Characterizing the Invasive Tumor Front of Aggressive Uterine Adenocarcinoma and Leiomyosarcoma

Characterizing the invasive tumor front of aggressive uterine adenocarcinoma and leiomyosarcoma

Sabina Sanegre1, Núria Eritja2, 3, 4, 5, 6, Carlos de Andrea2, 7, Juan Díaz-Martín2, 8, 9, 10, 11, Ángel Díaz-Lagares2, 12, 13, María Amalia Jácome14, Carmen Salguero-Aranda8, 9, 10, 11, David García-Ros7, Ben Davidson15, 16, 17, Rafael López2, 13, 12, Ignacio Melero2, 18, Samuel Navarro2, 1, Santiago Ramon Y Cajal2, 19, Enrique De Álava2, 8, 9, 10, 11, Xavier Matias-Guiu2, 3, 5, 6, 4*, Rosa Noguera2, 20, 1*

1Institute of Health Research (INCLIVA), Spain, 2Centro de Investigaciónació Biomédica en Red del Cáncer (CIBERONC), Spain, 3Department of Pathology, University Hospital ArArnau de Vilanova, Spain, 4Bellvitge University Hospital, Spain, 5Universitatt de Lleida, Spain, 6Universityniversity of Barcelona, Spain, 7University Clinic of Navarra, Spain, 8Institutetute of Biomedicine of Seville (IBIS), Spain,Sp 9Virgen del Rocío University Hospital, Spain, 10Consejo Superior de Investigaciones CientífiCientíficas (CSIC), Spain, 11Sevilla University, 12 13 Spain, University Clinical Hospital of Santiago, Spain,Spa Health Research Institute of Santiago de Compostelapostela (IDIS), Spain, 14Facultyculty of ScieScience, University of A Coruña, Spain, 15Institute of Clinical Medicine,ne, Faculty of Medicine,Me University of Oslo, Norway, 16Department of Pathology, Oslo University Hospital,l, Norway,Nor 17Norwegian Radium Hospital, Oslo University Hospital, Norway, 18Departamento de DDermatología, Clínica Universidad de Navarra, Spain, 19Department of Pathology, Vall d'Hebron University Hospital, Spain, 20Department of Pathology, University of Valencia, Spain

InSubmitted review to Journal: Frontiers in Cell and Developmental Biology Specialty Section: Cell Adhesion and Migration Article type: Original Research Article Manuscript ID: 670185 Received on: 20 Feb 2021 Journal website link: www.frontiersin.org

Conflict of interest statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest

Author contribution statement

RN, and MX contributed to conception and design of the study. SS wrote the first draft of the manuscript. SS, NE, CA, JD, AD, EA, XM and RN wrote sections of the manuscript. BD, RL, SR, EA, XM provided clinicopathological data. NE provided the collaborative consortium with microdissected samples. SS and RN designed and performed the morphometric characterization of the reticulin fibers. CA and DR performed the multiplex staining and analysis. JD, CS and EA evaluated the transcriptomic profile. AD and AJ performed the genome-wide methylation analysis. All authors contributed to manuscript revision, read, and approved the submitted version.

Keywords

tumor-host interface, Tumor Microenvironment, Extracellular Matrix, Reticularticular fibers, immune cells,c Gene Expression, epigenetic profiles

Abstractt

Word count:nt: 200 The tumor-host-host interface is vitally importantimport in malignant cell progression and metastasis. Tumor cell interactions with resident and infiltratingting host cells and with the surrounding extracellular matrix and secreted factors ultimately determine the fate of the tumor. Hereinin we focus on the invasive tumor front, making an in-depth characterization of reticular fiber scaffolding, infiltrating immune cells,ells gene expression and epigenetic profiles of classified aggressive primary uterine adenocarcinomas (ADC=24) and leiomyosarcomasIn (LMS=11). Sections review of formalin-fixed samples before and after microdissection were scanned and studied. Reticular fiber architecture and immune cell infiltration were analyzed by automatized algorithms in the same small regions of interest. Despite morphometric resemblance between reticular fibers and high presence of macrophages, we found some variance in other immune cell populations and distinctive gene expression and cell adhesion-related methylation signatures. Although no evident overall differences in immune response were detected at the gene expression and methylation level, impaired antimicrobial humoral response might be involved in LMS spread. Similarities found at the invasive tumor front of uterine ADC and LMS could facilitate use of common biomarkers and therapies, and additionally, molecular and architectural characterization of the invasive front of uterine malignancies may provide additional prognostic information beyond established prognostic factors.

Contribution to the field

We present a comprehensive study of the tumor-host interface in uterine tumors. We have characterized this interface, the invasive tumor front, at different levels. We know now that the fiber architectural pattern at the invasive front, as well as the immune infiltrate, epigenomic landscape and trancrptomic profile share a lot in common in the two uterine tumors that we have studied. We have also unveiled a very narrow differential gene signature regulated by epigenetic mechanisms. We hope our findings lead to new perspectives for biomarker development and provide additional information over the current prognostic factors.

Funding statement

This research was supported by grants from ISCIII and ERDF (PI17/01558 and PI20/01107), by CIBERONC (contracts CB16/12/00484, CB16/12/0328, CB16/12/00363, CB16/12/00364, CB16/12/00481, CB16/12/00231) and by Grupos Coordinados Estables from Asociación Española Contra el Cáncer (AECC). AD is funded by a contract “Juan Rodés” from ISCIII (JR17/00016). The funders had no involvement in the research process nor in the preparation and submission of the article.

Ethics statements

Studies involving animal subjects Generated Statement: No animal studies are presented in this manuscript.

Studies involving human subjects Generated Statement: The studies involving human participants were reviewed and approved by Comité Coordinador de Ética de la Investigación Biomédica de Andalucía: Deciphering the site-specific tumor microenvironment of advanced uterine tumors (Biobank code: S1800086) and Comité de Ética de la Investigación with the codes CEIC-1892, CEIC-2083 and CEIC-1858.. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.

Inclusion of identifiable human data Generated Statement: No potentially identifiable human images or data is presented in this study.

In review

Data availability statement

Generated Statement: The authors acknowledge that the data presented in this study must be deposited and made publicly available in an acceptable repository, prior to publication. Frontiers cannot accept a manuscript that does not adhere to our open data policies.

In review

Characterizing the invasive tumor front of aggressive uterine adenocarcinoma and leiomyosarcoma

1 Sanegre S1,2, Eritja N1,3, de Andrea C1,4, Diaz-Martin J1,5, Diaz-Lagares A1,6, Jácome A7, 2 Salguero-Aranda C1,5, García Ros D4, Davidson B8, Lopez R1,9,10, Melero I1,4, Navarro S1,2, 3 Ramon y Cajal S1,11, de Alava E1,5, Matias-Guiu X1,3, †,*, Noguera R1,2, †, *

4 1Cancer CIBER (CIBERONC), Madrid. 5 2Pathology Department, Medical School, University of Valencia-INCLIVA, Valencia. 6 3Pathology Department, Hospital U Arnau de Vilanova and Hospital U de Bellvitge, Universities of 7 Lleida and Barcelona, IRBLLEIDA, IDIBELL. 8 4Clínica Universidad de Navarra, University of Navarra, Pamplona,amplona, SpainSpain.. 9 5Institute of Biomedicine of Sevilla (IBiS), Virgen del Rocio University HospHospitHospital/CSIC/University of 10 Sevilla/CIBERONC, 41013 Seville,e, Spain. 11 6Cancer Epigenomics,cs, Translational Medical Oncology Group (Oncomet), HHealth Research Institute 12 of Santiagoo (IDIS), University Clinical Hospital of Santiago (CH(C(CHUS/SERGAS), 15760 Santiago de 13 Compostela,tela, Spain.Spain 14 7Departmententnt of Mathematics, MODES grogroup, CITIC, Faculty of Science, Universidade da Coruña, A 15 Coruña, Spain.painin. 16 8 Institute off CliniClinClinical Medicine, Faculty of Medicine, University of Oslo, 0424 Oslo, Norway; 17 Department Inof Pathology, Oslo reviewUniversity Hospital, Norwegian Radium Hospital, 0310 Oslo, 18 Norway. 19 9Translational Medical Oncology Group (Oncomet), Health Research Institute of Santiago (IDIS), 20 University Clinical Hospital of Santiago (CHUS/SERGAS), 15760 Santiago de Compostela, Spain. 21 10Roche-Chus Joint Unit, Translational Medical Oncology Group (Oncomet), Health Research 22 Institute of Santiago (IDIS), 15760 Santiago de Compostela, Spain. 23 11Department of Pathology, Vall d'Hebron University Hospital, Autonoma University of Barcelona. 24 25 †These authors contributed equally to this work. 26 27 *Correspondence: Rosa Noguera ([email protected]) and Xavier Matias-Guiu 28 ([email protected])

29 Keywords: tumor-host interface, tumor microenvironment, extracellular matrix, reticular 30 fibers, immune cells, gene expression, epigenetic profiles

31 Running Title: Uterine-myometrial tumor interface

32 Uterine-myometrial tumor interface

34 Abstract

35 The tumor-host interface is vitally important in malignant cell progression and metastasis. Tumor cell 36 interactions with resident and infiltrating host cells and with the surrounding extracellular matrix and 37 secreted factors ultimately determine the fate of the tumor. Herein we focus on the invasive tumor 38 front, making an in-depth characterization of reticular fiber scaffolding, infiltrating immune cells, 39 gene expression and epigenetic profiles of classified aggressive primary uterine adenocarcinomas 40 (ADC=24) and leiomyosarcomas (LMS=11). Sections of formalin-fixed samples before and after 41 microdissection were scanned and studied. Reticular fiber architecture and immune cell infiltration 42 were analyzed by automatized algorithms in the same small regions of interest. Despite 43 morphometric resemblance between reticular fibers and high presence of macrophages, we found 44 some variance in other immune cell populations and distinctive gene expression and cell adhesion- 45 related methylation signatures. Although no evident overall differences in immune response were 46 detected at the gene expression and methylation level, impairedpaired antimicroantimicrobial humoral response 47 might be involved in LMS spread. Similarities foundd at the invasive tumor frofront of uterine ADC and 48 LMS could facilitate use of common biomarkersomarkers and therapies, and addiadditioadditionally, molecular and 49 architectural characterization off the invasive frontont of uterine malignancies mamay provide additional 50 prognostic informationn beyond establishedshed prognostic factors.

51 1 Introductiontroductiontrodu

52 Endometrialiall adenocarcinadenocarcinoadenocarcinomama (ADC) is the fourth most common cancer among women in the 53 Western world,rld, with an estimeestimated incidence at 10-20 per 100,000 women. Although prognosis is 54 favorable fororr patientspatient identified with low-grade tumors and early-stage disease, outcomes for patients 55 with high-gradeIn and metastatic/recurrent review tumors remain poor (Ferlay et al. 2010). ADC patients 56 presenting with systemic disease and/or overt metastasis represent a therapeutic challenge nowadays. 57 The World Health Organization (WHO) tumor classification distinguishes several histopathological 58 types of ADC, based particularly on microscopic appearance (Cree et al. 2020): 1) endometrioid 59 carcinoma, low grade (grades I-II) or high grade (grade III), 2) serous carcinoma, 3) clear cell 60 carcinoma, 4) mixed carcinoma, 5) undifferentiated carcinoma, 6) carcinosarcoma, 7) neuroendocrine 61 carcinomas, and 8) other unusual types. These types have different histological and molecular 62 features, precursor lesions and natural history (Matias-guiu et al. 2001; Yeramian et al. 2013). They 63 are also stratified by Tumor Cancer Genome Atlas (TCGA)-based molecular classification into four 64 risk groups combining gene encoding DNA polymerase ε (POLE) mutational analysis with IHC 65 analysis of p53 and mismatch repair (MMR) proteins (PMS-2 and MSH-6) (Getz et al. 2013). This 66 provides additional prognostic information to complement the microscopic features. 67 Uterine leiomyosarcoma (LMS) accounts for 1% of all uterine malignancies, with an annual 68 incidence rate of 0.4-0.9 per 100,000 women. Although rare, this cancer can be extremely aggressive 69 and is known to be generally unresponsive to radiation or chemotherapy. Patient survival is highly 70 dependent on speed of diagnosis and treatment, falling to 14% at 5 years post-diagnosis for 71 metastatic LMS. Like other forms of sarcoma, LMS spreads to other parts of the body via the 72 bloodstream rather than the lymphatic system. From a molecular viewpoint, LMS contains complex 73 karyotypes, with numerous chromosomal aberrations and frequent deletions affecting chromosomal 74 arms 2p, 2q, 10q and 13q, as well as amplifications on 1p, 5q and 8q, and precise alterations in TP53, 75 RB1, PTEN, MED12, YWHAE and VIPR2 (Cuppens et al. 2018).

2 This is a provisional file, not the final typeset article Uterine-myometrial tumor interface

76 There is growing interest in understanding the molecular features involved in myometrial invasion, a 77 highly valuable parameter due to its direct association with poor prognosis and limited therapeutic 78 response in ADC and LMS. From a histopathological perspective, a variety of invasion patterns 79 frequently coexist in different areas of the same tumor (Euscher et al. 2013). Moreover, tumors from 80 diverse sites and histological appearances show varied frequency of these different patterns. In fact, 81 several tumor types show specific invasion patterns not seen in other types of cancer. These 82 differences can be explained by the fact that cancer cell invasion is currently viewed as an adaptive 83 and heterogeneous process (Friedl and Alexander 2011) involving key processes such as cytoskeleton 84 dynamics, cell adhesion plasticity and mechanotransduction of external stimuli.

85 Type-specific new stroma generated at the site of active tumor invasion, the invasive tumor front 86 (ITF), is crucial in tumor growth and invasion processes (Giatromanolaki, Sivridis, and Koukourakis 87 2007). Structures surrounding tumors such as mature collagen and smooth or striated muscle act as a 88 barrier to tumor invasion. However, tumor cells can disrupt the continuity of such structures by 89 remodeling the immediate stroma of the tumor to carve out paths for invasion.nvasio In fact, tumor-derived 90 extracellular matrix (ECM) is biochemically distinct in its composition and is stiffer than normal 91 ECM. This new associated ECM compartment, richh in crosscross-linked-linked collagen IIIII (reticular fibers), has 92 been proposed as a prospective markerr of early stromal invasion in incincipiincipient tumors (Sivridis, 93 Giatromanolaki, and Koukourakisakis 2004)2004).. Use of a multiplex immunolabellingimmunolabellin ppanel is essential as it 94 enables different cellll subpopulationss to be identified in one tissue sectisectiosection (Gorris et al. 2018). 95 Multiplexedd analysis can also simultaneously measure expression oof distinct markers in a single cell 96 as well as spatial associations between immune cell subpopulations.subsu We applied this technology to 97 evaluate thee complecomplex immune environmentenvironmen of the ITF of ADC and LMS. A previously developed 98 and validatedatedd multiplex immunolabimmunolabelimmunolabelling panel was used to simultaneously assess the phagocytic cell 99 marker CD68688 of macrophages,macrophagmacropha CD3+ and CD8+ T cells, and CD20+ B lymphocytes in a single FFPE 100 tissue (Abengozar-MuelangozangozIn et al. review2020; Salas-Benito et al. 2021). 101 Changes in activated-leukocyte cell adhesion molecule (ALCAM) expression at the endometrial 102 tumoral cell surface (Devis et al. 2018) and increased expression of cytoplasmic Cyclin D1(Fusté et 103 al. 2016) have been reported to allow dissemination and invasion of endometrial neoplastic cells. 104 Epigenetic mechanisms play an important role in regulating gene expression during many biological 105 processes (Sharma, Kelly, and Jones 2009). DNA methylation is the most widely studied epigenetic 106 modification, produced by adding a methyl group (CH3) to the 5′ carbon of cytosines in cytosine- 107 phosphate-guanine (CpG) dinucleotides to generate 5-methylcytosine (5mC) (Bao-Caamano, 108 Rodriguez-Casanova, and Diaz-Lagares 2020). Deregulation of this epigenetic mechanism has major 109 implications for cancer development and progression (Diaz-Lagares et al. 2016). In this context, 110 recent genome-wide analyses have revealed striking alterations in the methylation profile of uterine 111 ADC and LMS (Kommoss et al. 2020; B. Zhang et al. 2014; Vargas et al. 2021).

112 Cancers can occur anywhere in the body, always showing clear coevolution between tumor cells and 113 the tumor microenvironment. Specific differences in the tumor microenvironment at different 114 locations may play a role in tumor growth, metastatic progression, and therapy responses (Oliver et 115 al. 2018; C. Zhang and Yu 2019). Unlike tumor cells, tumor microenvironment elements are 116 genetically stable, and thus represent an attractive therapeutic target. Tumor invasion is a dynamic 117 process facilitated by bidirectional interactions between tumor cells and the microenvironment, and 118 such bidirectional communication is particularly intense at the ITF. Uterine LMS and ADC are 119 different tumor types occurring in the same organ, both of which infiltrate the myometrium during 120 local progression. The main objective of the present study is to analyze several aspects of 121 microenvironment response at the ITF in these two tumors types, looking for similarities and

Uterine-myometrial tumor interface

122 differences that could provide insight into potential new common therapeutic approaches, an aim that 123 to our best knowledge has not previously been addressed.

124

125 2 Materials and Methods

126 2.1 Patients, sample description and case selection

127 A search for ADC and LMS cases was conducted in the study institutions. Selection criteria included 128 available pathology reports, representative sections of the ITF, histologically-proven distant 129 metastasis, and acceptable pre-analytical conditions. Cases were reviewed by a panel of expert 130 gynaecological pathologists from the institutions involved. In total, 24 ADC and 11 LMS fulfilled all 131 criteria and were included in the study. 132 The study used formalin-fixed and paraffin-embedded (FFPE) tissueue samplesamples from 24 ADC and 11 133 LMS obtained from five Spanish hospitals (Hospital Clínicoico de Valencia; HosHospital Virgen del Rocio, 134 Seville; Hospital Universitari Vall d’Hebron and Hospital Universitari dde Bellvitge, Barcelona; 135 Hospital Universitari Arnau de Vilanova,lanova, Lleida) and the Norwegian RaRadium Hospital, Oslo 136 University Hospital, Oslo, NorwayNorway.. Tumors were classified following the mmost recent WHO criteria, 137 and were surgicallylly staged and graded according to FIGO (Internation(International Federation of Gynecology 138 and Obstetrics)tetrics)etrics) staging and grading systems. The study was approveda by the local research ethics 139 committee,ee, and specific informed consent was obtaineobtaobtained. 140 Whole slideidee FFPE tissue sections of 5µm of selected ADC and LMS were stained with H&E 141 (hematoxylinlinn and eeosin)osin) andan examined by the centralized expert group of pathologists to select the 142 representativeve areaareareas to include in the study. The interface between tumor tissue and adjacent 143 myometriumIn was microdissected review under the microscope. Microscopic images were obtained using a 144 digital slide scanner [Pannoramic 250 FLASH II 2.0 (3D Histec)].

145 All ADC were endometrioid type, and were classified according to the Cancer Genome Atlas 146 (TCGA) surrogate (0% POLE mutated, 58.5% non-specific molecular profile, 29% mismatch repair 147 deficient tumors, and 12.5% p53 abnormal). All LMS were conventional type high grade tumors. 148 149 2.2 Selection of regions of interest

150 Sections of FFPE tissue samples were scanned before and after microdissection. Whole sections 151 including the interface between tumor tissue and adjacent myometrium areas (the ITF) were used for 152 morphometric analysis. ITF microdissected tissue was employed for transcriptomic and epigenomic 153 studies (Supplementary Figure 1A). The amount of ITF microdissected tissue varied from case to 154 case but median width was 5 mm (±1.66 mm) and median length 15 mm (±4.91 mm). The amount of 155 tumor versus myometrium is shown in Supplementary Table 1. 156 Serial ADC and LMS whole slides were used for histomorphometric analysis of reticular fibers and 157 multiplex immunofluorescence-based immune profiling. ITF regions were identified in stained tissue 158 (Gomori and multiplex immunofluorescence) by extrapolation of previous H&E selected regions. A 159 region of interest (ROI) of 5x4 mm for each sample was identified within the ITF stained area. These 160 5x4 mm regions were used to correlate the results between genomic, epigenetic and morphometric 161 studies (Supplementary Figure 1A). To achieve in-depth characterization of the reticular fibers and 162 immune infiltrate of the ITF, we further broke down the 5x4 mm ROIs into 1 mm2 ROIs representing

4 This is a provisional file, not the final typeset article Uterine-myometrial tumor interface

163 the following categories: a) tumor (ADC or LMS with absence of myometrium), b) myometrium 164 (excluding infiltrated myometrium from the analysis) and c) balanced representation of the invasion 165 front (containing 50% tumor and 50% myometrium). Supplementary Figure 1 shows a schematic 166 representation of the procedure. Cases and ROIs where the algorithm failed because of poor or 167 excessive staining were excluded from the analysis, as were samples with unsatisfactory 168 segmentation. The number of 5x4 mm and 1 mm2 ROIs of each case in Gomori and multiplex 169 immunofluorescence stained tissue are shown in Supplementary Table 1.

170 2.3 Histomorphometric analysis of reticular fibers 171 The architecture of ADC and LMS reticular fibers stained using Gomori’s method was uncovered. 172 All samples were digitized with the whole-slide scanner Ventana iScan HT (Roche) at 20x with a 173 resolution of 0.46 µm/pixel. We used the open-source digital pathology software QuPath for sample 174 visualization and identification of ROIs (Bankhead et al. 2017). 5x4 mm and 1 mm2 ROIs were 175 exported to ImageJ (Schneider, Rasband, and Eliceiri 2012) and these wereere saved as TIFF for image 176 analysis. 177 In this study we employed an advanced morphometric methodology based on a probabilistic method 178 for the automatic segmentation of reticularr fibers. We used Gomoripath, a mmorphometric tool for 179 easy segmentation of reticular fibersbers developed by researchers from InclivIncliva BBiomedical Research 180 Institute/University of Valencia, Universityrsity of Castilla La Mancha and the AndAndalucía Public Health 181 System (Espinosa-Arandanosa-Aranda et al.l. 2016)2016).. Briefly, the algorithm is basedb on nominal logistic regression 182 applied to the optical density of the histopathological imiimages. The optical density is calculated 183 iteratively,y, allowing reticularetireticularr fibersibers to be enhanceenhanced. Subsequently,iw a logit model is used to calculate 184 the probabilitybilityility of each pixel belobelonginbelonging to the structure of reticular fibers depending on their 185 topology, thusus creating a prprobability map for the entire image. The ROC (Receiver Operating 186 Characteristic)tic)ic) curvcurve generated by the model finally obtained an AUC (area under the ROC curve) of 187 0.9563 in reticularIn fiber detection. review 188 Fifteen morphometric parameters defining the histological organization of reticular fibers were 189 calculated for each fiber detected and the mean for each sample calculated. Morphometric parameters 190 were extracted to characterize the size and shape of the morphometric variables at the ITF. In 191 addition, the algorithm measured the stained area of the tissue analyzed (excluding holes and 192 damaged tissue) allowing us to determine the number of fibers per mm2 (density) and the percentage 193 of fiber-stained area (%SA) (taking into account the sum of the areas of all fibers). Morphometric 194 parameters defining the histological organization of the reticular fiber networks have been explained 195 elsewhere (Tadeo et al. 2016). The mean of each parameter of similar ADC and LMS ROIs were 196 calculated for comparison (Figure 1-3). 197

198 2.1 Multiplex immunofluorescence cell phenotype 199 We next sought to investigate the ADC and LMS myeloid and lymphocytic contexture in the FFPE 200 tissue samples. A multiplex immunofluorescence panel was used to enable simultaneous examination 201 of several cellular markers, including the phagocytic cell marker CD68 of macrophages, CD3+ and 202 CD8+ T cells, and CD20+ B lymphocytes. Multiplex immunofluorescence development and 203 validation workflow and protocols were implemented as previously described (Abengozar-Muela et 204 al. 2020; Schalper et al. 2019; Salas-Benito et al. 2021)Briefly, 5 µm sections of FFPE tissue were 205 deparaffinized and antigen retrieval was performed using DAKO PT-Link heat-induced antigen 206 retrieval with low pH (pH6) or high pH (pH 9) target retrieval solution (DAKO). Each tissue section 207 was subjected to five successive rounds of antibody staining, each round consisting of protein

Uterine-myometrial tumor interface

208 blocking with antibody diluent/block (Akoya Biosciences ARD1001EA), incubation with primary 209 antibody, Opal Polymer anti mouse/rabbit HRP (Akoya Biosciences ARH1001EA), followed by 210 tyramide signal amplification (TSA) with Opal fluorophores (Akoya Biosciences) diluted 1:100 in 211 1X plus amplification diluent (Akoya Biosciences FP1498). The myeloid and lymphoid cell panel 212 included CD68 (Mouse monoclonal, clone PG-M1, ready-to-use, Agilent IR613), CD3 (Rabbit 213 polyclonal, IgG, ready-to-use, Agilent IR503), CD8 (Mouse monoclonal, clone C8/144B, ready-to- 214 use, Agilent IR623), CD20 (Mouse monoclonal, IgG2α, clone L26, ready-to-use, Roche 760-2531), 215 cytokeratin (Mouse monoclonal, clone AE1/AE3, diluted 1:100, Agilent M3515). Finally, in the last 216 round, nuclei were counterstained with spectral DAPI (Akoya Biosciences FP1490) and sections 217 mounted with Faramount Aqueous Mounting Medium (Dako S3025). 218 Each whole-tissue section was scanned on a Vectra-Polaris Automated Quantitative Pathology 219 Imaging System (Akoya Biosciences). Tissue imaging and spectral unmixing were performed using 220 InForm software (version 2.4.8, Akoya Biosciences), as previously described described (Abengozar- 221 Muela et al. 2020; Salas-Benito et al. 2021) Image analysis was performed on 5x4 mm and 1 mm2 222 ROIs using the open-source digital pathology software QuPath versionn 0.2.3, as previously described 223 (Abengozar-Muela et al. 2020). In short, cell segmentation based on nuclear ddetection was performed 224 on QuPath using StarDist 2D algorithm, a methodhod that localizes nuclei via star-convex polygons, 225 incorporated into QuPath software by scripting. A random trees algorithm classifier was trained 226 separately for each cell markerarker by an experiencedperienced pathologist annotatannotating the tumor regions. 227 Interactive feedbackk on cell classificationcation performance is provided duduring training in the form of 228 mark-up image,mage, significantly improving the accuracy of mmamachine learning-based phenotyping 229 (Bankheadad et al. 2017; AbengozarAbengozar-Muela-Muela et al. 2020). All phenotyping and subsequent 230 quantificationsationsons were performed blinded to tthe sample identity. Cells close to the border of the images 231 were removedoveded to reduce the riskisk oof artifacts. Based on the fluorescence panels, cells were further 232 subclassifieded as CD68+, CD3+,C CD8+, and CD20+. Cells negative for these markers were defined as 233 other cell types.Inpes The mean of thereview frequency of each cell marker of similar ADC and LMS ROIs were 234 calculated for comparison (Figure 3-4). 235 236 237 2.5 Transcriptomic Profiling

238 Total RNA for gene expression assays was prepared from 5µm FFPE tissue sections of 239 microdissected ITF using the Agencourt FormaPure kit (A33341; Beckman Coulter, Indianapolis, IN, 240 USA) and following the manufacturer’s instructions. RNA concentration was determined with Qubit 241 4 Fluorometer and Qubit® RNA HS Reagent (Thermo Fisher Scientific, Waltham, MA, USA). RNA 242 samples passing the quality control evaluation (ADC=19, LMS=11) were selected. 243 Transcriptomic profiling was performed with HTG EdgeSeq Precision Immuno-Oncology Panel, 244 which interrogates 1,392 genes involved in tumor/immune interaction 245 (https://www.htgmolecular.com/assets/htg/publications/GL-HTG-EdgeSeq-Precision-Immuno- 246 Oncology-Panel-GeneListAnnotated_01.pdf). HTG EdgeSeq Chemistry was employed to synthetize 247 the RNA-Seq library. Briefly, target capture was performed by hybridizing the mRNA with Nuclease 248 Protection Probes (NPPs). The S1 nuclease was added to the mix, producing a stoichiometric amount 249 of target mRNA/NPP duplexes. This reaction was blocked by enzyme heat denaturation of S1. The 250 samples were randomized before inclusion in the HTG EdgeSeq system to reduce potential biases in 251 the run. Each hybridized sample was used as template to set up PCR reactions with specially 252 designed tags, sharing common sequences that are complementary to both 5’-end and 3’-sequences 253 of the probes, and common adaptors required for cluster generation on an Illumina sequencing 254 platform. In addition, each tag contains a unique barcode used for sample identification and

6 This is a provisional file, not the final typeset article Uterine-myometrial tumor interface

255 multiplexing. After PCR amplification, a clean-up procedure was performed using Agencourt 256 AMPure XP (Beckman Coulter). The library was quantified by quantitative PCR, using KAPA 257 Library Quantification (Roche), according to the manufacturer’s instructions. All samples and 258 controls were quantified in triplicate and no template control was included in any run. Library 259 denaturation was performed by first adding 2N NaOH, followed by addition of 2N HCl. The PhiX 260 was spiked in at a 5% (concentration of 12.5 pM). Normalized libraries were sequenced by NGS. 261 One demultiplexed FASQ file per sample was retrieved from the sequencer for data processing. HTG 262 EdgeSeq host software performed the alignment the FASTQ files to the probe list, then results were 263 parsed, and the output obtained as a read counts matrix. 264 Raw counts normalization and differential expression analysis were calculated using DESeq2 R 265 package (1.30.0). Sample outliers were identified through variance stabilizing transformation. iDEP 266 v0.92 (http://ge-lab.org/idep/) was used for pathway analysis of normalized expression values from 267 RNA-Seq data (Ge, Son, and Yao 2018). 268 269 2.6 Genome-wide DNA methylation analysis

270 Total genomic DNA from 10 µm FFPE tissue sectionsions of microdissectemicrodissected ITFIT (ADC=24, LMS=11) 271 was isolated using the AllPrep DNA/RNANA/RNA FFPE Kit (Qiagen) and followinfollowingfollow the manufacturer’s 272 instructions. All DNA samplesmples were quantifiedified by fluorometric method usinusing tthe Qubit 1X dsDNA 273 HS (High-Sensitivity)tivity) Assay Kitt (ThermoFisher) and were also checcheckedcheck for suitability for FFPE 274 restorationon following the Infinium HD FFPE QC Assay (Illum(Il(Illumina). DNA samples (100-250 ng) that 275 passed thishiss quality control evaluation (ADC=22,(ADC=22 LMS=9)L were selected for bisulfite conversion 276 using the EZ DNA Methylation kit (Zymo(Zym Research) and were moved on to the FFPE Restore 277 protocol (Illumina).Illumina).umina). The restorirestorinrestoring step was followed by Infinium HD FFPE methylation assay for 278 hybridizationonn with InfInfinInfinium MethylationEPIC BeadChips, which cover over 850,000 CpG sites along 279 the human Ingenome (Moran, reviewArribas, and Esteller 2016). Whole-genome amplification and 280 hybridization were performed on the BeadChips and followed by single-base extension and analysis 281 on a HiScan (Illumina) to assess the cytosine methylation states. Image intensities were extracted 282 using GenomeStudio (V2011.1) Methylation module (1.9.0) software from Illumina. Data quality 283 control was assessed with GenomeStudio and BeadArray Controls Reporter, based on the internal 284 control probes present on the array. The methylation score of each CpG from samples that passed this 285 quality control (ADC=21, LMS=9) was represented as β-value, and previously normalized for color 286 bias adjustment, background level adjustment and quantile normalization across arrays. Probes and 287 sample filtering involved a two-step process for removing SNPs and unreliable β-values with a high 288 detection P-value > 0.01. After this filtering step, the remaining CpGs were considered valid for the 289 study. Non-parametric Wilcoxon tests were applied to determine differentially methylated CpGs 290 (DMCpGs), which were considered significant with a false discovery rate (FDR) below 5%. All 291 statistical analyses were performed in the R statistical environment (v.3.6.1). The enrichment analysis 292 of biological pathways for the methylation profiles were evaluated by gene ontology (GO) using 293 GENECODIS (Tabas-Madrid, Nogales-Cadenas, and Pascual-Montano 2012).

294 3 Results

295 3.1 Reticulin fiber scaffolding in LMS and ADC is similar

296 Several parameters of reticular fibers were assessed in 83.3% ADC and 81.8% LMS (20 of 24 ADC 297 and 9 of 11 LMS) (Supplementary Table 1). Interestingly, we found a high degree of similarity 298 between the 5x4 mm ROIs of ADC and LMS ITFs (Figure 1A, B); in fact, the only significant 299 differences observed in the cases studied were in area size and deformity of the individual or

Uterine-myometrial tumor interface

300 meshwork fibers. However, we next performed a more restrictive analysis, excluding one case of 301 each group based not only on algorithm quality control but on subjective assessment by two 302 independent scientists (Figure 1C). In this case, we analyzed a more homogeneous sample group and 303 observed no significant size or shape-related differences between the two tumor types. Large ADC 304 ROIs with reticular fibers occupying a higher proportion of stained area (higher %SA) than LMS 305 were detected.

306 To test the robustness of the algorithm and assess the representativeness of the 1 mm2 ROIs, we 307 compared four ROIs of 1 mm2 inside the 5x4 mm against four ROIs of 1mm2 along the ITF for both 308 ADC and LMS. We found no significant differences between ROIs (data not shown) and therefore 309 accepted the 1x1 mm areas as representative of the tumor, myometrium, and selected field of ITF. 310 Only regions meeting the quality control parameters were included in the analysis. In total, we 311 compared n=21 vs n=34 tumor ROIs, n=39 vs n=24 myometrium ROIs and n=79 vs n=35 selected 312 ITF ROIs of ADC and LMS respectively (Supplementary Table 1). We employed the same 313 morphometric feature extraction procedure as above, also performingg the sstatistical Student test to 314 compare the mean of each acquired parameter by patient sample. In the 1m1mm2 ITF ROIs the two 315 tumor types presented a high number of significantlyntlyly differentt parametersameters (7(7/(7/15) (Figure 2A). ADC 316 fibers appeared larger (area=363.2µm2 vs 231.9µm2)2) and thicker (width=16(width=16.21 µm vs 13.6µm) than 317 in LMS. However, because ADCDC has significantlytly lower density of fibers pper µm2 (83.2 vs 257.48 318 fibers/µm2) when comparedompared to LLMS,MS,S, smaller %SA of the tissue (2.73% vs 5.84%) was shown. 319 Furthermore,re, ADC reticular fibers appeared smoother than the wwawavy ones in LMS, as indicated by 320 the perimetermeter ratio (0.62 vs 0.68). The higher values for vertices in ADC compared to LMS (6.04 vs 321 5.56) suggestggestest that reticular fibers have grgreategreater branching in ADC. Finally, the fractal dimension 322 revealed thathatat reticular fibers are moremo haphazardly arranged in ADC than in LMS (1.44 vs 1.40). On 323 the other hand,and,nd, the myomemyometmyometrium of ADC and LMS displayed no significant differences in any of the 324 parameters studiestudistudied (Figure 2B). Interestingly, only one shape parameter (deformity) was significantly 325 different betweenIn ADC vs LMSreview (1932.4 vs 1345.7) (Figure 2C). This observation is a more reliable 326 indication than the 5x4 mm ROIs results of a high degree of similarity between the reticular fiber 327 scaffolding at the ITF of ADC and LMS.

328 To further characterize LMS and ADC, tumor, myometrium and selected ITF were compared against 329 each other for each tumor type. In contrast to the great similarity between 1 mm2 ROIs of different 330 tissue areas in LMS, we found that ADC exhibited multiple differences when ITF was compared 331 against tumor or myometrium (Table 1). There was higher deposition of fibers, indicated by 332 increased fiber density, in ITF compared to tumor (307.10 vs 83.28) leading to an increased %SA 333 (7.44 vs 2.73). Reticular fibers at the ADC also appeared thinner at ITF than in tumor (width=13.03 334 vs 16.22), were linearized (aspect=4.20 vs 3.42) and presented less branching (vertices=5.68 vs 6.04). 335 Finally, when comparing 1 mm2 ROIs of ADC ITF vs myometrium, the fibers appeared bigger at the 336 ITF (area=300.56 vs 270.29), were thicker (width=13.03 vs 11.71) and longer (height=48.36 vs 337 43.02) and therefore had a larger perimeter (88.25 vs 72.83). In addition, the fibers at the ADC ITF 338 appeared less linearized (aspect=4.20 vs 4.15) and branched (vertices= 5.68 vs 5.54) (Table 1A). 339 Interestingly, in LMS the only significant difference when comparing both ITF and tumor vs 340 myometrium was fiber shape (shape=ITF 16.17 and tumor 21.05 vs myometrium 11.04) (Table 1B).

341 3.2 Heterogeneous Immune Environment at the ITF

342 In order to characterize immune infiltrate at the ITF, we used six-color multiplex immunostaining 343 (CD20, CD3, CD8, CD68, CK and DAPI) to estimate four different immune infiltrate subpopulations 344 in the same 5x4 mm ROIs of ADCs and LMS, used for reticular fiber analysis in sequential cuts in

8 This is a provisional file, not the final typeset article Uterine-myometrial tumor interface

345 large areas of ADCs and LMS. Immune infiltrate was assessed in 95.8% ADC and in 81% of LMS 346 (23 out of 24 and 9 out of 11, respectively) (Supplementary Table 1). Overall, approximately 347 1,595,232 cells were counted and evaluated by digital pathology. Although the LMS had a lower 348 number of total cells, the two tumor types displayed the same amount of total immune infiltrate (956 349 for ADC and 991 for LMS) (Figure 3). However, we observed diverse immune cellular compositions 350 at the ITF, finding a clearly heterogeneous distribution of B and T lymphocytes and macrophages in 351 the different tumors analyzed. Comparing the two tumors, B lymphocytes (CD20+) appeared in 352 lowest numbers out of the total immune population, being statistically smaller in LMS, while there 353 was a significant increase in macrophages, which represented the highest immune population in 354 LMS. Conversely, T lymphocytes (CD3+) emerged as the predominant immune infiltrate in ADC. 355 The relative frequency of CD20, CD3, CD8, and CD68-positive cells within each tumor reflected 356 that ADCs and LMS exhibited cellular heterogeneity regarding immune distribution, whereas the 357 total amount of infiltrate remained the same in 5x4 mm ITF (Figure 3).

358 Aiming for a more thorough characterization of the immune infiltrate,e, and given the heterogeneity 359 observed, we proceed to analyze the 1x1 mm ROIs, dividing the rregionse between tumor, 360 myometrium and ITF, choosing the same regionss used for reticular fiber anaanalysis (Supplementary 361 Figure 1). After statistical comparison, we observedd that LMS had less totatotal iimmune infiltrate than 362 ADC (860 cells/ mm2 vs 276 cells/mmells/mm2) iinn thee tumor areas, and in fact, fewfewer overall cells with 363 markers were identifieded in LMS (Figuregure 4)4);; however, only CD8 became sstatistically smaller (132 364 cells/ mm2 vs 12 cells/ mm2). This observation suggests that LMSLML tumor areas are colder, or less 365 abundantt in terms of immune infiltrate, than ADC (Figure 4). Regarding ADC and LMS 366 myometrium,ium,m, no significant differences in ddistrdistribution of CD20, CD3, CD8 and CD68-positive cells 367 were foundnd between the two tumortumo types, as expected, with a low number of total cells in the 368 myometriumm near to LMLMS (Figure 4). However, the 1x1 mm ITF regions displayed significant 369 differences betweenbetwebetw CD20, CD3 and CD8 positive cells per mm2, being higher in ADC. Macrophage 370 remained theIn most abundant existingreview cells in both ITF tumors (Figure 4). Interestingly, the 1mm2 ITF 371 ROIs results were more accurate than5x4 mm ITF. The CD20 population still remained the lowest of 372 all in both tumors, being significantly smaller in LMS. While CD68-positive cells were the 373 predominant population in LMS, CD3 was the highest in ADC and again significantly different 374 compared to LMS.

375 3.3 Transcriptional profiling reveals a possible role for antimicrobial peptides in immune 376 response at the ITF of ADC

377 Differential gene expression analyses revealed statistically significant upregulation of 142 genes and 378 downregulation of 97 genes when comparing ADC (n=19) versus LMS ITF (n=11) (Supplementary 379 Table 1). Slight upregulation of T cell and B cell markers was observed in ADC ITF, but no 380 differences were detected in myeloid expression markers (Figure 5A) in accordance with the 381 multiplex immunofluorescence findings. Parametric analysis of gene set enrichment using GO 382 showed the activation of several biological processes related to cell adhesion, epithelial cell 383 development and kidney morphogenesis in ADC ITF, likely resulting from their intrinsic tumor 384 morphology (Figure 5B). Indeed, ADC ITF showed increased expression of E-cadherin, EPCAM and 385 several keratins and integrins (Figure 5C). 386 Interestingly, antimicrobial humoral response appeared to be activated in ADC ITF. By querying the 387 curated reactome database, upregulation of antimicrobial peptides was also observed, along with 388 innate immune response and neutrophil degranulation (Figure 5D). ECM proteoglycans seemed to be 389 upregulated in the LMS ITF, as expected given their mesenchymal phenotype. 390

Uterine-myometrial tumor interface

391 3.4 Genome-wide DNA methylation analysis identifies an epigenetic signature of ITF 392 related to cell adhesion and extracellular matrix

393 The DNA methylation profile of the primary ITF was compared in ADC (n=21) and LMS (n=9) 394 using the Infinium MethylationEPIC array (850K) (Figure 6A). The scatter plot of this epigenomic 395 comparison revealed that the ITF of both tumor types share a similar general DNA methylation 396 pattern (R2=0.946) (Figure 6B). In this sense, most of the 665,840 valid CpGs analyzed had similar 397 DNA methylation levels, while a small proportion of CpGs (3.5%, 23,296 CpGs) showed significant 398 differences (p<0.05; FDR<5%) between ADC and LMS. These differentially methylated CpGs 399 (DMCpGs) were distributed in several regions of the genome (Figure 6C), including CpG islands 400 (CGIs) and promoters. We found an epigenetic signature of 2,971 DMCpGs (corresponding to 1,253 401 genes) located in CGIs of promoters, which were able to differentiate the ITF of ADC and LMS 402 (Figure 7A). Of note, the GO analysis of this epigenetic signature revealed enrichment of 403 differentially methylated genes between ADC and LMS related to cell adhesion and ECM 404 organization pathways, among others. However, no evident enrichment of immune response 405 pathways was detected (Figure 7B).

406 3.5 Gene expression profiling and DNAA methylationation identify a 2020-gene-genegen epigenetic signature 407 characteristic of ITF

408 As hypermethylation of promotersoters has been demonstrated to induinduce gene silencing in several tumor 409 types, includingcluding uterine neoplasmneoplasms (W. Zhang et al. 20182018), we decided to evaluate the association 410 between thee gene expression and methylation profprofiling of promoter CGIs previously found in the TIF 411 of primaryy uterine ADC and LMSLMS. We found 20 genes whose methylation status correlated with 412 significant changes in gegene reviewreexpression (Figure 8A, Supplementary Table 3); specifically, 9 413 hypermethylatedlated ggenes showed downregulated gene expression, while 11 hypomethylated genes 414 were overexpressed.InI Of note, reviewrevie these 20 genes (corresponding to 58 CpGs) represent a signature able to 415 clearly differentiate the ITF of primary uterine ADC and LMS (Figure 8B). Importantly, some of the 416 genes of this signature are related to immune system, cell adhesion and tumor stroma.

417

418 4 Discussion

419 In the uterus, ADC and LMS account for a significant proportion of malignant tumors associated 420 with distant metastasis. As tumors they are very different, with distinct morphological and molecular 421 features, and also a differing cell of origin. However, both occur in the uterus, and each of them 422 invades the myometrium as the initial step for tumor progression. Although dissimilar in many 423 aspects, it is feasible that they share some adaptive responses while invading the myometrium, or 424 alternatively, they could produce a similar response to tumor invasion in the myometrium. It has been 425 suggested that we need a paradigm shift to focus our attention on the tumor microenvironment, rather 426 than only on tumor cells.

427 The process of stromatogenesis refers to the formation of new stroma in the active sites of tumor 428 invasion and metastasis. It is proposed that the formation of this new stroma is generated and 429 governed by the interplay between tumor cells and the complicity of adjacent activated fibroblasts. 430 This new stroma disrupts the normal structures of the ECM and its continuity and favors easy 431 penetration by tumor cells, thus helping their migration and immune cell infiltration among other 432 cells (Horrée et al. 2007; Bremnes et al. 2011; Giatromanolaki, Sivridis, and Koukourakis 2007). The 433 ECM composition and spatial organization of collagen fibers in the newly formed stroma is decisive

10 This is a provisional file, not the final typeset article Uterine-myometrial tumor interface

434 not only for development of several neoplasia but for tumor prognosis (Acerbi et al. 2015; Xu et al. 435 2019; Devendra et al. 2018; Kondo 2011). In fact, like type I collagen, reticulin distribution is 436 increased in many cancers such as head and neck squamous cell cancer, breast, pancreas and 437 colorectal cancer (Nissen, Karsdal, and Willumsen 2019) and increases proliferation, migration and 438 metastasis in pancreatic and glioblastoma cancer cells as well as in invasive prostate cancer (Menke 439 et al. 2001; Chintala et al. 1996; Kanematsu et al. 2004). Thus, in-depth characterization of 440 morphometric features of reticular fibers (collagen type III) in invasive tumors could provide 441 advanced insight into mechanisms underlying the invasion and galvanize targeted strategies for 442 precision medicine (Kanner et al. 2013; Walke and Bhagat 2017)

443 Although several aspects of the ITF have been characterized in other types of tumors, comparison 444 between them is very complicated due to differing quantification methods. Furthermore, the exact 445 size of the true invasion front is also under debate (Horrée et al. 2007) and little is known about the 446 histomorphometry of the fibers at the ITF of ADC and LMS. When analyzing the 5x4 mm ITF areas 447 of ADC and LMS we found very few significant differences betweeneen morphometricmo parameters, 448 indicating a striking similarity between the ITF of the twowo tumor types. In restrictive analysis, 449 choosing an ADC sample group with a more homogeneousogeneouseneous architecture (based(base on subjective criteria) 450 all the significant differences disappeared,eared, suggestingesting that an inclusivinclusive analysis reflects the 451 architectural heterogeneity of the ADADCC tumors. We also circumvented the ississue of variable tissue 452 amounts when analyzingzing 5x4 mm ITFF ROIs by increasing the magnificatiomagnification of the ROIs to 1 mm2 453 within the previously studied 5x4 mm ROIs in which we enensured 50% each of tumor and 454 myometrium.rium. After ththee necessary controls, we found thattha the area of the fibers in the 1 mm2 ITF was 455 no longerr significantlyignificantly different between ADCA and LMS, but the differences in deformity remained, 456 indicating thehe potential value of thithis parameter as a differential characteristic between the two types 457 of tumors. We hypothesizedhypothesizehypothesiz that the area of the fibers in the 5x4 mm areas could vary due to an 458 uneven contributiontributitribu of the tumor to the analysis, given that the area was no longer different in the 1 459 mm2 ITF ROIs.In The algorithm review showed notably high sensitivity when measuring morphometric 460 parameters, since in the 1 mm2 areas of myometrium we did not observe any significant differences 461 between the two tumors, as expected. However, the large number of differences observed between 1 462 mm2 areas of pure tumor tissue reflects the ability to detect differences between negligible reticular 463 fiber features of the tumors under subjective criteria.

464 At a microscopic level, ROIs of 1 mm2 were sufficient for comparison of reticular fiber 465 morphometric parameters between tumor, myometrium and ITF of ADC and LMS. Furthermore, 1 466 mm2 ROIs allowed us to characterize the microarchitecture of the reticular fibers in within the same 467 type of tumor. Interestingly, comparisons between tumor, myometrium and ITF in LMS revealed a 468 high degree of similarity and little disruption of tissue architecture. Conversely, in ADC we observed 469 a very high number of significantly different morphometric features comparing the above-mentioned 470 ROIs (Table 1), suggesting that continuity of ECM architecture is much more disrupted in ADC than 471 LMS tumors. Regardless of statistical difference, observing the mean of morphometric parameters at 472 the ITF of ADC, we found that almost all values fell within the range of myometrium and tumor 473 means, except for density (along with increased %SA) and aspect (linearization) of fibers (Table 1). 474 Since the areas were carefully chosen to represent 50% each of myometrium and tumor components, 475 we hypothesized that most changes observed at the ADC ITF could be attributable to the contribution 476 of adjacent tissue. However, we cannot exclude the possibility of orchestrated ECM remodeling by 477 tumor and non-tumoral cells to favor tumor cell migration (Horrée et al. 2007; Giatromanolaki, 478 Sivridis, and Koukourakis 2007; Bremnes et al. 2011). In fact, during tumor progression, cancer- 479 associated fibroblasts are key players in dysregulated collagen turnover, leading to tumor fibrosis as 480 evidenced by increased collagen depositions in the immediate periphery of the tumor (Pankova et al.

Uterine-myometrial tumor interface

481 2016). Consistent with our results, we observed an increase of fiber density at the ADC ITF 482 compared to tumor. In addition, the collagens are often crosslinked and linearized leading to 483 stiffening of the tissue in several tumors (Nissen, Karsdal, and Willumsen 2019) which we also 484 observed in ADC ITF as displayed by a significantly longer appearance (linearization) of the reticular 485 fibers at the ADC ITF. ECM stiffening could in turn elicit behavioral effects on surrounding tumor 486 cells regarding cell proliferation, differentiation, gene expression, migration, invasion, metastasis and 487 survival, all hallmarks of cancer (Pickup, Mouw, and Weaver 2014), and which correlate with worse 488 patient outcomes patient (Kehlet et al. 2016; Leeming et al. 2011). Nevertheless, there are still 489 several features of the fiber topology to take into account for future studies that would require smaller 490 regions, such as analysis of fiber orientation with respect to the tumor-myometrium boundary, which 491 would provide additional value for prognosis (Conklin et al. 2011; Xu et al. 2019; Acerbi et al. 492 2015).

493 Through multiplex cellular analyses we could demonstrate that the ADC ITF contains a 494 heterogeneous immune environment compared to the one in LSM. We found variance between 495 different CD8+ T cell and macrophage subpopulations of different ITF in ADC and LSM. This 496 suggests diverse myeloid and lymphocyte compositionsions of ITF immune envirenvironments, with CD8+ T 497 cell and macrophage density possibly influenced byy the presence of certain ECM components, such 498 as collagens fibers, which could prevent interactionsons with other immune cellcells.s 499 It has been shown that reciprocalocal interaction between the ECM aandw tumor and non-tumor cells such 500 as myofibroblastsbroblasts determines recruitment, activation, iwiewand reprogramming of stromal, inflammatory, 501 and immunene cells (Sanegre et al. 2020)2020).. DesDespiteview the great similarity of reticulin scaffolding at the ITF 502 of ADC andd LMS, we observed a differentiald eview immune infiltrate pattern, as informed by CD20, CD3 503 and CD8 markers.arkers. We thuthus cannot attribute a direct effect of reticulin architecture on the immune 504 infiltrate compositionompos right at the ITF. However, there could be a relationship between the 505 morphometryIn of reticular fibers revrerevie and the total amount of infiltrate, since the immune population is the 506 same in both tumor types. A dense architectural barrier can be imposed not only by reticular fibers 507 but by ECM elements such as other collagens, hyaluronic acid, and laminins. Whereas high 508 molecular weight hyaluronic acid provides structural integrity and suppresses the immune response 509 by increasing regulatory T cell activity, laminins prevent transmigration and polarize leucocytes 510 (Chanmee, Ontong, and Itano 2016; Bollyky et al. 2009; Simon and Bromberg 2017) ECM 511 remodeling enzymes such as metalloproteinases, matrikines and versicans act as cytokines and 512 chemokines promoting IL expression and T cell chemoattractants, polarizing and activating the 513 immune cells (Thomas, Edelman, and Stultz 2007; Hope et al. 2016). In fact, although an up- 514 regulation of collagen genes in LMS tumors can be observed, it is still not clear whether these genes 515 are responsible for immune density distribution. Another plausible explanation for leucocyte 516 accumulation at the ITF of ADC could be that reticulin morphometric description suggests that ADC 517 tumor areas seem stiffer than LMS tumors, as indicated by the larger area and width and greater 518 branching and fractal dimension of the fibers (although fewer fibers can be counted). Stiffer tumors 519 could lead to a physical barrier, thus increasing leucocyte concentration at the ITF. In addition, ECM 520 stiffness can induce chromosomal rearrangements and transcriptional changes in several diseases 521 (López-Carrasco et al. 2020; Simi, Pang, and Nelson 2018) and many stiffness-sensitive genes may 522 respond to stiffness in nonlinear ways (Darnell, Gu, and Mooney 2018).

523 Transcriptomic profiling of the ITF of ADC and LMS suggests that activated neutrophils secreting 524 antimicrobial peptides (AMP) may play a role in the immune response to oncogenesis in the context 525 of endometrial tumors. Indeed, antimicrobial humoral response seems to be inactivated in LMS, in 526 stark contrast to ADC, in which there is an upregulation of this signature. AMPs have been described

12 This is a provisional file, not the final typeset article Uterine-myometrial tumor interface

527 as important actors in angiogenesis and modulation of the immune response, via stimulation of 528 chemokines and chemotaxis of leukocytes, and may also exert cytotoxic activity against tumor cells 529 (Al-Rayahi and Sanyi 2015; Jin and Weinberg 2019). In fact, LCN2, one of the top upregulated 530 genes in the ITF of ADCs (Supplementary Table 2), codes for neutrophil gelatinase-associated 531 lipocalin, which has been reported to be associated with aggressive features of endometrial ADC 532 (Mannelqvist et al. 2012; Cymbaluk-PA Oska et al. 2019) but is also described as an invasiveness 533 and metastasis suppressor in other contexts (Tong et al. 2008; Lim et al. 2007; Lee et al. 2006; 534 Santiago-Sánchez et al. 2020). Another highly upregulated antimicrobial peptide, DEFB1 535 (Supplementary Table 2), belongs to a family of cytotoxic peptides made by neutrophils. This gene 536 has recently been identified as a chromosome 8p tumor-suppressor gene, downregulated in several 537 malignancies such as salivary gland tumors, oral squamous cell carcinoma, colon and liver cancer 538 (Álvarez, Martínez Velázquez, and Prado Montes de Oca 2018; Sun et al. 2019) Loss of orthologous 539 murine defense-1 in mice enhances nickel sulfate-induced LMS and causes mouse kidney cells to 540 exhibit increased susceptibility to HPV-16 E6/7-induced neoplastic transformation (Sun et al. 2019). 541 DEFB1 also induces formation of neutrophil extracellular traps (NETs) (Álvarez, Martínez 542 Velázquez, and Prado Montes de Oca 2018), mesh-like structures composed of cytosolic and granule 543 proteins which are assembled on a scaffold of decondensedcondensed chromatin fibfibers in a process termed 544 NETosis, which has been proposed to protect tumoror cells from T cellcell-- or natural killer (NK) cell- 545 mediated toxicity (Teijeira et al. 2020). Thiss should be further studied bby developing a specific 546 antibody panel to visualize and measuresure NET formation in tissue. OthOther upregulated antimicrobial 547 genes such as the chemokine CXCL3 could contribute to recrecruitmentw of neutrophils at the ITF. 548 Altogether,er, it is tempting to speculate that impaired anantimicrobialiew humoral response could contribute 549 to local or distant spread of uterine LMS.eview view 550 Epigenomic analysis showshowed areview generally similar methylation pattern in the ITF of ADC and LMS. 551 However, when we focused our analysis on CGI promoters, we observed a different methylation 552 profile in theIn ITF of the two revrevire tumor types. Interestingly, part of these differentially methylated genes 553 were involved in biological processes related to cell adhesion and ECM organization. Of note, both 554 biological pathways play an major role in invasion and migration of several types of tumors, 555 including uterine neoplasms (Abal et al. 2007; Yadav et al. 2020), suggesting that epigenetic 556 disruption of these pathways could influence the metastatic properties of ADC and LMS. In addition, 557 combining transcriptomic and epigenomic analysis, we observed an association between the 558 methylation status of CGI promoters and expression in a set of 20 genes, implying epigenetic 559 regulation of these genes in ITF. Importantly, this 20-gene epigenetic signature was able to clearly 560 differentiate between ITF of ADC and LMS. One of these genes was ITGB3, an integrin involved in 561 cell adhesion and ECM whose deregulation is associated with cancer metastasis metastasis (Hu et al. 562 2018; Langsenlehner et al. 2006) Another epigenetically regulated gene identified in this signature 563 was CXCL16, a chemokine-derived peptide implicated in antimicrobial and antitumoral response 564 (Valdivia-Silva, Medina-Tamayo, and Garcia-Zepeda 2015). Cancer patients with high levels of 565 CXCL16 expression have shown more elevated levels of CD4(+) and CD8(+) tumor-infiltrating 566 lymphocytes and NK, leading to improved disease prognosis (Hojo et al. 2007)(Valdivia-Silva, 567 Medina-Tamayo, and Garcia-Zepeda 2015). However, this chemokine can also induce invasion and 568 metastasis via several mechanisms, such as the recruitment of mesenchymal stem cells into the tumor 569 (Jung et al. 2013). In addition, CXCL16 contributes to ectopic endometrial stromal cell migration and 570 invasion (Peng, Ma, and Lin 2019). We also detected other epigenetically regulated genes in this 571 signature, including CD70, IRAK, IL20RA, THBD, TACSTD2, and MST1R, with implications related 572 to the tumor immune microenvironment (Kumar et al. 2020; Jain, Kaczanowska, and Davila 2014; 573 Gao et al. 2021; Gu et al. 2020; Rossi et al. 2014; Babicky et al. 2019). In our study, TACSTD2 was 574 overexpressed and hypomethylated in the ITF of ADC in contrast to LMS. Interestingly, other studies

Uterine-myometrial tumor interface

575 have found highly expressed TACSTD2 in aggressive endometrial carcinomas associated with tumor 576 invasion (Bignotti et al. 2011), and it has been suggested as an attractive target for cancer 577 immunotherapy in biologically aggressive endometrial carcinomas (Bignotti et al. 2012). 578 Collectively, these data indicate that although the ITFs of ADC and LMS have a similar global 579 methylation status, there are also some significant epigenetic differences that could influence the 580 invasive, migration, and tumor microenvironment properties of the two types of uterine neoplasms.

581 In conclusion, our results demonstrate that ADC and LMS share several morphometric and molecular 582 features at the interface between tumor cells and myometrium, but also an intertumoral 583 heterogeneity. There was a high degree of between-tumor ITF similarity in reticulin scaffolding, 584 macrophage infiltration and global DNA methylation pattern. However, some differences were also 585 found. LMS had fewer T and B lymphocytes; additionally, neutrophil degranulation releasing 586 antimicrobial peptides was observed in ADCs, but not in LMS. Furthermore, gene expression 587 profiling and DNA methylation identified a 20-gene epigenetic signature characteristic of ITF. It 588 seems that the two tumors exhibit certain remarkable similarities whenhen invinvading the myometrium 589 despite being different neoplasms with profound morphologicologic and molemolecular differences, and 590 originating from different precursor cells. These findingsindings open up the posspospossibility of investigating 591 common therapeutic strategies based on these similaritiesrit iinn the ITF microenvirmicroenvmicroenvironment. 592

593 5 Conflictct of Interest

594 The authorsorsrs declare that the research was conducconducted in the absence of any commercial or financial 595 relationshipsipss that could be construed as a potential conflict of interest. 596 In review 597 6 Author Contributions

598 RN, and MX contributed to conception and design of the study. SS wrote the first draft of the 599 manuscript. SS, NE, CA, JD, AD, EA, XM and RN wrote sections of the manuscript. BD, RL, SR, 600 EA, XM provided clinicopathological data. NE provided the collaborative consortium with 601 microdissected samples. SS and RN designed and performed the morphometric characterization of 602 the reticulin fibers. CA and DR performed the multiplex staining and analysis. JD, CS and EA 603 evaluated the transcriptomic profile. AD and AJ performed the genome-wide methylation analysis. 604 All authors contributed to manuscript revision, read, and approved the submitted version.

605 7 Funding

606 This research was supported by grants from ISCIII and ERDF (PI17/01558 and PI20/01107), by 607 CIBERONC (contracts CB16/12/00484, CB16/12/0328, CB16/12/00363, CB16/12/00364, 608 CB16/12/00481, CB16/12/00231) and by Grupos Coordinados Estables from Asociación Española 609 Contra el Cáncer (AECC). AD is funded by a contract “Juan Rodés” from ISCIII (JR17/00016). The 610 funders had no involvement in the research process nor in the preparation and submission of the 611 article. 612

613 8 Acknowledgments

14 This is a provisional file, not the final typeset article Uterine-myometrial tumor interface

614 The authors are grateful to Aitor Carretero (research collaborator of INCLIVA-Instituto de 615 Investigación Sanitaria) for the initial analysis of reticulin fibers and Nuria Santonja (medical 616 specialist in anatomic pathology at the Hospital General Universitario of Valencia) for providing 617 leiomyosarcomas samples, and to Aida Bao-Caamano (Epigenomics Unit, IDIS) for her technical 618 support with genome-wide DNA methylation analysis. The authors also thank Kathryn Davies for 619 English corrections. 620 621 9 References 622 623 Abal, M., M. Llauradá, A. Dolla, M. Monge, E. Colas, M. González, M. Rigau, et al. 2007. 624 “Molecular Determinants of Invasion in Endometrial Cancer.” Clinical and Translational 625 Oncology 9 (5): 272–77. https://doi.org/10.1007/s12094-007-0054-z. 626 Abengozar-Muela, Marta, María Villalba Esparza, David Garcia-Ros, Cindy Estefanía Vásquez, José 627 I. Echeveste, Miguel Angel Idoate, Maria D. Lozano, Ignacio Melero, and Carlos E. de Andrea. 628 2020. “Diverse Immune Environments in Human Lung Tuberculosisosis GraGranulomas Assessed by 629 Quantitative Multiplexed Immunofluorescence.” Modernern Pathology 33 ((1(12): 2507–19. 630 https://doi.org/10.1038/s41379-020-0600-6. 631 Acerbi, I., L. Cassereau, I. Dean, Q. Shi,, A. Au, C. Park,ark, Y. Y. CChen,hen, J. LiphardLiphaLiphardt, E. S. Hwang, and 632 V. M. Weaver. 2015. “Humanuman Breast Cancerncer Invasion and Aggression CCorrelatesCorr with ECM 633 Stiffening and Immune Cell Infiltration.”ltration.” Integrative Biology (United(Unit Kingdom)K 7 (10): 1120– 634 34. https://doi.org/10.1039/c5ib00040h.ps://doi.org/10.1039/c5ib00040h 635 Al-Rayahi,hi, Izzat A.M., and Raghad H.H. Sanyi. 2015. ““The Overlapping Roles of Antimicrobial 636 Peptidesideses and Complement in RecruitmRecruitmenRecruitment and Activation of Tumor-Associated Inflammatory 637 Cells.”.” Frontiers in ImmunologImmunoloImmunology. Frontiers Media S.A. 638 https://doi.org/10.3389/fimmu.2015.00002.//doi.org/10.338doi.org/10.33 639 Álvarez, ÁngelngelIn HH., Moisés Martínezreview Velázquez, and Ernesto Prado Montes de Oca. 2018. “Human β- 640 Defensin 1 Update: Potential Clinical Applications of the Restless Warrior.” International 641 Journal of Biochemistry and Cell Biology. Elsevier Ltd. 642 https://doi.org/10.1016/j.biocel.2018.09.007. 643 Babicky, Michele L., Megan M. Harper, Jeffery Chakedis, Alex Cazes, Evangeline S. Mose, Dawn 644 V. Jaquish, Randall P. French, et al. 2019. “MST1R Kinase Accelerates Pancreatic Cancer 645 Progression via Effects on Both Epithelial Cells and Macrophages.” Oncogene 38 (28): 5599– 646 5611. https://doi.org/10.1038/s41388-019-0811-9. 647 Bankhead, Peter, Maurice B. Loughrey, José A. Fernández, Yvonne Dombrowski, Darragh G. 648 McArt, Philip D. Dunne, Stephen McQuaid, et al. 2017. “QuPath: Open Source Software for 649 Digital Pathology Image Analysis.” Scientific Reports 7 (1): 1–7. 650 https://doi.org/10.1038/s41598-017-17204-5. 651 Bao-Caamano, Aida, Aitor Rodriguez-Casanova, and Angel Diaz-Lagares. 2020. “Epigenetics of 652 Circulating Tumor Cells in Breast Cancer.” In Advances in Experimental Medicine and Biology, 653 1220:117–34. Springer. https://doi.org/10.1007/978-3-030-35805-1_8. 654 Bignotti, Eliana, Antonella Ravaggi, Chiara Romani, Marcella Falchetti, Silvia Lonardi, Fabio 655 Facchetti, Sergio Pecorelli, et al. 2011. “Trop-2 Overexpression in Poorly Differentiated 656 Endometrial Endometrioid Carcinoma: Implications for Immunotherapy with HRS7, a 657 Humanized Anti-Trop-2 Monoclonal Antibody.” International Journal of Gynecological 658 Cancer 21 (9): 1613–21. https://doi.org/10.1097/IGC.0b013e318228f6da. 659 Bignotti, Eliana, Laura Zanotti, Stefano Calza, Marcella Falchetti, Silvia Lonardi, Antonella Ravaggi, 660 Chiara Romani, et al. 2012. “Trop-2 Protein Overexpression Is an Independent Marker for 661 Predicting Disease Recurrence in Endometrioid Endometrial Carcinoma.” BMC Clinical 662 Pathology 12 (1). https://doi.org/10.1186/1472-6890-12-22.

Uterine-myometrial tumor interface

663 Bollyky, Paul L., Ben A. Falk, Rebecca P. Wu, Jane H. Buckner, Thomas N. Wight, and Gerald T. 664 Nepom. 2009. “Intact Extracellular Matrix and the Maintenance of Immune Tolerance: High 665 Molecular Weight Hyaluronan Promotes Persistence of Induced CD4+CD25+ Regulatory T 666 Cells.” Journal of Leukocyte Biology 86 (3): 567–72. https://doi.org/10.1189/jlb.0109001. 667 Bremnes, Roy M., Tom Dønnem, Samer Al-Saad, Khalid Al-Shibli, Sigve Andersen, Rafael Sirera, 668 Carlos Camps, Inigo Marinez, and Lill Tove Busund. 2011. “The Role of Tumor Stroma in 669 Cancer Progression and Prognosis: Emphasis on Carcinoma-Associated Fibroblasts and Non- 670 Small Cell Lung Cancer.” Journal of Thoracic Oncology 6 (1): 209–17. 671 https://doi.org/10.1097/JTO.0b013e3181f8a1bd. 672 Chanmee, Theerawut, Pawared Ontong, and Naoki Itano. 2016. “Hyaluronan: A Modulator of the 673 Tumor Microenvironment.” Cancer Letters. Elsevier Ireland Ltd. 674 https://doi.org/10.1016/j.canlet.2016.02.031. 675 Chintala, Shravan Kumar, Raymond Sawaya, Ziya Levent Gokaslan, and Jasti Sambasiva Rao. 1996. 676 “The Effect of Type III Collagen on Migration and Invasion of Human Glioblastoma Cell Lines 677 in Vitro.” Cancer Letters 102 (1–2): 57–63. https://doi.org/10.1016/0304-3835(96)04163-8.16/0304 678 Conklin, Matthew W., Jens C. Eickhoff, Kristin M. Riching,, Carolyn A. PehlkPehlke, Kevin W. Eliceiri, 679 Paolo P. Provenzano, Andreas Friedl, and Patriciatriciacia J. Keely. 2011. “Align“Aligne“Aligned Collagen Is a 680 Prognostic Signature for Survival in Human Breasteast Carcinoma.” Americanmeric Journal of Pathology 681 178 (3): 1221–32. https://doi.org/10.1016/j.ajpath.2010.11.076./doi.org/10.1016/j.ajpath.2010.11.076./j.ajpath.2010.11.07 682 Cree, Ian A., Valeriee A. White, B. Iciarar Indave, and Dilani Lokuhetty. 22020. “Revising the WHO 683 Classification:fication:ation: Female Genital Tract TumoTumours.”uursrs.” HistopatholoHistopathologyistopathol . Blackwell Publishing Ltd. 684 https://doi.org/10.1111/his.13977.ps://doi.org/10.1111/his.13977.s://doi.org/10.1111/his.13977. 685 Cuppens, Tine, Matthieu Moisse, Jeroen DeprDeDepreeuw, Daniela Annibali, Eva Colas, Antonio Gil- 686 Moreno,no,o, Jutta Huvila, et al. 201202018. “Integrated Genome Analysis of Uterine Leiomyosarcoma to 687 Identifyfy Novel DriveDriver Genes and Targetable Pathways.” International Journal of Cancer 142 688 (6): 1230–43.30In4 https://doi.org/10.1002/ijc.31129. review 689 Cymbaluk-PA Oska, Aneta, Anita Chudecka-Gaaz, Ewa Pius-Sadowska, Bogusaw Machaliaski, 690 Agnieszka Sompolska-Rzechua, Sebastian Kwiatkowski, and Janusz Menkiszak. 2019. “The 691 Role of Lipocalin-2 Serum Levels in the Diagnostics of Endometrial Cancer.” Cancer 692 Biomarkers 24 (3): 315–24. https://doi.org/10.3233/CBM-181942. 693 Darnell, Max, Luo Gu, and David Mooney. 2018. “RNA-Seq Reveals Diverse Effects of Substrate 694 Stiffness on Mesenchymal Stem Cells.” Biomaterials 181 (October): 182–88. 695 https://doi.org/10.1016/j.biomaterials.2018.07.039. 696 Devendra, Alrani, Kochli C. Niranjan, Acharya Swetha, and Hallikeri Kaveri. 2018. “Histochemical 697 Analysis of Collagen Reorganization at the Invasive Front of Oral Squamous Cell Carcinoma 698 Tumors.” Journal of Investigative and Clinical Dentistry 9 (1). 699 https://doi.org/10.1111/jicd.12283. 700 Devis, Laura, Elena Martinez-Garcia, Cristian P. Moiola, Maria Teresa Quiles, Maria Antonia Arbos, 701 Tomita Vasilica Stirbat, Françoise Brochard-Wyart, et al. 2018. “ALCAM Shedding at the 702 Invasive Front of the Tumor Is a Marker of Myometrial Infiltration and Promotes Invasion in 703 Endometrioid Endometrial Cancer.” Oncotarget 9 (24): 16648–64. 704 https://doi.org/10.18632/oncotarget.24625. 705 Diaz-Lagares, Angel, Ana B. Crujeiras, Paula Lopez-Serra, Marta Soler, Fernando Setien, Ashish 706 Goyal, Juan Sandoval, et al. 2016. “Epigenetic Inactivation of the P53-Induced Long Noncoding 707 RNA TP53 Target 1 in Human Cancer.” Proceedings of the National Academy of Sciences of 708 the United States of America 113 (47): E7535–44. https://doi.org/10.1073/pnas.1608585113. 709 Espinosa-Aranda, Jose luis, Ismael Serrano, M. Milagro Fernández-Carrobles, Irene Tadeo, Rosa 710 Noguera, Rebeca Burgos, and Gloria Bueno. 2016. “XXXIV Congreso Anual de La Sociedad 711 Española de Ingeniería Biomédica.” In Modelo Logit de Regresión Nominal Para La Detección

16 This is a provisional file, not the final typeset article Uterine-myometrial tumor interface

712 de Fibras de Reticulina En Histopatología, 451–54. 713 Euscher, Elizabeth, Patricia Fox, Roland Bassett, Hayma Al-Ghawi, Rouba Ali-Fehmi, Denise 714 Barbuto, Bojana Djordjevic, et al. 2013. “The Pattern of Myometrial Invasion as a Predictor of 715 Lymph Node Metastasis or Extrauterine Disease in Low-Grade Endometrial Carcinoma.” 716 American Journal of Surgical Pathology 37 (11): 1728–36. 717 https://doi.org/10.1097/PAS.0b013e318299f2ab. 718 Ferlay, Jacques, Hai Rim Shin, Freddie Bray, David Forman, Colin Mathers, and Donald Maxwell 719 Parkin. 2010. “Estimates of Worldwide Burden of Cancer in 2008: GLOBOCAN 2008.” 720 International Journal of Cancer 127 (12): 2893–2917. https://doi.org/10.1002/ijc.25516. 721 Friedl, Peter, and Stephanie Alexander. 2011. “Cancer Invasion and the Microenvironment: Plasticity 722 and Reciprocity.” Cell. Cell. https://doi.org/10.1016/j.cell.2011.11.016. 723 Fusté, Noel P., Esmeralda Castelblanco, Isidre Felip, Maria Santacana, Rita Fernández-Hernández, 724 Sònia Gatius, Neus Pedraza, et al. 2016. “Characterization of Cytoplasmic Cyclin D1 as a 725 Marker of Invasiveness in Cancer.” Oncotarget 7 (19): 26979–91. 726 https://doi.org/10.18632/oncotarget.8876. 727 Gao, Wenjuan, Huiping Wen, Luyu Liang, Xiaoli Dong, Renlenle Du, Wei ZhouZhou, Xuehui Zhang, 728 Chunze Zhang, Rong Xiang, and Na Li. 2021.1. “IL20RA Signaling EnhancEnhanEnhances Stemness and 729 Promotes the Formation of an Immunosuppressivemunosuppressiveive Microenvironment in BBreast Cancer.” 730 Theranostics 11 (6): 2564–80.80. https://doi.org/10.7150/thno.45280.org/10.7150/thno.45280. 731 Ge, Steven Xijin, Eunun Wo Son, and Runan Yao. 2018. “IDEP: An IntegrIntegrated Web Application for 732 Differentialerentialential Expression and Pathway Analysis of RNARNA-Seq-Seq Data.” D BMC Bioinformatics 19 (1). 733 https://doi.org/10.1186/s12859-018-2486-6.ps://doi.org/10.1186/s12859s://doi.org/10.1186/s128 -00118-244886-6. 734 Getz, Gad,d, Stacey B. Gabriel, Kristian CibuCibulskCibulskis, Eric Lander, Andrey Sivachenko, Carrie Sougnez, 735 Mike Lawrence, et al. 2013. “In“I“Integrated Genomic Characterization of Endometrial Carcinoma.” 736 Naturee 497 (7447): 67–73.6 https://doi.org/10.1038/nature12113. 737 Giatromanolaki,Inlakilak Alexandra, Efthimiosreview Sivridis, and Michael I. Koukourakis. 2007. “The Pathology 738 of Tumor Stromatogenesis.” Cancer Biology and Therapy. Landes Bioscience. 739 https://doi.org/10.4161/cbt.6.5.4198. 740 Gorris, Mark A. J., Altuna Halilovic, Katrin Rabold, Anne van Duffelen, Iresha N. Wickramasinghe, 741 Dagmar Verweij, Inge M. N. Wortel, Johannes C. Textor, I. Jolanda M. de Vries, and Carl G. 742 Figdor. 2018. “Eight-Color Multiplex Immunohistochemistry for Simultaneous Detection of 743 Multiple Immune Checkpoint Molecules within the Tumor Microenvironment.” The Journal of 744 Immunology 200 (1): 347–54. https://doi.org/10.4049/jimmunol.1701262. 745 Gu, Fei Fei, Kai Zhang, Li Li Ma, Yang Yang Liu, Chang Li, Yue Hu, Qi Fan Yang, et al. 2020. 746 “The Superior Ability of Human BDCA3+ (CD141+) Dendritic Cells (DCs) to Cross-Present 747 Antigens Derived From Necrotic Lung Cancer Cells.” Frontiers in Immunology 11 (June). 748 https://doi.org/10.3389/fimmu.2020.01267. 749 Hojo, Shozo, Keiichi Koizumi, Koichi Tsuneyama, Yoshihisa Arita, Zhengguo Cui, Kanna 750 Shinohara, Takayuki Minami, et al. 2007. “High-Level Expression of Chemokine CXCL16 by 751 Tumor Cells Correlates with a Good Prognosis and Increased Tumor-Infiltrating Lymphocytes 752 in Colorectal Cancer.” Cancer Research 67 (10): 4725–31. https://doi.org/10.1158/0008- 753 5472.CAN-06-3424. 754 Hope, Chelsea, Simon Foulcer, Justin Jagodinsky, Sarah X. Chen, Jeffrey L. Jensen, Sanjay Patel, 755 Catherine Leith, et al. 2016. “Immunoregulatory Roles of Versican Proteolysis in the Myeloma 756 Microenvironment.” Blood 128 (5): 680–85. https://doi.org/10.1182/blood-2016-03-705780. 757 Horrée, Nicole, Paul J. van Diest, Daisy M.D.S. Sie-Go, and A. Peter M. Heintz. 2007. “The Invasive 758 Front in Endometrial Carcinoma: Higher Proliferation and Associated Derailment of Cell Cycle 759 Regulators.” Human Pathology 38 (8): 1232–38. 760 https://doi.org/10.1016/j.humpath.2007.01.008.

Uterine-myometrial tumor interface

761 Hu, Lei, Ming De Zang, He Xiao Wang, Bao Gui Zhang, Zhen Qiang Wang, Zhi Yuan Fan, Huo Wu, 762 et al. 2018. “G9A Promotes Gastric Cancer Metastasis by Upregulating ITGB3 in a SET 763 Domain-Independent Manner Article.” Cell Death and Disease 9 (3). 764 https://doi.org/10.1038/s41419-018-0322-6. 765 Jain, Ajay, Sabina Kaczanowska, and Eduardo Davila. 2014. “IL-1 Receptor-Associated Kinase 766 Signaling and Its Role in Inflammation, Cancer Progression, and Therapy Resistance.” Frontiers 767 in Immunology. Frontiers Media S.A. https://doi.org/10.3389/fimmu.2014.00553. 768 Jin, Ge, and Aaron Weinberg. 2019. “Human Antimicrobial Peptides and Cancer.” Seminars in Cell 769 and Developmental Biology. Elsevier Ltd. https://doi.org/10.1016/j.semcdb.2018.04.006. 770 Jung, Younghun, Jin Koo Kim, Yusuke Shiozawa, Jingcheng Wang, Anjali Mishra, Jeena Joseph, 771 Janice E. Berry, et al. 2013. “Recruitment of Mesenchymal Stem Cells into Prostate Tumours 772 Promotes Metastasis.” Nature Communications 4. https://doi.org/10.1038/ncomms2766. 773 Kanematsu, Akihiro, Akira Marui, Shingo Yamamoto, Makoto Ozeki, Yoshiaki Hirano, Masaya 774 Yamamoto, Osamu Ogawa, Masashi Komeda, and Yasuhiko Tabata. 2004. “Type I Collagen 775 Can Function as a Reservoir of Basic Fibroblast Growth Factor.” JournalJourna of Controlled Release 776 99 (2): 281–92. https://doi.org/10.1016/j.jconrel.2004.07.008.07.008. 777 Kanner, William A., Catherine I. Barry, Chandra N. Smart, David P. FrishbergFrishberFrishberg, Scott W. Binder, and 778 Mark R. Wick. 2013. “Reticulin andnd NM23 Stainingning in the Interpretation oof Lymph Nodal 779 Nevus Rests.” American Journal of Dermatopathologyatopathology 35 (4):(4): 452–57.452452–557.7. 780 https://doi.org/10.1097/DAD.0b013e318272c104.10.1097/DAD.0b013e3013e318272c10418272c104.. 781 Kehlet, S. N.,., R. SanzSanz-Pamplona,-Pamplona,ona, S. Brix, D. J. Leeming, M. A. KKarsdal, and V. Moreno. 2016. 782 “Excessivexcessivecessive Collagen Turnover Products Are ReleasedReleasRelea during Colorectal Cancer Progression and 783 Elevatedateded in Serum from Metastatic CoColorColorectal Cancer Patients.” Scientific Reports 6 (July). 784 https://doi.org/10.1038/srep30599.//doi.org/10.1038/srep305doi.org/10.1038/srep30 785 Kommoss, Felixelix K.F., DamianDamDa Stichel, Daniel Schrimpf, Mark Kriegsmann, Basile Tessier-Cloutier, 786 Aline TalhoTalhTalhouk,In Jessica N. reviewMcAlpine, et al. 2020. “DNA Methylation-Based Profiling of Uterine 787 Neoplasms: A Novel Tool to Improve Gynecologic Cancer Diagnostics.” Journal of Cancer 788 Research and Clinical Oncology 146 (1): 97–104. https://doi.org/10.1007/s00432-019-03093-w. 789 Kondo, Fukuo. 2011. “Assessment of Stromal Invasion for Correct Histological Diagnosis of Early 790 Hepatocellular Carcinoma.” International Journal of Hepatology 2011: 1–7. 791 https://doi.org/10.4061/2011/241652. 792 Kumar, Sandeep, Sunil Kumar Singh, Navin Viswakarma, Gautam Sondarva, Rakesh Sathish Nair, 793 Periannan Sethupathi, Matthew Dorman, et al. 2020. “Rationalized Inhibition of Mixed Lineage 794 Kinase 3 and CD70 Enhances Life Span and Antitumor Efficacy of CD8 + T Cells.” Journal for 795 ImmunoTherapy of Cancer 8 (2). https://doi.org/10.1136/jitc-2019-000494. 796 Langsenlehner, Uwe, Wilfried Renner, Babak Yazdani-Biuki, Tanja Eder, Thomas C. Wascher, 797 Bernhard Paulweber, Heimo Clar, Günter Hofmann, Hellmut Samonigg, and Peter Krippl. 2006. 798 “Integrin Alpha-2 and Beta-3 Gene Polymorphisms and Breast Cancer Risk.” Breast Cancer 799 Research and Treatment 97 (1): 67–72. https://doi.org/10.1007/s10549-005-9089-4. 800 Lee, Ho Jeong, Eun Kyoung Lee, Kong Ju Lee, Soon Won Hong, Yeup Yoon, and Jang Seong Kim. 801 2006. “Ectopic Expression of Neutrophil Gelatinase-Associated Lipocalin Suppresses the 802 Invasion and Liver Metastasis of Colon Cancer Cells.” International Journal of Cancer 118 803 (10): 2490–97. https://doi.org/10.1002/ijc.21657. 804 Leeming, D.J., M. Koizumi, P. Qvist, V. Barkholt, C. Zhang, K. Henriksen, I. Byrjalsen, and M.A. 805 Karsdal. 2011. “Serum N-Terminal Propeptide of Collagen Type I Is Associated with the 806 Number of Bone Metastases in Breast and Prostate Cancer and Correlates to Other Bone Related 807 Markers.” Biomarkers in Cancer 3 (January): BIC.S6484. https://doi.org/10.4137/bic.s6484. 808 Lim, Ratana, Nuzhat Ahmed, Niels Borregaard, Clyde Riley, Razan Wafai, Erik W. Thompson, 809 Michael A. Quinn, and Gregory E. Rice. 2007. “Neutrophil Gelatinase-Associated Lipocalin

18 This is a provisional file, not the final typeset article Uterine-myometrial tumor interface

810 (NGAL) an Early-Screening Biomarker for Ovarian Cancer: NGAL Is Associated with 811 Epidermal Growth Factor-Induced Epithelio-Mesenchymal Transition.” International Journal of 812 Cancer 120 (11): 2426–34. https://doi.org/10.1002/ijc.22352. 813 López-Carrasco, Amparo, Susana Martín-Vañó, Rebeca Burgos-Panadero, Ezequiel Monferrer, Ana 814 P. Berbegall, Beatriz Fernández-Blanco, Samuel Navarro, and Rosa Noguera. 2020. “Impact of 815 Extracellular Matrix Stiffness on Genomic Heterogeneity in MYCN-Amplified Neuroblastoma 816 Cell Line.” Journal of Experimental and Clinical Cancer Research 39 (1). 817 https://doi.org/10.1186/s13046-020-01729-1. 818 Mannelqvist, Monica, Ingunn M. Stefansson, Elisabeth Wik, Kanthida Kusonmano, Maria B. Raeder, 819 Anne M. Øyan, Karl Henning Kalland, Marsha A. Moses, Helga B. Salvesen, and Lars A. 820 Akslen. 2012. “Lipocalin 2 Expression Is Associated with Aggressive Features of Endometrial 821 Cancer.” BMC Cancer 12 (May). https://doi.org/10.1186/1471-2407-12-169. 822 Matias-guiu, Xavier, Lluis Catasus, Elena Bussaglia, Helena Lagarda, Arnald Garcia, Cristina Pons, 823 Josefina Muoz, Rosmary Argüelles, Pilar Machin, and Jaime Prat. 2001. “Molecular Pathology 824 of Endometrial Hyperplasia and Carcinoma.” Human Pathology 32 (6): 569–77. 825 https://doi.org/10.1053/hupa.2001.25929. 826 Menke, Andre, Christian Philippi, Roger Vogelmann,ann,n, Bjoern Seidel, Manfred PP. Lutz, Guido Adler, 827 and Doris Wedlich. 2001. “Down-RegulationRegulation off ee-Cadherin-Cadherindherin Gene ExpreExpressiExpression by Collagen Type 828 i and Type Iii in Pancreatictic Cancer Cell Lines.Lines.” CCancerancer Research 61 (8):(8) 353508–17. 829 https://pubmed.ncbi.nlm.nih.gov/11309315/.d.ncbi.nlm.nih.gov/11309315/./11309315 830 Moran, Sebastian,ebastian,astian, Carles Arribas, and Manel Esteller. 2016. “Va“Validationli of a DNA Methylation 831 Microarraycroarray for 850,000 CpG Sites of the Human GGenome Enriched in Enhancer Sequences.” 832 Epigenomicsgenomicsnomics 8 ((3):3): 38389–99.9–99.9. https://doihttps://doi.orhttps://doi.org/10.2217/epi.15.114. 833 Nissen, Neeleell I., Morten Karsdal, anand Nicholas Willumsen. 2019. “Collagens and Cancer Associated 834 Fibroblastsblastsasts in the ReReactive Stroma and Its Relation to Cancer Biology.” Journal of Experimental 835 and ClinicalinicainicIn Cancer Research review. BioMed Central Ltd. https://doi.org/10.1186/s13046-019-1110- 836 6. 837 Oliver, Amanda J., Peter K.H. Lau, Ashleigh S. Unsworth, Sherene Loi, Phillip K. Darcy, Michael H. 838 Kershaw, and Clare Y. Slaney. 2018. “Tissue-Dependent Tumor Microenvironments and Their 839 Impact on Immunotherapy Responses.” Frontiers in Immunology. Frontiers Media S.A. 840 https://doi.org/10.3389/fimmu.2018.00070. 841 Pankova, Daniela, Yulong Chen, Masahiko Terajima, Mark J. Schliekelman, Brandi N. Baird, 842 Monica Fahrenholtz, Li Sun, et al. 2016. “Cancer-Associated Fibroblasts Induce a Collagen 843 Cross-Link Switch in Tumor Stroma.” Molecular Cancer Research 14 (3): 287–95. 844 https://doi.org/10.1158/1541-7786.MCR-15-0307. 845 Peng, Yaoming, Junyan Ma, and Jun Lin. 2019. “Activation of the CXCL16/CXCR6 Axis by TNF-α 846 Contributes to Ectopic Endometrial Stromal Cells Migration and Invasion.” Reproductive 847 Sciences 26 (3): 420–27. https://doi.org/10.1177/1933719118776797. 848 Pickup, Michael W, Janna K Mouw, and Valerie M Weaver. 2014. “The Extracellular Matrix 849 Modulates the Hallmarks of Cancer.” EMBO Reports 15 (12): 1243–53. 850 https://doi.org/10.15252/embr.201439246. 851 Rossi, Edmund A., Diane L. Rossi, Thomas M. Cardillo, Chien Hsing Chang, and David M. 852 Goldenberg. 2014. “Redirected T-Cell Killing of Solid Cancers Targeted with an Anti- 853 CD3/Trop-2-Bispecific Antibody Is Enhanced in Combination with Interferon-α.” Molecular 854 Cancer Therapeutics 13 (10): 2341–51. https://doi.org/10.1158/1535-7163.MCT-14-0345. 855 Salas-Benito, Diego, Enrique Conde, Ibon Tamayo-Uria, Uxua Mancheño, Edurne Elizalde, David 856 Garcia-Ros, Jose M. Aramendia, et al. 2021. “The Mutational Load and a T-Cell Inflamed 857 Tumour Phenotype Identify Ovarian Cancer Patients Rendering Tumour-Reactive T Cells from 858 PD-1+ Tumour-Infiltrating Lymphocytes.” British Journal of Cancer.

Uterine-myometrial tumor interface

859 https://doi.org/10.1038/s41416-020-01218-4. 860 Sanegre, Sabina, Federico Lucantoni, Rebeca Burgos-Panadero, Luis de La Cruz-Merino, Rosa 861 Noguera, and Tomás Álvaro Naranjo. 2020. “Integrating the Tumor Microenvironment into 862 Cancer Therapy.” Cancers. MDPI AG. https://doi.org/10.3390/cancers12061677. 863 Santiago-Sánchez, Ginette S., Valentina Pita-Grisanti, Blanca Quiñones-Díaz, Kristyn Gumpper, 864 Zobeida Cruz-Monserrate, and Pablo E. Vivas-Mejía. 2020. “Biological Functions and 865 Therapeutic Potential of Lipocalin 2 in Cancer.” International Journal of Molecular Sciences 21 866 (12): 1–15. https://doi.org/10.3390/ijms21124365. 867 Schalper, Kurt A., Maria E. Rodriguez-Ruiz, Ricardo Diez-Valle, Alvaro López-Janeiro, Angelo 868 Porciuncula, Miguel A. Idoate, Susana Inogés, et al. 2019. “Neoadjuvant Nivolumab Modifies 869 the Tumor Immune Microenvironment in Resectable Glioblastoma.” Nature Medicine 25 (3): 870 470–76. https://doi.org/10.1038/s41591-018-0339-5. 871 Schneider, Caroline A., Wayne S. Rasband, and Kevin W. Eliceiri. 2012. “NIH Image to ImageJ: 25 872 Years of Image Analysis.” Nature Methods. Nat Methods. https://doi.org/10.1038/nmeth.2089. 873 Sharma, Shikhar, Theresa K. Kelly, and Peter A. Jones. 2009. “Epigeneticsnetics in Cancer.” 874 Carcinogenesis. Carcinogenesis. https://doi.org/10.1093/carcin/bgp220.93/carcin/bgp220. 875 Simi, Allison K., Mei Fong Pang, and Celeste M. Nelson.elson. 2018. “Extracellular“Extracellula Matrix Stiffness 876 Exists in a Feedback Loop That Drivesrives Tumor Progression.” In AdvancesAddvancesvance iin Experimental 877 Medicine and Biology, 1092:57–67.092:57–67. Springernger New York LLC. https://dohttps://doi.orhttps://doi.org/10.1007/978-3- 878 319-95294-9_4. 879 Simon, Thomas,homas,mas, and Jonathan S. Bromberg. 2017. “Regulation ofo the Immune System by Laminins.” 880 Trendsnds in ImmunologyImmunol . ElElsevierlseviersevier LtLtd.d. hhttps://doi.org/10.1016/j.it.2017.06.002.ttps://doi.orgttps://doi.or 881 Sivridis, Efthimios,fthimios, Alexandra GiatromanolGiatromanolakGiatromanolaki, and Michael I. Koukourakis. 2004. 882 “‘Stromatogenesis’omatogenesis’matogenesis’ and TumorTumo Progression.” International Journal of Surgical Pathology. 883 Westminsterminsternster PublicaPublicatPublications Inc. https://doi.org/10.1177/106689690401200101. 884 Sun, Carrie InQRQQ., Rebecca S. Arnold,review Chia Ling Hsieh, Julia R. Dorin, Fei Lian, Zhenghong Li, and 885 John A. Petros. 2019. “Discovery and Mechanisms of Host Defense to Oncogenesis: Targeting 886 the β-Defensin-1 Peptide as a Natural Tumor Inhibitor.” Cancer Biology and Therapy 20 (6): 887 774–86. https://doi.org/10.1080/15384047.2018.1564564. 888 Tabas-Madrid, Daniel, Ruben Nogales-Cadenas, and Alberto Pascual-Montano. 2012. “GeneCodis3: 889 A Non-Redundant and Modular Enrichment Analysis Tool for Functional Genomics.” Nucleic 890 Acids Research 40 (W1). https://doi.org/10.1093/nar/gks402. 891 Tadeo, Irene, Ana P. Berbegall, Victoria Castel, Purificación García-Miguel, Robert Callaghan, Sven 892 Påhlman, Samuel Navarro, and Rosa Noguera. 2016. “Extracellular Matrix Composition 893 Defines an Ultra-High-Risk Group of Neuroblastoma within the High-Risk Patient Cohort.” 894 British Journal of Cancer 115 (4): 480–89. https://doi.org/10.1038/bjc.2016.210. 895 Teijeira, Álvaro, Saray Garasa, María Gato, Carlos Alfaro, Itziar Migueliz, Assunta Cirella, Carlos de 896 Andrea, et al. 2020. “CXCR1 and CXCR2 Chemokine Receptor Agonists Produced by Tumors 897 Induce Neutrophil Extracellular Traps That Interfere with Immune Cytotoxicity.” Immunity 52 898 (5): 856-871.e8. https://doi.org/10.1016/j.immuni.2020.03.001. 899 Thomas, Amelia H., Elazer R. Edelman, and Collin M. Stultz. 2007. “Collagen Fragments Modulate 900 Innate Immunity.” Experimental Biology and Medicine 232 (3): 406–11. 901 https://doi.org/10.3181/00379727-232-2320406. 902 Tong, Zhimin, Ajaikumar B. Kunnumakkara, Huamin Wang, Yoichi Matsuo, Parmeswaran 903 Diagaradjane, Kuzhuvelil B. Harikumar, Vijaya Ramachandran, et al. 2008. “Neutrophil 904 Gelatinase-Associated Lipocalin: A Novel Suppressor of Invasion and Angiogenesis in 905 Pancreatic Cancer.” Cancer Research 68 (15): 6100–6108. https://doi.org/10.1158/0008- 906 5472.CAN-08-0540. 907 Valdivia-Silva, Julio, Jaciel Medina-Tamayo, and Eduardo A. Garcia-Zepeda. 2015. “Chemokine-

20 This is a provisional file, not the final typeset article Uterine-myometrial tumor interface

908 Derived Peptides: Novel Antimicrobial and Antineoplasic Agents.” International Journal of 909 Molecular Sciences. MDPI AG. https://doi.org/10.3390/ijms160612958. 910 Vargas, Ana Cristina, Lesley Ann Gray, Christine L. White, Fiona M. Maclean, Peter Grimison, 911 Nima Mesbah Ardakani, Fiona Bonar, et al. 2021. “Genome Wide Methylation Profiling of 912 Selected Matched Soft Tissue Sarcomas Identifies Methylation Changes in Metastatic and 913 Recurrent Disease.” Scientific Reports 11 (1). https://doi.org/10.1038/s41598-020-79648-6. 914 Walke, Ashwini, and Tushar Bhagat. 2017. “To Study the Length, Thickness, Density and Staining 915 Intensity of Reticulin Fibers in Various Grades of Oral Squamous Cell Carcinoma.” Dental, 916 Oral and Craniofacial Research 4 (2): 1–4. https://doi.org/10.15761/DOCR.1000243. 917 Xu, Shuaishuai, Huaxiang Xu, Wenquan Wang, Shuo Li, Hao Li, Tianjiao Li, Wuhu Zhang, Xianjun 918 Yu, and Liang Liu. 2019. “The Role of Collagen in Cancer: From Bench to Bedside.” Journal of 919 Translational Medicine. BioMed Central Ltd. https://doi.org/10.1186/s12967-019-2058-1. 920 Yadav, Vijesh Kumar, Tzong Yi Lee, Justin Bo Kai Hsu, Hsien Da Huang, Wei Chung Vivian Yang, 921 and Tzu Hao Chang. 2020. “Computational Analysis for Identification of the Extracellular 922 Matrix Molecules Involved in Endometrial Cancer Progression.” PLoS ONEO 15 (4). 923 https://doi.org/10.1371/journal.pone.0231594. 924 Yeramian, A., G. Moreno-Bueno, X. Dolcet, L. Catasus,atasus,sus, M. Abal, E. Colas, JJ. RReventos, J. Palacios, 925 J. Prat, and X. Matias-Guiu. 2013. “EndometrialEndometriall CarcinoCarcinoma:ma: Molecular AlAlterations Involved in 926 Tumor Development andd Progression.” OncogeneOncogene. Oncogene.Oncogene 927 https://doi.org/10.1038/onc.2012.76.10.1038/onc.2012.76.2.76. 928 Zhang, Bo,o, Xiao Yun Xing, Jing Li, Rebecca F. Lowdon, Yan ZhoZhZhou, Nan Lin, Baoxue Zhang, et al. 929 2014.4. “Comparative DNA MethylomeMethylomMethylo e Analysis of Endometrial Carcinoma Reveals Complex 930 and Distinctistinct Deregulation of Cancer PromPromotersPr and Enhancers.” BMC Genomics 15 (1). 931 https://doi.org/10.1186/1471-2164-15-868.//doi.org/10.1186/1471doi.org/10.1186/1471 2 932 Zhang, Chenyu,enyu,yu, and DihuDihua Yu. 2019. “Suppressing Immunotherapy by Organ-Specific Tumor 933 Microenvironments:envirInnvi What review Is in the Brain?” Cell and Bioscience. BioMed Central Ltd. 934 https://doi.org/10.1186/s13578-019-0349-0. 935 Zhang, Wei, Jo Hsin Chen, Tianjiao Shan, Irene Aguilera-Barrantes, Li Shu Wang, Tim Hui Ming 936 Huang, Janet S. Rader, Xiugui Sheng, and Yi Wen Huang. 2018. “MiR-137 Is a Tumor 937 Suppressor in Endometrial Cancer and Is Repressed by DNA Hypermethylation.” Laboratory 938 Investigation 98 (11): 1397–1407. https://doi.org/10.1038/s41374-018-0092-x. 939 940

941

942

943 Figure 1: Histomorphometric characterization of reticular fibres in adenocarcinomas (ADC) and 944 leiomyosarcomas (LMS) 5x4 mm region of interest (ROI). (A) Representative image of fibre segmentation in 5x4 mm 945 invasive tumor front (ITF) ROIs in ADC and LMS. Reticular fibres are highlighted in red. (B) Comparison of 946 morphometric parameters obtained after reticular fibre segmentation between ADC and LMS 5x4 mm ITF ROIs. Mean, 947 standard deviation (SD) and p values (* >0.05, **>0.01, ***>0.001) are shown. Total fibres, area of the core and the sum 948 of the total area of the fibres were measured to calculate the density (number of fibres/mm2) and percentage of stained 949 area (% SA). Area in μm2. Length and width in μm. C) Same comparison as in B with more restrictive criteria for case 950 selection.

951 Figure 2: Histomorphometric characterization of reticular fibres in adenocarcinomas (ADC) and 952 leiomyosarcomas (LMS) 1x1 mm regions of interest (ROIs). Representative image of fibre segmentation in 1x1 mm 953 tumor ROIs in ADC and LMS. Comparison of morphometric parameters obtained after reticular fibre segmentation 954 between ADC and LMS in (A) Intratumor 1x1 mm ROIs. (B) 1x1 mm myometrium ROIs and (C) 1x1 mm invasive

Uterine-myometrial tumor interface

955 tumor front (ITF) ROIs in ADC and LMS. Reticular fibres are highlighted in red. Mean, standard deviation (SD) and p 956 values (* >0.05, **>0.01, ***>0.001) are shown. Total fibres, area of the core and the sum of the total area of fibres were 957 measured to calculate de density (number of fibres/mm2) and percentage of stained area (% SA). Area in μm2. Length and 958 width in μm.

959 Table 1: Histomorphometric characterization of reticular fibres between 1x1 mm regions of interest (ROIs) of 960 tumor, myometrium and invasive tumor front (ITF) in within adenocarcinomas (ADC) and leiomyosarcomas 961 (LMS). (A) Comparison of morphometric parameters obtained after reticular fibre segmentation between myometrium 962 and ITF, tumor and ITF, and myometrium and tumor in 1x1 mm ADC ROIs. (B) Comparison of morphometric 963 parameters obtained after reticular fibre segmentation between myometrium and ITF, tumor and ITF, and myometrium 964 and tumor in 1x1 mm LMS ROIs. Mean, standard deviation (SD) and p values (* >0.05, **>0.01, ***>0.001) are shown. 965 Total fibres, area of the core and the sum of the total area of the fibres was measured to calculate the density (number of 966 fibres/mm2) and percentage of stained area (% SA). Area in μm2. Length and width in μm.

myometrium ITF myometrium ITF

p p mean Mean value value Fibers 263,94 246,26 - Fibers 245,38 257,15 - Area Core (mm2) 0,97 0,96 * Area Core (mm2)m2) 00,92,92 0,96 - Density 270,29 307,10 - Densitynsitysity 277,25277,25 268,95 - Fiber Total Area (mm2) 0,05 0,07 - FFiberiberer Total Area ((mm2)mm2) 00,00,05,05 0,06 - % SA 5,97 7,44 * % SASA 5,5,9494 6,47 - Area 227,58 300,55 * ArAreaea 215215,28,28 248,33 - Width 11,71 13,02 * WWidthidth 1111,69 12,34 - Height 43,02 48,35 * HHeighteight 42,57 46,73 - Angle -50,86-50,86 -50,84-50,84 - AngleA -55,72 -48,86 - Roundness 0,1300,,13 00,13,13 * RRoundness 0,14 0,13 - Aspect 4,154,4,115 4,204,4,220 - v Aspect 4,63 4,44 - Perimeter Ratioio 0,670,67 0,66060 eev- Perimeter Ratio 0,68 0,68 - Perimeter 72,8372 83 rre88,24 * Perimeter 67,34 83,88 - Deformity 1193,94 1932,43 ** Deformity 1018,52 1345,71 - Shape Inn24,14 review21,02 - Shape 11,04 16,75 * Vertices 5,54 5,67 * Vertices 5,60 5,63 - Fractal Dimension 1,39 1,39 - Fractal Dimension 1,40 1,39 -

tumor ITF tumor ITF p p mean mean value value Fibers 79,07 246,27 - Fibers 248,74 257,15 - Area Core (mm2) 0,93 0,96 - Area Core (mm2) 0,97 0,96 - Density 83,28 307,1 * Density 257,48 268,95 - Fiber Total Area (mm2) 0,02 0,07 - Fiber Total Area (mm2) 0,05 0,06 - % SA 2,73 7,44 *** % SA 5,84 6,47 - Area 363,23 300,56 - Area 231,91 248,33 - Width 16,21 13,03 *** Width 13,65 12,34 - Height 49,54 48,36 - Height 44,49 46,73 - Angle -50,50 -50,84 - Angle -49,74 -48,86 - Roundness 0,13 0,13 - Roundness 0,14 0,13 - Aspect 3,42 4,2 * Aspect 3,82 4,44 - Perimeter Ratio 0,63 0,66 - Perimeter Ratio 0,68 0,68 - Perimeter 106,88 88,25 - Perimeter 78,03 83,88 - Deformity 2156,74 1932,44 - Deformity 1579,96 1345,71 - Shape 13,10 21,02 - Shape 21,05 16,76 - Vertices 6,041 5,68 *** Vertices 5,57 5,63 - Fractal Dimension 1,45 1,39 - Fractal Dimension 1,40 1,39 -

myometrium tumor myometrium tumor p p mean mean value value Fibers 263,94 79,07 *** Fibers 245,38 248,74 - Area Core (mm2) 0,97 0,93 - Area Core (mm2) 0,92 0,97 -

22 This is a provisional file, not the final typeset article Uterine-myometrial tumor interface

Density 270,28 83,28 *** Density 277,25 257,48 - Fiber Total Area (mm2) 0,05 0,02 *** Fiber Total Area (mm2) 0,05 0,05 - % SA 5,97 2,73 ** % SA 5,94 5,84 - Area 227,58 363,23 ** Area 215,28 231,91 - Width 11,71 16,21 *** Width 11,69 13,65 - Height 43,02 49,53 - Height 42,57 44,49 - Angle -50,86 -50,50 - Angle -55,72 -49,74 - Roundness 0,14 0,13 - Roundness 0,14 0,14 - Aspect 4,15 3,42 - Aspect 4,63 3,82 - Perimeter Ratio 0,66 0,63 * Perimeter Ratio 0,68 0,68 - Perimeter 72,83 106,88 - Perimeter 67,34 78,03 - Deformity 1193,94 2156,73 ** Deformity 1018,52 1579,96 - Shape 24,14 13,10 - Shape 11,04 21,05 * Vertices 5,54 6,04 *** Vertices 5,60 5,57 - Fractal Dimension 1,39 1,44 ** Fractal Dimension 1,40 1,40 - 967

968 Figure 3: Immune infiltrate characterization at the invasive tumor front (ITF).ITF). (A)( Representative multiplex 969 immunofluorescence images of approximately 5x4 mm regions of interest (ROIs)ROIs) of an adenoadenocarcinoma (ADC) (4x4) and 970 a leiomyosarcoma (LMS) (5x3 mm). Colour code as follows: DAPI (blue), CD20 (yellow)(yellow),, CD68 (green), CD8 (red), 971 CD3 (orange), cytokeratin (cyan). (B) The mean of the densitieses (number(number ofof cells/mmcells/mm2) for eeaceach cell marker is shown for 972 ADC and LMS as well as the mean for thee total cells and thee total immune infiltrate. P vvalues (* >0.05, **>0.01, 973 ***>0.001) for statistical comparisonon between the meansns of ADC and LMS are displayed.

974 Figure 4: Immunemune infiltrate characterizationcterization in 1x1 mm regions of interestinteres (ROIs) in adenocarcinomas (ADC) and 975 leiomyosarcomasarcomasrcomas (LMS). (A) Representativeepresentative multiplex images of 1x1 mm ROIs of intra tumor, myometrium and 976 invasive tumormororr front (ITF) in ADC and LMS. Colour code asa follows: DAPI (blue), CD20 (yellow), CD68 (green), CD8 977 (red), CD3 (orange),range), cytokeratin (cyan). (B) The meanm of the densities (number of cells/mm2) for each cell marker in each 978 region is shownownwn for ADC and LMS as well as the mean for the total cells and the total immune infiltrate. P values (* 979 >0.05, **>0.01,01,, ***>0.001)***>0.001) forf statistical comparison between the means of intratumor, myometrium and ITF in ADC 980 and LMS are displayed.displaydisplaIn review 981 Figure 5. Transcriptomic profiling of invasive tumor front (ITF) in adenocarcinomas (ADC) and leiomyosarcomas 982 (LMS) samples. (A) RNA-seq expression levels (as counts per million, log2 scale) of cellular markers for macrophages 983 (CD68), T cells (CD3 and CD8) and B lymphocytes (CD20). (B) Parametric analysis of gene set enrichment using Gene 984 Ontology biological processes dataset. Red and blue indicate activated and suppressed pathways, respectively. (C) RNA- 985 seq expression levels (as counts per million, log2 scale) of genes related to cell adhesion and epithelial cell development. 986 Increased gene expression in ADC vs LMS was significant in all the markers. (D) Parametric analysis of gene set 987 enrichment using curated reactome dataset. Red and blue indicates activated and suppressed pathways, respectively.

988 Figure 6: Genome-wide DNA methylation analysis in invasive tumor front (ITF) of primary adenocarcinomas 989 (ADC) and leiomyosarcomas (LMS). (A) Schematic flow chart used to identify differentially methylated CpGs 990 (DMCpGs) in primary ITF between ADC and LMS. (B) Scatter plot representing mean normalized levels of DNA 991 methylation (β-values) in primary ITF of ADC and LMS. (C) Genomic distribution of 23,296 differentially methylated 992 CpGs (DMCpGs) in primary ITF of ADC and LMS, in relation to their respective location regarding CpG context and 993 gene region.

994 Figure 7. DNA methylation signature of promoter CGIs in invasive tumor front (ITF) of primary 995 adenocarcinomas (ADC) and leiomyosarcomas (LMS). (A) Supervised hierarchical clustering of the most variable 996 CpGs (2,971 CpGs and 1,253 genes; FDR<5%) from island and promoter regions between the primary ITF of ADC and 997 LMS. (B) Gene ontology (GO) analysis of the biological process categories for the 1,253 differentially methylated genes 998 at CpG island and promoters between the primary ITF of ADC and LMS.

999 Figure 8. Association between gene expression and DNA methylation profiling of promoter CGIs in invasive 1000 tumor front (ITF) of primary adenocarcinomas (ADC) and leiomyosarcomas (LMS). (A) Venn diagrams showing 1001 the differentially methylated and differentially expressed genes between the primary ITF of ADC and LMS. The name of 1002 the 9 hypermethylated and downregulated genes in ADC respect to LMS is indicated in red, and the name of the 11 1003 hypomethylated and upregulated genes in ADC respect to LMS is represented in green. (B) Supervised hierarchical

Uterine-myometrial tumor interface

1004 clustering of the CpGs from promoter CGIs (corresponding to 20 genes) whose methylation status correlated with 1005 significant changes in gene expression. β-values represent the normalized levels of DNA methylation.

1006 Supplementary Table 1: Case selection and regions of interest (ROIs) used in each characterization. Case 1007 identification is shown by tumor ID. CE- belong to adenocarcinomas (ADC) and LMS- to leiomyosarcomas (LMS). The 1008 percentage of tumor/myometrium tissues for genomic and epigenetic profiling was evaluated in H&E stained samples. 1009 The total numbers of ROIs and tumors used for all the studies are shown. Invasive tumor front (ITF).

1010 Supplementary Figure 1: In-depth characterization of the invasive tumor front (ITF) in adenocarcinomas (ADC) 1011 and leiomyosarcomas (LMS). A tumor biopsy from an adenocarcinoma (ADC) is shown. (A) Schematic pipeline for 1012 tumor ITF area selection for histomorphometric analysis, immune infiltrate, transcriptomics and methylation analysis. (B) 1013 Selection from a 5x4 mm region of interest (ROI) of 1x1 mm ROIs of invasive tumor front (ITF), myometrium and 1014 tumor. The same criteria was applied for ADC and LMS.

1015 Supplementary Table 2: Transcript expression profile of the invasive tumor front (ITF) in adenocarcinomas 1016 (ADC) and leiomyosarcomas (LMS).

1017 Supplementary Table 3: DNA methylation levels at promoter CGIs of 20 differentiallydiff methylated and 1018 differentially expressed genes associated to primary ITF in ADC and LMS.

1019

1020 1021 1022 In review

24 This is a provisional file, not the final typeset article Figure 1.TIF Figure 2.JPEG Figure 3.TIF Figure 4.TIF Figure 5.TIF Figure 6.JPEG Figure 7.JPEG Figure 8.JPEG