Extracellular Matrix Regulates Fibroblast Heterogeneity and Tumorigenesis

University of Pennsylvania ScholarlyCommons

Publicly Accessible Penn Dissertations

2017

Extracellular Matrix Regulates Fibroblast Heterogeneity And Tumorigenesis

Diana Leigh Avery University of Pennsylvania, [email protected]

Follow this and additional works at: https://repository.upenn.edu/edissertations

Part of the Cell Biology Commons, Oncology Commons, and the Pharmacology Commons

Recommended Citation Avery, Diana Leigh, "Extracellular Matrix Regulates Fibroblast Heterogeneity And Tumorigenesis" (2017). Publicly Accessible Penn Dissertations. 2173. https://repository.upenn.edu/edissertations/2173

This paper is posted at ScholarlyCommons. https://repository.upenn.edu/edissertations/2173 For more information, please contact [email protected]. Extracellular Matrix Regulates Fibroblast Heterogeneity And Tumorigenesis

Abstract Heterogeneous activated fibroblasts that deposit and remodel extracellular matrix (ECM) comprise desmoplasia, a key regulator of tumor development. The divergent outcomes in response to varied therapies targeting intratumoral desmoplasia underscore the pressing need to delineate the intricate role of a heterogeneous stroma in tumorigenesis. Fibroblast activation protein (FAP) and alpha-smooth muscle actin (αSMA) identify distinct, yet overlapping, activated fibroblast subsets in myriad tumor types, fibrosis, and wound healing. FAPHi reactive fibroblasts and αSMAHi myofibroblasts can exert divergent influences on tumor progression. However, the factors that drive this phenotypic heterogeneity and the unique functional roles of these distinct phenotypes are not yet understood. By comparing fibroblast activation on fibronectin- and collagen I-coated hydrogels of varying stiffness, I demonstrated that a convergence of ECM composition, elasticity, and TGF-β signaling governs activated fibroblast phenotypic heterogeneity. Furthermore, by characterizing gene expression signatures, I revealed potentially unique roles of activated fibroblast subsets in tissue remodeling.

In addition to its utility as a marker of activated fibroblasts, FAP also constitutes a highly attractive drug target due to its selective expression in the tumor microenvironment and its pro-tumorigenic proteolytic activity. FAP-mediated proteolysis of collagen I and III fragments promotes collagen clearance. However, the potential mechanistic link between FAP’s role in collagen proteolysis and tumorigenesis is not yet known. I investigated the role of FAP-dependent collagen proteolysis and ECM remodeling in tumorigenesis using both a spontaneous oncogenic Kras-driven lung tumor model and fibroblast-derived matrices. I found that Fap genetic deletion promotes intratumoral fibrillar collagen accumulation and attenuates lung tumor multiplicity, size, progression, and proliferation. Furthermore, I obtained preliminary evidence that Fap-/- lung fibroblasts produce ECM with enhanced fibrillar collagen accumulation, which directly impedes both tumor cell proliferation and motility. Taken together, my research underscores the essential role of ECM remodeling in regulating key aspects of tumorigenesis, such as intratumoral fibroblast heterogeneity, proliferation, and invasion.

Degree Type Dissertation

Degree Name Doctor of Philosophy (PhD)

Graduate Group Pharmacology

First Advisor Ellen Pur�

Keywords Biomechanics, Desmoplasia, Extracellular Matrix, Fibroblast, Tumor Microenvironment

Subject Categories Cell Biology | Oncology | Pharmacology

This dissertation is available at ScholarlyCommons: https://repository.upenn.edu/edissertations/2173 EXTRACELLULAR MATRIX REGULATES FIBROBLAST HETEROGENEITY AND

TUMORIGENESIS

Diana Avery

A DISSERTATION

Pharmacology

Presented to the Faculties of the University of Pennsylvania

Partial Fulfillment of the Requirements for the

Degree of Doctor of Philosophy

2017

Supervisor of Dissertation

______

Ellen Puré, Ph.D.

Professor of Biomedical Sciences

Graduate Group Chairperson

______

Julie Blendy, Ph.D.

Professor of Pharmacology

Dissertation Committee

Rebecca G. Wells, M.D., Associate Professor of Medicine

Richard K. Assoian, Ph.D., Professor of Pharmacology

Sandra W. Ryeom, Ph.D., Associate Professor of Cancer Biology

Edna Cukierman, Ph.D., Associate Professor of Cancer Biology, Fox Chase Cancer Center

DEDICATION

To my family for their love and support

ACKNOWLEDGMENT

I would first like to thank my advisor, Ellen Puré, for her support and guidance over the past few years, which helped me develop as both a scientist and scholar. I would like to thank members of the Puré lab (past and present): Priya Govindaraju,

Rachel Blomberg, Mia Krolikoski, Leslie Hopper, James Monslow, Allyson Lieberman,

Albert Lo, Michael Leibowitz, Toru Kimura, John Chojnowski, Lisa Chang, Irene

Crichton, Michele Jacob, Sonali Majumdar, and Ming-Hui Fan, for stimulating scientific discussion and for making the past few years so enjoyable. I would like to thank Paola

Castagnino of the Biomechanics Core Facility of the Institute for Translational Medicine and Therapeutics (ITMAT) for providing hydrogels and technical support. I would also like to thank Colleen McEntee and Sarah Rauers as well as our collaborators, Susan

Volk, Becky Brisson, and Chelsea Burgwin, for all their assistance.

I would like to acknowledge my thesis committee: Rebecca Wells (Chair),

Richard Assoian, Sandra Ryeom, and Edna Cukierman, for their expertise and insight.

I want to thank my family and friends for their support and advice. I especially want to thank my parents, sisters, brothers-in-law, niece, and nephew for their support and love. I also want to thank my classmates of the Pharmacology Graduate Group:

Sima Patel, Abby Christian, Mansi Shinde, Bridgin Lee, Jackie St. Louis, Kevin Patel,

Mike Chiorazzo, Natalie Daurio, Nishita Shastri, John O’Donnell, and Brian Weiser. You all made my time in graduate school more enjoyable by providing friendship and support.

Finally, I would like to thank the Pharmacology Graduate Group, especially Sarah

Squire and Julie Blendy, for this experience.

iii

ABSTRACT

EXTRACELLULAR MATRIX REGULATES FIBROBLAST HETEROGENEITY AND

TUMORIGENESIS

Diana Avery

Ellen Puré

Heterogeneous activated fibroblasts that deposit and remodel extracellular matrix

(ECM) comprise desmoplasia, a key regulator of tumor development. The divergent outcomes in response to varied therapies targeting intratumoral desmoplasia underscore the pressing need to delineate the intricate role of a heterogeneous stroma in tumorigenesis. Fibroblast activation protein (FAP) and alpha-smooth muscle actin

(αSMA) identify distinct, yet overlapping, activated fibroblast subsets in myriad tumor types, fibrosis, and wound healing. FAPHi reactive fibroblasts and αSMAHi myofibroblasts can exert divergent influences on tumor progression. However, the factors that drive this phenotypic heterogeneity and the unique functional roles of these distinct phenotypes are not yet understood. By comparing fibroblast activation on fibronectin- and collagen I- coated hydrogels of varying stiffness, I demonstrated that a convergence of ECM composition, elasticity, and TGF-β signaling governs activated fibroblast phenotypic heterogeneity. Furthermore, by characterizing gene expression signatures, I revealed potentially unique roles of activated fibroblast subsets in tissue remodeling.

iv of collagen I and III fragments promotes collagen clearance. However, the potential mechanistic link between FAP’s role in collagen proteolysis and tumorigenesis is not yet known. I investigated the role of FAP-dependent collagen proteolysis and ECM remodeling in tumorigenesis using both a spontaneous oncogenic Kras-driven lung tumor model and fibroblast-derived matrices. I found that Fap genetic deletion promotes intratumoral fibrillar collagen accumulation and attenuates lung tumor multiplicity, size, progression, and proliferation. Furthermore, I obtained preliminary evidence that Fap-/- lung fibroblasts produce ECM with enhanced fibrillar collagen accumulation, which directly impedes both tumor cell proliferation and motility. Taken together, my research underscores the essential role of ECM remodeling in regulating key aspects of tumorigenesis, such as intratumoral fibroblast heterogeneity, proliferation, and invasion.

TABLE OF CONTENTS

DEDICATION ...... II

ACKNOWLEDGMENT ...... III

ABSTRACT...... IV

LIST OF FIGURES ...... VIII

LIST OF TABLES ...... X

CHAPTER ONE: INTRODUCTION ...... 1

The analogy between tumor stroma generation and wound healing ...... 1

Fibroblast activation protein (FAP): Structure and function ...... 2

FAP as a therapeutic target in the tumor microenvironment ...... 4

FAP and alpha smooth muscle actin (αSMA) identify distinct, yet overlapping, activated fibroblast subsets in actively remodeling tissues...... 5

Extracellular matrix (ECM) proteolysis, mechanics, and architecture regulate tumorigenesis...... 8

Intratumoral stroma as a therapeutic target ...... 9

Targeting FAPHi stromal cells in cancer ...... 11

Targeting FAP proteolytic activity in cancer ...... 12

Dissertation goals ...... 13

CHAPTER TWO: EXTRACELLULAR MATRIX DIRECTS PHENOTYPIC HETEROGENEITY OF ACTIVATED FIBROBLASTS ...... 14

Introduction ...... 14

Results ...... 15

Discussion ...... 30

Future Directions ...... 34

Methods ...... 36

CHAPTER THREE: THE ROLE OF FAP-DEPENDENT ECM REMODELING IN LUNG TUMORIGENESIS ...... 42

Introduction ...... 42

Results ...... 44

Discussion ...... 62

Methods ...... 64

CHAPTER FOUR: CONCLUSIONS AND FUTURE DIRECTIONS ...... 71

Regulation and functionality of heterogenous fibroblast subsets in actively remodeling tissues ...... 71

Fibroblast heterogeneity: A determinant of stroma-targeted therapy efficacy ...... 75

Investigating the mechanisms underlying ECM-dependent generation of FAPHi versus αSMAHi myofibroblasts ...... 77

FAP promotes primary tumor growth and metastasis in diverse tumor types ...... 80

ECM: A key player in tumorigenesis ...... 81

REFERENCES ...... 82

vii

LIST OF FIGURES

Figure 1. Analogous evolution of ECM composition in wound healing and tumorigenesis.

Figure 2. FAP functional domains.

Figure 3. FAP plays a role in the ordered proteolytic processing of collagen.

Figure 4. Substratum directs activated fibroblast phenotypic heterogeneity.

Figure 5. ECM composition and elasticity govern activated fibroblast phenotypic heterogeneity.

Figure 6. Fibroblast differentiation in response to TGF-β is governed by ECM composition and elasticity.

Figure 7. Neither ROCK nor FAK activity are required to induce the FAPHi reactive fibroblast phenotype.

Figure 8. Gene expression profiling indicates that FAPHi reactive fibroblasts predominantly synthesize and proteolyze ECM, while αSMAHi myofibroblasts mediate contraction.

Figure 9. A convergence of ECM composition, elasticity, and TGF-β signaling governs activated fibroblast phenotypic heterogeneity.

Figure 10. Fap+/+, Fap+/-, and Fap-/- mice exhibit equivalent lung tumor burden in the LSL- KrasG12D/+;Tgfbr2flox/flox model of lung adenocarcinoma.

Figure 11. Fap genetic deletion attenuates lung tumor burden in 12-week-old KrasLA2/+ mice.

Figure 12. Fap genetic deletion attenuates tumor multiplicity, size, and total tumor burden in the lungs of 18-week-old KrasLA2/+ mice.

Figure 13. Fap genetic deletion restrains tumor progression in the KrasLA2/+ model of lung tumorigenesis.

Figure 14. Fap genetic deletion restrains lung tumor proliferation.

Figure 15. Fibrillar collagen accumulates at the tumor-stromal interface in KrasLA2/+;Fap-/- tumors.

Figure 16. Possible mechanisms by which FAP-dependent collagen proteolysis may regulate ECM remodeling, and consequently, tumor growth.

Figure 17. Lung tumor cells proliferate equally well in the presence of collagen I digested by MMP-1 alone or by MMP-1 and FAP. viii

Figure 18. Fap-/- lung fibroblasts deposit FDMs with enhanced fibrillar collagen accumulation.

Figure 19. Fap-/- FDMs restrain lung tumor cell proliferation.

Figure 20. Fap-/- FDMs restrain tumor cell motility.

Figure 21. Working model depicting the proposed regulation and functionality of FAPHi and αSMAHi fibroblast subsets in the injury response.

Figure 22. Proposed mechanisms underlying ECM-dependent generation of FAPHi versus αSMAHi fibroblasts in actively remodeling tissues.

LIST OF TABLES

Table 1. Fibroblasts cultured on plastic are resistant to induction of Fap expression.

Table 2. Raw data from the Qiagen RT² Profiler Mouse Fibrosis PCR Array.

CHAPTER ONE: INTRODUCTION

The analogy between tumor stroma generation and wound healing

Heterogeneous activated fibroblasts that deposit and remodel extracellular matrix

(ECM) comprise desmoplasia, a key regulator of tumor development. Tumor stroma resembles granulation tissue to such an extent that many researchers characterize tumors as “wounds that never heal” or “overhealing wounds”1–4. Of particular relevance, both intratumoral desmoplasia and granulation tissue evolve from a provisional fibrin- and fibronectin (FN)-rich ECM into mature collagenous ECM1 (Fig. 1). Bidirectional regulatory mechanisms between activated fibroblasts and ECM orchestrate this transition. Activated fibroblasts are identified by diverse markers, such as alpha smooth muscle actin (αSMA), fibroblast activation protein (FAP), platelet-derived growth factor receptor-β (PDGFRβ), PDGFRα, prolyl 4-hydroxylase, and discoidin domain-containing receptor 25,6. Thus far, most research has focused on the mechanisms that drive

αSMAHi myofibroblast differentiation. ECM stiffness, growth factors (in particular, TGF-β), and ECM components (such as EDA-FN) collectively promote myofibroblast differentiation7–10. However, the regulatory mechanisms that govern generation of other activated fibroblast subsets remain mostly undefined.

Figure 1. Analogous evolution of ECM composition in wound healing and tumorigenesis. In acute injury, plasma extravasation from damaged blood vessels forms a FN-rich provisional matrix. Highly permeable tumor microvasculature similarly promotes plasma FN deposition. In both granulation tissue and tumors, activated fibroblasts gradually replace provisional FN-rich ECM with mature, collagenous ECM.

FAP: Structure and function

FAP is a marker of a subset of activated fibroblasts in the injury response11, fibrotic and inflammatory conditions12–14, and myriad tumor types15–19. Of all the members of its prolyl peptidase subfamily (S9B) of serine proteases, FAP shares highest amino acid sequence homology (~50%) with dipeptidyl peptidase IV (DPPIV)20,21. Based on its amino acid sequence, FAP exhibits the structural features of a cell surface type II integral membrane protein, with a short cytoplasmic tail, a transmembrane segment, and a large extracellular domain20. X-ray crystallography illustrated that FAP resides at the cell surface as a homodimer. Each extracellular subunit of the FAP homodimer contains

2 a seven-bladed β-propeller domain that overhangs an α/β hydrolase domain (Fig. 2).

The FAP catalytic triad, composed of residues Ser624, Asp702, and His734, resides at the interface of these two domains22.

Figure 2. FAP functional domains. Cytoplasmic domain (Cyto; yellow); transmembrane domain (TM, black); linker region (red); β-propeller domain (blue); catalytic domain (green).

FAP functions not only as a dipeptidyl peptidase, but also an endopeptidase, which distinguishes it from all other members of its prolyl peptidase subfamily18,20,22. No study has yet definitively determined the physiologic substrates of FAP. However, recent in vitro evidence has implicated type I and III collagen fragments as FAP substrates. The in vitro digest of collagen by FAP requires prior cleavage by matrix metalloproteinase-1

(MMP-1). Subsequently, FAP digests the resulting ¼ -C-terminal and ¾ -N-terminal collagen fragments into smaller peptides23 (Fig. 3). Further corroborating FAP as a mediator of collagen catabolism, FAP enzymatic activity expedites collagen turnover24.

Therefore, FAP deficiency exacerbates collagen accumulation in murine models of lung cancer and fibrosis24,25. Additional putative substrates of FAP endopeptidase activity

3 include FGF2126–28 and α2-antiplasmin23, implicating FAP as a regulator of metabolism and fibrinolysis, respectively.

Figure 3. FAP plays a role in the ordered proteolytic processing of collagen. The digest of fibrillar collagen (I or III) by FAP requires prior cleavage by MMP collagenases. Subsequently, FAP (in conjunction with other gelatinases) digests the resulting ¼ -C-terminal and ¾ -N-terminal collagen fragments into smaller peptides.

FAP as a therapeutic target in the tumor microenvironment

Among the array of secreted and cell surface proteases in the tumor microenvironment, FAP constitutes a highly attractive therapeutic target for several reasons. First, its low levels of expression in most normal adult tissue and selective expression on carcinoma-associated fibroblasts (CAFs) minimize the likelihood of adverse events due to on-target, off-tumor effects of FAP-targeted therapies. Second, its expression on the cell surface of CAFs and its proteolytic function render it amenable to targeting via diverse therapeutic approaches. Third, a variety of murine tumor models have demonstrated an active role of FAPHi CAFs and FAP proteolytic activity in the promotion of tumorigenesis. Lastly, when assessed clinically, higher FAP expression correlated with poorer clinical outcome in myriad solid tumor types, including gastric cancer29, colorectal cancer30–33, pancreatic adenocarcinoma34–36, clear cell renal cell carcinoma37,38, oral squamous cell carcinoma39, breast cancer40,41, osteosarcoma42, non- small cell lung cancer43, rectal cancer44, and hepatocellular carcinoma45. 4

FAP and αSMA identify distinct, yet overlapping, activated fibroblast subsets in actively remodeling tissues.

The discovery of FAP has facilitated the identification of heterogeneous subsets of activated fibroblasts. After transient expression in some fetal mesenchymal tissues during embryonic development46,47, the quiescent mesenchyme of most normal adult organs exhibits little or no FAP expression15,18. However, activated fibroblasts exhibit high FAP expression in conditions associated with active tissue remodeling, such as the injury response11, fibrotic and inflammatory conditions12–14, and myriad tumor types15–19.

As discussed herein, FAP and αSMA identify spatiotemporally distinct, yet overlapping, subsets of activated fibroblasts in virtually all assessed physiological and pathophysiological settings.

FAP and αSMA identify spatially and temporally distinct subsets of activated fibroblasts in granulation tissue. In various injury models, FAP expression increases relatively rapidly in response to wounding. For example, in response to bleomycin- or radiation-induced injury, lungs exhibit near peak Fap gene expression (at least five-fold upregulation) as early as 24 hours post-injury24. In contrast, Acta2 (αSMA) gene expression remains low at early timepoints after bleomycin-induced lung injury (days 1 and 3), despite an early increase in Col1a1 gene expression in pulmonary fibroblasts at day 348. Fibroblasts exhibit peak Acta2 expression between days 7 and 14 post-injury48.

Similarly, in a rat model of myocardial infarction (MI), FAP expression increases at day 3 and peaks around day 7 post-MI11. Most FAP+ cells show strong co-expression of prolyl-

4-hydroxylase and vimentin, signifying an activated and collagen synthetic fibroblast phenotype. However, only a minority of FAP+ fibroblasts co-express αSMA11.

Analogous to the expression patterns observed in granulation tissue, FAP and

αSMA identify overlapping, but spatially distinct fibroblast populations in various fibrotic conditions. In liver cirrhosis, stellate cells express high FAP levels in the periseptal area of regenerative nodules (the presumptive tissue remodeling interface), but low levels in the portal fibrous septa. In direct contrast, stellate cells express virtually no αSMA in regenerative nodules, but elevated αSMA levels in the portal fibrous septa13.

Interestingly, FAP, but not αSMA, expression directly correlates with the severity of liver fibrosis and the necroinflammatory score49. In idiopathic pulmonary fibrosis (IPF), fibroblast foci at the tissue remodeling interface (areas of ongoing injury and repair) similarly exhibit high FAP, but low αSMA, expression.

Inflammatory conditions similarly exhibit distinct patterns of FAP and αSMA expression. For example, in Crohn's disease-associated intestinal fibrosis, collagen-rich fibrotic areas show high αSMA, but low FAP, expression. Meanwhile, fibroblasts adjacent to fibrotic areas exhibit high FAP expression50. In a comparable fashion, smooth muscle cells (SMCs) in human atherosclerotic aortic plaques upregulate FAP expression predominantly in collagen-poor regions. In fact, Masson staining for collagen and FAP immunohistochemistry (IHC) in adjacent tissue sections demonstrated that

FAP expression is significantly enhanced in thin-cap (< 65 μm) versus thick-cap (> 65

μm) human coronary fibroatheromata51.

These divergent patterns of FAP and αSMA expression also arise during tumorigenesis. Reactive stroma exhibits high FAP expression in virtually all carcinomas52 as well as melanoma53,54, glioblastoma19, and sarcoma42. Depending on tumor type and stage, the degree of overlap between FAP and αSMA varies markedly.

The spatiotemporal regulation of FAP and αSMA expression in intratumoral stroma has 6 been studied most extensively in pancreatic ductal adenocarcinoma (PDAC). CD45-

PDGFRα+ stromal cells in pancreatic intraepithelial neoplasia express FAP, but rarely

αSMA55. However, upon progression to PDAC, the frequency of αSMAHi myofibroblasts increases55. In human and murine PDAC, intratumoral stroma displays striking heterogeneity, including FAPHiαSMALow, FAPLowαSMAHi, and FAPHiαSMAHi activated fibroblast subsets16,55–57. Interestingly, analysis of the spatial distribution of these fibroblast subsets in human and murine PDAC revealed that FAPHiαSMAHi fibroblasts often form periglandular rings surrounding tumor cell nests, while FAPHiαSMALow fibroblasts predominantly localize further away from the tumor-stromal interface. Thus, stromal cell composition evolves throughout the course of PDAC, akin to the spatiotemporal separation of FAPHi and αSMAHi activated fibroblast subsets seen in the injury response.

As in PDAC, both breast and lung cancer microenvironments foster the generation of heterogeneous stromal cell populations. In multiple breast cancer subtypes

(including hormone receptor-positive, HER2/neu-positive, and triple negative breast cancer), the majority of stromal cells (~85%) express FAP, while only a minority (~30%) express αSMA. Consistent with the spatial distribution seen in PDAC, αSMAHi myofibroblasts often localize to concentric rings surrounding EpCAM+ tumor cell nests in human breast cancer17. In non-small cell lung cancer (NSCLC), IHC analysis of tumor specimens from a large cohort (n=536) of patients demonstrated no correlation between

FAP and αSMA expression in both lung adenocarcinoma and squamous cell carcinoma, indicating that FAP and αSMA identify distinct pulmonary CAF subsets58. Interestingly, a thorough characterization of other tumor and stromal cell markers in these same tumor specimens illustrated that, of all markers assessed, only collagen (as visualized with

Masson’s trichrome staining) positively correlated with αSMA expression. Meanwhile,

FAP expression and collagen showed no correlation58, coinciding with the expression patterns of FAP, αSMA, and collagen in fibrotic and inflammatory disorders as discussed above.

ECM proteolysis, mechanics, and architecture regulate tumorigenesis.

Importantly, activated fibroblasts synthesize and remodel intratumoral ECM. The biochemical, mechanical, and architectural properties of ECM coordinately regulate tumor evolution. ECM degradation influences tumorigenesis through a variety of different mechanisms. Proteolysis of collagen I can alter integrin engagement by exposing matricryptic RGD motifs. For example, β1 and αvβ3 integrins preferentially bind native fibrillar collagen and exposed RGD motifs, respectively59,60. This differential integrin binding, in turn, can regulate tumor cell behavior. α2β1 ligation of fibrillar collagen arrests melanoma cell proliferation61, while αvβ3 ligation of degraded collagen promotes melanoma cell survival and proliferation62,63. Collagen proteolysis also promotes cancer cell invasion through ECM64,65 and regulates tumor neovascularization by liberating latent growth factors (such as VEGF66 and TGF-β67) and releasing type IV collagen- derived anti-angiogenic fragments68. Furthermore, high and low molecular weight hyaluronan (HA; a glycosaminoglycan that accumulates in tumors) exert anti- and pro- tumorigenic influences, respectively69,70. For example, the extremely high molecular mass of HA in naked mole-rats (Heterocephalus glaber) mediates their inherent cancer resistance71.

ECM mechanics also play a fundamental role in tumor evolution. For example, lysyl oxidase (LOX)-mediated collagen crosslinking and cytoskeletal contractility both heighten intratumoral ECM stiffness, which promotes breast tumor proliferation and 8 invasion via integrin-mediated signaling72–75. High ECM stiffness generated by myofibroblast contraction also releases latent growth factors, such as TGF-β76,77.

Furthermore, tensile forces expose cryptic epitopes in FN (a crucial ECM component of the pre-metastatic niche78), which triggers fibrillogenesis79 and promotes cellular adhesion80. Lastly, HA deposition augments interstitial fluid pressure, which restricts intratumoral drug delivery in PDAC81.

ECM architecture also directs tumor cell behavior. Collagen organization at the tumor-stromal interface regulates invasion, as evidenced by tumor-associated collagen signatures that correlate with the degree of murine mammary tumor invasiveness.

Collagen fibers that align radially from tumors facilitate local tumor growth and invasion82. In vitro studies have demonstrated how CAFs orchestrate this pro-invasive

ECM remodeling. As CAFs invade a type I collagen-rich ECM, they generate aligned collagen fibers, via a mechanism dependent on MMP activity and Rho-mediated ECM contraction75,83. Carcinoma cells travel along these linear collagen-containing ECM fibers at high velocities, as demonstrated via intravital imaging of metastatic primary tumors84,85. Moreover, the density and orientation of intratumoral ECM fibers regulates antitumor immunity by controlling T cell migration86.

Intratumoral stroma as a therapeutic target

Considering the essential role of ECM in tumor evolution, many laboratories have developed therapies that target intratumoral desmoplasia. Several therapeutic approaches that deplete or target activated stromal cells hinder disease progression in pre-clinical models of cancer and fibrosis. For example, depleting FAPHi fibroblasts impedes primary and metastatic outgrowth in a variety of different tumor models

(discussed at length in the next section). Furthermore, the synthetic Vitamin D receptor 9 agonist calcipotriol antagonizes pro-fibrotic TGF-β/SMAD-dependent signaling by displacing SMAD3 from the promoters of target genes. In murine models of hepatic injury, calcipotriol treatment inhibits hepatic stellate cell activation, and consequently, abrogates liver fibrosis87. Calcipotriol treatment also reverts the pro-fibrotic and pro- inflammatory phenotype of activated pancreatic stellate cells. As a result, adjuvant administration of calcipotriol enhances the delivery and efficacy of gemcitabine, slows tumor growth, and prolongs survival by 57% compared to gemcitabine monotherapy in the LSL-KrasG12D/+; LSL-Trp53R172H/+, Pdx-1-Cre (KPC) model of PDAC88.

However, not all stroma-targeted therapeutic strategies achieve positive outcomes. After finding that paracrine sonic hedgehog (SHH) ligands secreted by neoplastic cells promote desmoplasia89,90, a pre-clinical study demonstrated that short- term treatment with an inhibitor of SHH-mediated signaling (IPI-926; a smoothened inhibitor) improves gemcitabine delivery in KPC tumors by decreasing collagen content, while increasing intratumoral vascular density91. However, concurrent administration of smoothened inhibitors with chemotherapy did not improve survival in clinical trials92.

Thereafter, follow-up studies discovered that long-term inhibition of SHH signaling actually shortens survival of KPC mice93,94. Further analysis demonstrated that while inhibition of SHH signaling restrains myofibroblast proliferation, it accelerates epithelial cell proliferation, thereby promoting pancreatic intraepithelial neoplasia formation (the precursor lesion before the onset of PDAC) and PDAC progression93,94. Around the same time, another group found that depletion of proliferating αSMAHi myofibroblasts (via systemic ganciclovir administration in αSMA-thymidine kinase transgenic mice) yields adverse outcomes in multiple autochthonous murine models of PDAC56. This mode of stromal depletion reduces intratumoral stiffness and collagen content; however, FAP-

10 expressing fibroblasts remain abundant56. αSMAHi myofibroblast depletion shortens the lifespan of KPC mice, likely through its enhancement of epithelial-mesenchymal transition, cancer stem cell prevalence, and immunosuppression56. Altogether, the dichotomous outcomes in response to varied stroma-targeted therapies emphasize the urgent need to delineate the intricate role of a heterogeneous stroma in tumorigenesis.

Targeting FAPHi stromal cells in cancer

Taking full advantage of the selective expression of FAP on the cell surface of intratumoral stromal cells, several laboratories have successfully curtailed tumor growth and metastasis in murine models by depleting FAPHi CAFs via diverse therapeutic strategies. Our laboratory demonstrated that FAPHi CAF depletion via FAP-targeted chimeric antigen receptor (FAP-CAR) T cells curbs tumor growth in autochthonous and transplanted tumor models via immune-dependent and -independent mechanisms. In highly immunogenic tumors, FAP-CAR T cells enhance anti-tumor immunity by augmenting endogenous CD8+ T cell responses95. In non-immunogenic, but highly desmoplastic tumors, FAP-CAR T cells restrain tumor growth by diminishing intratumoral vascularization and desmoplasia, including prevalent ECM components, such as FN,

HA, collagen, and versican29. FAP-targeted vaccines attain similar anti-tumor efficacy.

FAP-transduced whole-cell tumor vaccines enhance intratumoral CD8+ T cell infiltration and suppress proliferation and differentiation of immunosuppressive immune cell populations, such as M2 macrophages, myeloid-derived suppressor cells, and regulatory

T cells96. Vaccination with FAP-transfected dendritic cells or DNA-based FAP vaccines suppresses tumor growth, reduces pulmonary metastatic burden, and prolongs survival in a variety of transplanted tumor models97,98. Moreover, combining FAPHi CAF depletion

(FAP-CAR T cells or FAP-targeted vaccines) with chemotherapy achieves additive or

11 synergistic anti-tumor efficacy and prolongs survival16,98,99, in part by increasing intratumoral drug uptake98.

Other laboratories have exploited the unique proteolytic activity of FAP to design

CAF-targeted prodrugs and protoxins. For example, FAP-activated peptide toxins contain FAP-cleavable sequences conjugated to promelittin protoxins. This therapeutic approach restricts melittin cytolytic activity mainly to the tumor site, thereby inhibiting tumor growth with minimal toxicity to the host animal100. In a similar fashion, FAP- activated prodrugs achieve comparable anti-tumor efficacy and mitigate toxicity compared to commonly used chemotherapeutic agents101–103. Moreover, nanoparticle- based photoimmunotherapy (nano-PIT) only depletes FAPHi stromal cells at the tumor site. This approach features a FAP-specific single chain variable fragment conjugated to apoferritin, a compact nanoparticle protein cage used as a photosensitizer carrier. By selectively administering photoirradiation at the tumor site, nano-PIT exclusively eliminates FAPHi intratumoral CAFs. Comparable to aforementioned FAPHi fibroblast depletion strategies, this therapeutic approach enhances intratumoral cytotoxic T cell infiltration, suppresses tumor growth, and prolongs survival104.

Targeting FAP proteolytic activity in cancer

In addition to its utility as a marker of activated fibroblasts, FAP also constitutes a highly attractive drug target due to its selective expression in the tumor microenvironment and its pro-tumorigenic proteolytic activity. In our laboratory, we have found that FAP accelerates tumor growth and metastasis in a variety of transplanted and autochthonous murine tumor models. In an inducible K-rasG12D-driven autochthonous murine model of lung cancer, both Fap genetic deletion and pharmacological inhibition of

FAP enzymatic activity reduced lung tumor burden25. The diminished lung tumor burden 12 in FAP-deficient mice was associated with excess collagen accumulation and dysregulated integrin-mediated signaling25. Recent studies also demonstrated that FAP cleaves collagen I and III fragments, and that FAP-mediated collagen proteolysis promotes collagen clearance23,24. However, the potential mechanistic link between FAP’s role in promoting collagen proteolysis and tumorigenesis is not known.

Dissertation goals

Overall, the research described in this dissertation aimed to delineate bidirectional regulatory mechanisms between heterogeneous activated fibroblast subsets and ECM, and how these features of desmoplasia coordinately regulate tumorigenesis. Specifically, the goals of this dissertation were to: 1) identify features of actively remodeling tissues that drive FAPHi versus αSMAHi fibroblast activation, 2) unravel potentially unique functional roles of FAPHi and αSMAHi activated fibroblasts subsets in the tumor microenvironment, and 3) explore the mechanisms by which FAP proteolytic activity (especially its role in ECM remodeling) promotes tumor proliferation and metastasis.

CHAPTER TWO: EXTRACELLULAR MATRIX DIRECTS PHENOTYPIC

HETEROGENEITY OF ACTIVATED FIBROBLASTS

Introduction

In development and homeostasis, fibroblasts synthesize extracellular matrix

(ECM), which regulates cell differentiation and behavior. As sentinel cells, fibroblasts respond to injury and neoplasia by differentiating to highly secretory, proliferative, or contractile phenotypes. This heterogeneous activated fibroblast population orchestrates injury resolution. However, in fibrosis and cancer, the aberrant accumulation of activated fibroblasts and ECM disrupts organ function7–10,105.

Fibroblast activation protein (FAP) is a marker of activated fibroblasts in the injury response11, fibrotic and inflammatory conditions12–14, and myriad tumor types15–19. In most epithelial-derived tumors, FAP and alpha smooth muscle actin (αSMA; the canonical myofibroblast marker) identify distinct, yet overlapping, fibroblast subsets16,17,56–58. Importantly, FAPHi fibroblasts and αSMAHi myofibroblasts diverge in their impact on tumorigenesis. FAPHi fibroblasts accelerate the growth of multiple tumor types by promoting intratumoral desmoplasia, angiogenesis, and immunosuppression16,95,106,107. In contrast, αSMAHi myofibroblasts can curb pancreatic cancer aggressiveness by restraining epithelial-mesenchymal transition, cancer stem cell generation, and immunosuppression56. FAP and αSMA similarly mark distinct fibroblast subsets in fibrotic conditions, including idiopathic pulmonary fibrosis and liver cirrhosis12,13. However, the conditions regulating this phenotypic heterogeneity and the unique functional roles of these phenotypically distinct subsets are not yet known.

Disentangling these complexities will clarify the roles of heterogeneous fibroblast subsets in physiological and pathophysiological settings.

ECM composition and mechanics exhibit spatiotemporal heterogeneity in acute injury, fibrosis, and cancer108–113. Notably, both granulation tissue and intratumoral desmoplasia evolve from elastic, fibronectin (FN)-rich to stiff, collagen I (COL 1)-rich

ECM1–3,112,114. Therefore, we hypothesized that ECM properties directly regulate fibroblast heterogeneity. To test this hypothesis, we assessed fibroblast activation on substrata of defined stiffness and composition. Gene expression analysis revealed distinct signatures, suggesting unique functional roles of FAPHi fibroblasts and αSMAHi myofibroblasts in tissue remodeling. These insights will help foster the rational design and clinical application of stromal cell-targeted therapies in cancer and fibrosis.

Results

Substratum directs activated fibroblast phenotypic heterogeneity.

Primary murine adult pulmonary fibroblasts were cultured on tissue culture- treated plastic for 2 passages, which generated a heterogeneous population of activated fibroblasts with variable levels of FAP and αSMA (data not shown). These fibroblasts were reseeded on plastic and the impact of soluble mediators thought to promote fibroblast activation5,6,115 on Fap and Acta2 gene expression was assessed. Consistent with prior studies10, treatment with a variety of fibrotic and inflammatory mediators (in 1% serum to minimize exogenous growth factors) modulated Acta2 expression.

Unexpectedly, none of these mediators modulated Fap expression, except ascorbic acid, which promoted Fap expression at both the mRNA (Table 1 and Fig. 4A) and protein level (Fig. 4B).

Table 1. Fibroblasts cultured on plastic are resistant to induction of Fap expression. QRT-PCR analysis of Fap and Acta2 gene expression in fibroblasts (seeded in 1% serum on tissue culture-treated plastic) in response to 4-day treatment with indicated soluble mediators. The change in gene expression is indicated by the multiplier relative to the untreated control; NC denotes no change.

Ascorbic acid (Vitamin C), an essential cofactor for lysyl and prolyl hydroxylation, promotes stable deposition of collagen (Fig. 4C) by ensuring proper folding of its triple helical structure116. Thus, we hypothesized that ascorbic acid regulates FAP expression by promoting ECM deposition. To test this hypothesis, fibroblasts were cultured on FN- and fibrillar collagen-rich fibroblast-derived matrices (FDMs; Fig. 4D). Interestingly, relative to culture on plastic, fibroblasts cultured on FDMs markedly up-regulated Fap

16 gene expression (Fig. 4E). Flow cytometric analysis further demonstrated that culture on

FDMs versus plastic enriched for FAPHi fibroblasts (Fig. 4F). Moreover, a concomitant reduction in αSMAHi fibroblasts was observed on FDMs versus plastic (Fig. 4F). These data demonstrate that varying substrata can enrich for phenotypically distinct subsets of activated fibroblasts.

Figure 4. Substratum directs activated fibroblast phenotypic heterogeneity. QRT-PCR (A) and representative flow cytometric analysis (B) of FAP and αSMA expression in fibroblasts cultured in 10% serum on tissue culture-treated plastic in the 17 presence or absence of 75 μg/ml ascorbic acid (Vit. C) for 4 days. Data were compiled from 4 independent experiments and bar graphs depict the mean +/- SEM. (C) Collagen levels (as measured via hydroxyproline content) in FDMs deposited by fibroblasts in the presence or absence of 75 μg/ml Vit. C. Data were compiled from 2 independent experiments and bar graphs depict the mean +/- SEM. (D) Representative IF staining of FN and two-photon second harmonic generation imaging of fibrillar collagen in lung FDMs. QRT-PCR (E) and representative flow cytometric analysis (F) of FAP and αSMA expression in fibroblasts in 10% serum on tissue culture-treated plastic or FDM for 4 days. Data were compiled from 4 independent experiments and bar graphs depict the mean +/- SEM; ** p < 0.01.

ECM composition and elasticity govern activated fibroblast phenotypic heterogeneity.

Compared to plastic, FDMs constitute a more physiologically relevant substratum with respect to multiple parameters, including ECM compliance, architecture, and composition117. To delineate the roles of ECM elasticity and composition in driving activated fibroblast heterogeneity, we employed polyacrylamide hydrogels (where ECM ligand and elasticity can be independently controlled118). We primarily utilized 2 and 20 kilopascal (kPa) hydrogels, which encompasses the range of stiffness found in pathophysiological conditions, including tumors and lung fibrosis110,111. Hydrogels were coated with FN or COL I to simulate early versus late stages, respectively, of wound repair, fibrosis, and tumorigenesis1–4.

FN elasticity impacted fibroblast morphology, with reduced cell spreading and cytoskeletal organization after 72 hours of culture on 2 versus 20 kPa FN-coated hydrogels (Fig. 5A), consistent with previous reports119,120. Compared to 20 kPa FN- coated hydrogels, 2 kPa FN-coated hydrogels promoted higher FAP and lower αSMA expression, at the mRNA (Fig. 5B) and protein (Fig. 5C) level. Across the pathophysiological stiffness range, Fap gene expression inversely correlated, while

Acta2 gene expression directly correlated with FN stiffness (Fig. 5D, top panel). The full spectrum of activated fibroblast phenotypic differentiation (FAPHiαSMALow, FAPHiαSMAHi,

18 and FAPLowαSMAHi subsets) was observed on 2, 5, 12, and 20 kPa FN-coated hydrogels, as evidenced by flow cytometric analysis at the single cell level (Fig. 5D, bottom panel).

However, our data clearly illustrate a shift in prevalence from the FAPHiαSMALow reactive fibroblast phenotype to the FAPLowαSMAHi myofibroblast phenotype with increasing FN stiffness (Fig. 5D, bottom panel).

Figure 5. ECM composition and elasticity govern activated fibroblast phenotypic heterogeneity. Representative phalloidin staining of the actin cytoskeleton (A) and Fap and Acta2 gene expression (B) in fibroblasts cultured in 10% serum on 2 versus 20 kPa FN- or COL I- coated hydrogels for 72 hours. Data were compiled from 4 independent experiments and

19 bar graphs depict the mean +/- SEM. (C) Representative flow cytometric analysis, including quantification of relative median fluorescent intensities (MFI) for FAP and αSMA expression in fibroblasts cultured in 10% serum on 2 kPa (blue) versus 20 kPa (red) FN-coated hydrogels for 72 hours. Data were compiled from 3 independent experiments and bar graphs depict the mean +/- SEM. (D) QRT-PCR (top) and flow cytometric analysis (bottom) of FAP and αSMA expression in fibroblasts cultured in 10% serum on 2, 5, 12, and 20 kPa FN-coated hydrogels for 72 hours. Data were compiled from 2 independent experiments and bar graphs depict the mean +/- SEM. (E) Representative flow cytometric analysis, including quantification of relative MFI for FAP and αSMA expression in lung fibroblasts cultured in 10% serum on 2 kPa (blue) versus 20 kPa (red) COL I-coated hydrogels for 72 hours. Data were compiled from 3 independent experiments and bar graphs depict the mean +/- SEM; * p < 0.05; ** p < 0.01; *** p < 0.001; **** p < 0.0001.

Unlike FN-coated hydrogels, both 2 and 20 kPa COL I-coated hydrogels promoted cell spreading and actin stress fiber formation after 72 hours of culture (Fig.

5A). Moreover, both 2 and 20 kPa COL I-coated hydrogels promoted low Fap and high

Acta2 gene expression (Fig. 5B), thereby enriching for FAPLowαSMAHi myofibroblasts

(Fig. 5E). Taken together, these data indicate that an integrated response to ECM composition and elasticity governs activated fibroblast heterogeneity.

Fibroblast differentiation in response to transforming growth factor beta (TGF-β) is governed by ECM composition and elasticity.

To dissect the role of growth factors in regulating fibroblast heterogeneity, fibroblasts were cultured in low (1%) serum (to minimize exogenous growth factors) on

FN- or COL I-coated hydrogels (2 or 20 kPa) for 72 hours. ECM stiffness regulated

Acta2 expression, even in 1% serum (Fig. 6A). In contrast, Fap expression was comparably low in fibroblasts on all substrata in 1% serum (Fig. 6A), unlike in 10% serum (Fig. 5). Serum contains appreciable levels of TGF-β, a pro-fibrotic cytokine thought to promote fibroblast activation. Therefore, we investigated the role of TGF-β- mediated signaling in ECM-dependent regulation of FAP expression by using both gain- and loss-of-function approaches. Addition of recombinant, active TGF-β to 1% serum 20 was sufficient to promote the FAPHiαSMALow reactive fibroblast phenotype on 2 kPa FN- coated hydrogels. Conversely, TGF-β promoted the FAPLowαSMAHi myofibroblast phenotype on 2 kPa COL I-coated hydrogels and 20 kPa hydrogels coated with either

FN or COL I (Fig. 6A-B).

To determine whether TGF-β was necessary for the response to serum, fibroblasts in 10% serum were treated with calcipotriol, a vitamin D receptor (VDR) agonist that inhibits TGF-β/SMAD signaling via genomic competition87. The expression of Cyp24a1, a VDR target gene, was measured as a positive control for calcipotriol efficacy. Interestingly, calcipotriol induced higher Cyp24a1 expression in fibroblasts cultured on 2 versus 20 kPa FN-coated hydrogels (Fig. 6C), indicating that high ECM stiffness can confer resistance to calcipotriol. Furthermore, calcipotriol attenuated high

Fap and Acta2 gene expression in fibroblasts on 2 and 20 kPa FN-coated hydrogels, respectively (Fig. 6C). These findings demonstrate that VDR agonists can curtail both activated fibroblast phenotypes.

Figure 6. Fibroblast differentiation in response to TGF-β is governed by ECM composition and elasticity. QRT-PCR (A) and flow cytometric analysis (B) of FAP and αSMA expression in fibroblasts that were seeded in 1% serum on 2 versus 20 kPa COL I- or FN-coated hydrogels, allowed to adhere overnight, and subsequently treated with 2 ng/ml TGF-β for 48 (A) or 72 hours (B). Data were compiled from 4 independent experiments and bar graphs depict the mean +/- SEM. (C) QRT-PCR analysis of Fap, Acta2, and Cyp24a1 gene expression in fibroblasts that were seeded in 10% serum on FN-coated hydrogels (2 or 20 kPa), allowed to adhere overnight, and subsequently treated with calcipotriol (Cal; 100 or 500 nM) or vehicle control (-) for 48 hours. Data were compiled from 3 independent experiments and bar graphs depict the mean +/- SEM; * p < 0.05; ** p < 0.01; *** p < 0.001; **** p < 0.0001.

Neither ROCK nor FAK activity are required to induce the FAPHi reactive fibroblast phenotype.

Both Rho kinase (ROCK) and focal adhesion kinase (FAK) activity can promote

αSMAHi myofibroblast differentiation. ROCK activity facilitates the actin cytoskeletal reorganization necessary for myofibroblast differentiation in response to ECM stiffening121. FAK activity facilitates TGF-β-induced myofibroblast differentiation122.

Furthermore, FN-mediated FAK activation depends on stiffness, since exposure of the cryptic FN synergy site requires mechanical tension123. Therefore, we assessed whether

ROCK or FAK activity similarly regulate FAPHi reactive fibroblast differentiation. FAPHi fibroblasts (on 2 kPa FN) and αSMAHi myofibroblasts (on 20 kPa FN) were treated with a

ROCK (Y-27632) or FAK (PF573228) inhibitor. As expected, ROCK inhibition attenuated actin stress fiber formation (Fig. 7A) and Acta2 expression (Fig. 7B) on 20 kPa FN- coated hydrogels. In contrast, ROCK inhibition had no effect on Fap expression on 2 or

20 kPa FN-coated hydrogels (Fig. 7B). Treatment with PF573228 diminished FAK activity, as indicated by reduced phospo-FAKTyr397 levels (Fig. 7C). However, FAK inhibition did not revert either the FAPHi or αSMAHi phenotype on 2 or 20 kPa FN, respectively (Fig. 7D). Thus, FAK activity does not appear to be required for the induction of either activated fibroblast phenotype in response to changes in FN elasticity.

Figure 7. Neither ROCK nor FAK activity are required to induce the FAPHi reactive fibroblast phenotype. Phalloidin staining of the actin cytoskeleton (A) and qRT-PCR analysis of Fap and Acta2 gene expression (B) in fibroblasts seeded in 10% serum on 2 versus 20 kPa FN-coated hydrogels, allowed to adhere overnight, and subsequently treated with 10 μM Y-27632 or vehicle control (-) for 48 hours. Data were compiled from 3 independent experiments and bar graphs depict the mean +/- SEM; ** p < 0.01. Immunoblot for p-FAKTyr397 and total FAK protein levels (C) and quantification (D; representative of 2 independent experiments) of qRT-PCR analysis of Fap and Acta2 gene expression in fibroblasts that were seeded (and subsequently cultured for 72 hours) in 10% serum supplemented with PF-573228 (1 or 5 μM) or vehicle control (-). 24

Gene expression profiling indicates that FAPHi reactive fibroblasts predominantly synthesize and proteolyze ECM, while αSMAHi myofibroblasts mediate contraction.

To identify potentially unique functional roles of fibroblast subsets in tissue remodeling, we compared gene expression profiles of fibroblasts cultured in 10% serum on 2 (enriched for FAPHi reactive fibroblasts) versus 20 (enriched for αSMAHi myofibroblasts) kPa FN-coated hydrogels. FAPHi fibroblasts exhibited a more ECM proteolytic phenotype, with higher gene expression of most matrix metalloproteinases

(Mmps) and tissue plasminogen activator (Plat; Fig. 8A). Meanwhile, gene expression of most tissue inhibitors of metalloproteinases (Timps) was comparable or reduced in

FAPHi compared to αSMAHi fibroblasts (Fig. 8A). FAPHi fibroblasts also exhibited higher gene expression of many ECM and matricellular constituents, including collagens I

(Col1a1) and III (Col3a1), decorin (Dcn), extra domain A FN (EDA-Fn1), thrombospondin

2 (Thbs2), and osteopontin (Spp1; Fig. 8A-B). Collectively, these results implicate FAPHi fibroblasts in the dynamic ECM deposition and turnover characteristic of desmoplastic tissue.

On the other hand, FAPHi fibroblasts showed lower gene expression of constituents of the actomyosin contractile apparatus, such as smooth muscle myosin heavy chain (Myh11; Fig. 8B) and beta actin (Actb; Fig. 8C). Furthermore, the gene expression of mediators known to promote fibroblast contractility, including caveolin-1

(Cav1) and endothelin (Edn1), was attenuated in FAPHi compared to αSMAHi fibroblasts

(Fig. 8C). The reduced gene expression of most integrin subunits in FAPHi fibroblasts

(Fig. 8C) also indicates reduced contractile capacity, considering the necessity of integrins for fibroblasts to exert contractile force on ECM112. The gene expression of lysyl oxidase (Lox), an enzyme that crosslinks collagen, was also reduced in FAPHi fibroblasts

(Fig. 8C). Taken together, these results suggest that FAPHi reactive fibroblasts have a diminished capacity to contract, crosslink, and, consequently, stiffen ECM compared to

αSMAHi myofibroblasts.

FAPHi reactive fibroblasts and αSMAHi myofibroblasts also diverged with respect to their gene expression of growth factors, inflammatory mediators, and regulators of cell cycle progression. Compared to αSMAHi myofibroblasts, FAPHi fibroblasts expressed lower mRNA levels of autocrine growth factors that can promote myofibroblast differentiation and expansion, including connective tissue growth factor (Ctgf), platelet derived growth factor α (Pdgfa), Tgfb2, and Tgfb3 (Fig. 8D). Conversely, FAPHi fibroblasts expressed higher mRNA levels of hepatocyte growth factor (Hgf), a paracrine growth factor known to promote epithelial proliferation and survival (Fig. 8D). FAPHi fibroblasts also expressed higher mRNA levels of most inflammatory mediators assessed in our fibrosis array, however, they were expressed at low levels, exhibited relatively high variability, and did not achieve statistical significance at a confidence level of 95% (Table 2). Lastly, FAPHi fibroblasts exhibited lower gene expression of pro- proliferative transcription factors, such as jun oncogene (Jun; Fig. 8D) and myelocytomatosis oncogene (Myc; Fig. 8D), and mediators of cell cycle progression, such as cyclin A2 (Ccna2; Fig. 8B), suggestive of a less proliferative phenotype.

Figure 8. Gene expression profiling indicates that FAPHi reactive fibroblasts predominantly synthesize and proteolyze ECM, while αSMAHi myofibroblasts mediate contraction. Comparison of gene expression profiles, assessed via Qiagen Fibrosis Gene Expression Array (A, C, and D) and independent qRT-PCR analyses (B), of fibroblasts cultured in 10% serum on 2 (FAPHi fibroblasts) or 20 kPa (αSMAHi fibroblasts) FN-coated hydrogels for 72 hours. Relative gene expression (FAPHi/αSMAHi fibroblasts) of ECM proteases (A), ECM protease inhibitors (A), ECM and matricellular components (A and B), contractile mediators (B and C), integrin subunits (C), growth factors (D), and proliferative mediators (B and D). Data were compiled from 3 independent experiments and bar graphs depict the mean +/- SEM; * p < 0.05; ** p < 0.01; *** p < 0.001.

Table 2 (Part 1). Raw data from the Qiagen RT² Profiler Mouse Fibrosis PCR Array. Gene expression values (normalized to β2 microglobulin mRNA levels) in fibroblasts cultured in 10% serum on 2 or 20 kPa FN-coated hydrogels for 72 hours. ND denotes no detectable signal.

Table 2 (Part 2). Raw data from the Qiagen RT² Profiler Mouse Fibrosis PCR Array. Gene expression values (normalized to β2 microglobulin mRNA levels) in fibroblasts cultured in 10% serum on 2 or 20 kPa FN-coated hydrogels for 72 hours. ND denotes no detectable signal.

Discussion

In summary, we delineated features of remodeling tissues that direct activated fibroblast heterogeneity. In our proposed model (Fig. 9), a convergence of ECM composition, elasticity, and TGF-β signaling governs FAPHiαSMALow, FAPHiαSMAHi, and

FAPLowαSMAHi fibroblast phenotypic heterogeneity. These data help explain the activated fibroblast heterogeneity observed in vivo, where FAP and αSMA identify distinct, yet overlapping, fibroblast subsets. This heterogeneity likely stems from the spatiotemporally dynamic variation in ECM composition and elasticity in injury, fibrosis, and cancer1,108–113. Furthermore, our gene expression profiling data indicate potentially unique functionality of these fibroblast subsets in tissue remodeling. FAPHi reactive fibroblasts exhibited an ECM synthetic and proteolytic phenotype, while αSMAHi myofibroblasts exhibited a contractile and proliferative phenotype.

Figure 9. A convergence of ECM composition, elasticity, and TGF-β signaling governs activated fibroblast phenotypic heterogeneity. In FN-rich ECMs of low stiffness, TGF-β promotes the FAPHiαSMALow reactive fibroblast phenotype. Conversely, TGF-β promotes the FAPLowαSMAHi myofibroblast phenotype in the context of stiff and/or COL I-rich ECMs. Gene expression profiling indicates unique 30 functionality of these divergent fibroblast subsets in tissue remodeling, with FAPHi reactive fibroblasts exhibiting an ECM synthetic and proteolytic phenotype, and αSMAHi myofibroblasts exhibiting a contractile and proliferative phenotype.

Extrapolating our model to the well-characterized progression of acute injury resolution109 illuminates ECM-dependent regulation and diverse functionality of fibroblast subsets in tissue remodeling. In early wounds, the highly elastic, FN-rich environment likely promotes FAPHi fibroblast differentiation. The gene expression profile of FAPHi fibroblasts suggests that they orchestrate early phases of wound resolution by secreting:

(1) ECM proteases to facilitate cell migration into the wound, (2) paracrine growth factors to support re-epithelialization, and (3) ECM and matricellular components to replace provisional ECM with COL 1-rich connective tissue. Thereafter, the stiff, COL 1-rich environment of late granulation tissue likely promotes αSMAHi myofibroblast differentiation. The gene expression profile of αSMAHi myofibroblasts suggests that they orchestrate late phases of wound resolution, including fibro-proliferation (via secretion of autocrine growth factors) and wound closure (via contraction). We are now conducting studies utilizing murine models of tissue injury to assess this proposed model of temporal regulation and diverse functionality of FAPHi and αSMAHi fibroblasts in wound resolution.

Comparable regulation and functionality of FAPHi and αSMAHi fibroblast subsets likely occur in tumors, considering the parallels between the wound healing response and tumorigenesis1–3. Previous studies have characterized diverse mechanisms by which carcinoma-associated fibroblasts promote tumor growth. However, the diverse roles of fibroblast subsets in tumorigenesis remained unclear. Our gene expression analysis suggests that FAPHi and αSMAHi fibroblasts diverge in their tumor-promoting

31 functions. For example, αSMAHi myofibroblasts express higher mRNA levels of Lox, an enzyme that stiffens ECM by crosslinking collagen, which promotes focal adhesions,

PI3K signaling, and consequently, tumor invasion72. αSMAHi myofibroblasts also exhibit higher gene expression of Cav1, which promotes Rho-dependent contraction, ECM alignment and stiffening, and consequently, tumor invasion124. Meanwhile, FAPHi fibroblasts exhibit higher gene expression of many tumor-promoting ECM proteases, including Fap itself and most Mmps. Extensive research has previously demonstrated fundamental roles of these stromal proteases in tumor invasion125, angiogenesis25,126, proliferation25,127,128, and epithelial-mesenchymal transition129,130. Moreover, compared to

αSMAHi myofibroblasts, FAPHi fibroblasts exhibit higher gene expression of Hgf, a previously described mechanism by which stromal fibroblasts promote oncogenesis and resistance to targeted cancer therapies131,132.

Importantly, our study addresses the ongoing controversy regarding intratumoral fibroblasts as therapeutic targets133,134. A number of laboratories have employed diverse approaches to deplete FAPHi fibroblasts in murine tumor models16,95–100,104. Virtually all these studies illustrated that depletion of FAPHi fibroblasts disrupts desmoplastic stroma and impedes tumor progression, corroborating the data presented herein and the correlation between high FAP expression and poor prognosis in myriad solid tumor types29–45. On the contrary, the net effect of αSMAHi myofibroblasts in tumor progression remains obfuscated due to conflicting reports. Although many studies have demonstrated a pro-tumorigenic role for αSMAHi myofibroblasts135,136, one recent study utilizing αSMA-thymidine kinase (tk) transgenic mice demonstrated that depletion of proliferating αSMAHi myofibroblasts (via systemic ganciclovir administration) accelerated pancreatic ductal adenocarcinoma aggressiveness56. Importantly, αSMAHi myofibroblast

32 depletion did not diminish the prevalence of FAPHi fibroblasts56, a population that possesses unique tumor-promoting capabilities. In contrast, FAPHi fibroblast depletion also eliminated most intratumoral αSMAHi myofibroblasts16. Thus, FAPHi fibroblast depletion may have demonstrated superior anti-tumor efficacy than αSMAHi myofibroblast depletion because it eliminated heterogeneous, tumor-promoting populations of activated fibroblasts. Alternately, the net benefit of FAPHi fibroblast depletion may have overcome the potential detrimental effect of αSMAHi myofibroblast depletion. The heterogeneity of the input population for our experiments suggests that either: (1) varying substrata supports the survival and/or growth of divergent fibroblast subsets, or (2) activated fibroblasts possess plasticity enabling them to adopt the

FAPHiαSMALow, FAPHiαSMAHi, or FAPLowαSMAHi phenotype, depending on substratum.

Future studies should investigate the lineage, interdependence, and plasticity of these activated fibroblast subsets in vivo. FAPHi fibroblasts may facilitate the generation and/or recruitment of αSMAHi myofibroblasts, or myofibroblast precursors may express FAP.

Unraveling the complex regulation of fibroblast heterogeneity will hasten the successful application of therapies that target fibroblasts. In fact, our findings may help explain recent failures in clinical trials of a few molecularly targeted approaches. For example, simtuzumab (an antibody that blocks LOXL2 enzymatic activity) failed in clinical trials for colorectal adenocarcinoma137, metastatic pancreatic adenocarcinoma138, and idiopathic pulmonary fibrosis139,140. The clinical use of simtuzumab aimed to curb tumor and fibrosis progression by reducing ECM stiffness and myofibroblast generation.

Our finding that FAPHi fibroblasts prevail at low stiffness might help explain the inefficacy of simtuzumab, considering the tumor- and fibrosis-promoting propensities of FAPHi fibroblasts. The attenuation of myofibroblast generation via inhibitors of sonic hedgehog

33 signaling may have failed in clinical trials for similar reasons91,93,94. Intriguingly, calcipotriol-mediated blockade of TGF-β/SMAD signaling (currently being evaluated in clinical trials) has shown promise in pre-clinical models of both pancreatic ductal adenocarcinoma and liver cirrhosis87,88. We have now demonstrated that calcipotriol can attenuate fibroblast activation to both the FAPHi and αSMAHi phenotype, which helps explain the efficacy of calcipotriol in treating both fibrosis and cancer. Overall, our findings underscore the importance of comprehensive monitoring of heterogenous fibroblast populations in response to stroma-targeted therapies to identify optimal treatment modalities.

Future Directions

Importantly, these data are the first (to our knowledge) to disentangle the regulation and potentially unique functionality of heterogeneous activated fibroblast subsets. Future research should build on these results to more precisely delineate the unique roles of activated fibroblast subsets.

Although our work illustrated key roles of FN, COL 1, and TGF-β in regulating activated fibroblast phenotypic heterogeneity, other soluble factors and ECM components present in actively remodeling tissue likely participate in this phenotypic regulation. Fibroblasts cultured on plastic were resistant to induction of Fap expression in response to treatment with various soluble mediators thought to promote fibroblast activation. However, a screen should be performed on 2 and 20 kPa FN-coated hydrogels to determine whether these soluble factors regulate Fap expression under more physiologically relevant conditions (as observed with TGF-β). Furthermore, our studies only utilized hydrogels coated with FN or COL 1 to simulate early versus late stages, respectively, of wound repair, fibrosis, and tumorigenesis1–4. Follow-up studies 34 should assess activated fibroblast heterogeneity on 2 and 20 kPa hydrogels coated with other prevalent ECM components in tumors, fibrosis, and granulation tissue, such as fibrin, EDA-FN, and collagen III (COL 3). Unlike COL 1, COL 3 expression peaks at early stages of wound repair108,141,142. Furthermore, COL 3-deficient excisional wounds contract at accelerated rates and COL 3-deficient fibroblasts display increased myofibroblast differentiation142. Thus, considering its role in restraining myofibroblast differentiation, COL 3 (unlike COL 1) likely promotes the FAPHiαSMALow phenotype at 2 kPa. If true, this result could explain how fibrillar collagen-rich FDMs promote the FAPHi phenotype. Considering the higher gene expression of Col3a1 versus Col1a1 in FAPHi fibroblasts (Table 2), COL3 likely constitutes the majority of fibrillar collagen accumulating in FDMs. Therefore, further analysis (IF or immunoblot) should characterize the fibrillar collagen in FDMs.

Future studies should also further elucidate the signaling pathways driving this phenotypic heterogeneity. Notably, cell toxicity was observed at the doses of PF573228

(> 10 μM) required to fully abrogate FAK activity. To definitively rule out FAK activity in the regulation of either activated fibroblast phenotype, future experiments should utilize adenoviruses encoding FAK mutants, including FRNK (a dominant-negative, FAK- related nonkinase), FAK397F (a non-phosphorylatable FAK mutant), and CD2-FAK (a constitutively active FAK). Moreover, although our studies illustrated that TGF-β is sufficient to promote the FAPHi fibroblast phenotype, further validation is warranted to determine whether TGF-β signaling is necessary for this phenotype. Thus, our results with calcipotriol (which does not exclusively target TGF-β/SMAD signaling) need to be corroborated using either TGF-β type II receptor antagonists or TGF-β neutralizing antibodies.

Lastly, follow-up experiments should compare the unique functionality of heterogeneous activated fibroblasts. Some of the differences between FAPHi (on 2 kPa

FN) and αSMAHi fibroblasts (on 20 kPa FN) identified in my gene expression array may solely be stiffness-dependent, rather than consistent differences between FAPHi and

αSMAHi fibroblasts (regardless of context). Therefore, expression of key genes identified in my array (such as Col1a1, Col3a1, Spp1, Mmps, Hgf, Ctgf, and Edn1) should be compared on 2 kPa FN (FAPHi), 20 kPa FN (αSMAHi), 2 kPa COL 1 (αSMAHi), and 20 kPa COL 1 (αSMAHi). Furthermore, the unique functionality of FAPHi and αSMAHi fibroblasts should be assessed by comparing their contractility (using atomic force microscopy), proliferation, collagenase and/or gelatinase activity (using DQ gelatin or collagen), and secretion of growth factors, ECM components, and inflammatory cytokines. Taken together, these experiments would elucidate how the functional roles of activated fibroblasts evolve in response to changes in ECM composition and stiffness in actively remodeling tissues.

Methods

Isolation of murine lung fibroblasts

Lungs were harvested from 10- to 16-week-old male or female C57BL/6 (Charles

River) mice. Lungs were finely minced, then digested in a cocktail of collagenases I, II, and IV (Worthington; each at 250 μg/ml in DMEM) at 37°C for 2 hours. The collagenase solution was quenched with an equivalent volume of 10% heat-inactivated fetal calf serum (FCS) DMEM. The digested tissue was sequentially strained through 70 and 40

μm cell strainers then spun at 1200 r.p.m. for 10 minutes at 4°C. The cell pellet was re- suspended in 10% FCS DMEM (containing HEPES, L-glutamine, penicillin-streptomycin, fungizone, and gentamicin) and plated on tissue culture-treated plastic (polystyrene).

Cells were maintained at 37°C with 5% CO2 in a humidified incubator. The next day, cell monolayers were washed several times with PBS to remove non-adherent cells. Cells were maintained in 10% FCS DMEM on tissue culture-treated plastic for two passages.

Second passage fibroblasts were used for all experiments.

Fibroblast-derived matrices (FDMs)

Decellularized FDMs were generated as described previously143, using murine lung fibroblasts cultured at confluence in the presence of 75 μg/ml ascorbic acid for 8 days.

Hydrogels

Polyacrylamide hydrogels of varying stiffness were prepared and coated overnight at 4°C with bovine FN (EMD Millipore; 5 μg/ml) or COL I (PureCol; Advanced

BioMatrix; 10 μg/ml) as described previously118. Hydrogels were rinsed twice with PBS and unreacted N-hydroxysuccinimide was blocked with 1 mg/ml heat-inactivated bovine serum albumin in serum-free DMEM for 30 minutes at 37°C. Hydrogels were rinsed twice with PBS before cell seeding.

RNA isolation and quantitative real-time polymerase chain reaction (qRT-PCR)

Cells were rinsed with PBS, lysed in 1 ml Trizol (Invitrogen), and processed for

RNA extraction according to manufacturer’s instructions. After quality assessment by agarose gel electrophoresis, reverse transcription was performed following the standard protocol for the TaqMan Reverse Transcription kit (Applied Biosystems). Transcript levels were assayed via the StepOnePlus Real-Time PCR System (Applied Biosystems) using SYBR Green (Invitrogen). mRNA levels for all genes of interest were normalized to

Hprt1 mRNA levels. Primer sequences: Fap-F, 5’-cacctgatcggcaatttgtg; Fap-R, 5’-

37 cccattctgaaggtcgtagatgt; Acta2-F, 5’-ccagagcaagagagggatcct; Acta2-R, 5’- tgtcgtcccagttggtgatg; Hprt1-F, 5’-tgacactggtaaaacaatgca; Hprt1-R, 5’- ggtccttttcaccagcaagct; Tnc-F, 5’-ggacttacgggtgtctgaaacc; Tnc-R, 5’- tgaggcggtaacgatcaaact; Cyp24a1-F, 5’-ccagcggctagagatcaaac; Cyp24a1-R, 5’- cacgggcttcatgagtttct; Spp1-F, 5’-ccctcgatgtcatccctgtt; Spp1-R, 5’-tgccctttccgttgttgtc;

Myh11-F, 5’-gacaactcctctcgctttgg; Myh11-R, 5’-gctctccaaaagcaggtcac; Ccna2-F, 5’- tggatggcagttttgaatcacc; Ccna2-R, 5’-ccctaaggtacgtgtgaatgtc; EDA-Fn1-F, 5’- cgagccctgaggatggaat; EDA-Fn1-R, 5’-agctctgcagtgtcgtcttcac; Has2-F, 5’- cggtcgtctcaaattcatctg; Has2-R, 5’-acaatgcatcttgttcagctc.

Fibrosis gene expression array

RNA was extracted as described above. RNA purification was performed following the standard protocol for the RNeasy Mini Kit with on-column DNase digestion

(Qiagen). After quality assessment by agarose gel electrophoresis, cDNA was synthesized (using 1 µg RNA per sample) with the RT2 First Strand Kit (Qiagen). Both

RNA and cDNA quality were confirmed using the Murine RT² Profiler Quality Control

Array (Qiagen). Transcript levels were assessed with the RT² Profiler Mouse Fibrosis

PCR Array (Qiagen) using RT2 Sybr Green ROX qPCR Mastermix (Qiagen). RNA levels for all genes of interest were normalized to β2 microglobulin (β2M) mRNA levels.

Flow cytometry

For hydrogel (Figs. 5-6) and Vit. C (Fig 4B) experiments, cells were trypsinized to obtain a single cell suspension. For FDM experiments (Fig. 4F), cells (on plastic or FDM) were trypsinized, and then incubated with 500 μg/ml collagenase I (Worthington) for 30 minutes at 37°C to obtain a single cell suspension. Flow cytometric analyses were

38 performed on an LSR-Fortessa using FACSDiva software (BD Bioscience) and data were analyzed using FlowJo (Tree Star).

Antibodies used for flow cytometry

Biotinylated anti-FAP (Clone 73-3) antibody was generated in-house, and detected using Streptavidin-APC (BioLegend). The specificity of the anti-FAP antibody was verified based on reactivity with cells from wild-type but not FAP-null mice. Dead cells were excluded by using the Live/Dead Fixable Violet Dead Cell Stain Kit (Molecular

Probes). Staining with anti-αSMA antibody (Sigma, clone1A4, FITC) was performed after permeabilizing cell suspensions with the Cytofix/Cytoperm Fixation/Permeabilization Kit

(BD Biosciences), as per the manufacturer’s instructions.

Treatment with inhibitors or soluble factors

For indicated experiments, cells were treated with recombinant human active

TGF-β (R&D), PF573228 (Tocris), Y-27632 (Sigma), calcipotriol (Tocris), L-ascorbic acid

(Vit. C; Fisher), basic fibroblast growth factor (bFGF; Peprotech), interleukin (IL)-1β

(Peprotech), IL-3 (Peprotech), interferon gamma (IFNγ; R&D), IL-6 (Peprotech), tumor necrosis factor alpha (TNFα; Peprotech), high molecular weight hyaluronan (HMW HA;

Healon), low molecular weight hyaluronan (LMW HA; ICN Biochemicals), periostin/OSF-

2 (R&D), buthionine-sulfoximine (BSO; Sigma), tiron (Sigma), trichostatin A (Sigma), pleiotrophin (R&D), angiotensin II (Sigma), and platelet-derived growth factor-bb

(PDGFbb; Calbiochem).

Immunoblot

Total cell lysates were prepared from fibroblasts cultured on hydrogels by placing coverslips face down on 4X NuPAGE LDS Sample Buffer (Invitrogen; supplemented with

10 mM 2-mercaptoethanol) for 2 minutes. Cell lysates were then heated at 70°C for 10

minutes. Equal amounts of extracted protein were resolved on NuPAGE 4-12% Bis-Tris

gels (Invitrogen) and the fractioned proteins were transferred onto PVDF membranes.

Membranes were probed with antibodies to FAK (Cell Signaling, 3285) and phospo-

FAKTyr397 (Cell Signaling, 3283), followed by HRP-conjugated goat anti-rabbit secondary

antibody (Sigma, A0545).

Microscopy

Imaging of fibrillar collagen in decellularized fibroblast-derived matrices was

based on second harmonic generation utilizing a Leica SP5 confocal/multiphoton

microscope (Leica Microsystems) at 20X magnification, as described previously144.

Immunofluorescent (IF) images were obtained with a Nikon Eclipse Ti-E inverted

microscope at 10X (FN) or 20X (phalloidin) magnification.

Hydroxyproline assay

Collagen content in FDMs produced by lung fibroblasts (in the presence or

absence of 75 μg/ml ascorbic acid) was quantified by measuring hydroxyproline levels

(normalized to cell number as measured by calcein AM levels), as previously

described145.

For FN IF, FDMs produced by lung fibroblasts were incubated with rabbit anti-FN antibody (Sigma, Clone F3648), followed by Alexa Fluor 658 goat anti-rabbit antibody

(Invitrogen). For visualization of the actin cytoskeleton, lung fibroblasts on 2 and 20 kPa hydrogels were fixed with 4% paraformaldehyde, permeabilized with 0.1% Triton-X-100, then stained with Alexa Fluor 594-conjugated phalloidin (Molecular Probes) and DAPI.

Statistics

All results are reported as mean (of the indicated number of independent experiments) +/- standard error of the mean (SEM). Statistical analysis was performed using one-way analysis of variance (ANOVA) with the Tukey’s multiple comparison test or two-tailed Student’s t test (Prism 7.0, GraphPad Software). Asterisks denote statistical significance: **** p < 0.0001, *** p < 0.001, ** p < 0.01, * p < 0.05.

CHAPTER THREE: THE ROLE OF FAP-DEPENDENT ECM REMODELING IN LUNG

TUMORIGENESIS

Introduction

The desmoplastic reaction, a complex process involving the recruitment and differentiation of CAFs together with extensive ECM remodeling, plays a crucial role in tumor development6,64,112,146. However, the mechanisms by which desmoplasia regulates tumor development remain mostly undefined. Delineating these mechanisms will enable the refinement of integrative cancer therapy, whereby neoplastic and stromal cells are simultaneously targeted, potentially resulting in synergistic anti-tumor efficacy.

In virtually all carcinomas, a subset of CAFs express FAP, a cell surface serine protease with both endopeptidase and dipeptidyl peptidase activity18,20,22. The selective expression of FAP in the tumor microenvironment makes it an attractive drug target15–19.

Our laboratory previously demonstrated that both Fap genetic deletion and pharmacological inhibition of FAP enzymatic activity attenuate tumor burden and proliferation in an inducible, autochthonous lung tumor model (the Lox-Stop-Lox (LSL)

KrasG12D/+ model). However, the LSL-KrasG12D/+ model has notable drawbacks. The tumors in this model exhibit only modest intratumoral desmoplasia, making it difficult to study ECM remodeling. Furthermore, the necessary viral administration for tumor initiation induces an inflammatory response, which can influence tumorigenesis.

Therefore, in this study, we extended our analyses by comparing lung tumor development and intratumoral desmoplasia in two additional models, the LSL-

KrasG12D/+;Tgfbr2flox/flox model model and the KrasG12D latent allele 2 (KrasLA2/+) model.

Highly desmoplastic lung tumors develop in the LSL-KrasG12D/+;Tgfbr2flox/flox model, making it an attractive model to study intratumoral ECM remodeling. Meanwhile, the 42 spontaneous nature of tumor initiation in the KrasLA2/+ model avoids the potentially confounding influence of viral administration.

The diminished lung tumor burden in FAP-deficient LSL-KrasG12D/+ mice was associated with excess collagen accumulation and dysregulated integrin-mediated signaling25. Recent studies also demonstrated that FAP cleaves MMP-1-digested collagen I and III, and that FAP-mediated collagen proteolysis promotes cellular uptake and turnover of collagen23,24. Therefore, considering the importance of desmoplasia in tumorigenesis, we hypothesized that FAP-dependent collagen proteolysis and ECM remodeling directly promote tumor cell proliferation and invasion. To test this hypothesis, we employed FDMs as a reductionist approach to assess the direct effect of FAP- dependent ECM remodeling on tumor cell proliferation and migration.

We demonstrated that Fap genetic deletion promotes intratumoral fibrillar collagen accumulation and attenuates lung tumor multiplicity, size, progression, and proliferation in the KrasLA2/+ model. Furthermore, we obtained preliminary evidence that

Fap-/- lung fibroblasts produce ECM with enhanced fibrillar collagen accumulation, which directly impedes both tumor cell proliferation and motility. These studies validate FAP as a drug target in the treatment of lung cancer and help explain how FAP promotes lung tumor growth, further clarifying the mechanisms by which desmoplasia regulates tumorigenesis.

Results

Fap+/+, Fap+/-, and Fap-/- mice exhibit equivalent lung tumor burden in the LSL-

KrasG12D/+;Tgfbr2flox/flox model of lung adenocarcinoma.

We sought to investigate the role of FAP-dependent ECM remodeling in tumorigenesis by utilizing a highly desmoplastic lung tumor model. For this purpose, we employed the LSL-KrasG12D/+Tgfbr2flox/flox model, characterized by highly invasive tumors with abundant desmoplasia147 (Fig. 10A). In this model, lung tumorigenesis is initiated by intranasal administration of adenovirus expressing cre recombinase (Ad-cre), which causes activation of oncogenic Kras (KrasG12D) and loss of TGF-β type II receptor in targeted cells. Surprisingly, however, LSL-KrasG12D/+;Tgfbr2flox/flox;Fap+/+, LSL-

KrasG12D/+;Tgfbr2flox/flox;Fap+/-, and LSL-KrasG12D/+;Tgfbr2flox/flox;Fap-/- mice exhibited nearly equivalent lung tumor burden (Fig. 10B-C), distinguishing it from other lung, breast, and pancreatic autochthonous tumor models that we have assessed in our lab to date.

Therefore, we deemed the LSL-KrasG12D/+TGFBR2flox/flox lung cancer model an unsuitable model to delineate the mechanisms by which FAP-mediated ECM remodeling promotes tumorigenesis.

Figure 10. Fap+/+, Fap+/-, and Fap-/- mice exhibit equivalent lung tumor burden in the LSL-KrasG12D/+;Tgfbr2flox/flox model of lung adenocarcinoma. Representative Dichrome staining (A), H&E staining (B), and quantification of tumor burden (C) in LSL-KrasG12D/+;Tgfbr2flox/flox;Fap+/+ (n=14 mice), LSL- KrasG12D/+;Tgfbr2flox/flox;Fap+/- (n=5 mice), and LSL-KrasG12D/+;Tgfbr2flox/flox;Fap-/- (n=9 mice) 45 lung tissue sections (8 weeks after initiating tumorigenesis via intranasal administration of Ad-cre). Bar graph depicts the mean +/- SEM.

Fap genetic deletion restrains lung tumor progression and growth in the spontaneous KrasLA2/+ model of lung tumorigenesis.

The inducible LSL-KrasG12D/+ model has notable caveats, including the promiscuous targeting of varied cell types in the lung by intranasal administration.

Furthermore, the inflammatory response induced by Ad-cre administration likely exerts a profound and confounding influence on tumor development. Therefore, we investigated the role of FAP in tumorigenesis in the spontaneous KrasLA2/+ lung tumor model148. In this model, mice harbor latent, oncogenic KrasG12D that spontaneously activates in vivo via genetic recombination. KrasLA2/+ mice develop lung tumors with 100% penetrance that evolve from hyperplasia to adenoma (or carcinoma, in some cases), reminiscent of human non-small cell lung cancer148. Tumor burden was compared in KrasLA2/+;Fap+/+,

KrasLA2/+;Fap+/-, and KrasLA2/+;Fap-/- mice at 12 and 18 weeks of age. At 12 weeks,

KrasLA2/+;Fap-/- mice exhibited diminished lung tumor burden compared to KrasLA2/+;Fap+/+ and KrasLA2/+;Fap+/- mice. However, a gene dosage effect was not observed in this model, as KrasLA2/+;Fap+/+ and KrasLA2/+;Fap+/- lungs had nearly equivalent tumor burden

(Fig. 11A-B). Therefore, given the high variability of lung tumor burden in this model

(reflecting the stochastic nature of tumor initiation), data were combined from

KrasLA2/+;Fap+/+ and KrasLA2/+;Fap+/- mice for most analyses, thereby focusing the comparison on KrasLA2/+;Fap+ (KrasLA2/+;Fap+/+ and KrasLA2/+;Fap+/-) versus KrasLA2/+;Fap-

(KrasLA2/+;Fap-/-) mice. Lung tumor burden was reduced by 37% in KrasLA2/+;Fap- versus

KrasLA2/+;Fap+ mice (Fig. 11C). The multiplicity of lung tumors did not differ between

KrasLA2/+;Fap- and KrasLA2/+;Fap+ mice (Fig. 11D). Instead, this difference in tumor burden

46 was attributed to a reduced frequency of large tumors (> 0.5 mm2) in KrasLA2/+;Fap- mice

(Fig. 11E). These results suggest that FAP promotes tumor growth, but not initiation, in this model.

Figure 11. Fap genetic deletion attenuates lung tumor burden in 12-week-old KrasLA2/+ mice. 47

Representative H&E staining (A) and quantification of % tumor/lung area (B-C), tumor number (D), and incidence of large tumors (E; % tumors > 0.5 mm2) in KrasLA2/+;Fap+/+ (n=20 mice), KrasLA2/+;Fap+/- (n=13 mice), and KrasLA2/+;Fap-/- (n=18 mice) lung tissue sections. Combined KrasLA2/+;Fap+/+ and KrasLA2/+;Fap+/- (depicted as Fap+), n = 33 mice. Bar graphs depict mean +/- SEM; * p < 0.05; ** p < 0.01.

The role of FAP at later stages of lung tumorigenesis was investigated by comparing the tumor burden in KrasLA2/+;Fap+/+, KrasLA2/+;Fap+/-, and KrasLA2/+;Fap-/- mice at 18 weeks of age. Of note, 5 mice (1 KrasLA2/+;Fap+/+, 3 KrasLA2/+;Fap+/-, and 1

KrasLA2/+;Fap-/-) in the allocated cohort of mice for this analysis (a total of 12

KrasLA2/+;Fap+/+, 13 KrasLA2/+;Fap+/-, and 11 KrasLA2/+;Fap-/- mice) succumbed before reaching 18 weeks of age, and were therefore excluded from analysis. Similar to the 12- week timepoint, KrasLA2/+;Fap-/- mice exhibited diminished lung tumor burden compared to KrasLA2/+;Fap+/+ and KrasLA2/+;Fap+/- mice at the 18-week timepoint; meanwhile,

KrasLA2/+;Fap+/+ and KrasLA2/+;Fap+/- mice had nearly equivalent lung tumor burden (Fig.

12A-B). Taken together, KrasLA2/+;Fap- mice exhibited a 50% decrease in lung tumor burden compared to KrasLA2/+;Fap+ (KrasLA2/+;Fap+/+ and KrasLA2/+;Fap+/- combined) mice at 18 weeks of age (Fig. 12C). At this later timepoint, diminished lung tumor burden in

KrasLA2/+;Fap- mice was attributed to reductions in tumor multiplicity, average tumor size, and incidence of large tumors (> 0.5 mm2) (Fig. 12D-F). Consistent with the 12 week timepoint, the increased frequency of large tumors at 18 weeks of age in KrasLA2/+;Fap+ mice suggests that FAP promotes lung tumor growth.

Figure 12. Fap genetic deletion attenuates tumor multiplicity, size, and total tumor burden in the lungs of 18-week-old KrasLA2/+ mice. Representative H&E staining (A) and quantification of % tumor/lung area (B-C), tumor number (D), average tumor area (E; in mm2), and incidence of large tumors (F; % tumors > 0.5 mm2) in KrasLA2/+;Fap+/+ (n = 11 mice), KrasLA2/+;Fap+/- (n = 10 mice), KrasLA2/+;Fap-/- (n = 10 mice) lung tissue sections. Combined KrasLA2/+;Fap+/+ and KrasLA2/+;Fap+/- (depicted as Fap+), n = 21 mice. Bar graphs depict mean +/- SEM. * p < 0.05; ** p < 0.01; *** p < 0.001.

The role of FAP in lung tumor progression was assessed by comparing the frequency of hyperplasias, adenomas, and adenocarcinomas in KrasLA2/+;Fap+/+ versus

KrasLA2/+;Fap-/- lung tissue sections. At both 12 and 18 weeks of age, KrasLA2/+;Fap+/+ and

KrasLA2/+;Fap-/- lungs harbored the same number of hyperplasias, further substantiating the conclusion that FAP does not regulate tumor initiation in this model. However, at 18 weeks of age, the incidence of adenomas was significantly reduced in KrasLA2/+;Fap-/- versus KrasLA2/+;Fap+/+ lungs (Fig. 13). These results indicate that Fap genetic deletion restrains lung tumor progression and the incidence of adenomas. The frequency of adenocarcinomas at these timepoints was too low to ascertain the role of FAP in regulating adenoma-to-adenocarcinoma progression.

Figure 13. Fap genetic deletion restrains tumor progression in the KrasLA2/+ model of lung tumorigenesis. Quantification of the frequency of hyperplasias, adenomas, and adenocarcinomas in lung tissue sections from 12- and 18-week-old KrasLA2/+;Fap+/+ and KrasLA2/+;Fap-/- mice.

12-week-old KrasLA2/+;Fap+/+, n = 10 mice; 12-week-old KrasLA2/+;Fap-/-, n = 9 mice; 18- week-old KrasLA2/+;Fap+/+, n = 11 mice; 18-week-old KrasLA2/+;Fap-/-, n = 11 mice. Bar graph depicts the mean +/- SEM; * p < 0.05.

To further investigate the role of FAP in regulating lung tumorigenesis in the

KrasLA2/+ model, intratumoral proliferation was assessed. Using Ki67 immunohistochemistry (IHC), the proliferative indices of KrasLA2/+;Fap+/+, KrasLA2/+;Fap+/-, and KrasLA2/+;Fap-/- tumors were measured at the 18-week timepoint. Compared to

KrasLA2/+;Fap+/+ and KrasLA2/+;Fap+/- tumors, KrasLA2/+;Fap-/- tumors exhibited approximately a 50% reduction in proliferation at the 18-week timepoint (Fig. 14A-B).

Reflecting the comparable tumor burden in KrasLA2/+;Fap+/+ and KrasLA2/+;Fap+/- lungs, no difference in proliferation was observed in KrasLA2/+;Fap+/+ versus KrasLA2/+;Fap+/- tumors.

The attenuated proliferative indices of KrasLA2/+;Fap-/- tumors help explain the decreased tumor burden, incidence of large tumors, and average tumor size in KrasLA2/+;Fap-/- mice at this timepoint. Thus, Fap genetic deletion slows tumor growth in the KrasLA2/+ model by curtailing intratumoral proliferation.

Figure 14. Fap genetic deletion restrains lung tumor proliferation. Representative images (A) and quantification (B) of Ki67 IHC staining of KrasLA2/+;Fap+/+ (n = 153 tumors, pooled from 11 mice), KrasLA2/+;Fap+/- (n = 98 tumors, pooled from 10 mice), and KrasLA2/+;Fap-/- (n = 86 tumors, pooled from 10 mice) lung tumors from 18- week-old mice. Proliferative index was calculated by the ratio of Ki-67+ to hematoxylin+ nuclei. Bar graph depicts the mean +/- SEM; **** p < 0.0001. 52

To investigate the role of FAP-mediated ECM remodeling in lung tumorigenesis, the desmoplastic response was compared in KrasLA2/+;Fap+/+ versus KrasLA2/+;Fap-/- tumors. Picro Sirius Red staining of lung tumors was visualized under polarized light to image intratumoral fibrillar collagen. Within KrasLA2/+;Fap+/+ and KrasLA2/+;Fap-/- tumor nests, very little fibrillar collagen was detected. Instead, most fibrillar collagen localized to the tumor-stromal interface at the tumor periphery. In these areas, more fibrillar collagen tended to accumulate in KrasLA2/+;Fap-/- versus KrasLA2/+;Fap+/+ tumors (Fig. 15).

Figure 15. Fibrillar collagen accumulates at the tumor-stromal interface in KrasLA2/+;Fap-/- tumors. Representative images (acquired under brightfield or polarized light settings) of KrasLA2/+;Fap+/+ (n = 6 mice) and KrasLA2/+;Fap-/- (n = 6 mice) lung tumor sections (from 18-week-old mice) stained with Picro Sirius Red to visualize fibrillar collagen structure. White arrows point at fibrillar collagen accumulation at the tumor-stromal interface.

Fap-/- fibroblast-derived matrices (FDMs) restrain tumor cell proliferation and motility.

Considering the correlation between enhanced collagen accumulation and reduced proliferation observed in Fap-/- tumors in a variety of different tumor models 53

(including the spontaneous KrasLA2/+ model, the inducible LSL-KrasG12D/+ model, and various transplanted tumor models25), we hypothesized that FAP-dependent ECM remodeling regulates tumor proliferation directly or indirectly (for example, by regulating retention, presentation, and/or activation of soluble factors). The degradation of collagen may expose cryptic binding sites that influence tumor cell behavior. Alternately, the promotion of collagen internalization by FAP proteolysis may remove modulating growth signals from the ECM. Lastly, collagen fragmentation may dictate how fibroblasts contract and remodel ECM, which consequently influences the mechanical and/or architectural properties of intratumoral ECM (Fig. 16).

Figure 16. Possible mechanisms by which FAP-dependent collagen proteolysis may regulate ECM remodeling, and consequently, tumor growth.

A variety of different in vitro experimental approaches were employed to test these potential mechanisms. First, we tested the hypothesis that FAP-mediated proteolysis of ¼ and ¾ collagen fragments eliminates bioactive fragments or generates products with distinct bioactivities. Purified type I collagen was first digested in vitro with recombinant MMP-1 alone or sequentially by MMP-1, then recombinant FAP extracellular domain (Fig. 17A). Murine lung carcinoma cells, including Lewis lung

54 carcinoma (LLC1; derived from a spontaneous lung carcinoma in a C57/BL6 mouse149) and K-ras mutant lung adenocarcinoma (LKR13; derived from a KrasLA1/+ mouse148,150) cell lines, were seeded on tissue culture-treated plastic and treated with either MMP-1- digested- or MMP-1 and FAP-digested type I collagen. Treatment with variable concentrations of these collagen fragments did not modulate proliferation of either cell line (Fig. 17B). LKR13 cells were also cultured on type I collagen gels in the presence or absence of MMP-1 +/- FAP. However, addition of MMP-1 and/or FAP did not alter tumor proliferation (Fig. 17C). Taken together, these experiments suggest that FAP-mediated collagen proteolysis does not directly impact tumor proliferation by generating or degrading bioactive type I collagen fragments.

Figure 17. Lung tumor cells proliferate equally well in the presence of collagen I digested by MMP-1 alone or by MMP-1 and FAP. LLC or LKR13 lung adenocarcinoma cells were seeded on plastic, treated with or without 5, 10, or 20 μg/ml of type I collagen fragments (A; digested by MMP-1 alone or sequentially with MMP-1 and FAP), and counted after 72 hours of treatment (B). (C) LKR13 cells were seeded on 1 mg/ml type I collagen gels, treated with or without MMP-1 (500 ng/ml) +/- FAP (3 nM), and counted after 72 hours of treatment. Bar graphs depict the mean +/- SEM of three replicate wells per condition.

Recent studies have demonstrated that Fap+/+ primary lung fibroblasts internalize type I collagen more efficiently than Fap-/- fibroblasts, indicating that FAP participates

56 directly in fibroblast-dependent collagen catabolism and clearance24. Considering these recent findings, we hypothesized that FAP-dependent collagen proteolysis facilitates tumor-promoting, fibroblast-dependent ECM remodeling. To test this hypothesis, we compared the composition and influence on tumor cell behavior of FDMs deposited by

Fap+/+ versus Fap-/- lung fibroblasts.

Using two-photon second harmonic generation imaging, a two-fold enhancement of fibrillar collagen accumulation was observed in ECM deposited by Fap-/- versus Fap+/+ fibroblasts (Fig. 18A-B). Furthermore, Fap-/- lung FDMs exhibited higher hydroxyproline content (a surrogate assessment of total collagen levels) than Fap+/+ FDMs (Fig. 18C). In contrast, our preliminary assessment of FN deposition showed roughly equivalent levels in Fap-/- versus Fap+/+ FDMs, suggesting that FAP may selectively regulate fibrillar collagen, rather than total ECM accumulation (Fig. 18D). Furthermore, FDM-producing

Fap+/+ and Fap-/- fibroblasts exhibited roughly equivalent patterns of αSMA expression, suggesting that these differences in collagen accumulation do not result from varying states of myofibroblast activation between isolated Fap+/+ and Fap-/- fibroblast populations under these culture conditions (Fig. 18E). Rather, the impaired capacity of

Fap-/- fibroblasts to degrade and turn over fibrillar collagen24 likely explains the enhanced collagen accumulation in Fap-/- FDMs. Thus, Fap-/- FDMs recapitulate the collagen accumulation observed in Fap-/- versus Fap+/+ lungs in murine models of lung tumorigenesis and fibrosis24,25. These results validate this in vitro system as a model to study the effect of collagen accumulation (due to Fap deletion) on tumor cell behavior.

Figure 18. Fap-/- lung fibroblasts deposit FDMs with enhanced fibrillar collagen accumulation. Representative images (A) and quantification (B) of fibrillar collagen (visualized via two- photon SHG microscopy) and quantification of total collagen (C; assessed via hydroxyproline assay) in Fap-/- versus Fap+/+ FDMs. Results were compiled from 4 independent experiments and the bar graphs depicts the means of experiments +/- SEM; * p < 0.05; ** p < 0.01. Representative images of FN (D) and αSMA (E) IF staining of Fap+/+ versus Fap-/- FDMs (before de-cellularization). N = 2 FDMs per genotype.

To investigate the role of FAP-dependent fibroblast-mediated ECM remodeling in the regulation of tumor cell proliferation, LKR13 lung adenocarcinoma cells were cultured on Fap+/+ versus Fap-/- FDMs. LKR13 proliferation on FDMs (in 1% serum) was assessed following an 8-hour pulse with 5-ethynyl-2-deoxyuridine (EdU), a thymidine analog incorporated into DNA during active DNA synthesis151. Compared to Fap+/+

FDMs, Fap-/- FDMs hindered LKR13 proliferation, as indicated by a reduction in cell number (measured via Hoescht staining of nuclei) and EdU uptake (Fig. 19A). Pooling data from 3 independent experiments showed that tumor cell count and proliferation were attenuated by 40% and 20%, respectively, on Fap-/- versus Fap+/+ FDMs (Fig. 19B).

Statistical significance will need to be determined with additional experimental replicates.

Similarly, additional lung tumor cell lines should be investigated to definitively establish that Fap-/- FDMs consistently restrain lung tumor cell proliferation.

Figure 19. Fap-/- FDMs restrain lung tumor cell proliferation. LKR13 tumor cells were cultured in 1% serum on Fap+/+ versus Fap-/- FDMs for 24 hours (with 10 μM EdU for the final 8 hours). Quantification of cell count (left; average number of Hoechst+ nuclei per 10X field) and proliferative index (right; % Edu+ / Hoechst + nuclei) from a representative experiment (A) and compiled data from 3 independent experiments (B). Graphs depict the means +/- SEM (n = 3 FDMs in A; n = 3 independent experiments in B); * p < 0.05; ** p < 0.01.

Prior studies demonstrated that CAFs can enhance the invasion of carcinoma cells via a mechanism dependent on MMP activity and Rho-mediated contraction of the 59 matrix75. Furthermore, another study found that FAP-overexpressing NIH-3T3 fibroblasts produce highly aligned ECMs that enhanced the directionality and velocity of pancreatic cancer cell migration152. Therefore, we compared tumor cell motility on Fap+/+ versus

Fap-/- pulmonary FDMs to investigate the potential role of FAP-dependent ECM remodeling in lung tumor invasion and metastasis. Lung (LKR13) and metastatic human breast (MDA-MB-231153) adenocarcinoma cell lines were utilized to simulate intrapulmonary migration of primary and metastatic tumor cells, respectively. Preliminary experiments demonstrated that Fap-/- versus Fap+/+ lung FDMs impeded the trajectory

(total distance traveled), but not the directionality, of lung and breast tumor cell migration

(Fig. 20A-B). These experiments provide preliminary evidence that FAP-mediated ECM remodeling can directly promote tumor cell migration.

Figure 20. Fap-/- FDMs restrain tumor cell motility. Trajectory (left; in microns) and directionality (right; displacement/trajectory) of MDA-MB- 231 (A) or LKR13 (B) migration were tracked on Fap+/+ versus Fap-/- FDMs for 6 hours. Bar graphs depict mean +/- SEM; * p < 0.05; **** p < 0.0001. In A, n = 100 cells per condition (compiled from 3 FDMs) and results depict 1 representative experiment (of 2 experiments). In B, n = 30 cells per condition (compiled from 3 FDMs) and results depict data from 1 experiment.

Discussion

In summary, I have shown that Fap genetic deletion curbs tumor growth in the spontaneous KrasLA2/+ model, but not the highly desmoplastic LSL-KrasG12D/+Tgfbr2flox/flox model of lung tumorigenesis. Highly aggressive LSL-KrasG12D/+Tgfbr2flox/flox tumors147 may have overcome the tumor suppressive microenvironment evident in Fap-/- lungs in other tumor models, including the LSL-KrasG12D/+ model25 and the KrasLA2/+ model. Of note, tumor burden was assessed relatively late (8 weeks post-tumor initiation), and differences in tumor growth kinetics in Fap+/+, Fap+/-, and Fap-/- lungs may have been missed. Therefore, assessment of tumor burden at earlier timepoints in this model is warranted.

In the KrasLA2/+ model, the diminished lung tumor burden in KrasLA2/+Fap-/- mice at

18 weeks of age was attributed to reductions in tumor multiplicity, size, progression, and proliferation. Notably, an attenuation in lung tumor multiplicity was observed in

KrasLA2/+;Fap- versus KrasLA2/+;Fap+ mice at 18, but not 12, weeks of age. Intrapulmonary metastatic dissemination may explain this discrepancy. Definitively proving this hypothesis would require karyotyping of individual tumors. In KrasLA2/+;Fap+ versus

KrasLA2/+;Fap- lungs, the comparable number of total tumors at the 12-week timepoint as well as the comparable number of hyperplasias at both the 12- and 18-week timepoints suggests that FAP does not regulate tumor initiation in this model. However, assessment at earlier timepoints is warranted to more definitively validate this conclusion. Furthermore, future experiments should assess intratumoral EdU incorporation to more definitively establish that Fap genetic deletion restrains intratumoral proliferation.

Picro Sirius Red staining revealed increased fibrillar collagen accumulation at the tumor-stromal interface of KrasLA2/+Fap-/- versus KrasLA2/+Fap+/+ tumors, corroborating the enhanced collagen accumulation (as assessed via hydroxyproline assay) in Fap-/- versus

Fap+/+ lungs in the LSL-KrasG12D/+ lung tumor model25. Using FDMs as a reductionist approach, my preliminary data suggest that Fap-/- pulmonary fibroblasts accumulate a fibrillar collagen-rich ECM that directly attenuates tumor cell proliferation and migration, compared to growth on Fap+/+ FDM. These results offer further support for my hypothesis that FAP-dependent collagen proteolysis and ECM remodeling directly promote tumor cell proliferation and invasion.

Both in vivo (KrasLA2/+ mice) and in vitro (FDMs) models require further analyses to more carefully delineate FAP-dependent changes to ECM (including its architecture, mechanics, biochemical modifications, and composition) and ECM-dependent signaling

(including integrin- and discoidin domain receptor-mediated pathways) to uncover specific mechanisms by which FAP-dependent ECM modification regulates tumorigenesis. For example, the fibrillar collagen that accumulates in Fap-/- tumors and

FDMs should be characterized. Closer examination of the gene expression profile of

FAPHi fibroblasts (Table 2) revealed approximately two-fold higher levels of Col3a1 than

Col1a1. Thus, COL 3 might constitute the majority of the fibrillar collagen accumulating in Fap-/- ECM. A recent study from our collaborators (the laboratory of Dr. Susan Volk) demonstrated that COL 3 suppresses primary tumor growth and metastasis144.

Specifically, in an orthotopic model of breast cancer, they demonstrated accelerated primary tumor growth and heightened pulmonary metastatic burden in Col3a1+/- versus

Col3a1+/+ mice. Furthermore, in vitro experiments demonstrated that COL 3 directly impairs breast cancer cell invasion and that breast cancer cells proliferate more on

Col3a1-/- versus Col3a1+/+ FDMs144. Thus, impaired collagen degradation and clearance in Fap-/- tumors might directly impede tumor proliferation and invasion by causing the excessive accumulation of tumor-inhibitory COL 3.

Taken together, these studies offer further support for FAP as a drug target for the treatment of lung cancer. Importantly, these studies also highlight the complexity of intratumoral desmoplasia in the regulation of tumorigenesis. As discussed in chapter 1, intratumoral ECM can promote or impede tumorigenesis, often depending on its composition and degradative state. Thus, rather than eliminate intratumoral desmoplasia, novel stroma-targeted therapeutic strategies might aim to enrich for the tumor-inhibitory components of ECM for optimal cancer treatment, and perhaps even prevention.

Methods

Animals

Fap-/- mice (Fap-luciferase; FapLuc/Luc) were generated by modifying a targeting construct provided by Dr. Andreas Schnapp at Boehringer-Ingelheim. LA2–KrasG12D/+ mice (KrasLA2/+; provided by Dr. Joseph Kissil) were crossed with FAPLuc/Luc mice to generate littermate KrasLA2/+FapLuc/Luc, KrasLA2/+FapLuc/+, and KrasLA2/+Fap+/+ mice. LSL-

KrasG12D/+Tgfbr2flox/flox (provided by Dr. Steven Albelda) were crossed with FapLuc/Luc mice to generate littermate LSL-KrasG12D/+;Tgfbr2flox/flox;Fap+/+, LSL-KrasG12D/+;Tgfbr2flox/flox;

Fap+/Luc, and LSL-KrasG12D/+;Tgfbr2flox/flox;FapLuc/Luc mice.

Genotyping

The following primer sequences were used for genotyping. Fap: 5’-

CAGCTGTGCTTTATGGGTTT, 5’-CCCTCGAAGGCCTACTAATC, and 5’-

ATCTCTCGTGGGATCATTGT. KrasLA2: 5’-TGCACAGCTTAGTGAGACCC, 5’-

GACTGCTCTCTTTCACCTCC, 5’-GGAGCAAAGCTGCTATTGGC. Tgfbr2flox: 5’-

TAAACAAGGTCCGGAGCCCA, 5’-ACTTCTGCAAGAGGTCCCCT. LSL-KrasG12D: 5’-

CCATGGCTTGAGTAAGTCTGC, 5’- CGCAGACTGTAGAGCAGCG.

Tumor models

For the KrasLA2/+ tumor model, male and female mice were sacrificed at 12 or 18 weeks of age. For the LSL-KrasG12D/+Tgfbr2flox/flox model, tumor initiation was induced in

12-to-14-week-old male or female mice via intranasal administration of 100 uL MEM containing 5x107 plaque forming units of adenovirus expressing cre recombinase

(University of Iowa). Approximately 15 minutes before viral administration, 0.5 μl CaCl2 was added to improve gene transfer147. LSL-KrasG12D/+Tgfbr2flox/flox mice were sacrificed 8 weeks after initiating tumorigenesis via adenoviral administration. In both tumor models, after mice were euthanized, the pulmonary circulation was perfused with PBS via intracardiac injection. Lungs were then fixed via intratracheal instillation of Prefer

(Anatech) and resected. All 5 lung lobes were fixed overnight in Prefer, dehydrated in

70% ethanol, and embedded in paraffin. 4 μm sections were cut from 3 distinct cross- sectional areas of lung tissue.

Assessment of lung tumor burden

For each sample, 4 μm sections (from 3 distinct cross-sectional areas of the lung) were stained with hematoxylin and eosin. Images were obtained at 2X magnification using an Olympus BX51 upright microscope and image stitching was performed using Photoshop (Adobe). Quantification of tumor burden (percent tumor-to- lung area), tumor number, and tumor area was performed using FIJI software.

Dichrome staining

Tissue sections were de-paraffinized, rehydrated, and incubated with Bouin’s solution (Rowley Biochemicals) overnight at room temperature (RT). After washing slides with running tap H2O to remove yellow color, slides were stained with Biebrich

Scarlet-Acid Fuschin (Rowley Biochemicals) for 5 minutes, washed with deionized H2O for 5 minutes, and then placed in Phosphotungstic/Phosphomolybdic Acid (Rowley

Biochemicals) solution for 20 minutes at RT. Sections were stained with Aniline Blue

(Rowley Biochemicals) for 10 minutes, dipped in 1% acetic acid at RT, and rinsed in deionized H2O. Slides were then dehydrated, cleared in Clear-Rite 3, and mounted in

Cytoseal XYL. Images were obtained at 10X magnification using an Olympus BX51 upright microscope.

Ki67 immunohistochemistry

After de-paraffinizing and re-hydrating tissue sections, antigen retrieval was performed in 10 mM sodium citrate buffer (pH 7.4) for 20 minutes at 95°C. Endogenous peroxidase activity was quenched with 3% hydrogen peroxide for 15 minutes at RT.

Sections were blocked with 10% normal goat serum in 1% BSA for 15 minutes at RT.

Endogenous avidin and biotin were blocked using the avidin-biotin kit (Vector Labs).

Sections were incubated with Ki67 antibody (Abcam, clone SP6, 1:100) overnight at 4°C, followed by a 1 hour incubation with biotinylated goat anti-rabbit antibody (Vector Labs,

BA-1000) at RT. Bound antibodies were detected with horseradish peroxidase using the

Vectastain Elite ABC kit (Vector Laboratories) and sections were counterstained with hematoxylin. Slides were then dehydrated, cleared in Clear-Rite 3, and mounted in

Cytoseal XYL. Digital images of the entirety of every tumor in each tissue section were taken at 20X magnification using an Olympus BX51 upright microscope and image

66 stitching was performed using Photoshop (Adobe). Quantification was performed using

FIJI software. Proliferative index was calculated by the ratio of Ki-67+ cells to hematoxylin+ cells.

Picro Sirius Red staining

Tissue sections were de-paraffinized, rehydrated, and stained with 0.1% Picro

Sirius Red solution (Direct Red-80 in picric acid) at RT. Slides were then washed in acidifed water (0.5% glacial acetic acid), dehydrated, cleared in Clear-Rite 3, and mounted in Cytoseal XYL. Images were obtained under brightfield or polarized light at

10X magnification using a Nikon Eclipse E600 microscope.

Collagen digests

Rat tail type I collagen gels (BD Biosciences) were sequentially digested with recombinant human MMP-1 (R&D Systems) and/or recombinant murine FAP extracellular domain (ECD), as described previously24. Collagen digests were prepared for electrophoresis by adding 4X NuPAGE sample buffer and reducing agent and heating at 70°C for 10 minutes. Equivalent volumes of the collagen digests were resolved on a 4–12% Bis-Tris gel (Invitrogen) with MES running buffer. Undigested, acid-solubilized rat tail type 1 collagen was also included as a reference for intact collagen. The gel was stained with SimplyBlue SafeStain (Invitrogen), washed in water, and photographed.

Proliferation assays with collagen fragments

LLC and LKR13 lung adenocarcinoma cells were seeded in 10% serum in tissue culture-treated 96-well plates with or without 5, 10, or 20 μg/ml MMP-1- or MMP-1/FAP- digested collagen fragments. 3 replicates were seeded per condition. 72 hours later,

67 cells were trypsinized, quenched with 10% serum, and enumerated using a Z Series

Coulter Counter (Beckman Coulter Life Sciences).

Proliferation assays on collagen gels with recombinant MMP-1 +/- FAP ECD

96-well plates were coated with 30 μL of 1 mg/ml neutralized rat tail type I collagen (BD Biosciences) and incubated for 1 hour at 37°C to facilitate collagen polymerization. LKR13 cells were seeded on collagen gels in 10% serum +/- MMP-1

(500 ng/ml) +/- FAP ECD (3 nM). 3 replicates were seeded per condition. 72 hours later, wells were washed in PBS and then incubated in 150 μl collagenase solution [1.5 mg/ml type IV collagenase (Worthington), 2 mM EDTA (Gibco), and 200 μg/ml DNase (Roche) in DMEM) for 20 minutes at 37°C to obtain a single cell suspension. Cells were counted using a Z Series Coulter Counter.

Lung fibroblast isolation (digest method)

See description of method in chapter 2.

Fibroblast-derived matrices (FDMs)

See description of method in chapter 2.

2-photon second harmonic generation imaging of fibrillar collagen

Imaging of fibrillar collagen in decellularized fibroblast-derived matrices was based on second harmonic generation (SHG) utilizing a Leica SP5 confocal/multiphoton microscope (Leica Microsystems) as described previously144.

Hydroxyproline assay

See description of method in chapter 2.

Immunofluorescence

After depositing FDMs for 8 days, lung fibroblasts were fixed with 4% paraformaldehyde, permeabilized with 0.05% Triton-X-100, blocked with 5% BSA, and incubated with rabbit anti-FN antibody (Sigma, Clone F3648) and anti-αSMA antibody

(Sigma, clone1A4, FITC), followed by AF658 goat anti-rabbit antibody (Invitrogen) and

DAPI.

EdU proliferation assay on FDMs

Before seeding tumor cells, FDMs were blocked with heat-denatured BSA for 1 hour at 37°C and then washed twice with PBS. LKR13 lung adenocarcinoma cells were seeded on FDMs (3 replicate wells per condition) in 1% serum and allowed to adhere overnight at 37°C. The next day, 10 μM EdU was added. After an 8-hour incubation with

EdU at 37°C, cells were fixed with 4% PFA. EdU+ and total nuclei were detected using the Click-iT EdU Imaging Kit (Invitrogen) and Hoechst staining, respectively. Images were obtained in 5 random 20X fields per well (for quantification of proliferative index) and 5 random 10X fields per well (for quantification of cell count) using an Olympus

CKX41 inverted microscope. Quantification was performed using FIJI software.

Proliferative index was calculated by the ratio of EdU+ nuclei to total Hoechst+ nuclei and cell count was measured as the average number of Hoescht+ nuclei per 10X field.

Time lapse motility assays on FDMs

Before seeding tumor cells, FDMs were blocked with heat-denatured BSA for 1 hour at 37°C and then washed twice with PBS. LKR13 or MDA-MB-231 were seeded at low confluence in 10% FCS DMEM (3 replicate wells per condition) and allowed to adhere to FDMs overnight at 37°C. The next day, the 6-well plate was put on an environmentally controlled microscope stage and allowed to equilibrate for at least 30 minutes. Cell movements were then recorded with a Leica DMI4000 microscope by 69 acquiring a phase contrast image every 10 minutes for 6 hours in 10-12 randomly chosen 10X fields per well. Volocity software was used to track migration of individual cells, including the net path distance (D, μm), path trajectory (T, μm), and directionality

(calculated by the D/T ratio). As recommended by established protocols154, cells that underwent cell division or collided with other cells were excluded from analysis, since these events can interfere with the rate and directionality of cell migration.

CHAPTER FOUR: CONCLUSIONS AND FUTURE DIRECTIONS

In this dissertation, I delineated the regulation and potentially unique functionality of heterogeneous activated fibroblast subsets. My studies revealed that ECM mechanics and composition regulate fibroblast activation to the FAPHi reactive fibroblast versus

αSMAHi myofibroblast phenotype. Moreover, I defined gene expression signatures for

FAPHi and αSMAHi fibroblasts, which suggest that these subsets may exert unique functional roles in tissue remodeling. Lastly, my studies provide novel insights into the mechanisms by which FAP proteolytic activity regulates ECM composition to promote tumorigenesis.

Regulation and functionality of heterogenous fibroblast subsets in actively remodeling tissues

The divergent outcomes in response to varied therapies targeting CAFs

(discussed at length in chapter one) emphasize the urgent need to delineate the unique roles of CAF subsets. The data obtained in this dissertation address this gap in knowledge. Specifically, we sought to uncover the unique regulation and functionality of

αSMAHi and FAPHi CAFs in tumorigenesis.

Importantly, the emergence of αSMAHi and FAPHi fibroblast subsets is not restricted to tumorigenesis. Rather, activated fibroblast heterogeneity arises in virtually all conditions associated with active tissue remodeling, including fibrosis and wound healing. Since tumors are often characterized as “wounds that never heal”1, we hypothesized that similar mechanisms direct fibroblast heterogeneity in both wound healing and cancer. In acute injury, plasma extravasation from damaged blood vessels deposits a provisional matrix, consisting mostly of fibrin and FN. This provisional matrix

71 facilitates migration of diverse cell types into the wound. The injury response is highly regulated, and consists of three sequential phases: (1) inflammation, (2) tissue formation

(including re-epithelialization, ECM deposition, and neovascularization), and (3) ECM remodeling and contraction. During the process of wound healing, provisional matrix gradually evolves into collagen-rich connective tissue109. Analogous evolution of the

ECM occurs during tumorigenesis. Highly permeable tumor microvasculature promotes plasma fibrinogen and FN deposition. As in granulation tissue, intratumoral desmoplasia evolves from provisional fibrin and FN-rich ECM into mature, collagenous ECM1,2.

Therefore, we hypothesized that the transformation of a provisional FN-rich ECM into mature COL 1-rich ECM directly regulates the spatiotemporal heterogeneity of activated fibroblasts noted in these contexts. Using FN- and COL1-coated hydrogels of varying stiffness, we discovered that ECM composition, stiffness, and TGF-β coordinately regulate activated fibroblast heterogeneity. On soft, FN-rich ECMs, TGF-β promotes the FAPHiαSMALow reactive fibroblast phenotype. Conversely, TGF-β promotes the FAPLowαSMAHi myofibroblast phenotype on stiff and/or COL 1-rich ECMs (Fig. 9).

Comparable regulation and diversification of functionality of fibroblast subsets likely occurs in injury and tumorigenesis. For clarification, we center our working model on the acute injury response, due to its well-characterized and highly regulated progression (Fig. 21). In early wounds, the soft, FN-rich environment likely promotes

FAPHi fibroblast differentiation. The gene expression profile of FAPHi fibroblasts suggests that they orchestrate early phases of wound resolution by secreting: (1) ECM proteases to facilitate cell migration into the wound, (2) paracrine growth factors to support re- epithelialization, and (3) ECM and matricellular components to replace provisional ECM with COL1-rich connective tissue. Thereafter, the stiff, COL1-rich environment of late 72 granulation tissue likely promotes αSMAHi myofibroblast differentiation. Contractile,

αSMAHi myofibroblasts orchestrate the late phases of wound resolution, including fibro- proliferation (by secreting autocrine growth factors) and wound closure.

Figure 21. Working model depicting the proposed regulation and functionality of FAPHi and αSMAHi fibroblast subsets in the injury response.

Recent evidence from our laboratory substantiates this proposed model. We observed spatiotemporal regulation of fibroblast heterogeneity in cutaneous wound healing. Temporally, FAP expression peaks around day 7 post-injury, while αSMA expression peaks around day 11. Spatially, most FAPHi activated fibroblasts reside in the hypodermis, and most αSMAHi myofibroblasts reside in the dermis (Govindaraju, et al.; manuscript in preparation). Furthermore, we observed comparable spatial heterogeneity of smooth muscle cell (SMC) populations in atherosclerotic plaques. αSMAHi SMCs

73 localize mostly to the collagen-rich fibrous cap, while FAPHi SMCs accumulate primarily in the lipid-rich plaque interior (Monslow, et al.; manuscript in preparation). These results coincide with the divergent spatiotemporal patterns of FAP and αSMA expression observed in diverse contexts (discussed at length in chapter 1), such as in varied tumor types (including breast17, pancreatic55,57, and lung58), fibrotic conditions (including idiopathic pulmonary fibrosis12 and liver cirrhosis12), and after acute injury (including post-bleomycin-induced lung injury24,48 and post-myocardial infarction11). Across these diverse contexts, αSMAHi, but not FAPHi, fibroblasts usually reside in collagen-rich areas, which substantiates our working model12,50,51,58. Moreover, the emergence of FAPHi fibroblasts usually precedes that of αSMAHi myofibroblasts, further corroborating our working model11,24,48,55. Considering these findings in conjunction with recent evidence that FAPHi fibroblast depletion eliminates most αSMAHi myofibroblasts, but not vice versa16, it seems that FAPHi fibroblasts may play a key role in the generation of αSMAHi myofibroblasts.

Future work should further investigate our proposed model in vivo. Correlations between activated fibroblast heterogeneity and ECM composition could be assayed by immunofluorescence staining for FAP, αSMA, FN, and COL1 in tumors and granulation tissue. Combining FAP and αSMA immunofluorescence with atomic force microscopy could assess the relationship between ECM stiffness and fibroblast heterogeneity.

Lastly, studies could demonstrate in vivo that modifying ECM regulates fibroblast heterogeneity. For these studies, intratumoral stiffness can be modified by administering

β-aminopropionitrile (BAPN) or LOX inhibitors to tumor-bearing mice. Moreover, injecting collagenase intratumorally could degrade fibrillar collagen. I predict that BAPN, LOX inhibitors, and collagenase treatments would all enrich for FAPHi fibroblasts.

Our gene expression profiling data also highlight potentially unique tumor- modulatory capacities of FAPHi fibroblasts and αSMAHi myofibroblasts. The heightened capacity of αSMAHi myofibroblasts to contract and crosslink ECM suggests that they further amplify intratumoral stiffness. Given the ample evidence that ECM stiffness promotes tumor growth and invasion73,74,112,155, it may therefore appear counterintuitive that FAPHi fibroblasts (which prevail under low stiffness conditions) seem to more profoundly and consistently promote tumorigenesis than αSMAHi myofibroblasts16,55,56,95,96,100,102. Our gene expression data helps explain this apparent conundrum, since FAPHi fibroblasts exhibit higher gene expression of many tumor- promoting ECM proteases, growth factors, and ECM and matricellular components (as discussed in chapter 2). Furthermore, coinciding with a recent report of a pro- tumorigenic, pro-inflammatory, and αSMALow fibroblast phenotype57, our data suggest that FAPHi fibroblasts may exhibit a more pro-inflammatory phenotype. However, follow- up studies will need to more definitively establish this pro-inflammatory phenotype by simulating other features of the tumor microenvironment.

Taken together, these studies are the first to precisely delineate features of actively remodeling tissues that drive activated fibroblast heterogeneity. These mechanistic insights will help foster the rational design of therapeutic strategies that target activated fibroblasts in cancer and fibrosis.

Fibroblast heterogeneity: A determinant of stroma-targeted therapy efficacy

Importantly, our results may help explain the dichotomous results obtained so far with different stroma-targeted therapies, such as sonic hedgehog (SHH) pathway inhibitors and VDR agonists. In PDAC, SHH pathway inhibition accelerates tumor growth and attenuates survival by changing the balance between the epithelial and stromal 75 compartments. Specifically, SHH inhibition accelerates the proliferation of epithelial- derived tumor cells, but depletes αSMAHi myofibroblasts and type I collagen93,94. Our gene expression data illustrated that, compared to αSMAHi myofibroblasts, FAPHi reactive fibroblasts express higher mRNA levels of paracrine growth factors supportive of epithelial proliferation (such as Egf and Hgf), and lower mRNA levels of autocrine growth factors (such as Ctgf, Tgfb2, Tgfb3, and Pdgfa). Since SHH inhibition depletes intratumoral αSMAHi myofibroblasts and type I collagen, it may enrich for FAPHi reactive fibroblasts that preferentially support epithelial (i.e. tumor cell) proliferation. Therefore, future studies should assess changes in intratumoral fibroblast heterogeneity (including the relative abundance of αSMAHi and FAPHi fibroblast populations) in response to treatment with inhibitors of SHH signaling.

Furthermore, our results have important implications for understanding VDR agonist-mediated blockade of TGF-β/SMAD signaling as a therapeutic strategy87,88. We found that treatment with calcipotriol (a VDR agonist) induced higher expression of

Cyp24a1 (a VDR target gene) in fibroblasts cultured on 2 versus 20 kPa FN-coated hydrogels. Furthermore, calcipotriol fully abrogated high Fap expression on 2 kPa hydrogels, but only partially attenuated high Acta2 expression on 20 kPa hydrogels.

These results indicate that FAPHi fibroblasts are more sensitive to calcipotriol and that high ECM stiffness may confer resistance to calcipotriol treatment. Intriguingly, in murine models of chronic pancreatitis and PDAC, calcipotriol treatment attenuates the expression of many ECM components, inflammatory cytokines, and growth factors, but does not significantly alter Acta2 expression88. Therefore, future studies should assess changes to intratumoral fibroblast heterogeneity (including the relative abundance of

αSMAHi and FAPHi fibroblast populations) in response to treatment with calcipotriol.

Since recent studies suggest that αSMAHi myofibroblasts may restrain tumor aggressiveness56, preferential targeting of tumor-promoting FAPHi fibroblasts may help explain the efficacy of calcipotriol in pre-clinical tumor models.

Investigating the mechanisms underlying ECM-dependent generation of FAPHi versus αSMAHi myofibroblasts

Fully understanding the regulation of heterogenous fibroblast populations will foster the rational design of stroma-targeted therapies in cancer. Therefore, future studies should delineate the mechanisms underlying the ECM-dependent switch between the FAPHi reactive fibroblast and αSMAHi myofibroblast phenotypes.

Surprisingly, neither FAK nor ROCK activity regulates this phenotypic transition. Based on the evidence discussed herein, I propose that changes in ECM composition and mechanics directs fibroblast heterogeneity by regulating ligation of specific integrins (Fig.

22).

Accessibility of integrin-binding sites in FN depends on mechanical tension.

Under low stiffness conditions, cells can bind the constitutively exposed RGD site in FN, but not the sequestered PHSRN sequence (also known as the synergy site). Increasing mechanical tension stretches FN molecules, and consequently exposes the cryptic

PHSRN motif80,123. Importantly, exposure of the PHSRN sequence regulates integrin engagement. αvβ3 binds RGD, but not PHSRN motifs. In contrast, full α5β1 ligation requires β1 and α5 binding of RGD and PHSRN, respectively80. Consequently, substratum with only accessible RGD sites favors αvβ3 binding, while substratum with

156 exposed RGD and PHSRN motifs favors α5β1 binding . Thus, FN stiffness may dictate

αvβ3 versus α5β1 engagement in lung fibroblasts by regulating the accessibility of the

PHSRN motif. 77

A recent study on the regulation of mesenchymal stem cell (MSC) differentiation corroborates this hypothesis157. This study demonstrated that modulating the stiffness of

FN-coated hydrogels controls MSC differentiation by regulating αvβ3 versus α5β1 integrin ligation. Specifically, the folded conformation of FN on low stiffness hydrogels preferentially engaged αvβ3 integrins, which promoted adipogenic differentiation. In contrast, the stretched conformation of FN on high stiffness hydrogels preferentially

157 engaged α5β1 integrins, which promoted osteogenic differentiation . Consequently, I propose a similar mechanism underlying fibroblast differentiation, in which ligation of

Hi Hi αvβ3 promotes FAP fibroblast differentiation and ligation of α5β1 promotes αSMA myofibroblast differentiation (Fig. 22). Follow-up studies can test this proposed mechanism by treating lung fibroblasts on FN-coated hydrogels of variable stiffness with

αvβ3 and α5β1 integrin antagonists.

Future studies should also explore how COL 1 promotes αSMAHi myofibroblast differentiation, even under low stiffness conditions. COL 1’s fibrillar structure may confer added stiffness to the underlying hydrogel substratum, which we could test by comparing the stiffness of uncoated, FN-coated, and COL1-coated hydrogels via atomic force microscopy. In this light, collagen I deposition may similarly stiffen actively remodeling tissues in vivo, thereby promoting myofibroblast differentiation. Alternately, the constitutively accessible GFOGER sequence in COL 1’s triple helical structure may underlie COL 1-induced myofibroblast differentiation, since the GFOGER sequence can

123 fully activate α2β1 integrin at any stiffness . α2β1 integrin recognition of COL 1 through the GFOGER motif depends on COL 1 triple helical structure123,158. As MMPs degrade

COL I, they alter integrin specificity by unwinding COL I triple helices159 and exposing cryptic RGD sites59. As a result, β1 and αvβ3 integrins preferentially bind native fibrillar

78 collagen (via the GFOGER sequence) and exposed RGD sites, respectively59,60.

Therefore, COL 1-induced αSMAHi myofibroblast differentiation on low stiffness hydrogels may depend on its triple helical structure. Future studies should compare fibroblast differentiation on low stiffness hydrogels coated with fibrillar or denatured

(either monomeric or proteolyzed) COL 1. I predict that αvβ3 ligation of RGD sites in denatured collagen would promote FAPHi fibroblast differentiation, while β1 ligation of fibrillar collagen would promote αSMAHi myofibroblast differentiation (Fig. 22).

Figure 22. Proposed mechanisms underlying ECM-dependent generation of FAPHi versus αSMAHi fibroblasts in actively remodeling tissues.

Considering these hypothesized mechanisms in the context of granulation tissue helps clarify this proposed model of ECM-dependent regulation of fibroblast heterogeneity (Fig. 22). A FN-rich, low stiffness, and highly proteolytic environment characterizes early granulation tissue. Under these conditions, both FN and proteolyzed

COL 1 present RGD binding sites that preferentially engage αvβ3, which likely promotes

FAPHi reactive fibroblast differentiation. Meanwhile, a COL 1-rich, high stiffness environment characterizes late granulation tissue. Under stiff conditions, FN molecules

present both RGD and PHSRN motifs, which preferentially engage α5β1 integrins. Intact, fibrillar COL 1 molecules present GFOGER motifs, which preferentially engage β1 integrins. Under these conditions, preferential ligation of β1 integrins likely promotes

Hi αSMA myofibroblast differentiation. Meanwhile, a heterogeneous mixture of αvβ3 and β1 ligands in the ECM may promote the FAPHiαSMAHi intermediate phenotype. Performing the experiments described above will test the mechanisms proposed in this model.

FAP promotes primary tumor growth and metastasis in diverse tumor types

In this dissertation, we validated FAP as a drug target in lung cancer by demonstrating that FAP promotes tumor progression and proliferation in the spontaneous, KrasLA2/+ model of lung tumorigenesis. These results fully corroborate a prior report from our laboratory that similarly demonstrated a pro-tumorigenic role of FAP in the LSL-KrasG12D/+ lung cancer model25. Taken together, these studies demonstrate that FAP proteolytic activity itself constitutes an important mechanism by which FAPHi reactive fibroblasts promote tumorigenesis25. Given its role in the ordered proteolytic processing of collagen I and III23, we also tested the hypothesis that FAP-mediated ECM remodeling directly promotes tumor cell proliferation and invasion. We found that FAP regulates fibrillar collagen accumulation both in vivo and in vitro. Furthermore, we

80 obtained preliminary evidence that compared to Fap+/+ FDMs, fibrillar collagen-rich Fap-/-

FDMs directly attenuate lung tumor cell proliferation and migration.

Recently, our laboratory has also demonstrated that FAP accelerates primary tumor initiation and growth in the MMTV-HER2/neu model of breast cancer (Blomberg, et al., in preparation) and the KPC model of pancreatic ductal adenocarcinoma (Lo, et al., in preparation). Furthermore, FAP promotes colonization and metastatic outgrowth of transplanted pancreatic, breast, and pulmonary adenocarcinoma cell lines in the lung as well as metastatic dissemination to multiple visceral organs in the KPC model (Lo, et al., in preparation). Taken together, these studies validate FAP as an attractive drug target for the treatment of a variety of tumor types. Future studies should further characterize the fibrillar collagen that accumulates in Fap-/- tumors and FDMs. An accumulation of tumor-inhibitory COL 3 could help explain how Fap genetic deletion restrains tumor growth, invasion, and metastasis.

ECM: A key player in tumorigenesis

The work in this dissertation underscores the essential role of ECM remodeling in tumor evolution. We demonstrated that ECM regulates key aspects of tumorigenesis, such as CAF heterogeneity, proliferation, and invasion. Notably, these studies illustrate how cells integrate mechanical, biochemical, and architectural cues from ECM to regulate their differentiation and behavior. Collectively, these findings constitute important advances in understanding intratumoral fibroblast heterogeneity and desmoplasia.

REFERENCES

1. Dvorak, H. F. Tumors: wounds that do not heal. N. Engl. J. Med. 315, 1650–1659 (1986). 2. Dvorak, H. F. Tumors: wounds that do not heal--Redux. Cancer Immunol. Res. 3, 1–11 (2015). 3. Schafer, M. & Werner, S. Cancer as an overhealing wound: an old hypothesis revisited. Nat. Rev. Mol. Cell Biol. 9, 628–638 (2008). 4. Rybinski, B., Franco-Barraza, J. & Cukierman, E. The wound healing, chronic fibrosis, and cancer progression triad. Physiol. Genomics 46, 223–244 (2014). 5. Kalluri, R. The biology and function of fibroblasts in cancer. Nat. Rev. Cancer 16, 582–598 (2016). 6. Öhlund, D., Elyada, E. & Tuveson, D. Fibroblast heterogeneity in the cancer wound. J. Exp. Med. 211, 1503–1523 (2014). 7. Tomasek, J. J., Gabbiani, G., Hinz, B., Chaponnier, C. & Brown, R. A. Myofibroblasts and mechano-regulation of connective tissue remodelling. Nat. Rev. Mol. Cell Biol. 3, 349–363 (2002). 8. Hinz, B. et al. Recent developments in myofibroblast biology: paradigms for connective tissue remodeling. Am. J. Pathol. 180, 1340–1355 (2012). 9. Hinz, B. et al. The myofibroblast: one function, multiple origins. Am. J. Pathol. 170, 1807–1816 (2007). 10. Serini, G. & Gabbiani, G. Mechanisms of myofibroblast activity and phenotypic modulation. Exp. Cell Res. 250, 273–283 (1999). 11. Tillmanns, J. et al. Fibroblast activation protein alpha expression identifies activated fibroblasts after myocardial infarction. J. Mol. Cell. Cardiol. 87, 194–203 (2015). 12. Acharya, P. S., Zukas, A., Chandan, V., Katzenstein, A.-L. A. & Puré, E. Fibroblast activation protein: a serine protease expressed at the remodeling interface in idiopathic pulmonary fibrosis. Hum. Pathol. 37, 352–360 (2006). 13. Levy, M. T. et al. Fibroblast activation protein: a cell surface dipeptidyl peptidase and gelatinase expressed by stellate cells at the tissue remodelling interface in human cirrhosis. Hepatology 29, 1768–78 (1999). 14. Bauer, S. et al. Fibroblast activation protein is expressed by rheumatoid myofibroblast-like synoviocytes. Arthritis Res. Ther. 8, R171 (2006). 15. Garin-Chesa, P., Old, L. J. & Rettig, W. J. Cell surface glycoprotein of reactive stromal fibroblasts as a potential antibody target in human epithelial cancers. Proc. Natl. Acad. Sci. U. S. A. 87, 7235–7239 (1990). 16. Lo, A. et al. Tumor-promoting desmoplasia is disrupted by depleting FAP- expressing stromal cells. Cancer Res. 75, 2800–2810 (2015). 17. Tchou, J. et al. Fibroblast activation protein expression by stromal cells and tumor-associated macrophages in human breast cancer. Hum. Pathol. 44, 2549– 2557 (2013). 18. Park, J. E. et al. Fibroblast activation protein, a dual specificity serine protease expressed in reactive human tumor stromal fibroblasts. J. Biol. Chem. 274, 36505–36512 (1999). 19. Busek, P. et al. Fibroblast activation protein alpha is expressed by transformed and stromal cells and is associated with mesenchymal features in glioblastoma. Tumor Biol. 37, 13961–13971 (2016). 82

20. Scanlan, M. J. et al. Molecular cloning of fibroblast activation protein alpha, a member of the serine protease family selectively expressed in stromal fibroblasts of epithelial cancers. Proc. Natl. Acad. Sci. 91, 5657–5661 (1994). 21. Rosenblum, J. S. & Kozarich, J. W. Prolyl peptidases: a serine protease subfamily with high potential for drug discovery. Curr. Opin. Chem. Biol. 7, 496–504 (2003). 22. Aertgeerts, K. et al. Structural and kinetic analysis of the substrate specificity of human fibroblast activation protein alpha. J. Biol. Chem. 280, 19441–19444 (2005). 23. Christiansen, V. J., Jackson, K. W., Lee, K. N. & McKee, P. A. Effect of fibroblast activation protein and alpha2-antiplasmin cleaving enzyme on collagen types I, III, and IV. Arch. Biochem. Biophys. 457, 177–86 (2007). 24. Fan, M.-H. et al. Fibroblast activation protein (FAP) accelerates collagen degradation and clearance from lungs in mice. J. Biol. Chem. 291, 8070–8089 (2016). 25. Santos, A. M., Jung, J., Aziz, N., Kissil, J. L. & Puré, E. Targeting fibroblast activation protein inhibits tumor stromagenesis and growth in mice. J. Clin. Invest. 119, 3613–3625 (2009). 26. Gillum, M. P. & Potthoff, M. J. FAP finds FGF21 easy to digest. Biochem. J 473, 1125–1127 (2016). 27. Zhen, E. Y., Jin, Z., Ackermann, B. L., Thomas, M. K. & Gutierrez, J. A. Circulating FGF21 proteolytic processing mediated by fibroblast activation protein. Biochem. J 473, 605–614 (2016). 28. Dunshee, D. R. et al. Fibroblast activation protein cleaves and inactivates fibroblast growth factor 21. J. Biol. Chem. 291, 5986–5996 (2016). 29. Wen, X. X. Fibroblast activation protein-α-positive fibroblasts promote gastric cancer progression and resistance to immune checkpoint blockade. Oncol. Res. 25, 629–640 (2016). 30. Iwasa, S. et al. Increased expression of seprase, a membrane-type serine protease, is associated with lymph node metastasis in human colorectal cancer. Cancer Letters 227, 229–236 (2005). 31. Chen, L., Qiu, X., Wang, X. & He, J. FAP positive fibroblasts induce immune checkpoint blockade resistance in colorectal cancer via promoting immunosuppression. Biochem. Biophys. Res. Commun. 487, 8–14 (2017). 32. Henry, L. R. et al. Clinical implications of fibroblast activation protein in patients with colon cancer. Clin. Cancer Res. 13, 1736–1741 (2007). 33. Wikberg, M. L. et al. High intratumoral expression of fibroblast activation protein (FAP) in colon cancer is associated with poorer patient prognosis. Tumor Biol. 34, 1013–1020 (2013). 34. Cohen, S. J. et al. Fibroblast activation protein and its relationship to clinical outcome in pancreatic adenocarcinoma. Pancreas 37, 154–158 (2008). 35. Kawase, T. et al. Fibroblast activation protein-α-expressing fibroblasts promote the progression of pancreatic ductal adenocarcinoma. BMC Gastroenterol. 15, 109 (2015). 36. Patsouras, D. et al. Fibroblast activation protein and its prognostic significance in correlation with vascular endothelial growth factor in pancreatic adenocarcinoma. Mol. Med. Rep. 11, 4585–90 (2015). 37. Errarte, P. et al. The expression of fibroblast activation protein in clear cell renal cell carcinomas is associated with synchronous lymph node metastases. PLoS One 11, e0169105 (2016). 83

38. López, J. I. et al. Fibroblast activation protein predicts prognosis in clear cell renal cell carcinoma. Hum. Pathol. 54, 100–105 (2016). 39. Wang, H. et al. Downregulation of FAP suppresses cell proliferation and metastasis through PTEN/PI3K/AKT and Ras-ERK signaling in oral squamous cell carcinoma. Cell Death Dis. 5, e1155 (2014). 40. Yu, H., Yang, J., Li, Y. & Jiao, S. The expression of fibroblast activation protein-α in primary breast cancer is associated with poor prognosis. Chinese J. Cell. Mol. Immunol. 31, 370–4 (2015). 41. Jia, J., Martin, T., Ye, L. & Jiang, W. FAP-α (Fibroblast activation protein-α) is involved in the control of human breast cancer cell line growth and motility via the FAK pathway. BMC Cell Biol. 15, 16 (2014). 42. Yuan, D. et al. Overexpression of fibroblast activation protein and its clinical implications in patients with osteosarcoma. J. Surg. Oncol. 108, 157–162 (2013). 43. Liao, Y., Ni, Y., He, R., Liu, W. & Du, J. Clinical implications of fibroblast activation protein-α in non-small cell lung cancer after curative resection: a new predictor for prognosis. J. Cancer Res. Clin. Oncol. 139, 1523–1528 (2013). 44. Saigusa, S. et al. Cancer-associated fibroblasts correlate with poor prognosis in rectal cancer after chemoradiotherapy. Int. J. Oncol. 38, 655–63 (2011). 45. Ju, M.-J. et al. Peritumoral activated hepatic stellate cells predict poor clinical outcome in hepatocellular carcinoma after curative resection. Am. J. Clin. Pathol. 131, 498–510 (2009). 46. Niedermeyer, J. et al. Targeted disruption of mouse fibroblast activation protein. Mol. Cell. Biol. 20, 1089–94 (2000). 47. Niedermeyer, J. et al. Expression of the fibroblast activation protein during mouse embryo development. Int. J. Dev. Biol. 45, 445–7 (2001). 48. Zhang, K., Rekhter, M. D., Gordon, D. & Phan, S. H. Myofibroblasts and their role in lung collagen gene expression during pulmonary fibrosis. A combined immunohistochemical and in situ hybridization study. Am. J. Pathol. 145, 114–25 (1994). 49. Levy, M., McCaughan, G., Marinos, G. & Gorrell, M. Intrahepatic expression of the hepatic stellate cell marker fibroblast activation protein correlates with the degree of fibrosis in hepatitis C virus infection. Liver Int. 22, 93–101 (2002). 50. Scharl, M. et al. Hallmarks of epithelial to mesenchymal transition are detectable in Crohn’s disease associated intestinal fibrosis. Clin. Transl. Med. 4, (2015). 51. Brokopp, C. E. et al. Fibroblast activation protein is induced by inflammation and degrades type I collagen in thin-cap fibroatheromata. Eur. Heart J. 32, 2713–22 (2011). 52. Dolznig, H. et al. Characterization of cancer stroma markers: In silico analysis of an mRNA expression database for fibroblast activation protein and endosialin. Cancer Immun. 5, (2005). 53. Huber, M. A. et al. Fibroblast activation protein: differential expression and serine protease activity in reactive stromal fibroblasts of melanocytic skin tumors. J. Invest. Dermatol. 120, 182–188 (2003). 54. Huber, M. A. et al. Expression of stromal cell markers in distinct compartments of human skin cancers. J. Cutan. Pathol. 33, 145–55 (2006). 55. Feig, C. et al. Targeting CXCL12 from FAP-expressing carcinoma- associated fibroblasts synergizes with anti–PD-L1 immunotherapy in pancreatic cancer. Proc. Natl. Acad. Sci. U. S. A. 110, 20212–20217 (2013). 56. Özdemir, B. C. et al. Depletion of carcinoma-associated fibroblasts and fibrosis 84

induces immunosuppression and accelerates pancreas cancer with reduced survival. Cancer Cell 25, 719–734 (2014). 57. Öhlund, D. et al. Distinct populations of inflammatory fibroblasts and myofibroblasts in pancreatic cancer. J. Exp. Med. 214, 579–596 (2017). 58. Kilvaer, T. K. et al. Cancer associated fibroblasts in stage I-IIIA NSCLC: prognostic impact and their correlations with tumor molecular markers. PLoS One 10, e0134965 (2015). 59. Davis, G. E., Bayless, K. J., Davis, M. J. & Meininger, G. A. Regulation of tissue injury responses by the exposure of matricryptic sites within extracellular matrix molecules. Am. J. Pathol. 156, 1489–1498 (2000). 60. Davis, G. E. Affinity of integrins for damaged extracellular matrix: αvβ3 binds to denatured collagen type I through RGD sites. Biochem. Biophys. Res. Commun. 182, 1025–1031 (1992). 61. Henriet, P., Zhong, Z.-D., Brooks, P. C., Weinberg, K. I. & Declerck, Y. A. Contact with fibrillar collagen inhibits melanoma cell proliferation by up-regulating p27 KIP1. Proc. Natl. Acad. Sci. U. S. A. 97, 10026–10031 (2000). 62. Petitclerc, E. et al. Integrin αVβ3 promotes M21 melanoma growth in human skin by regulating tumor cell survival. Cancer Res. 59, 2724–2730 (1999). 63. Montgomery, A. M. P., Reisfeld, R. A. & Cheresh, D. A. Integrin αvβ3 rescues melanoma cells from apoptosis in three-dimensional dermal collagen. Proc. Natl. Acad. Sci. U. S. A. 91, 8856–8860 (1994). 64. Egeblad, M., Rasch, M. G. & Weaver, V. M. Dynamic interplay between the collagen scaffold and tumor evolution. Curr. Opin. Cell Biol. 22, 697–706 (2010). 65. Sabeh, F., Shimizu-Hirota, R. & Weiss, S. J. Protease-dependent versus - independent cancer cell invasion programs: three-dimensional amoeboid movement revisited. J. Cell Biol. 185, 11–19 (2009). 66. Bergers, G. et al. Matrix metalloproteinase-9 triggers the angiogenic switch during carcinogenesis. Nat. Cell Biol. 2, 737–744 (2000). 67. Yu, Q. & Stamenkovic, I. Cell surface-localized matrix metalloproteinase-9 proteolytically activates TGF-beta and promotes tumor invasion and angiogenesis. Genes Dev. 14, 163–76 (2000). 68. Kalluri, R. Angiogenesis: Basement membranes: structure, assembly and role in tumour angiogenesis. Nat. Rev. Cancer 3, 422–433 (2003). 69. Puré, E. & Assoian, R. K. Rheostatic signaling by CD44 and hyaluronan. Cell. Signal. 21, 651–655 (2009). 70. Monslow, J., Govindaraju, P. & Puré, E. Hyaluronan - a functional and structural sweet spot in the tissue microenvironment. Front. Immunol. 6, (2015). 71. Tian, X. et al. High-molecular-mass hyaluronan mediates the cancer resistance of the naked mole rat. Nature 499, 346–349 (2013). 72. Levental, K. R. et al. Matrix crosslinking forces tumor progression by enhancing integrin signaling. Cell 139, 891–906 (2009). 73. Provenzano, P. P., Inman, D. R., Eliceiri, K. W. & Keely, P. J. Matrix density- induced mechanoregulation of breast cell phenotype, signaling and gene expression through a FAK-ERK linkage. Oncogene 28, 4326–43 (2009). 74. Paszek, M. J. et al. Tensional homeostasis and the malignant phenotype. Cancer Cell 8, 241–254 (2005). 75. Gaggioli, C. et al. Fibroblast-led collective invasion of carcinoma cells with differing roles for RhoGTPases in leading and following cells. Nat. Cell Biol. 9, 1392–1400 (2007). 85

76. Wipff, P.-J., Rifkin, D. B., Meister, J.-J. & Hinz, B. Myofibroblast contraction activates latent TGF-β1 from the extracellular matrix. J. Cell Biol. 179, 1311–1323 (2007). 77. Munger, J. S. et al. The integrin alpha v beta 6 binds and activates latent TGF beta 1: a mechanism for regulating pulmonary inflammation and fibrosis. Cell 96, 319–28 (1999). 78. Kaplan, R. N. et al. VEGFR1-positive haematopoietic bone marrow progenitors initiate the pre-metastatic niche. Nature 438, 820–827 (2005). 79. Carraher, C. L. & Schwarzbauer, J. E. Regulation of matrix assembly through rigidity-dependent fibronectin conformational changes. J. Biol. Chem. 288, 14805– 14 (2013). 80. Friedland, J. C., Lee, M. H. & Boettiger, D. Mechanically activated integrin switch controls α5β1 function. Science 323, 642–644 (2009). 81. Provenzano, P. P. et al. Enzymatic targeting of the stroma ablates physical barriers to treatment of pancreatic ductal adenocarcinoma. Cancer Cell 21, 418– 429 (2012). 82. Provenzano, P. P. et al. Collagen reorganization at the tumor-stromal interface facilitates local invasion. BMC Med. 4, 38 (2006). 83. Shieh, A. C., Rozansky, H. A., Hinz, B. & Swartz, M. A. Tumor cell invasion is promoted by interstitial flow-induced matrix priming by stromal fibroblasts. Cancer Res. 71, 790–800 (2010). 84. Condeelis, J. & Segall, J. E. Intravital imaging of cell movement in tumours. Nat. Rev. Cancer 3, 921–930 (2003). 85. Wang, W. et al. Single cell behavior in metastatic primary mammary tumors correlated with gene expression patterns revealed by molecular profiling. Cancer Res. 62, 6278–6288 (2002). 86. Salmon, H. et al. Matrix architecture defines the preferential localization and migration of T cells into the stroma of human lung tumors. J. Clin. Invest. 122, 899–910 (2012). 87. Ding, N. et al. A vitamin D receptor/SMAD genomic circuit gates hepatic fibrotic response. Cell 153, 601–613 (2013). 88. Sherman, M. H. et al. Vitamin D receptor-mediated stromal reprogramming suppresses pancreatitis and enhances pancreatic cancer therapy. Cell 159, 80– 93 (2014). 89. Bailey, J. M. et al. Sonic hedgehog promotes desmoplasia in pancreatic cancer. Clin. Cancer Res. 14, 5995–6004 (2008). 90. Yauch, R. L. et al. A paracrine requirement for hedgehog signalling in cancer. Nature 455, 406–411 (2008). 91. Olive, K. P. et al. Inhibition of hedgehog signaling enhances delivery of chemotherapy in a mouse model of pancreatic cancer. Science 324, 1457–1461 (2009). 92. Amakye, D., Jagani, Z. & Dorsch, M. Unraveling the therapeutic potential of the Hedgehog pathway in cancer. Nat. Med. 19, 1410–1422 (2013). 93. Lee, J. J. et al. Stromal response to Hedgehog signaling restrains pancreatic cancer progression. Proc. Natl. Acad. Sci. 111, E3091–E3100 (2014). 94. Rhim, A. D. et al. Stromal elements act to restrain, rather than support, pancreatic ductal adenocarcinoma. Cancer Cell 25, 735–747 (2014). 95. Wang, L.-C. S. et al. Targeting fibroblast activation protein in tumor stroma with chimeric antigen receptor T cells can inhibit tumor growth and augment host 86

immunity without severe toxicity. Cancer Immunol. Res. 2, 154 LP-166 (2014). 96. Chen, M. et al. A whole-cell tumor vaccine modified to express fibroblast activation protein induces antitumor immunity against both tumor cells and cancer-associated fibroblasts. Sci. Rep. 5, 14421 (2015). 97. Lee, J., Fassnacht, M., Nair, S., Boczkowski, D. & Gilboa, E. Tumor immunotherapy targeting fibroblast activation protein, a product expressed in tumor-associated fibroblasts. Cancer Res. 65, 11156–11163 (2005). 98. Loeffler, M., Krüger, J. A., Niethammer, A. G. & Reisfeld, R. A. Targeting tumor- associated fibroblasts improves cancer chemotherapy by increasing intratumoral drug uptake. J. Clin. Invest. 116, 1955–1962 (2006). 99. Fang, J. et al. A potent immunotoxin targeting fibroblast activation protein for treatment of breast cancer in mice. Int. J. Cancer 138, 1013–1023 (2016). 100. LeBeau, A. M., Brennen, W. N., Aggarwal, S. & Denmeade, S. R. Targeting the cancer stroma with a fibroblast activation protein-activated promelittin protoxin. Mol. Cancer Ther. 8, 1378–1386 (2009). 101. Brennen, W. N. et al. Pharmacokinetics and toxicology of a fibroblast activation protein (FAP)-activated prodrug in murine xenograft models of human cancer. Prostate 74, 1308–19 (2014). 102. Akinboye, E. S., Brennen, W. N., Rosen, D. M., Bakare, O. & Denmeade, S. R. Iterative design of emetine-based prodrug targeting fibroblast activation protein (FAP) and dipeptidyl peptidase IV DPPIV using a tandem enzymatic activation strategy. Prostate 76, 703–714 (2016). 103. Brennen, W. N. et al. Targeting carcinoma-associated fibroblasts within the tumor stroma with a fibroblast activation protein-activated prodrug. JNCI J. Natl. Cancer Inst. 104, 1320–1334 (2012). 104. Zhen, Z. et al. Protein nanocage mediated fibroblast-activation protein targeted photoimmunotherapy to enhance cytotoxic T cell infiltration and tumor control. Nano Lett. 17, 862–869 (2017). 105. Servais, C. & Erez, N. From sentinel cells to inflammatory culprits: cancer- associated fibroblasts in tumour-related inflammation. J. Pathol. 229, 198–207 (2013). 106. Yang, X. et al. FAP promotes immunosuppression by cancer-associated fibroblasts in the tumor microenvironment via STAT3–CCL2 signaling. Cancer Res. 76, 4124 LP-4135 (2016). 107. Kraman, M. et al. Suppression of antitumor immunity by stromal cells expressing fibroblast activation protein–α. Science 330, 827–830 (2010). 108. Hurme, T., Kalimo, H., Sandberg, M., Lehto, M. & Vuorio, E. Localization of type I and III collagen and fibronectin production in injured gastrocnemius muscle. Lab. Invest. 64, 76—84 (1991). 109. Singer, A. J. & Clark, R. A. F. Cutaneous wound healing. N. Engl. J. Med. 341, 738–746 (1999). 110. Liu, F. et al. Feedback amplification of fibrosis through matrix stiffening and COX- 2 suppression. J. Cell Biol. 190, 693–706 (2010). 111. Plodinec, M. et al. The nanomechanical signature of breast cancer. Nat Nano 7, 757–765 (2012). 112. Yu, H., Mouw, J. K. & Weaver, V. M. Forcing form and function: biomechanical regulation of tumor evolution. Trends Cell Biol. 21, 47–56 (2011). 113. Kubow, K. E. et al. Mechanical forces regulate the interactions of fibronectin and collagen I in extracellular matrix. Nat. Commun. 6, 8026 (2015). 87

114. Goffin, J. M. et al. Focal adhesion size controls tension-dependent recruitment of alpha-smooth muscle actin to stress fibers. J. Cell Biol. 172, 259–68 (2006). 115. Puré, E. & Lo, A. Can targeting stroma pave the way to enhanced antitumor immunity and immunotherapy of solid tumors? Cancer Immunol. Res. 4, 269–278 (2016). 116. Pinnell, S. R. Regulation of collagen biosynthesis by ascorbic acid: a review. Yale J. Biol. Med. 58, 553–559 (1985). 117. Cukierman, E., Pankov, R., Stevens, D. R. & Yamada, K. M. Taking cell-matrix adhesions to the third dimension. Science 294, 1708–1712 (2001). 118. Cretu, A., Castagnino, P. & Assoian, R. Studying the effects of matrix stiffness on cellular function using acrylamide-based hydrogels. J. Vis. Exp. 2089 (2010). doi:10.3791/2089 119. Solon, J., Levental, I., Sengupta, K., Georges, P. C. & Janmey, P. A. Fibroblast adaptation and stiffness matching to soft elastic substrates. Biophys. J. 93, 4453– 4461 (2007). 120. Klein, E. A. et al. Cell cycle control by physiological matrix elasticity and in vivo tissue stiffening. Curr. Biol. 19, 1511–1518 (2009). 121. Huang, X. et al. Matrix stiffness–induced myofibroblast differentiation is mediated by intrinsic mechanotransduction. Am. J. Respir. Cell Mol. Biol. 47, 340–348 (2012). 122. Thannickal, V. J. et al. Myofibroblast differentiation by transforming growth factor- beta1 is dependent on cell adhesion and integrin signaling via focal adhesion kinase. J. Biol. Chem. 278, 12384–12389 (2003). 123. Seong, J. et al. Distinct biophysical mechanisms of focal adhesion kinase mechanoactivation by different extracellular matrix proteins. Proc. Natl. Acad. Sci. 110, 19372–19377 (2013). 124. Goetz, J. G. et al. Biomechanical remodeling of the microenvironment by stromal Caveolin-1 favors tumor invasion and metastasis. Cell 146, 148–163 (2011). 125. Boire, A. et al. PAR1 is a matrix metalloprotease-1 receptor that promotes invasion and tumorigenesis of breast cancer cells. Cell 120, 303–313 (2005). 126. Itoh, T. et al. Reduced angiogenesis and tumor progression in gelatinase A- deficient mice. Cancer Res. 58, 1048–1051 (1998). 127. Cheng, J. D. et al. Promotion of tumor growth by murine fibroblast activation protein, a serine protease, in an animal model. Cancer Res. 62, 4767–4772 (2002). 128. Cheng, J. D. et al. Abrogation of fibroblast activation protein enzymatic activity attenuates tumor growth. Mol. Cancer Ther. 4, 351–360 (2005). 129. Sternlicht, M. D. et al. The stromal proteinase MMP3/stromelysin-1 promotes mammary carcinogenesis. Cell 98, 137–146 (1999). 130. Radisky, D. C. et al. Rac1b and reactive oxygen species mediate MMP-3-induced EMT and genomic instability. Nature 436, 123–127 (2005). 131. Bhowmick, N. A. et al. TGF-ß signaling in fibroblasts modulates the oncogenic potential of adjacent epithelia. Science 303, 848–851 (2004). 132. Straussman, R. et al. Tumour micro-environment elicits innate resistance to RAF inhibitors through HGF secretion. Nature 487, 500–504 (2012). 133. Mueller, M. M. & Fusenig, N. E. Friends or foes - bipolar effects of the tumour stroma in cancer. Nat Rev Cancer 4, 839–849 (2004). 134. Gore, J. & Korc, M. Pancreatic cancer stroma: friend or foe? Cancer Cell 25, 711– 712 (2014). 88

135. De Wever, O., Demetter, P., Mareel, M. & Bracke, M. Stromal myofibroblasts are drivers of invasive cancer growth. Int. J. Cancer 123, 2229–2238 (2008). 136. Vong, S. & Kalluri, R. The role of stromal myofibroblast and extracellular matrix in tumor angiogenesis. Genes Cancer 2, 1139–1145 (2011). 137. Hecht, J. R. et al. A phase II, randomized, double-blinded, placebo-controlled study of simtuzumab or placebo in combination with FOLFIRI for the second line treatment of metastatic KRAS mutant colorectal adenocarcinoma. J. Clin. Oncol. 33, 243-e23 (2015). 138. Benson A., I. I. I. et al. 618PD A phase 2 randomized, double-blind, placebo controlled study of simtuzumab or placebo in combination with gemcitavine for the first line treatment of pancreatic adenocarcinoma. Ann. Oncol. 25, iv211-iv211 (2014). 139. Raghu, G. et al. Efficacy of simtuzumab versus placebo in patients with idiopathic pulmonary fibrosis: a randomised, double-blind, controlled, phase 2 trial. Lancet Respir. Med. 5, 22–32 (2017). 140. Meyer, K. C. Great expectations for simtuzumab in IPF fall short. Lancet Respir. Med. 5, 2–3 (2017). 141. Bateman, E. D., Turner-Warwick, M. & Adelmann-Grill, B. C. Immunohistochemical study of collagen types in human foetal lung and fibrotic lung disease. Thorax 36, 645–53 (1981). 142. Volk, S. W., Wang, Y., Mauldin, E. A., Liechty, K. W. & Adams, S. L. Diminished Type III Collagen Promotes Myofibroblast Differentiation and Increases Scar Deposition in Cutaneous Wound Healing. Cells Tissues Organs 194, 25–37 (2011). 143. Beacham, D. A., Amatangelo, M. D. & Cukierman, E. in Current Protocols in Cell Biology (John Wiley & Sons, Inc., 2001). doi:10.1002/0471143030.cb1009s33 144. Brisson, B. K. et al. Type III collagen directs stromal organization and limits metastasis in a murine model of breast cancer. Am. J. Pathol. 185, 1471–1486 (2015). 145. Blissett, A. R. et al. Regulation of collagen fibrillogenesis by cell-surface expression of kinase dead DDR2. J. Mol. Biol. 385, 902–911 (2009). 146. Pengfei Lu, Valerie Weaver, Z. W. ECM: a dynamic niche in cancer progression. J. Cell Biol. 196, 395–406 (2012). 147. Borczuk, A. C. et al. Progression of human bronchioloalveolar carcinoma to invasive adenocarcinoma is modeled in a transgenic mouse model of K-ras– induced lung cancer by loss of the TGF-β type II receptor. Cancer Res. 71, 6665– 6675 (2011). 148. Johnson, L. et al. Somatic activation of the K-ras oncogene causes early onset lung cancer in mice. Nature 410, 1111–1116 (2001). 149. Bertram, J. S. & Janik, P. Establishment of a cloned line of Lewis lung carcinoma cells adapted to cell culture. Cancer Lett. 11, 63–73 (1980). 150. DuPage, M. et al. Endogenous T cell responses to antigens expressed in lung adenocarcinomas delay malignant tumor progression. Cancer Cell 19, 72–85 (2011). 151. Salic, A. & Mitchison, T. J. A chemical method for fast and sensitive detection of DNA synthesis in vivo. Proc. Natl. Acad. Sci. 105, 2415–2420 (2008). 152. Lee, H.-O. et al. FAP-overexpressing fibroblasts produce an extracellular matrix that enhances invasive velocity and directionality of pancreatic cancer cells. BMC Cancer 245 (2011). 89

153. Cailleau, R., Young, R., Olivé, M. & Reeves, W. J. Breast Tumor Cell Lines From Pleural Effusions 2. JNCI J. Natl. Cancer Inst. 53, 661–674 (1974). 154. Cukierman, E. Cell migration analyses within fibroblast-derived 3-D matrices. Methods Mol. Biol. 294, 79–93 (2005). 155. Butcher, D. T., Alliston, T. & Weaver, V. M. A tense situation: forcing tumour progression. Nat. Rev. Cancer 9, 108–22 (2009). 156. Mao, Y. & Schwarzbauer, J. E. Accessibility to the fibronectin synergy site in a 3D matrix regulates engagement of alpha5beta1 versus alphavbeta3 integrin receptors. Cell Commun. Adhes. 13, 267–77 (2006). 157. Li, B., Moshfegh, C., Lin, Z., Albuschies, J. & Vogel, V. Mesenchymal stem cells exploit extracellular matrix as mechanotransducer. Sci. Rep. 3, 2425 (2013). 158. Leitinger, B. Transmembrane collagen receptors. Annu. Rev. Cell Dev. Biol. 27, 265–290 (2011). 159. Chung, L. et al. Collagenase unwinds triple-helical collagen prior to peptide bond hydrolysis. EMBO J. 23, 3020–30 (2004).