #: 108416

Anti-mouse CD45-BV510 (30-F11) BioLegend Cat #: 103138

Anti-mouse I-A/I-E-Pacific Blue (M5/114.15.2) BioLegend Cat #: 107619

Anti-mouse CD11b-Pacific Blue (M1/70) BioLegend Cat #: 101223

Rat IgG1, κ Isotype Ctrl-Pacific Blue (RTK2071) BioLegend Cat #: 400419

Rat IgG1, κ Isotype Ctrl-PE/Cy7 (RTK2071) BioLegend Cat #: 400415

Mouse IgG1, κ Isotype Ctrl-PE/Cy7 (MOPC-21) BioLegend Cat #: 400125

Mouse IgG1, κ Isotype Ctrl-PE (MOPC-21) BioLegend Cat #: 400111

Mouse IgG1, κ Isotype Ctrl-BV510 (MOPC-21) BioLegend Cat #: 400171

Mouse IgG1, κ Isotype Ctrl-PerCP/Cy5.5 BioLegend Cat #: 400149 (MOPC-21)

Rat IgG2a, κ Isotype Ctrl-PerCP/Cy5.5 BioLegend Cat #: 400531 (RTK2758)

Anti-mouse CD45-89Y (30-F11) Fluidigm Cat #: 3089005B

Anti-mouse Ly-6G-41Pr (1A8) Fluidigm Cat #: 3141008B

Anti-mouse CD11c-142Nd (N418) Fluidigm Cat #: 3142003B

127

Anti-mouse CD4-145Nd (RM4-5) Fluidigm Cat #: 3145002B

Anti-mouse F4/80-146Nd (BM8) Fluidigm Cat #: 3146008B

Anti-mouse Gr-1-147Sm (RB6-8C5) BioLegend/Fl Cat #: uidigm 108449/201147B

Anti-mouse CD11b-148Nd (M1/70) Fluidigm Cat #: 3148003B

Anti-mouse CD19-149Sm (6D5) Fluidigm Cat #: 3149002B

Anti-mouse CD25-150Nd (3C7) Fluidigm Cat #: 3150002B

Anti-mouse CD28-151Eu (37.51) Fluidigm Cat #: 3151005B

Anti-mouse CD3e-152Sm (145-2C11) Fluidigm Cat #: 3152004B

Anti-mouse CD274-153Eu (10F.9G2) Fluidigm Cat #: 3153016B

Anti-mouse CD152-154Sm (UC10-4B9) Fluidigm Cat #: 3154008B

Anti-mouse CD279-155Gd (RMP1-30) BioLegend/Fl Cat #: uidigm 109113/201155A

Anti-mouse CD335-156Gd (29A1.4) BioLegend/Fl Cat #: uidigm 137625/201156B

Anti-mouse Foxp3-158Gd (FJK-16s) Fluidigm Cat #: 3158003A

Anti-mouse RORgt-B2D-159Tb (B2D) Fluidigm Cat #: 3159019B

Anti-mouse CD62L-160Gd (MEL-14) Fluidigm Cat #: 3160008B

Anti-mouse Ki-67-161Dy (B56) Fluidigm Cat #: 3161007B

Anti-mouse Ly-6C-162Dy (HK1.4) Fluidigm Cat #: 3162014B

Anti-mouse CD197-164Dy (4B12) Fluidigm Cat #: 3164013A

Anti-mouse IFNg-165Ho (XMG1.2) Fluidigm Cat #: 3165003B

Anti-mouse IL-4-166Er (11B11) Fluidigm Cat #: 3166003B

Anti-mouse CD103-167Er (2E7) BioLegend/Fl Cat #: 121402/ uidigm 201167B

Anti-mouse CD8a-168Er (53-6.7) Fluidigm Cat #: 3168003B

Anti-mouse CD49b-170Er (HMa2) Fluidigm Cat #: 3170008B

Anti-mouse CD80-171Yb (16-10A1) Fluidigm Cat #: 3171008B

Anti-mouse CD86-172Yb (GL1) Fluidigm Cat #: 3172016B

128

Anti-mouse Granzyme B-173Yb (GB11) Fluidigm Cat #: 3173006B

Anti-mouse CD127-174Yb (A7R34) Fluidigm Cat #: 3174013B

Anti-mouse CD44-176Yb (IM7) BioLegend/Fl Cat #: uidigm 103051/201176B

Anti-mouse I-A/I-E-209Bi (M5/114.15.2) Fluidigm Cat #: 3209006B

Cell-ID™ Intercalator-Ir Fluidigm Cat #: 201192B

Cell-ID™ Cisplatin Fluidigm Cat #: 201064

H-2Ld MuLV gp70 Tetramer-APC MBL Cat #: TB-M521-2 International

IFN gamma Mouse ELISA Kit Thermo Fisher Cat #: BMS606 Scientific

Zombie Violet™ Fixable Viability Kit BioLegend Cat #: 423113

Zombie Aqua™ Fixable Viability Kit BioLegend Cat #: 423101

Zombie Green™ Fixable Viability Kit BioLegend Cat #: 423111

Cyto-Fast™ Fix/Perm Buffer Set BioLegend Cat #: 426803

True-Nuclear™ Transcription Factor Buffer Set BioLegend Cat #: 424401

IL-2 Mouse ELISA Kit Invitrogen Cat #: 88-7024-22

Mouse Fibrin Degradation Product D-Dimer LSBio Cat #: LS-F6179 ELISA Kit

Proteome Profiler Mouse Cytokine Array Kit, R&D systems Cat #: ARY006 Panel A

Dynabeads™ Untouched™ Human T Cells Kit Invitrogen Cat #: 11344D

Dynabeads™ Untouched™ Mouse T Cells Kit Invitrogen Cat #: 11413D

Dynabeads™ Human T-Activator CD3/CD28 for Invitrogen Cat #: 11161D T Cell Expansion and Activation

Dynabeads™ Mouse T-Activator CD3/CD28 for Invitrogen Cat #: 11452D T-Cell Expansion and Activation

SEE highly purified STA 100UG Toxin Cat #: ET404100UG Technologies Inc

TREATMENT SOURCE IDENTIFIER

129

InVivoMAb anti-mouse CD80 Bio X Cell Cat #: BE0024

InVivoMAb anti-mouse CD86 Bio X Cell Cat #: BE0025

InVivoPlus anti-mouse PD-1 Bio X Cell Cat #: BP0146

InVivoPlus anti-mouse CTLA-4 Bio X Cell Cat #: BP0032

InVivoPlus rat IgG2a isotype control Bio X Cell Cat #: BP0089

InVivoPlus polyclonal Armenian hamster IgG Bio X Cell Cat #: BP0091

REAGENT FOR DNA/RNA EXPERIMENTS SOURCE IDENTIFIER psiCHECK™-2 Vectors Promega Cat #: C8021 pPACKH1-XL packaging mix SBI Cat #: LV510A-1

Lipofectamine™ 3000 Transfection Reagent Invitrogen Cat #: L3000001 mirVana™ miRNA Isolation Kit Invitrogen Cat #: AM1560 mirVana Total RNA Isolation Kit Life Cat #: A27828 Technologies miScript II RT Kit Qiagen Cat #: 218161

LightCycler® 480 SYBR Green I Master Roche Cat #: 04707516001

PureLink™ Genomic DNA Mini Kit Invitrogen Cat #: K182001

Phusion® High-Fidelity DNA Polymerase New England Cat #: M0530S Biolabs

Q5® Site-Directed Mutagenesis Kit New England Cat #: E0554S Biolabs

In situ hybridization Kit BioChain Cat #: K2191020

Nuclear Fast Red™ Sigma Aldrich Cat #: N3020

Eukitt® Quick-hardening mounting medium Sigma Aldrich Cat #: 03989

SCRAMBLE-MIR MIRCURY LNA Probe Qiagen Cat #: YD00699004-BCG

U6, HSA/MMU/RNO MIRCURY LNA Probe Qiagen Cat #: YD00699002-BCG

HSA-MIR-424-5P MIRCURY LNA Probe Qiagen Cat #: YD00619854-BCG

130

PRIMERS AND SEQUENCES SEQUENCE IDENTIFIER (5’-3’) hsa-miR-424-5p CAGCAGCA NA ATTCATGTT TTGAA mmu-miR-322-5p CAGCAGCA NA ATTCATGTT TTGGA

U6 snRNA AAGGATGA NA CACGCAAA TTCG3

Human CD28 3’UTR miR-424 binding site TGCAGCA NA mutated sequence

Mouse Cd28 3’UTR miR-424 binding site TACGGAC NA mutated sequence

Human CD80 3’UTR miR-424 binding site Site 1: NA mutated sequence TATCATT Site 2: TATCATT Site 3: AGGTAAG Site 4: TCGTGCA

Mouse Cd80 3’UTR miR-424 binding site TCGTACT NA mutated sequence

OTHER REAGENTS SOURCE IDENTIFIER

Advanced DMEM F/12 Thermo Fisher Cat #: 12634-010 Scientific

Glutamax Life Cat #: 35050-061 Technologies

N-Acetylcysteine Sigma Aldrich Cat #: A9165-SG

Growth factor reduced Matrigel BD Cat #: 354230

Y-27632 (Rho Kinase inhibitor) EMD Cat #: 688001 Millipore

131

EGF Invitrogen Cat #: PMG8043

L-WRN cells ATCC Cat #: CRL-3276

SB202190 Sigma Aldrich Cat #: S7067

B27 supplement Invitrogen Cat #: 17504-044

N2 Supplement Invitrogen Cat #: 17502-048

TrypLE Express Invitrogen Cat #: 12605-036

Primocin Invivogen Cat #: ant-pm-1

Gastrin Sigma Aldrich Cat #: G9145

Nicotinamide Sigma Aldrich Cat #: N0636

A-83-01 Tocris Cat #: 2939

Prostaglandin E2 Cayman Cat #: 14010-1 Chemicals

D-Luciferin, Potassium Salt GOLDBIO Cat #: LUCK-1G

Matrigel™ Membrane Matrix Corning Cat #: 354234

H&E Staining Kit Abcam Cat #: ab245880

ProLong™ Gold Antifade Mountant Invitrogen Cat #: P36934

AR6 Buffer PerkinElmer Cat #: AR600250ML

AR9 Buffer PerkinElmer Cat #: AR900250ML

MICE SOURCE IDENTIFIER

Mouse: BALB/cJ The Jackson Stock No: 000651 Laboratory

Mouse: B6.129S2-Cd28tm1Mak/J The Jackson Stock No: 002666 Laboratory

132

Mouse: B6.129S4-Cd80tm1Shr Cd86tm2Shr/J The Jackson Stock No: 003610 Laboratory

Mouse: NOD.CB17-Prkdcscid/J The Jackson Stock No: 001303 Laboratory

Mouse: B6.CB17-Prkdcscid/SzJ The Jackson Stock No: 001913 Laboratory

Mouse: B6(Cg)-Tlr4tm1.2Karp/J The Jackson Stock No: 029015 Laboratory

Mouse: C57BL/6 The Charles Stock No: 027 River Laboratories

SOFTWARE SOURCE IDENTIFIER

GraphPad Prism 6 GraphPad https://www.graphpa Software d.com/

R NA https://www.r- project.org/

FlowJo BD https://www.flowjo. Biosciences com/

Morpheus The https://software.broa Broadinstitute dinstitute.org/morph eus/

CIBERSORT NA https://cibersort.stan ford.edu/

133

5. Chapter 5

Conclusions and Future Directions

134

Most cancer treatment regimens have targeted tumor cells through radiotherapy, chemotherapy, and surgical procedures. But recently, novel therapies have shifted from targeting tumor cells to enhancing the antitumor capabilities of the host immune system. Immunotherapy works with host immune cells, particularly the tumor-infiltrating lymphocytes (TILs), to attack tumor cells that were previously recognized as “self” antigens. Examples of immunotherapy include immune checkpoint blockade therapy (ICBT), antibody-based oncoprotein-targeted therapy, cancer vaccines, and adoptive immune cell transfer.

However, like all other cancer treatments, the efficacy of ICBT is limited by both intrinsic and acquired resistance. Indeed, ICBT is only effective in around half of the MSI-H subtype treatment-refractory CRC patients38. Unfortunately, most CRC patients (>85%) have MSS tumors39,40, which do not respond to ICBT that is used as an adjuvant treatment for late- stage treatment-refractory patients. The MSI-H tumors are generally associated with a higher mutational load and neoantigen number, thus are considered immunogenic194,195. Currently, the MSI/MSS phenotype is considered as a biomarker for ICBT response9,45. Results from a recent phase 2 clinical trial might change our perception that MSS tumors are not suitable for immunotherapy. The study demonstrated that all (20/20) MSI tumors and a portion of (4/15) MSS CRC tumors responded to ICBT when used as a neoadjuvant therapy226. The different ICBT response rates in the adjuvant and neoadjuvant conditions suggested that the tumor-immune cell interaction during CRC development and the traditional cancer treatment, such as surgery, chemotherapy, and oncogenic pathway targeted therapy, may affect ICBT efficacy in CRC patients. A mechanistic understanding of how CRC tumors become ICBT resistant and why the existing tumor-infiltrating immune cells are not functionally rescued by ICBT are critical for improving outcomes in both MSS- and MSI- CRC patients.

A significant impediment to progress in the cancer immunotherapy field is the availability of mouse models that recapitulate the heterogeneity of human malignancy and immune contexture within the tumor microenvironment242. So far, the best mouse model that faithfully recapitulates human cancer features is the patient-derived xenograft (PDX) models. However, only SCID mice and patient-matched humanized mice can serve as the host of PDX tumors due to the restriction of MHC compatibility. SCID mouse cannot provide immune microenvironment, excluding its use in cancer immunotherapy research. Humanized mice have an immune system. However, the MHC has to be matched between the immune cells of the humanized mice and tumors derived from patients. Our working experience with humanized mice 135 indicated that the high cost and complexity of establishing this model overwhelm the capacity of most research labs. These restrictions directly impeded the use of the humanized CRC model in cancer immunotherapy studies. The transgenic spontaneous tumor models provided another choice to recapitulate human tumors and study tumor pathogenesis. However, the unpredictable tumor formation site (small or large intestine), number, and stage (adenoma or adenocarcinoma) of spontaneous CRC mouse models limited their value in pre-clinical studies that require stable and comparable tumor baselines between all treatment arms. Therefore, our pre-clinical research primarily relies on the tumor cell line-derived syngeneic tumor models, which are routinely used in cancer immunotherapy studies.

Orthotopic tumor models growing in the tissue microenvironment of human malignancies have been widely used in pre-clinical studies. In the present thesis, we first contributed a feasible minimal invasive method to establish mouse orthotopic CRC with small animal endoscopy. This technique provided the foundation for other projects in our lab. We also demonstrated that tumors growing in the colorectal microenvironment are more immune responsive than the same background tumors growing in the subcutaneous tissue. The difference is probably due to the vibrant lymphatic structures in the colorectal tissue. Additionally, we validated the metastatic potential of surgically implanted mouse cecum tumors, which has been reported previously124,141,142. Collectively, the second chapter of our study pointed out that tumor location impacts immune response and disease progressing patterns in the tumor cell line-derived CRC mouse models. The subcutaneous model, which mimics a relatively sparse immune microenvironment, has its value in studying immunosuppressive mechanisms. Our orthotopic model, on the other hand, provides another useful option that better mimics CRC tumors with higher levels of immune infiltration in human patients. Moreover, the endoscopy-guided tumor implantation provided a platform to understand the CRC and colon microbiota interactions. Finally, the surgically implanted cecum tumors are more metastatic and can be used for mimicking late-stage CRC tumors. These results provided solid support for mouse model selection in following pre-clinical studies.

Although we have established and characterized mouse CRC models for different study aims, it is crucial to note that those tumor cell line-derived models cannot entirely recapitulate the pathogenesis of human tumors. Specifically, the cell line-derived tumors lack heterogeneity and the complex microenvironment of human CRC tumors, leading to different treatment responses than human tumors. Those drawbacks might impair the translational value of our work. In recent 136 years, standard protocols for establishing organoids from mouse spontaneous CRC tumors were established. Implantation of these tumor organoids either by surgical procedures or by endoscopy-guided injections have successfully established orthoptic tumors that are similar to human CRC243,244. Currently, we are working closely with our collaborators to isolate mouse tumor organoids for establishing this new generation immunocompetent orthotopic CRC model.

ICBT, such as anti-CTLA-4 and anti-PD-1/PD-L1, are predominantly used as second- and third-line therapies for treating patients with heavily pretreated tumors 9,104,143. The interactions between first-line therapy may influence tumor response to subsequently administered ICBTs due to tumor evolution and heterogeneity. In most patients with solid tumors, common interventions before ICBT include resection of primary tumors with concurrent resection of draining lymph nodes followed by administration of chemotherapies and/or targeted therapies9,146,147. However, minimal information is known about whether these interventions will impact tumor response to ICBT.

Figure 5.1. Graphic summary of Chapter 3.

The effects of tumor-draining lymph nodes (TdLNs) resection and a combination of cytotoxic chemotherapy on immune checkpoint blockade therapies are evaluated in this study. TdLNs 137 resection was adverse in eliciting an antitumor immune response in early-stage tumors, but not in late-stage tumors. Further, sequential treatments with cytotoxic chemotherapy and immunotherapy showed better tumor control compared to concurrent combinatorial therapies. This figure is reprinted with permission from Zhao et al., iScience, 2020. In the third chapter of our study, we investigated the effects of TdLNs on anti-tumor immune responses. Our data demonstrated that TdLNs are necessary for early-stage tumors to initiate a robust anti-tumor immune response. Resection of TdLNs prior to tumor cell inoculation significantly accelerated tumor development via weakening the anti-tumor immunity. However, TdLNs resection in late-stage tumor models didn’t affect anti-tumor immunity nor ICBT efficacy in recurrent tumors. Mechanistic studies revealed that TdLNs shift from an immunogenic environment to an immunotolerant environment during tumor progression. Meanwhile, the amount of tumor antigen-specific T-cells in peripheral lymphatic tissues accumulated during tumor progression. Loss of TdLNs could not significantly influence the number of tumor antigen- specific T-cells that are responsible for controlling tumor recurrence. These data suggested that resection of the immune-privileged TdLNs together with primary tumor during first-line surgical treatment may not attenuate the anti-tumor immune response for controlling tumor recurrence and determining second-line ICBT response.

In the second part of chapter 3, we studied whether the timing of chemotherapy and immunotherapy in combination would affect treatment response. It is widely accepted that chemotherapies have both immunogenic and immunosuppressive effects. Thus, we hypothesize that different chemo- and immunotherapy combination schedules have different efficacies. We showed that using 5-fluorouracil (5-FU) chemotherapy as induction therapy, followed by ICBT as a maintenance treatment, showed better responses than adding ICB concurrently with 5-FU. Immune profiling of tumors revealed that using 5-FU as induction treatment successfully stimulated tumor immunogenicity, while limited chemotherapy-induced T-cell depletion. However, concurrent 5-FU and ICBT combination failed to elicit T-cell responses due to the strong T-cell depletion effects of 5-FU. Collectively, our study showed that the balance between the immunostimulatory and immunosuppressive effects of traditional cytotoxic treatment would determine the chemo- and immunotherapy combination treatment response. These results will provide essential considerations for designing successful immunotherapy strategies in clinical conditions.

Notably, it is important to note that the differences between mouse tumor models and human cancers may limit the translational value of our study. Our pre-clinical studies were 138 mainly performed on mouse tumor models. Although mouse models are heavily used in pre- clinical studies, mouse tumor development, numbers of tumor-draining lymph nodes, and disease kinetics are different from human clinical conditions. Although we refined our surgical methods and mouse models to reflect human conditions closely, the differences in mouse anatomy and physiology may potentially limit the translational value of our conclusions. Besides, our study was limited only to the commonly used chemotherapies such as 5-FU in pre-clinical models; therefore, further clinical trials are needed to validate our findings acquired in pre-clinical settings.

The fourth chapter of the thesis addressed the molecular regulation of CRC immune response. Recent studies have demonstrated that the CD28-CD80/86 co-stimulatory pathways determine the efficacy of ICBT18,19. Administration of anti-PD-1 cannot revitalize T cells without the presence of functional CD28-CD80/86 signaling18,19. Our preliminary experiments and publications from other groups have shown that CRC tumors with larget amounts of T-cell infiltration do not necessarily respond to ICBT40,106,245. Therefore, we hypothesized that some of the CRC tumor-infiltrating T cells might not have an intact CD28-CD80/86 signaling, leading to ICBT resistance. Our data showed that the expression of CD28 on tumor-infiltrating T cells and CD80/86 on tumor-infiltrating DCs are highly variable and can be weak or absent on tumor- infiltrating T cells and DCs in both MSI- and MSS-CRC tumors. Mechanistically, hypoxia in the tumor microenvironment could stimulate tumor cell miR-424 expression, which is subsequently transferred to tumor-infiltrating immune cells via TEVs. The TEVs downregulated CD28 and CD80 expression on tumor-infiltrating immune cells via the transferred miR-424. Blocking the functional miR-424 in TEVs significantly reduced TEVs mediated tumor immune suppression via rescuing the CD28-CD80/86 expression. More importantly, tumors without functional miR-424 were more vulnerable to ICBT than tumors with high levels of functional miR-424.

139

Figure 5.2. Graphic summary of Chapter 4.

During tumor development, the expression of miR-424 increases, while the expression of CD28/CD80 decreases. Tumor cells derived miR-424 that are packaged in tumor cell-derived extracellular vesicles (TEVs) suppress the anti-tumor immune response by disrupting the T-cell costimulation. TEVs with blocked miR-424 (MTEVs) show strong immunogenicity, thus enhancing immune checkpoint blockades efficiency. Given that TEVs contain tumor-specific antigens, we sought to unleash the immunogenicity of TEVs by depleting their immunosuppressive factors, such as the functional miR-424. Notably, systemic administration of the functional miR-424 depleted TEVs (MTEVs) successfully induced an anti-tumor immune response in naïve mice and prevented tumor formation. Combination treatment of the MTEVs and ICBT resulted in better survival than the ICBT alone in advanced CRC mouse models. Several recent publications echoed our concept of using MTEVs to treat cancers. Irradiated or adjuvant-coated TEVs also showed strong immunogenicity in pre-clinical cancer models100,101,231,246. Combing the different TEVs modification methods to maximize the anti-tumor potential of MTEVs will be a potential method to convert ICBT resistant late-stage tumors to ICBT sensitive ones.

In summary, the present thesis promoted our understanding of CRC ICBT resistance through multiple aspects including reporting and characterizing the immunological and 140 pathological features of different CRC models, evaluating the impacts of first-line treatment on cancer immunotherapy, identifying a new mechanism causing T-cell anergy in the tumor microenvironment, and developing a novel modified TEVs based adjuvant treatment. Finally, to further expand our understanding of CRC immune regulation and translate our findings into clinical trials, we will develop and standardize the tumor organoids derived orthotopic CRC models for overcoming the restrictions of cell-line based tumor models. We will also strengthen the anti-tumor efficacy of our modified TEVs by removing multiple immunosuppressive cargos from them and incorporating immunostimulatory factors (such as adjuvant) into them.

141

Bibliography 1 DeVita, V. T., Jr. & Chu, E. A history of cancer chemotherapy. Cancer research 68, 8643- 8653, doi:10.1158/0008-5472.CAN-07-6611 (2008). 2 Gajewski, T. F., Schreiber, H. & Fu, Y. X. Innate and adaptive immune cells in the tumor microenvironment. Nature immunology 14, 1014-1022, doi:10.1038/ni.2703 (2013). 3 Schreiber, R. D., Old, L. J. & Smyth, M. J. Cancer immunoediting: integrating immunity's roles in cancer suppression and promotion. Science 331, 1565-1570, doi:10.1126/science.1203486 (2011). 4 Cerottini, J. C. & Brunner, K. T. Cell-mediated cytotoxicity, allograft rejection, and tumor immunity. Advances in immunology 18, 67-132, doi:10.1016/s0065-2776(08)60308-9 (1974). 5 Vesely, M. D., Kershaw, M. H., Schreiber, R. D. & Smyth, M. J. Natural innate and adaptive immunity to cancer. Annual review of immunology 29, 235-271, doi:10.1146/annurev- immunol-031210-101324 (2011). 6 O'Donnell, J. S., Teng, M. W. L. & Smyth, M. J. Cancer immunoediting and resistance to T cell-based immunotherapy. Nature reviews. Clinical oncology 16, 151-167, doi:10.1038/s41571-018-0142-8 (2019). 7 Dunn, G. P., Old, L. J. & Schreiber, R. D. The three Es of cancer immunoediting. Annual review of immunology 22, 329-360, doi:10.1146/annurev.immunol.22.012703.104803 (2004). 8 Le, D. T. et al. PD-1 Blockade in Tumors with Mismatch-Repair Deficiency. The New England journal of medicine 372, 2509-2520, doi:10.1056/NEJMoa1500596 (2015). 9 Le, D. T. et al. Mismatch repair deficiency predicts response of solid tumors to PD-1 blockade. Science 357, 409-413, doi:10.1126/science.aan6733 (2017). 10 Larkin, J. et al. Combined Nivolumab and Ipilimumab or Monotherapy in Untreated Melanoma. The New England journal of medicine 373, 23-34, doi:10.1056/NEJMoa1504030 (2015). 11 Robert, C. et al. Nivolumab in previously untreated melanoma without BRAF mutation. The New England journal of medicine 372, 320-330, doi:10.1056/NEJMoa1412082 (2015). 12 Robert, C. et al. Pembrolizumab versus ipilimumab in advanced melanoma (KEYNOTE- 006): post-hoc 5-year results from an open-label, multicentre, randomised, controlled, phase 3 study. The Lancet. Oncology 20, 1239-1251, doi:10.1016/S1470-2045(19)30388- 2 (2019). 13 Sharma, P., Hu-Lieskovan, S., Wargo, J. A. & Ribas, A. Primary, Adaptive, and Acquired Resistance to Cancer Immunotherapy. Cell 168, 707-723, doi:10.1016/j.cell.2017.01.017 (2017). 14 Kalbasi, A. & Ribas, A. Tumour-intrinsic resistance to immune checkpoint blockade. Nature reviews. Immunology 20, 25-39, doi:10.1038/s41577-019-0218-4 (2020). 15 Wei, S. C., Duffy, C. R. & Allison, J. P. Fundamental Mechanisms of Immune Checkpoint Blockade Therapy. Cancer discovery 8, 1069-1086, doi:10.1158/2159-8290.CD-18-0367 (2018). 16 Francisco, L. M., Sage, P. T. & Sharpe, A. H. The PD-1 pathway in tolerance and autoimmunity. Immunological reviews 236, 219-242, doi:10.1111/j.1600- 065X.2010.00923.x (2010).

142

17 Honda, T. et al. Tuning of antigen sensitivity by T cell receptor-dependent negative feedback controls T cell effector function in inflamed tissues. Immunity 40, 235-247, doi:10.1016/j.immuni.2013.11.017 (2014). 18 Hui, E. et al. T cell costimulatory receptor CD28 is a primary target for PD-1-mediated inhibition. Science 355, 1428-1433, doi:10.1126/science.aaf1292 (2017). 19 Kamphorst, A. O. et al. Rescue of exhausted CD8 T cells by PD-1-targeted therapies is CD28-dependent. Science 355, 1423-1427, doi:10.1126/science.aaf0683 (2017). 20 Fenwick, C. et al. Tumor suppression of novel anti-PD-1 antibodies mediated through CD28 costimulatory pathway. The Journal of experimental medicine 216, 1525-1541, doi:10.1084/jem.20182359 (2019). 21 Dong, H., Zhu, G., Tamada, K. & Chen, L. B7-H1, a third member of the B7 family, co- stimulates T-cell proliferation and interleukin-10 secretion. Nature medicine 5, 1365-1369, doi:10.1038/70932 (1999). 22 Latchman, Y. et al. PD-L2 is a second ligand for PD-1 and inhibits T cell activation. Nature immunology 2, 261-268, doi:10.1038/85330 (2001). 23 Muhlbauer, M. et al. PD-L1 is induced in hepatocytes by viral infection and by interferon- alpha and -gamma and mediates T cell apoptosis. Journal of hepatology 45, 520-528, doi:10.1016/j.jhep.2006.05.007 (2006). 24 Chen, J. et al. Interferon-gamma-induced PD-L1 surface expression on human oral squamous carcinoma via PKD2 signal pathway. Immunobiology 217, 385-393, doi:10.1016/j.imbio.2011.10.016 (2012). 25 Zhao, X. & Subramanian, S. Oncogenic pathways that affect antitumor immune response and immune checkpoint blockade therapy. Pharmacology & therapeutics 181, 76-84, doi:10.1016/j.pharmthera.2017.07.004 (2018). 26 Loke, P. & Allison, J. P. PD-L1 and PD-L2 are differentially regulated by Th1 and Th2 cells. Proceedings of the National Academy of Sciences of the United States of America 100, 5336-5341, doi:10.1073/pnas.0931259100 (2003). 27 Esensten, J. H., Helou, Y. A., Chopra, G., Weiss, A. & Bluestone, J. A. CD28 Costimulation: From Mechanism to Therapy. Immunity 44, 973-988, doi:10.1016/j.immuni.2016.04.020 (2016). 28 Linsley, P. S. et al. Human B7-1 (CD80) and B7-2 (CD86) bind with similar avidities but distinct kinetics to CD28 and CTLA-4 receptors. Immunity 1, 793-801, doi:10.1016/s1074- 7613(94)80021-9 (1994). 29 Lee, K. M. et al. Molecular basis of T cell inactivation by CTLA-4. Science 282, 2263-2266, doi:10.1126/science.282.5397.2263 (1998). 30 Schneider, H. et al. Reversal of the TCR stop signal by CTLA-4. Science 313, 1972-1975, doi:10.1126/science.1131078 (2006). 31 Barnes, M. J. et al. CTLA-4 promotes Foxp3 induction and regulatory T cell accumulation in the intestinal lamina propria. Mucosal immunology 6, 324-334, doi:10.1038/mi.2012.75 (2013). 32 Simpson, T. R. et al. Fc-dependent depletion of tumor-infiltrating regulatory T cells co- defines the efficacy of anti-CTLA-4 therapy against melanoma. The Journal of experimental medicine 210, 1695-1710, doi:10.1084/jem.20130579 (2013). 33 Wei, S. C. et al. Distinct Cellular Mechanisms Underlie Anti-CTLA-4 and Anti-PD-1 Checkpoint Blockade. Cell 170, 1120-1133 e1117, doi:10.1016/j.cell.2017.07.024 (2017).

143

34 Sharma, A. et al. Anti-CTLA-4 Immunotherapy Does Not Deplete FOXP3(+) Regulatory T Cells (Tregs) in Human Cancers. Clinical cancer research : an official journal of the American Association for Cancer Research 25, 1233-1238, doi:10.1158/1078-0432.CCR- 18-0762 (2019). 35 MacGregor, H. L. & Ohashi, P. S. Molecular Pathways: Evaluating the Potential for B7-H4 as an Immunoregulatory Target. Clinical cancer research : an official journal of the American Association for Cancer Research 23, 2934-2941, doi:10.1158/1078-0432.CCR- 15-2440 (2017). 36 Pasero, C. & Olive, D. Interfering with coinhibitory molecules: BTLA/HVEM as new targets to enhance anti-tumor immunity. Immunology letters 151, 71-75, doi:10.1016/j.imlet.2013.01.008 (2013). 37 Tinoco, R. et al. PSGL-1 Is an Immune Checkpoint Regulator that Promotes T Cell Exhaustion. Immunity 44, 1190-1203, doi:10.1016/j.immuni.2016.04.015 (2016). 38 Oliveira, A. F., Bretes, L. & Furtado, I. Review of PD-1/PD-L1 Inhibitors in Metastatic dMMR/MSI-H Colorectal Cancer. Frontiers in oncology 9, 396, doi:10.3389/fonc.2019.00396 (2019). 39 Kim, C. G. et al. Effects of microsatellite instability on recurrence patterns and outcomes in colorectal cancers. British journal of cancer 115, 25-33, doi:10.1038/bjc.2016.161 (2016). 40 Mlecnik, B. et al. Integrative Analyses of Colorectal Cancer Show Immunoscore Is a Stronger Predictor of Patient Survival Than Microsatellite Instability. Immunity 44, 698- 711, doi:10.1016/j.immuni.2016.02.025 (2016). 41 Lawrence, M. S. et al. Mutational heterogeneity in cancer and the search for new cancer- associated genes. Nature 499, 214-218, doi:10.1038/nature12213 (2013). 42 McGranahan, N. et al. Clonal neoantigens elicit T cell immunoreactivity and sensitivity to immune checkpoint blockade. Science 351, 1463-1469, doi:10.1126/science.aaf1490 (2016). 43 Van Allen, E. M. et al. Genomic correlates of response to CTLA-4 blockade in metastatic melanoma. Science 350, 207-211, doi:10.1126/science.aad0095 (2015). 44 Llosa, N. J. et al. The vigorous immune microenvironment of microsatellite instable colon cancer is balanced by multiple counter-inhibitory checkpoints. Cancer discovery 5, 43-51, doi:10.1158/2159-8290.CD-14-0863 (2015). 45 Mandal, R. et al. Genetic diversity of tumors with mismatch repair deficiency influences anti-PD-1 immunotherapy response. Science 364, 485-491, doi:10.1126/science.aau0447 (2019). 46 Gao, J. et al. Loss of IFN-gamma Pathway Genes in Tumor Cells as a Mechanism of Resistance to Anti-CTLA-4 Therapy. Cell 167, 397-404 e399, doi:10.1016/j.cell.2016.08.069 (2016). 47 Sucker, A. et al. Genetic evolution of T-cell resistance in the course of melanoma progression. Clinical cancer research : an official journal of the American Association for Cancer Research 20, 6593-6604, doi:10.1158/1078-0432.CCR-14-0567 (2014). 48 McGranahan, N. et al. Allele-Specific HLA Loss and Immune Escape in Lung Cancer Evolution. Cell 171, 1259-1271 e1211, doi:10.1016/j.cell.2017.10.001 (2017). 49 Peng, W. et al. Loss of PTEN Promotes Resistance to T Cell-Mediated Immunotherapy. Cancer discovery 6, 202-216, doi:10.1158/2159-8290.CD-15-0283 (2016).

144

50 George, S. et al. Loss of PTEN Is Associated with Resistance to Anti-PD-1 Checkpoint Blockade Therapy in Metastatic Uterine Leiomyosarcoma. Immunity 46, 197-204, doi:10.1016/j.immuni.2017.02.001 (2017). 51 Zhao, J. et al. Immune and genomic correlates of response to anti-PD-1 immunotherapy in glioblastoma. Nature medicine 25, 462-469, doi:10.1038/s41591-019-0349-y (2019). 52 Parsa, A. T. et al. Loss of tumor suppressor PTEN function increases B7-H1 expression and immunoresistance in glioma. Nature medicine 13, 84-88, doi:10.1038/nm1517 (2007). 53 Liu, C. et al. The superior efficacy of anti-PD-1/PD-L1 immunotherapy in KRAS-mutant non-small cell lung cancer that correlates with an inflammatory phenotype and increased immunogenicity. Cancer letters 470, 95-105, doi:10.1016/j.canlet.2019.10.027 (2020). 54 Sumimoto, H., Takano, A., Teramoto, K. & Daigo, Y. RAS-Mitogen-Activated Protein Kinase Signal Is Required for Enhanced PD-L1 Expression in Human Lung Cancers. PloS one 11, e0166626, doi:10.1371/journal.pone.0166626 (2016). 55 Zdanov, S. et al. Mutant KRAS Conversion of Conventional T Cells into Regulatory T Cells. Cancer immunology research 4, 354-365, doi:10.1158/2326-6066.CIR-15-0241 (2016). 56 Roh, W. et al. Integrated molecular analysis of tumor biopsies on sequential CTLA-4 and PD-1 blockade reveals markers of response and resistance. Science translational medicine 9, doi:10.1126/scitranslmed.aah3560 (2017). 57 Davoli, T., Uno, H., Wooten, E. C. & Elledge, S. J. Tumor aneuploidy correlates with markers of immune evasion and with reduced response to immunotherapy. Science 355, doi:10.1126/science.aaf8399 (2017). 58 Roberts, E. W. et al. Critical Role for CD103(+)/CD141(+) Dendritic Cells Bearing CCR7 for Tumor Antigen Trafficking and Priming of T Cell Immunity in Melanoma. Cancer cell 30, 324-336, doi:10.1016/j.ccell.2016.06.003 (2016). 59 Salmon, H. et al. Expansion and Activation of CD103(+) Dendritic Cell Progenitors at the Tumor Site Enhances Tumor Responses to Therapeutic PD-L1 and BRAF Inhibition. Immunity 44, 924-938, doi:10.1016/j.immuni.2016.03.012 (2016). 60 de Mingo Pulido, A. et al. TIM-3 Regulates CD103(+) Dendritic Cell Function and Response to Chemotherapy in Breast Cancer. Cancer cell 33, 60-74 e66, doi:10.1016/j.ccell.2017.11.019 (2018). 61 Binnewies, M. et al. Unleashing Type-2 Dendritic Cells to Drive Protective Antitumor CD4(+) T Cell Immunity. Cell 177, 556-571 e516, doi:10.1016/j.cell.2019.02.005 (2019). 62 Motz, G. T. et al. Tumor endothelium FasL establishes a selective immune barrier promoting tolerance in tumors. Nature medicine 20, 607-615, doi:10.1038/nm.3541 (2014). 63 Buckanovich, R. J. et al. Endothelin B receptor mediates the endothelial barrier to T cell homing to tumors and disables immune therapy. Nature medicine 14, 28-36, doi:10.1038/nm1699 (2008). 64 Spranger, S., Dai, D., Horton, B. & Gajewski, T. F. Tumor-Residing Batf3 Dendritic Cells Are Required for Effector T Cell Trafficking and Adoptive T Cell Therapy. Cancer cell 31, 711- 723 e714, doi:10.1016/j.ccell.2017.04.003 (2017). 65 Feig, C. et al. Targeting CXCL12 from FAP-expressing carcinoma-associated fibroblasts synergizes with anti-PD-L1 immunotherapy in pancreatic cancer. Proceedings of the National Academy of Sciences of the United States of America 110, 20212-20217, doi:10.1073/pnas.1320318110 (2013).

145

66 Cremasco, V. et al. FAP Delineates Heterogeneous and Functionally Divergent Stromal Cells in Immune-Excluded Breast Tumors. Cancer immunology research 6, 1472-1485, doi:10.1158/2326-6066.CIR-18-0098 (2018). 67 Salmon, H. et al. Matrix architecture defines the preferential localization and migration of T cells into the stroma of human lung tumors. The Journal of clinical investigation 122, 899-910, doi:10.1172/JCI45817 (2012). 68 Yang, X. et al. FAP Promotes Immunosuppression by Cancer-Associated Fibroblasts in the Tumor Microenvironment via STAT3-CCL2 Signaling. Cancer research 76, 4124-4135, doi:10.1158/0008-5472.CAN-15-2973 (2016). 69 Chen, L., Qiu, X., Wang, X. & He, J. FAP positive fibroblasts induce immune checkpoint blockade resistance in colorectal cancer via promoting immunosuppression. Biochemical and biophysical research communications, doi:10.1016/j.bbrc.2017.03.039 (2017). 70 Weber, R. et al. Myeloid-Derived Suppressor Cells Hinder the Anti-Cancer Activity of Immune Checkpoint Inhibitors. Frontiers in immunology 9, 1310, doi:10.3389/fimmu.2018.01310 (2018). 71 Kaneda, M. M. et al. PI3Kgamma is a molecular switch that controls immune suppression. Nature 539, 437-442, doi:10.1038/nature19834 (2016). 72 De Henau, O. et al. Overcoming resistance to checkpoint blockade therapy by targeting PI3Kgamma in myeloid cells. Nature 539, 443-447, doi:10.1038/nature20554 (2016). 73 Park, B., Yee, C. & Lee, K. M. The effect of radiation on the immune response to cancers. International journal of molecular sciences 15, 927-943, doi:10.3390/ijms15010927 (2014). 74 Bracci, L., Schiavoni, G., Sistigu, A. & Belardelli, F. Immune-based mechanisms of cytotoxic chemotherapy: implications for the design of novel and rationale-based combined treatments against cancer. Cell death and differentiation 21, 15-25, doi:10.1038/cdd.2013.67 (2014). 75 Pfirschke, C. et al. Immunogenic Chemotherapy Sensitizes Tumors to Checkpoint Blockade Therapy. Immunity 44, 343-354, doi:10.1016/j.immuni.2015.11.024 (2016). 76 Aoto, K. et al. Immunogenic tumor cell death induced by chemotherapy in patients with breast cancer and esophageal squamous cell carcinoma. Oncology reports 39, 151-159, doi:10.3892/or.2017.6097 (2018). 77 Hernandez, C., Huebener, P. & Schwabe, R. F. Damage-associated molecular patterns in cancer: a double-edged sword. Oncogene 35, 5931-5941, doi:10.1038/onc.2016.104 (2016). 78 O'Donnell, T. et al. Chemotherapy weakly contributes to predicted neoantigen expression in ovarian cancer. BMC cancer 18, 87, doi:10.1186/s12885-017-3825-0 (2018). 79 Alizadeh, D. & Larmonier, N. Chemotherapeutic targeting of cancer-induced immunosuppressive cells. Cancer research 74, 2663-2668, doi:10.1158/0008-5472.CAN- 14-0301 (2014). 80 Sharabi, A. B., Lim, M., DeWeese, T. L. & Drake, C. G. Radiation and checkpoint blockade immunotherapy: radiosensitisation and potential mechanisms of synergy. The Lancet. Oncology 16, e498-509, doi:10.1016/S1470-2045(15)00007-8 (2015). 81 Gupta, A. et al. Radiotherapy promotes tumor-specific effector CD8+ T cells via dendritic cell activation. Journal of immunology 189, 558-566, doi:10.4049/jimmunol.1200563 (2012).

146

82 Gameiro, S. R. et al. Radiation-induced immunogenic modulation of tumor enhances antigen processing and calreticulin exposure, resulting in enhanced T-cell killing. Oncotarget 5, 403-416, doi:10.18632/oncotarget.1719 (2014). 83 Cocucci, E. & Meldolesi, J. Ectosomes and exosomes: shedding the confusion between extracellular vesicles. Trends in cell biology 25, 364-372, doi:10.1016/j.tcb.2015.01.004 (2015). 84 Osaki, M. & Okada, F. Exosomes and Their Role in Cancer Progression. Yonago acta medica 62, 182-190, doi:10.33160/yam.2019.06.002 (2019). 85 Teo, B. H. & Wong, S. H. MHC class II-associated invariant chain (Ii) modulates dendritic cells-derived microvesicles (DCMV)-mediated activation of microglia. Biochemical and biophysical research communications 400, 673-678, doi:10.1016/j.bbrc.2010.08.126 (2010). 86 Melo, S. A. et al. Cancer exosomes perform cell-independent microRNA biogenesis and promote tumorigenesis. Cancer cell 26, 707-721, doi:10.1016/j.ccell.2014.09.005 (2014). 87 Abd Elmageed, Z. Y. et al. Neoplastic reprogramming of patient-derived adipose stem cells by prostate cancer cell-associated exosomes. Stem cells 32, 983-997, doi:10.1002/stem.1619 (2014). 88 Le, M. T. et al. miR-200-containing extracellular vesicles promote breast cancer cell metastasis. The Journal of clinical investigation 124, 5109-5128, doi:10.1172/JCI75695 (2014). 89 Tominaga, N. et al. Brain metastatic cancer cells release microRNA-181c-containing extracellular vesicles capable of destructing blood-brain barrier. Nature communications 6, 6716, doi:10.1038/ncomms7716 (2015). 90 Zhou, W. et al. Cancer-secreted miR-105 destroys vascular endothelial barriers to promote metastasis. Cancer cell 25, 501-515, doi:10.1016/j.ccr.2014.03.007 (2014). 91 Yokoi, A. et al. Malignant extracellular vesicles carrying MMP1 mRNA facilitate peritoneal dissemination in ovarian cancer. Nature communications 8, 14470, doi:10.1038/ncomms14470 (2017). 92 Ricklefs, F. L. et al. Immune evasion mediated by PD-L1 on glioblastoma-derived extracellular vesicles. Science advances 4, eaar2766, doi:10.1126/sciadv.aar2766 (2018). 93 Chen, G. et al. Exosomal PD-L1 contributes to immunosuppression and is associated with anti-PD-1 response. Nature 560, 382-386, doi:10.1038/s41586-018-0392-8 (2018). 94 Timaner, M. et al. Microparticles from tumors exposed to radiation promote immune evasion in part by PD-L1. Oncogene 39, 187-203, doi:10.1038/s41388-019-0971-7 (2020). 95 Poggio, M. et al. Suppression of Exosomal PD-L1 Induces Systemic Anti-tumor Immunity and Memory. Cell 177, 414-427 e413, doi:10.1016/j.cell.2019.02.016 (2019). 96 Kim, D. H. et al. Exosomal PD-L1 promotes tumor growth through immune escape in non- small cell lung cancer. Experimental & molecular medicine 51, 1-13, doi:10.1038/s12276- 019-0295-2 (2019). 97 Vignard, V. et al. MicroRNAs in Tumor Exosomes Drive Immune Escape in Melanoma. Cancer immunology research 8, 255-267, doi:10.1158/2326-6066.CIR-19-0522 (2020). 98 Kalluri, R. & LeBleu, V. S. The biology, function, and biomedical applications of exosomes. Science 367, doi:10.1126/science.aau6977 (2020). 99 Liu, H. et al. Dendritic cells loaded with tumor derived exosomes for cancer immunotherapy. Oncotarget 9, 2887-2894, doi:10.18632/oncotarget.20812 (2018).

147

100 Wan, C. et al. Irradiated tumor cell-derived microparticles mediate tumor eradication via cell killing and immune reprogramming. Science advances 6, eaay9789, doi:10.1126/sciadv.aay9789 (2020). 101 Diamond, J. M. et al. Exosomes Shuttle TREX1-Sensitive IFN-Stimulatory dsDNA from Irradiated Cancer Cells to DCs. Cancer immunology research 6, 910-920, doi:10.1158/2326-6066.CIR-17-0581 (2018). 102 Siegel, R. L., Miller, K. D. & Jemal, A. Cancer statistics, 2020. CA: a cancer journal for clinicians 70, 7-30, doi:10.3322/caac.21590 (2020). 103 Singh, P. P., Sharma, P. K., Krishnan, G. & Lockhart, A. C. Immune checkpoints and immunotherapy for colorectal cancer. Gastroenterology report 3, 289-297, doi:10.1093/gastro/gov053 (2015). 104 Topalian, S. L. et al. Safety, activity, and immune correlates of anti-PD-1 antibody in cancer. The New England journal of medicine 366, 2443-2454, doi:10.1056/NEJMoa1200690 (2012). 105 Brahmer, J. R. et al. Phase I study of single-agent anti-programmed death-1 (MDX-1106) in refractory solid tumors: safety, clinical activity, pharmacodynamics, and immunologic correlates. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 28, 3167-3175, doi:10.1200/JCO.2009.26.7609 (2010). 106 Pages, F. et al. International validation of the consensus Immunoscore for the classification of colon cancer: a prognostic and accuracy study. Lancet 391, 2128-2139, doi:10.1016/S0140-6736(18)30789-X (2018). 107 Palucka, A. K. & Coussens, L. M. The Basis of Oncoimmunology. Cell 164, 1233-1247, doi:10.1016/j.cell.2016.01.049 (2016). 108 Fesnak, A. D., June, C. H. & Levine, B. L. Engineered T cells: the promise and challenges of cancer immunotherapy. Nature reviews. Cancer 16, 566-581, doi:10.1038/nrc.2016.97 (2016). 109 Sharma, P. & Allison, J. P. The future of immune checkpoint therapy. Science 348, 56-61, doi:10.1126/science.aaa8172 (2015). 110 Borghaei, H. et al. Nivolumab versus Docetaxel in Advanced Nonsquamous Non-Small-Cell Lung Cancer. The New England journal of medicine 373, 1627-1639, doi:10.1056/NEJMoa1507643 (2015). 111 Motzer, R. J. et al. Nivolumab versus Everolimus in Advanced Renal-Cell Carcinoma. The New England journal of medicine 373, 1803-1813, doi:10.1056/NEJMoa1510665 (2015). 112 Xiao, Y. & Freeman, G. J. The microsatellite instable subset of colorectal cancer is a particularly good candidate for checkpoint blockade immunotherapy. Cancer discovery 5, 16-18, doi:10.1158/2159-8290.CD-14-1397 (2015). 113 Jure-Kunkel, M. et al. Synergy between chemotherapeutic agents and CTLA-4 blockade in preclinical tumor models. Cancer immunology, immunotherapy : CII 62, 1533-1545, doi:10.1007/s00262-013-1451-5 (2013). 114 Zhao, D. X. et al. Enhanced antitumor immunity is elicited by adenovirus-mediated gene transfer of CCL21 and IL-15 in murine colon carcinomas. Cellular immunology 289, 155- 161, doi:10.1016/j.cellimm.2014.03.020 (2014). 115 Lehmann, B. et al. Tumor location determines tissue-specific recruitment of tumor- associated macrophages and antibody-dependent immunotherapy response. Science immunology 2, doi:10.1126/sciimmunol.aah6413 (2017).

148

116 Zitvogel, L., Pitt, J. M., Daillere, R., Smyth, M. J. & Kroemer, G. Mouse models in oncoimmunology. Nature reviews. Cancer 16, 759-773, doi:10.1038/nrc.2016.91 (2016). 117 Masugi, Y. et al. Tumour CD274 (PD-L1) expression and T cells in colorectal cancer. Gut 66, 1463-1473, doi:10.1136/gutjnl-2016-311421 (2017). 118 Deschoolmeester, V., Baay, M., Lardon, F., Pauwels, P. & Peeters, M. Immune Cells in Colorectal Cancer: Prognostic Relevance and Role of MSI. Cancer microenvironment : official journal of the International Cancer Microenvironment Society 4, 377-392, doi:10.1007/s12307-011-0068-5 (2011). 119 Selby, M. J. et al. Preclinical Development of Ipilimumab and Nivolumab Combination Immunotherapy: Mouse Tumor Models, In Vitro Functional Studies, and Cynomolgus Macaque Toxicology. PloS one 11, e0161779, doi:10.1371/journal.pone.0161779 (2016). 120 Bibby, M. C. Orthotopic models of cancer for preclinical drug evaluation: advantages and disadvantages. European journal of cancer 40, 852-857, doi:10.1016/j.ejca.2003.11.021 (2004). 121 Mittal, V. K., Bhullar, J. S. & Jayant, K. Animal models of human colorectal cancer: Current status, uses and limitations. World journal of gastroenterology 21, 11854-11861, doi:10.3748/wjg.v21.i41.11854 (2015). 122 Bettenworth, D. et al. Endoscopy-guided orthotopic implantation of colorectal cancer cells results in metastatic colorectal cancer in mice. Clinical & experimental metastasis 33, 551-562, doi:10.1007/s10585-016-9797-7 (2016). 123 Beura, L. K. et al. Normalizing the environment recapitulates adult human immune traits in laboratory mice. Nature 532, 512-516, doi:10.1038/nature17655 (2016). 124 Furukawa, T. et al. A metastatic model of human colon cancer constructed using cecal implantation of cancer tissue in nude mice. Surgery today 23, 420-423, doi:10.1007/bf00309500 (1993). 125 Schackert, H. K. & Fidler, I. J. Development of an animal model to study the biology of recurrent colorectal cancer originating from mesenteric lymph system metastases. International journal of cancer 44, 177-181, doi:10.1002/ijc.2910440131 (1989). 126 Gould, S. E., Junttila, M. R. & de Sauvage, F. J. Translational value of mouse models in oncology drug development. Nature medicine 21, 431-439, doi:10.1038/nm.3853 (2015). 127 Housseau, F. & Llosa, N. J. Immune checkpoint blockade in microsatellite instable colorectal cancers: Back to the clinic. Oncoimmunology 4, e1008858, doi:10.1080/2162402X.2015.1008858 (2015). 128 Weide, B. et al. Baseline Biomarkers for Outcome of Melanoma Patients Treated with Pembrolizumab. Clinical cancer research : an official journal of the American Association for Cancer Research 22, 5487-5496, doi:10.1158/1078-0432.CCR-16-0127 (2016). 129 Fridman, W. H., Pages, F., Sautes-Fridman, C. & Galon, J. The immune contexture in human tumours: impact on clinical outcome. Nature reviews. Cancer 12, 298-306, doi:10.1038/nrc3245 (2012). 130 Pitt, J. M. et al. Resistance Mechanisms to Immune-Checkpoint Blockade in Cancer: Tumor-Intrinsic and -Extrinsic Factors. Immunity 44, 1255-1269, doi:10.1016/j.immuni.2016.06.001 (2016). 131 Restifo, N. P., Smyth, M. J. & Snyder, A. Acquired resistance to immunotherapy and future challenges. Nature reviews. Cancer 16, 121-126, doi:10.1038/nrc.2016.2 (2016). 132 de Visser, K. E., Eichten, A. & Coussens, L. M. Paradoxical roles of the immune system during cancer development. Nature reviews. Cancer 6, 24-37, doi:10.1038/nrc1782 (2006). 149

133 Cheng, M., Chen, Y., Xiao, W., Sun, R. & Tian, Z. NK cell-based immunotherapy for malignant diseases. Cellular & molecular immunology 10, 230-252, doi:10.1038/cmi.2013.10 (2013). 134 Schmohl, J. U., Gleason, M. K., Dougherty, P. R., Miller, J. S. & Vallera, D. A. Heterodimeric Bispecific Single Chain Variable Fragments (scFv) Killer Engagers (BiKEs) Enhance NK-cell Activity Against CD133+ Colorectal Cancer Cells. Targeted oncology 11, 353-361, doi:10.1007/s11523-015-0391-8 (2016). 135 Nishimura, H. & Honjo, T. PD-1: an inhibitory immunoreceptor involved in peripheral tolerance. Trends in immunology 22, 265-268, doi:10.1016/s1471-4906(01)01888-9 (2001). 136 Schwartz, R. H. Costimulation of T lymphocytes: the role of CD28, CTLA-4, and B7/BB1 in interleukin-2 production and immunotherapy. Cell 71, 1065-1068, doi:10.1016/s0092- 8674(05)80055-8 (1992). 137 Moynihan, K. D. et al. Eradication of large established tumors in mice by combination immunotherapy that engages innate and adaptive immune responses. Nature medicine 22, 1402-1410, doi:10.1038/nm.4200 (2016). 138 Li, G. et al. Successful chemoimmunotherapy against hepatocellular cancer in a novel murine model. Journal of hepatology 66, 75-85, doi:10.1016/j.jhep.2016.07.044 (2017). 139 Curran, M. A., Montalvo, W., Yagita, H. & Allison, J. P. PD-1 and CTLA-4 combination blockade expands infiltrating T cells and reduces regulatory T and myeloid cells within B16 melanoma tumors. Proceedings of the National Academy of Sciences of the United States of America 107, 4275-4280, doi:10.1073/pnas.0915174107 (2010). 140 Mosely, S. I. et al. Rational Selection of Syngeneic Preclinical Tumor Models for Immunotherapeutic Drug Discovery. Cancer immunology research 5, 29-41, doi:10.1158/2326-6066.CIR-16-0114 (2017). 141 Fiegle, E. et al. Dual CTLA-4 and PD-L1 Blockade Inhibits Tumor Growth and Liver Metastasis in a Highly Aggressive Orthotopic Mouse Model of Colon Cancer. Neoplasia 21, 932-944, doi:10.1016/j.neo.2019.07.006 (2019). 142 Zhu, G. et al. Pazopanib Inhibits Tumor Growth, Lymph-node Metastasis and Lymphangiogenesis of an Orthotopic Mouse of Colorectal Cancer. Cancer genomics & proteomics 17, 131-139, doi:10.21873/cgp.20173 (2020). 143 Robert, C. et al. Ipilimumab plus dacarbazine for previously untreated metastatic melanoma. The New England journal of medicine 364, 2517-2526, doi:10.1056/NEJMoa1104621 (2011). 144 Zhao, X. & Subramanian, S. Intrinsic Resistance of Solid Tumors to Immune Checkpoint Blockade Therapy. Cancer research 77, 817-822, doi:10.1158/0008-5472.CAN-16-2379 (2017). 145 Gide, T. N., Wilmott, J. S., Scolyer, R. A. & Long, G. V. Primary and Acquired Resistance to Immune Checkpoint Inhibitors in Metastatic Melanoma. Clinical cancer research : an official journal of the American Association for Cancer Research 24, 1260-1270, doi:10.1158/1078-0432.CCR-17-2267 (2018). 146 Lee, C. K. et al. Checkpoint Inhibitors in Metastatic EGFR-Mutated Non-Small Cell Lung Cancer-A Meta-Analysis. Journal of thoracic oncology : official publication of the International Association for the Study of Lung Cancer 12, 403-407, doi:10.1016/j.jtho.2016.10.007 (2017).

150

147 Rizvi, N. A. et al. Activity and safety of nivolumab, an anti-PD-1 immune checkpoint inhibitor, for patients with advanced, refractory squamous non-small-cell lung cancer (CheckMate 063): a phase 2, single-arm trial. The Lancet. Oncology 16, 257-265, doi:10.1016/S1470-2045(15)70054-9 (2015). 148 Fisher, B. & Fisher, E. R. Studies concerning the regional lymph node in cancer. I. Initiation of immunity. Cancer 27, 1001-1004 (1971). 149 Shu, S., Cochran, A. J., Huang, R. R., Morton, D. L. & Maecker, H. T. Immune responses in the draining lymph nodes against cancer: implications for immunotherapy. Cancer metastasis reviews 25, 233-242, doi:10.1007/s10555-006-8503-7 (2006). 150 Toki, M. I., Kumar, D., Ahmed, F. S., Rimm, D. L. & Xu, M. L. Benign Lymph Node Microenvironment is Associated with Response to Immunotherapy. Precision Clinical Medicine, doi:10.1093/pcmedi/pbaa003 (2020). 151 Karlsson, M. et al. Pilot study of sentinel-node-based adoptive immunotherapy in advanced colorectal cancer. Annals of surgical oncology 17, 1747-1757, doi:10.1245/s10434-010-0920-8 (2010). 152 Cochran, A. J. et al. Sentinel lymph nodes show profound downregulation of antigen- presenting cells of the paracortex: implications for tumor biology and treatment. Modern pathology : an official journal of the United States and Canadian Academy of Pathology, Inc 14, 604-608, doi:10.1038/modpathol.3880358 (2001). 153 Ito, M. et al. Tumor-derived TGFbeta-1 induces dendritic cell apoptosis in the sentinel lymph node. Journal of immunology 176, 5637-5643, doi:10.4049/jimmunol.176.9.5637 (2006). 154 Watanabe, S. et al. Tumor-induced CD11b+Gr-1+ myeloid cells suppress T cell sensitization in tumor-draining lymph nodes. Journal of immunology 181, 3291-3300, doi:10.4049/jimmunol.181.5.3291 (2008). 155 Murthy, V. et al. Tumor-draining lymph nodes demonstrate a suppressive immunophenotype in patients with non-small cell lung cancer assessed by endobronchial ultrasound-guided transbronchial needle aspiration: A pilot study. Lung cancer 137, 94- 99, doi:10.1016/j.lungcan.2019.08.008 (2019). 156 Munn, D. H. & Mellor, A. L. The tumor-draining lymph node as an immune-privileged site. Immunological reviews 213, 146-158, doi:10.1111/j.1600-065X.2006.00444.x (2006). 157 Khosravianfar, N. et al. Myeloid-derived Suppressor Cells Elimination by 5-Fluorouracil Increased Dendritic Cell-based Vaccine Function and Improved Immunity in Tumor Mice. Iranian journal of allergy, asthma, and immunology 17, 47-55 (2018). 158 Zhang, L. et al. Differential impairment of regulatory T cells rather than effector T cells by paclitaxel-based chemotherapy. Clinical immunology 129, 219-229, doi:10.1016/j.clim.2008.07.013 (2008). 159 Chester, C., Ambulkar, S. & Kohrt, H. E. 4-1BB agonism: adding the accelerator to cancer immunotherapy. Cancer immunology, immunotherapy : CII 65, 1243-1248, doi:10.1007/s00262-016-1829-2 (2016). 160 Buchan, S. L. et al. Antibodies to Costimulatory Receptor 4-1BB Enhance Anti-tumor Immunity via T Regulatory Cell Depletion and Promotion of CD8 T Cell Effector Function. Immunity 49, 958-970 e957, doi:10.1016/j.immuni.2018.09.014 (2018). 161 Vallejo, A. N., Brandes, J. C., Weyand, C. M. & Goronzy, J. J. Modulation of CD28 expression: distinct regulatory pathways during activation and replicative senescence. Journal of immunology 162, 6572-6579 (1999). 151

162 Lake, R. A., O'Hehir, R. E., Verhoef, A. & Lamb, J. R. CD28 mRNA rapidly decays when activated T cells are functionally anergized with specific peptide. International immunology 5, 461-466, doi:10.1093/intimm/5.5.461 (1993). 163 Linterman, M. A. et al. CD28 expression is required after T cell priming for helper T cell responses and protective immunity to infection. eLife 3, doi:10.7554/eLife.03180 (2014). 164 Liu, J. et al. Improved Efficacy of Neoadjuvant Compared to Adjuvant Immunotherapy to Eradicate Metastatic Disease. Cancer discovery 6, 1382-1399, doi:10.1158/2159-8290.CD- 16-0577 (2016). 165 Galluzzi, L., Buque, A., Kepp, O., Zitvogel, L. & Kroemer, G. Immunogenic cell death in cancer and infectious disease. Nature reviews. Immunology 17, 97-111, doi:10.1038/nri.2016.107 (2017). 166 Fend, L. et al. Immune Checkpoint Blockade, Immunogenic Chemotherapy or IFN-alpha Blockade Boost the Local and Abscopal Effects of Oncolytic Virotherapy. Cancer research 77, 4146-4157, doi:10.1158/0008-5472.CAN-16-2165 (2017). 167 Emens, L. A. & Middleton, G. The interplay of immunotherapy and chemotherapy: harnessing potential synergies. Cancer immunology research 3, 436-443, doi:10.1158/2326-6066.CIR-15-0064 (2015). 168 Marzo, A. L. et al. Tumor antigens are constitutively presented in the draining lymph nodes. Journal of immunology 162, 5838-5845 (1999). 169 Jeanbart, L. et al. Enhancing efficacy of anticancer vaccines by targeted delivery to tumor- draining lymph nodes. Cancer immunology research 2, 436-447, doi:10.1158/2326- 6066.CIR-14-0019-T (2014). 170 Fransen, M. F. et al. Tumor-draining lymph nodes are pivotal in PD-1/PD-L1 checkpoint therapy. JCI insight 3, doi:10.1172/jci.insight.124507 (2018). 171 Cochran, A. J. et al. Tumour-induced immune modulation of sentinel lymph nodes. Nature reviews. Immunology 6, 659-670, doi:10.1038/nri1919 (2006). 172 Shuang, Z.-Y. et al. The tumor-draining lymph nodes are immunosuppressed in patients with hepatocellular carcinoma. Translational Cancer Research 6, 1188-1196 (2017). 173 Wu, X. et al. PD-1(+) CD8(+) T cells are exhausted in tumours and functional in draining lymph nodes of colorectal cancer patients. British journal of cancer 111, 1391-1399, doi:10.1038/bjc.2014.416 (2014). 174 Kareva, I. A Combination of Immune Checkpoint Inhibition with Metronomic Chemotherapy as a Way of Targeting Therapy-Resistant Cancer Cells. International journal of molecular sciences 18, doi:10.3390/ijms18102134 (2017). 175 Wang, N., Wang, Z., Xu, Z., Chen, X. & Zhu, G. A Cisplatin-Loaded Immunochemotherapeutic Nanohybrid Bearing Immune Checkpoint Inhibitors for Enhanced Cervical Cancer Therapy. Angewandte Chemie 57, 3426-3430, doi:10.1002/anie.201800422 (2018). 176 Samanta, D. et al. Chemotherapy induces enrichment of CD47(+)/CD73(+)/PDL1(+) immune evasive triple-negative breast cancer cells. Proceedings of the National Academy of Sciences of the United States of America 115, E1239-E1248, doi:10.1073/pnas.1718197115 (2018). 177 Lutsiak, M. E. et al. Inhibition of CD4(+)25+ T regulatory cell function implicated in enhanced immune response by low-dose cyclophosphamide. Blood 105, 2862-2868, doi:10.1182/blood-2004-06-2410 (2005).

152

178 Michels, T. et al. Paclitaxel promotes differentiation of myeloid-derived suppressor cells into dendritic cells in vitro in a TLR4-independent manner. Journal of immunotoxicology 9, 292-300, doi:10.3109/1547691X.2011.642418 (2012). 179 Tesniere, A. et al. Immunogenic death of colon cancer cells treated with oxaliplatin. Oncogene 29, 482-491, doi:10.1038/onc.2009.356 (2010). 180 Lanier, L. L. et al. CD80 (B7) and CD86 (B70) provide similar costimulatory signals for T cell proliferation, cytokine production, and generation of CTL. Journal of immunology 154, 97- 105 (1995). 181 Chambers, B. J., Salcedo, M. & Ljunggren, H. G. Triggering of natural killer cells by the costimulatory molecule CD80 (B7-1). Immunity 5, 311-317 (1996). 182 Singh, N. P. et al. A novel approach to cancer immunotherapy: tumor cells decorated with CD80 generate effective antitumor immunity. Cancer research 63, 4067-4073 (2003). 183 Beyranvand Nejad, E. et al. Tumor Eradication by Cisplatin Is Sustained by CD80/86- Mediated Costimulation of CD8+ T Cells. Cancer research 76, 6017-6029, doi:10.1158/0008-5472.CAN-16-0881 (2016). 184 Kumar, V., Patel, S., Tcyganov, E. & Gabrilovich, D. I. The Nature of Myeloid-Derived Suppressor Cells in the Tumor Microenvironment. Trends in immunology 37, 208-220, doi:10.1016/j.it.2016.01.004 (2016). 185 Veglia, F., Perego, M. & Gabrilovich, D. Myeloid-derived suppressor cells coming of age. Nature immunology 19, 108-119, doi:10.1038/s41590-017-0022-x (2018). 186 Vincent, J. et al. 5-Fluorouracil selectively kills tumor-associated myeloid-derived suppressor cells resulting in enhanced T cell-dependent antitumor immunity. Cancer research 70, 3052-3061, doi:10.1158/0008-5472.CAN-09-3690 (2010). 187 Ghiringhelli, F. et al. Metronomic cyclophosphamide regimen selectively depletes CD4+CD25+ regulatory T cells and restores T and NK effector functions in end stage cancer patients. Cancer immunology, immunotherapy : CII 56, 641-648, doi:10.1007/s00262-006- 0225-8 (2007). 188 Cao, Z. et al. Antitumor and immunomodulatory effects of low-dose 5-FU on hepatoma 22 tumor-bearing mice. Oncology letters 7, 1260-1264, doi:10.3892/ol.2014.1856 (2014). 189 Scrimieri, F. et al. Murine leukemia virus envelope gp70 is a shared biomarker for the high-sensitivity quantification of murine tumor burden. Oncoimmunology 2, e26889, doi:10.4161/onci.26889 (2013). 190 Song, M. K., Park, M. Y. & Sung, M. K. 5-Fluorouracil-induced changes of intestinal integrity biomarkers in BALB/c mice. Journal of cancer prevention 18, 322-329 (2013). 191 Darvin, P., Toor, S. M., Sasidharan Nair, V. & Elkord, E. Immune checkpoint inhibitors: recent progress and potential biomarkers. Experimental & molecular medicine 50, 1-11, doi:10.1038/s12276-018-0191-1 (2018). 192 Hargadon, K. M., Johnson, C. E. & Williams, C. J. Immune checkpoint blockade therapy for cancer: An overview of FDA-approved immune checkpoint inhibitors. International immunopharmacology 62, 29-39, doi:10.1016/j.intimp.2018.06.001 (2018). 193 Sharma, P. & Allison, J. P. Dissecting the mechanisms of immune checkpoint therapy. Nature reviews. Immunology 20, 75-76, doi:10.1038/s41577-020-0275-8 (2020). 194 Salem, M. E. et al. Landscape of Tumor Mutation Load, Mismatch Repair Deficiency, and PD-L1 Expression in a Large Patient Cohort of Gastrointestinal Cancers. Molecular cancer research : MCR 16, 805-812, doi:10.1158/1541-7786.MCR-17-0735 (2018).

153

195 Zhao, X., May, A., Lou, E. & Subramanian, S. Genotypic and phenotypic signatures to predict immune checkpoint blockade therapy response in patients with colorectal cancer. Translational research : the journal of laboratory and clinical medicine 196, 62-70, doi:10.1016/j.trsl.2018.02.001 (2018). 196 Wolfers, J. et al. Tumor-derived exosomes are a source of shared tumor rejection antigens for CTL cross-priming. Nature medicine 7, 297-303, doi:10.1038/85438 (2001). 197 Horrevorts, S. K. et al. Glycan-Modified Apoptotic Melanoma-Derived Extracellular Vesicles as Antigen Source for Anti-Tumor Vaccination. Cancers 11, doi:10.3390/cancers11091266 (2019). 198 Sugiura, D. et al. Restriction of PD-1 function by cis-PD-L1/CD80 interactions is required for optimal T cell responses. Science 364, 558-566, doi:10.1126/science.aav7062 (2019). 199 Zhao, Y. et al. PD-L1:CD80 Cis-Heterodimer Triggers the Co-stimulatory Receptor CD28 While Repressing the Inhibitory PD-1 and CTLA-4 Pathways. Immunity 51, 1059-1073 e1059, doi:10.1016/j.immuni.2019.11.003 (2019). 200 Becht, E. et al. Immune and Stromal Classification of Colorectal Cancer Is Associated with Molecular Subtypes and Relevant for Precision Immunotherapy. Clinical cancer research : an official journal of the American Association for Cancer Research 22, 4057-4066, doi:10.1158/1078-0432.CCR-15-2879 (2016). 201 Prat, A. et al. Immune-Related Gene Expression Profiling After PD-1 Blockade in Non-Small Cell Lung Carcinoma, Head and Neck Squamous Cell Carcinoma, and Melanoma. Cancer research 77, 3540-3550, doi:10.1158/0008-5472.CAN-16-3556 (2017). 202 Bracken, C. P., Scott, H. S. & Goodall, G. J. A network-biology perspective of microRNA function and dysfunction in cancer. Nature reviews. Genetics 17, 719-732, doi:10.1038/nrg.2016.134 (2016). 203 Huber, V. et al. Tumor-derived microRNAs induce myeloid suppressor cells and predict immunotherapy resistance in melanoma. The Journal of clinical investigation 128, 5505- 5516, doi:10.1172/JCI98060 (2018). 204 Valadi, H. et al. Exosome-mediated transfer of mRNAs and microRNAs is a novel mechanism of genetic exchange between cells. Nature cell biology 9, 654-659, doi:10.1038/ncb1596 (2007). 205 Kopper, O. et al. An organoid platform for ovarian cancer captures intra- and interpatient heterogeneity. Nature medicine 25, 838-849, doi:10.1038/s41591-019-0422-6 (2019). 206 van de Wetering, M. et al. Prospective derivation of a living organoid biobank of colorectal cancer patients. Cell 161, 933-945, doi:10.1016/j.cell.2015.03.053 (2015). 207 Baba, Y. et al. HIF1A overexpression is associated with poor prognosis in a cohort of 731 colorectal cancers. The American journal of pathology 176, 2292-2301, doi:10.2353/ajpath.2010.090972 (2010). 208 Ghosh, G. et al. Hypoxia-induced microRNA-424 expression in human endothelial cells regulates HIF-alpha isoforms and promotes angiogenesis. The Journal of clinical investigation 120, 4141-4154, doi:10.1172/JCI42980 (2010). 209 King, H. W., Michael, M. Z. & Gleadle, J. M. Hypoxic enhancement of exosome release by breast cancer cells. BMC cancer 12, 421, doi:10.1186/1471-2407-12-421 (2012). 210 Zhang, D., Shi, Z., Li, M. & Mi, J. Hypoxia-induced miR-424 decreases tumor sensitivity to chemotherapy by inhibiting apoptosis. Cell death & disease 5, e1301, doi:10.1038/cddis.2014.240 (2014).

154

211 Daassi, D., Mahoney, K. M. & Freeman, G. J. The importance of exosomal PDL1 in tumour immune evasion. Nature reviews. Immunology, doi:10.1038/s41577-019-0264-y (2020). 212 Whiteside, T. L. Exosomes and tumor-mediated immune suppression. The Journal of clinical investigation 126, 1216-1223, doi:10.1172/JCI81136 (2016). 213 Cespedes, M. V. et al. Orthotopic microinjection of human colon cancer cells in nude mice induces tumor foci in all clinically relevant metastatic sites. The American journal of pathology 170, 1077-1085, doi:10.2353/ajpath.2007.060773 (2007). 214 Song, W. et al. Synergistic and low adverse effect cancer immunotherapy by immunogenic chemotherapy and locally expressed PD-L1 trap. Nature communications 9, 2237, doi:10.1038/s41467-018-04605-x (2018). 215 Terracina, K. P. et al. Development of a metastatic murine colon cancer model. The Journal of surgical research 199, 106-114, doi:10.1016/j.jss.2015.04.030 (2015). 216 Efremova, M. et al. Targeting immune checkpoints potentiates immunoediting and changes the dynamics of tumor evolution. Nature communications 9, 32, doi:10.1038/s41467-017-02424-0 (2018). 217 Kikuchi, T. et al. A subset of patients with MSS/MSI-low-colorectal cancer showed increased CD8(+) TILs together with up-regulated IFN-gamma. Oncology letters 18, 5977- 5985, doi:10.3892/ol.2019.10953 (2019). 218 Harding, F. A., McArthur, J. G., Gross, J. A., Raulet, D. H. & Allison, J. P. CD28-mediated signalling co-stimulates murine T cells and prevents induction of anergy in T-cell clones. Nature 356, 607-609, doi:10.1038/356607a0 (1992). 219 Li, Y. et al. MART-1-specific melanoma tumor-infiltrating lymphocytes maintaining CD28 expression have improved survival and expansion capability following antigenic restimulation in vitro. Journal of immunology 184, 452-465, doi:10.4049/jimmunol.0901101 (2010). 220 Weng, N. P., Akbar, A. N. & Goronzy, J. CD28(-) T cells: their role in the age-associated decline of immune function. Trends in immunology 30, 306-312, doi:10.1016/j.it.2009.03.013 (2009). 221 Xie, F., Xu, M., Lu, J., Mao, L. & Wang, S. The role of exosomal PD-L1 in tumor progression and immunotherapy. Molecular cancer 18, 146, doi:10.1186/s12943-019-1074-3 (2019). 222 Li, L. et al. Sequential expression of miR-182 and miR-503 cooperatively targets FBXW7, contributing to the malignant transformation of colon adenoma to adenocarcinoma. The Journal of pathology 234, 488-501, doi:10.1002/path.4407 (2014). 223 Moridikia, A., Mirzaei, H., Sahebkar, A. & Salimian, J. MicroRNAs: Potential candidates for diagnosis and treatment of colorectal cancer. Journal of cellular physiology 233, 901-913, doi:10.1002/jcp.25801 (2018). 224 Sarver, A. L. et al. Human colon cancer profiles show differential microRNA expression depending on mismatch repair status and are characteristic of undifferentiated proliferative states. BMC cancer 9, 401, doi:10.1186/1471-2407-9-401 (2009). 225 Mizuno, R. et al. PD-1 Primarily Targets TCR Signal in the Inhibition of Functional T Cell Activation. Frontiers in immunology 10, 630, doi:10.3389/fimmu.2019.00630 (2019). 226 Chalabi, M. et al. Neoadjuvant immunotherapy leads to pathological responses in MMR- proficient and MMR-deficient early-stage colon cancers. Nature medicine 26, 566-576, doi:10.1038/s41591-020-0805-8 (2020). 227 Iero, M. et al. Tumour-released exosomes and their implications in cancer immunity. Cell death and differentiation 15, 80-88, doi:10.1038/sj.cdd.4402237 (2008). 155

228 Kurywchak, P., Tavormina, J. & Kalluri, R. The emerging roles of exosomes in the modulation of immune responses in cancer. Genome medicine 10, 23, doi:10.1186/s13073-018-0535-4 (2018). 229 Bu, N. et al. Exosome-loaded dendritic cells elicit tumor-specific CD8+ cytotoxic T cells in patients with glioma. Journal of neuro-oncology 104, 659-667, doi:10.1007/s11060-011- 0537-1 (2011). 230 Gu, X., Erb, U., Buchler, M. W. & Zoller, M. Improved vaccine efficacy of tumor exosome compared to tumor lysate loaded dendritic cells in mice. International journal of cancer 136, E74-84, doi:10.1002/ijc.29100 (2015). 231 Zuo, B. et al. Alarmin-painted exosomes elicit persistent antitumor immunity in large established tumors in mice. Nature communications 11, 1790, doi:10.1038/s41467-020- 15569-2 (2020). 232 Sato, T. et al. Long-term expansion of epithelial organoids from human colon, adenoma, adenocarcinoma, and Barrett's epithelium. Gastroenterology 141, 1762-1772, doi:10.1053/j.gastro.2011.07.050 (2011). 233 Ruiz-Estevez, M. et al. Promotion of Myoblast Differentiation by Fkbp5 via Cdk4 Isomerization. Cell reports 25, 2537-2551 e2538, doi:10.1016/j.celrep.2018.11.006 (2018). 234 Yuan, C., Burns, M. B., Subramanian, S. & Blekhman, R. Interaction between Host MicroRNAs and the Gut Microbiota in Colorectal Cancer. mSystems 3, doi:10.1128/mSystems.00205-17 (2018). 235 Bolger, A. M., Lohse, M. & Usadel, B. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics 30, 2114-2120, doi:10.1093/bioinformatics/btu170 (2014). 236 Kim, D., Langmead, B. & Salzberg, S. L. HISAT: a fast spliced aligner with low memory requirements. Nature methods 12, 357-360, doi:10.1038/nmeth.3317 (2015). 237 Robinson, M. D., McCarthy, D. J. & Smyth, G. K. edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics 26, 139- 140, doi:10.1093/bioinformatics/btp616 (2010). 238 Langmead, B., Trapnell, C., Pop, M. & Salzberg, S. L. Ultrafast and memory-efficient alignment of short DNA sequences to the human genome. Genome biology 10, R25, doi:10.1186/gb-2009-10-3-r25 (2009). 239 Masella, A. P., Bartram, A. K., Truszkowski, J. M., Brown, D. G. & Neufeld, J. D. PANDAseq: paired-end assembler for illumina sequences. BMC bioinformatics 13, 31, doi:10.1186/1471-2105-13-31 (2012). 240 Anders, S., Pyl, P. T. & Huber, W. HTSeq--a Python framework to work with high- throughput sequencing data. Bioinformatics 31, 166-169, doi:10.1093/bioinformatics/btu638 (2015). 241 Newman, A. M. et al. Robust enumeration of cell subsets from tissue expression profiles. Nature methods 12, 453-457, doi:10.1038/nmeth.3337 (2015). 242 Olson, B., Li, Y., Lin, Y., Liu, E. T. & Patnaik, A. Mouse Models for Cancer Immunotherapy Research. Cancer discovery 8, 1358-1365, doi:10.1158/2159-8290.CD-18-0044 (2018). 243 O'Rourke, K. P. et al. Transplantation of engineered organoids enables rapid generation of metastatic mouse models of colorectal cancer. Nature biotechnology 35, 577-582, doi:10.1038/nbt.3837 (2017). 244 Roper, J. et al. In vivo genome editing and organoid transplantation models of colorectal cancer and metastasis. Nature biotechnology 35, 569-576, doi:10.1038/nbt.3836 (2017).

156

245 Song, Z., Yu, X., Cheng, G. & Zhang, Y. Programmed death-ligand 1 expression associated with molecular characteristics in surgically resected lung adenocarcinoma. J Transl Med 14, 188, doi:10.1186/s12967-016-0943-4 (2016). 246 Lin, W. et al. Radiation-induced small extracellular vesicles as "carriages" promote tumor antigen release and trigger antitumor immunity. Theranostics 10, 4871-4884, doi:10.7150/thno.43539 (2020).

157

Appendix According to the policy of Cancer Research (https://aacrjournals.org/content/3rd-party- permissions), as an author, I do not need to obtain permission for reusing this article (Intrinsic Resistance of Solid Tumors to Immune Checkpoint Blockade Therapy. Cancer Res. 2017 Feb 15;77(4):817-822.) as part of my dissertation.

According to the policy of Pharmacology & Therapeutics (https://www.elsevier.com/about/policies/copyright/permissions), as an author, I do not need to obtain permission for reusing this article (Oncogenic pathways that affect antitumor immune response and immune checkpoint blockade therapy. Pharmacol Ther. 2018 Jan;181:76-84.) as part of my dissertation.

According to the policy of Translational Research (https://www.elsevier.com/about/policies/copyright/permissions), as an author, I do not need to obtain permission for reusing this article (Genotypic and phenotypic signatures to predict immune checkpoint blockade therapy response in patients with colorectal cancer. Transl Res. 2018 Jun;196:62-70.) as part of my dissertation.

I have obtained the permission (Order Number: 4838430369383) from Springer eBook for reusing the book chapter (Novel Methods to Overcome Acquired Resistance to Immunotherapy. Current Applications for Overcoming Resistance to Targeted Therapies, 97-129.) as part of my dissertation.

According to the policy of Cancers (https://www.mdpi.com/authors/rights), I retain the copyright of the article: Acquired resistance to immune checkpoint blockade therapies. Cancers. Cancers, 2020 May 5;12(5):E1161. I do not need to obtain permission to reuse this article as part of my dissertation.

According to the policy of Oncotarget (https://www.oncotarget.com/editorial-policies/), I retain the copyright of the article: Tumor location impacts immune response in mouse models of colon cancer. Oncotarget. 2017 Jun 9;8(33):54775-54787. I do not need to obtain permission to reuse this article as part of my dissertation.

According to the policy of iScience (https://www.elsevier.com/about/open-science/open-access), I retain the copyright of the article: Chemotherapy but not the tumor-draining lymph nodes

158 determine the immunotherapy response in secondary tumors. iSCIENCE. 2020 May 22; 23(5): 101056. I do not need to obtain permission to reuse this article as part of my dissertation.

159