{"paper_id":"9565209c-371a-424c-9b37-d6cf05faf0a8","body_text":"Endometriosis is an estrogen-dependent inflammatory disease that results in\npelvic pain and/or infertility. It affects approximately 10% of reproductive age\nwomen ( 1 – 4 ) and is characterized by the presence of endometrial-like tissue\noutside the uterus where it elicits an inflammatory response ( 1 , 3 , 4 ). The eutopic endometrium of women with endometriosis\nhas been widely studied with regard to dysfunctionality of steroid hormone response,\nstem cell populations, and recruitment of immune populations for immune tolerance\nand overall tissue homeostasis and pregnancy success versus women without disease\n( 1 – 3 ). However, there are scant data about the function and phenotypes of\neutopic endometrial immune cells in women with and without endometriosis. As the\nendometrial immune niche involves multiple cell types with varying degrees of\nactivation and communications among immune and non-immune cells that dictate\nfunctionality of the tissue, characterizing the endometrial immune niche is of great\nrelevance to understanding endometrial function and dysfunction.\nUterine natural killer cells (uNK) secrete angiogenic factors that\ncontribute to the maturation of blood vessels having a role in embryo implantation\nand successful pregnancy ( 6 , 7 ). In healthy endometrium, their cytotoxic activity\ndiminishes during the secretory phase of the menstrual cycle which allows embryo\nimplantation ( 8 – 10 ). However, in infertile endometriosis patients, uNK\nhave high cytotoxicity in eutopic endometrium that could lead to an inhospitable\nenvironment for embryo implantation ( 11 ).\nOther immune cells, such as T regulatory cells (Treg), have been also described to\nbehave differently in the endometrium of women with endometriosis. In healthy\nendometrium, they increase in the proliferative and decrease during the secretory\nphase, with the latter creating an immune-tolerant environment allowing embryo\nimplantation. However, in infertile women with endometriosis, Treg are increased in\nthe peri-implantation endometrium, leading to an implantation failure ( 12 ).\nEndometriosis has been referred to as “a disease of the\nmacrophage” ( 13 ), based mainly on a\nreplete literature on the roles and functionality of this cell type in peritoneal\nfluid of women with disease and in establishment of endometriosis lesions and\nassociated processes of angiogenesis and fibrosis. Mφ are key effector cells\nin both innate and humoral immunity as they phagocytose pathogens, act as antigen\npresenting cells, and have a role in tissue regeneration, angiogenesis and wound\nhealing ( 14 ). In eutopic endometrium of women\nwithout endometriosis, their numbers vary throughout the menstrual cycle, increasing\nin the secretory and menstrual phases ( 15 ).\nThis increase may be attributed to their phagocytic properties and role in clearing\ncell debris and apoptotic cells during endometrial shedding ( 16 ). Cycle variation among endometrial Mφ does not\noccur in women with endometriosis ( 17 ),\nsuggesting that survival of shed and refluxed endometrial cells may be enhanced,\nenabling them to migrate to the peritoneal cavity and establish disease. Mφ\nare classified as either “classically activated” Mφ\n(Mφ1) or “alternatively activated” Mφ (Mφ2)\n( 14 ) and, depending on the\nmicroenvironment, they can switch from one type to the other ( 18 ). Mφ1 have a role in pro-inflammatory\nresponses; whereas, Mφ2 are involved in angiogenesis, anti-inflammatory\nreactions, and tissue repair ( 14 , 19 ). In healthy endometrium, the predominant\npopulation is Mφ2 ( 19 , 20 ), suggesting that the normal environment is\nanti-inflammatory. Taken together, most of the studies in eutopic endometrium have\nfocused on the number of immune cells in this tissue and how they fluctuate\nthroughout the cycle, but little is known about their functionality in women with\nendometriosis.\nHerein, we designed a novel flow cytometry panel to isolate Mφ1,\nMφ2, Treg and uNK from eutopic endometrium of women with endometriosis and\nthose with no evidence of disease. RNA High-Sequencing (RNA-Seq) was used to\nelucidate Mφ1 and Mφ2 phenotypes and possible functions in disease.\nOverall, the data support a phenotypic switch of the common anti-inflammatory\nMφ2 to the pro-inflammatory Mφ1 phenotype and a more exaggerated\npro-inflammatory phenotype of the Mφ1 population in women with\nendometriosis.\n\nEleven endometrial biopsies in the secretory phase were collected: 6\nfrom women with endometriosis (stage I-IV) and 5 from women with no evidence of\nendometriosis at the time of the surgery for benign gynecologic disorders. The\nmean age was 37 and 42 (23-49) years old, respectively. In order to evaluate if\nthe age could confound the results, a non-parametric t-test with a subsequent\nMann-Whitney test (p<0.05) was performed and no significant differences\nbetween groups were found (p=0.2641). Patients had not used hormonal therapy for\nat least three months prior to the study. Endometrial samples were obtained\nthrough the University of California San Francisco (UCSF) NIH Human Endometrial\nTissue and DNA Bank under approval of the UCSF Committee on Human Research\n(IRB#10-02786), and written informed consent was obtained from all participants.\nEndometrial tissue was digested as previously described ( 21 ). Briefly, it was minced mechanically and\nincubated for one hour at 37ºC in digestion media, which contained\ncollagenase type I and hyaluronidase ( 21 ). Subsequently, the single cell suspension was filtered using a\n40μm mesh to discard cell clumps, and single cells were cryopreserved in\nliquid nitrogen until use.\nA cytometry panel of 10 conjugated antibodies able to separate the\nimmune populations of interest (Mφ1, Mφ2, Treg and uNK) was\ndesigned. Specific membrane markers of resident and blood infiltrating immune\ncells were included to avoid sorting cells derived from the peripheral\ncirculation. The brightest colors were used for the markers with the lowest\nantigen density. Minimum overlapping of 11 colors (10 antibodies plus the\nlive/dead dye) was achieved. The markers and lasers used for each population are\nin  Table 1  and the gating strategy is in\n Figure 1.A . First, the cells were gated\nwith CD45 (leukocyte marker), conjugated with brilliant violet 605 (BV605). In\nthe case of Mφ, usually this population is a resident tissue population,\nthus no specific tissue markers were used. It is challenging to differentiate\nbetween the Mφ1 and Mφ2 subpopulations, as they have some common\nmarkers and have the ability to polarize from one type to the other. Our\nstrategy was as follows: for both types, CD14 marker conjugated with\nphycoerythrin (PE) was used. For Mφ2, CD163, a specific marker for this\npopulation, was used conjugated with PE-cyanin7 (PE-Cy7). As there are no\nspecific markers for Mφ1 and it was suspected that the concentration of\nactivated Mφ1 would be low in the endometrium (as Mφ2 are higher\nthan Mφ1 in normal endometrium ( 19 )), CD80 (activation marker) conjugated with peridinin chlorophyll\nprotein complex-cyanin 5.5. (PerCP-Cy5.5) antibody, a bright dye, was used. With\nregard to Treg, they express CD3 and CD4 markers. The most accepted specific\nmarker for Treg is Foxp3. This, is an intracellular marker and thus could not be\nused for sorting. However, they also express CD25. To be able to discern between\ntissue Treg and Treg deriving from circulation, CD69, an activation marker for\nTreg that is expressed in tissue, was included. Thus,\nCD3 + CD4 + CD25 + CD69 +  cells (tissue\nTreg) were isolated. The CD3 antibody was conjugated with ultraviolet B 737\n(BUV737), as was the CD4 antibody with ultraviolet B 395 (BUV395). CD25 has a\nlow antigen density, therefore, brilliant blue 515 (BB515) was used, which is\none of the brightest dyes. Then, CD69, a tissue activation Treg marker, was\nconjugated with allophycocyanin-cyanin7 (APC-Cy7), assuring that only resident\nTreg were isolated. For uNK cells, which are CD56 + , as it is known\nthat blood NK are CD16 + ; whereas, uNK are CD16 Low/− \n( 22 – 25 ), CD56 + CD16 −  were\ncollected. For CD56, brilliant violet 421 (BV421), was used and in the case of\nCD16, allophycocyanin (APC) conjugated antibody was used. Finally, to separate\nbetween live and dead cells, Aqua dye was used (sources of all antibodies are\nlisted in  Table 1 ).\nEndometrial samples were thawed at 37ºC. After centrifugation at\n1,300rpm for five minutes, the supernatant was discarded and the pellet was\nwashed with 1X PBS. After another centrifugation, cells were resuspended with 1X\nPBS + 5% bovine serum albumin (BSA) and incubated at room temperature for 30\nminutes. A minimum of 100,000 unlabeled cells were separated as a negative\ncontrol. Ten conjugated antibodies were used to label the samples ( Table 1 ). 1μl of antibody per\nmillion cells was used in all cases except for CD45 and CD4, where 2μl\nper million cells were needed for an optimal cell labelling. A solution of all\nFluorochromes Minus One (FMO) was prepared for each antibody to assess any\noverlap among the channels in the FACS instrument. After one hour of incubation\nat 4ºC in 1X PBS + 3% BSA and in the dark, cells were washed with 1X PBS\nand centrifuged for 5 minutes at 1,300 rpm. The pellet was resuspended with\n500μl of 1X PBS and labelled with 1μl of LIVE/DEAD™ Fixable\nAqua Dead cell labelling dye (ThermoFisher, Waltham, MA). On the other hand,\nUltraComp eBeads compensation magnetic beads (ThermoFisher, Waltham, MA, USA)\nwere labelled with each of the 10 antibodies following the manufacturer’s\ninstructions, to allow the correction of the spectral overlap between\nfluorochromes. Using the gating strategy ( Figure\n1.A ), each population was sorted in the FACS Aria Jabba the Hutt (BD\nBiosciences, East Rutherford, NJ, USA) instrument and collected in 1X PBS. Flow\ncytometry analysis of the sorted cells was performed using FlowJo.v10 software\n(FlowJo LLC, Ashland, OR, USA), and statistical analyses (Mann-Whitney test,\np-value<0.05) were conducted using GraphPad software (GraphPad Software\nInc, San Diego, CA, USA).\nAs low yields of cells were obtained after FACS ( Figure 1.B ), RNeasy micro kit (Qiagen, Hilden,\nGermany) was used to isolate RNA from Mφ populations (note yields from\nuNK and Treg cells were too low, and were not used for RNA-Seq, see  Methods ). From the 22-sorted Mφ\nsamples (Mφ1 and Mφ2 populations of each of the 11 endometrial\nsamples), RNA was extracted following the manufacturer’s instructions to\nperform the total RNA-seq library prep from samples containing >900 cells\n(9 samples). The library preparation from the remaining samples, which contained\nat least 20 cells, was performed directly from cells in 1X PBS. RNA was eluted\nin 10μl of RNase free-water, and the quality (RNA integrity numbers\n(RIN)) and concentration were measured using a Tapestation4200 System (Agilent,\nSanta Clara, CA, USA).\nSMART-Seq™ v4 Ultra™ low input RNA kit for sequencing\n(Clontech, Mountain View, CA, USA) was used to perform the RNA-seq library\npreparation. It allows RNA-Seq to be performed with very low concentrations of\nRNA or to use whole cells to preserve sample integrity. In total, library\npreparations for 22 samples (10 from control and 12 from endometriosis) were\nperformed. The quality of fastq files was tested using the FastQC (v0.11.5)\n( 26 ) and the Qualimap (rnaseq module\n– v2.2.1) software ( 27 ). Reads\nwere aligned with the STAR mapper (v2.5.2a) ( 28 ) to release 88 of the Homo sapiens ENSEMBL version of the genome\n(GRCh38/hg38 assembly) ( 29 ). A raw count\nof reads per gene was also obtained with STAR ( 28 ) . In order to overcome the heterogeneity between samples, first,\nsamples were removed from the analysis if they had <5 million uniquely\nmapped reads, and the remaining samples were downsampled to 30 million mapped\nreads when needed. The data have been deposited in NCBI GEO database (accession\nnumber  GSE130435 ). The R/Bioconductor package DESeq2 (v1.20.0) ( 30 – 32 )\nwas used to assess differential expression between experimental groups (Wald\nstatistical test + false discovery rate (FDR) correction). Statistically\nsignificant differentially expressed genes (DEG) were considered when\nFDR<0.05 and log fold change>2 (LogFC>2). Different\ncomparisons performed using Mφ populations are shown in  Table 2 . Biological significance analyses were\nconducted using Ingenuity Pathway Analyses (IPA) software (Ingenuity®\nSystems, Redwood City, CA, USA), and significant molecular functions were\nestablished with an activation Z-score> |2.00|.\n\nAfter FACS, low cell numbers were obtained ( Figure 1.B ) that subsequently guided further analyses.\nCytometry analyses from all the immune populations showed that CD45 + \ncells corresponded to an average of 6.8% of the total sample, in agreement with\nother studies wherein leukocytes comprise 10-20% of total endometrial cells\n( 33 – 38 ). No significant differences in CD45 + \ncells were observed between control and endometriosis groups ( Figure 1.C ). Statistical analyses comparing percentage\nof each sub-population between controls and endometriosis were performed, and no\nsignificant differences were observed except for Mφ1, which was\nsignificantly higher in the endometriosis group (p=0.0087) ( Figure 1.D ). Because resident tissue markers were\nincluded in the cytometry panel, contamination of immune populations from the\nperipheral circulation could also be calculated. In both the control and\nendometriosis groups, uNK (CD16 − ) were significantly higher\ncompared to blood NK (CD16 + ) ( Figure\n1.E ), demonstrating that there was almost no contamination with blood\nNK cells. The percentage of Treg coming from blood (CD69) was higher than tissue\nTreg (CD69 + ), although it was not significant ( Figure 1.E ). Due to the low number of uNK and Treg\ncells obtained, RNA-Seq was only performed in the Mφ populations, that\nhad significantly greater numbers of cells. Thus, from the 44 original\nFACS-sorted immune populations, 22 samples (Mφ populations) from\nendometriosis and control were used for the transcriptome study.\nRNA concentrations extracted from Mφ ranged between\n5-45ng/μl. After RNA-Seq and quality controls, we excluded any FastQ\nsequences for which the number of reads did not reach our threshold of five\nmillion reads/sequence. Thus, the populations analyzed were: 5 Mφ1\nendometriosis, 3 Mφ1 control, 6 Mφ2 endometriosis and 4 Mφ2\ncontrol. After statistical analysis, DEG (FDR<0.05 and LogFC≥2)\nwere found in all comparisons ( Table 2 ;\n Supplemental Table\n1 ).\nBiological significance of the DEG analyses revealed significant\nmolecular functions, relevant molecules secreted by Mφ1 and Mφ2,\nactivation/inhibition of upstream regulators and de-regulated networks in each\ncomparison (activation Z-score≥2.00) ( Table 3 ). The 25 top de-regulated networks are in  Supplemental Table 2 . Increase in\ncell-cell contact was observed along with repression of RNA molecular functions,\nwhen comparing Mφ1 endometriosis versus Mφ1 control\n( Comparison 4, \n Table 2 ). Increased cell-cell contact is\nconsistent with, e.g., increased adhesion to bacteria to accomplish bacterial\nengulfment. Top de-regulated networks showed overexpression of cellular\ndevelopment, growth and proliferation, and overexpression of immune\nresponse-related networks, such as infectious disease and antimicrobial and\ninflammatory responses ( Table 3 ). These\ndata indicate that Mφ1 in endometriosis have a more extensive\npro-inflammatory phenotype than Mφ1 in the control group.\nIn contrast, molecular functions upregulated in Mφ2 in\nendometriosis ( Comparison 5, \n Table 2 ) included an accumulation of\nCa 2+ , increase in carbohydrate transport, and internalization of\nbacteria ( Table 3 ). When comparing\nMφ1 of women with versus without endometriosis ( comparison\n4, \n Table 2 ), the upstream regulator\nTNFα was predicted to be increased ( Table\n3 ). Increased internalization of bacteria is consistent with\nphagocytic properties of the pro-inflammatory Mφ1 phenotype. The top\nnetworks in endometrial Mφ2 from women with endometriosis included\nderegulation of connective tissue disorders, endocrine system development and\nfunction, lipid metabolism, inflammatory disease/response, and drug metabolism\n( Table 3 ). These data overall\ndemonstrate that Mφ2 in eutopic endometrium of women with endometriosis\nhave a pro-inflammatory phenotype, compared to Mφ2 in control women.\n\nIn the current study, we developed a cytometry panel that allowed for\nseparating circulating immune cells and tissue resident cells and different immune\ncell types within human endometrium. Thus, the analyzed immune populations were\npurely tissue-activated resident cells devoid of contamination by circulating immune\ncells. One goal was to develop and optimize this panel for the current study.\nHowever, it will also have value for other researchers aiming to separate these\ntissue-specific populations, since it is a challenging panel to design due to the\nmultiple colors used and the possible overlap between channels.\nAfter cytometry analyses, where Mφ1 were found to be significantly\nhigher in endometriosis, Mφ were studied in more detail by transcriptomic\nanalyses. To our knowledge this is the first RNA-Seq dataset of Mφ in eutopic\nendometrium of women with endometriosis. Abnormal distribution of Mφ within\neutopic endometrium of women with disease could contribute to the aberrant\ndistribution of immune cells in the pelvic cavity and the abnormal development and\ngene expression of this tissue. While Mφ maintain organ homeostasis and\nfacilitate host defense and wound healing, they also underlie the pathogenesis of\nmany chronic inflammatory diseases ( 39 ).\nThe increased de-regulated molecular functions and networks in Mφ1 in\nendometrium of women with endometriosis indicate these cells have a more\npro-inflammatory phenotype than Mφ1 in the control group. In addition, a\nsignificantly higher number of sorted Mφ1 was observed in endometriosis\npatients ( Figure 1.D ), confirming a previous\nreport ( 40 ). Moreover, these results\nsuggesting that eutopic endometrium of women with endometriosis is more\npro-inflammatory than control endometrium, is consistent with findings from other\ngroups ( 17 , 41 ).\nAn unexpected finding herein was the pro-inflammatory phenotype exhibited by\nendometrial Mφ2 from women with endometriosis. Mφ2 in other tissues\ngenerally display an anti-inflammatory phenotype ( 39 ), and, importantly, Mφ are phenotypically plastic with regard\nto their polarization state depending on their microenvironment ( 42 ). Moreover, Mφ1 and Mφ2 gene expression\nsignatures often overlap, and the resultant phenotype depends on the tissue\nmicroenvironment ( 40 ). Thus, endometrial\nMφ2 of women with endometriosis could undergo polarization  in\nsitu  to Mφ1, adopting a pro-inflammatory phenotype, due to an\naltered environment. The paradigm of different subpopulations of Mφ is\ncontroversial in the immunology literature. Specifically, it is unclear whether\nthere are unique Mφ populations (as Mφ1, Mφ2) or if Mφ\ncomprise a unique population that alters its phenotype depending on environmental\ncues. Herein, we have referred to Mφ as two different subpopulations\n(Mφ1 and Mφ2), although the dynamics and mechanisms driving\npro-inflammatory and anti-inflammatory Mφ functional phenotypes remain to be\ndetermined.\nNotably, tumors take advantage of macrophage plasticity. For example, in the\nearly phases of cancer, high production of Mφ1 inflammatory mediators\nactivates the adaptive immune response capable of eliminating nascent neoplastic\ncells, and also support neoplastic transformation ( 40 ). In contrast, once the tumor is stablished, the main population of\nMφ is Mφ2, producing an anti-inflammatory environment, which allows\ntumor growth. Endometriosis it is not a malignancy, however, it shares some\ncharacteristics with cancers. In endometriotic lesions and peritoneal fluid of women\nwith endometriosis, e.g., Mφ2 are increased ( 43 ), indicating that, as in cancer, an anti-inflammatory environment\nprevails favoring development and growth of the endometriotic lesions. In addition,\nthat Mφ2 have a role in angiogenesis further supports this paradigm. Finally,\nMφ2 are also involved in nerve growth, suggesting they may also have a role\nin endometriosis-related pain ( 44 ).\nThe initial pro-inflammatory phenotype of Mφ in cancer increases NFKB\nand downstream events and increases transcription of pro-inflammatory cytokines such\nas TNFα, IL12, IL23, IL1β, IL6, and ROS. In the current study, the\nNFKB pathway was activated in Mφ1 of endometriosis, which does not occur in\nMφ1 of control women ( Table 3 ).\nIndeed, it has been described that the NFKB pathway is de-regulated in the eutopic\nendometrium of women with endometriosis ( 45 ),\nwhich also indicates that the microenvironment in endometrium of women with disease\nis more pro-inflammatory than heathy tissue.\nNotably, an increase of transport of carbohydrates was observed in\nMφ2 of women with endometriosis. It is known that glycolysis is high in\nMφ1 and is decreased in Mφ2 and Mφ polarization may derive from\na reprogramming of glucose metabolism ( 46 ).\nSeveral studies have suggested that altering nutrient availability or blocking\nspecific metabolic pathways skews the Mφ phenotype and alters their effector\nfunctions in chronic inflammatory diseases ( 47 ). In this regard, Mφ metabolism modulation could open a new\ntherapeutic window for treating inflammatory diseases including endometriosis.\nFinally, the upstream regulator TNFα was increased in IPA analysis\nwhen comparing Mφ1 of women with versus without endometriosis\n( comparison 4, \n Table 2 ), as well as increased\nCa 2+  accumulation was activated in Mφ2 ( Table 3 ). It has been noted that a transient increase of\nCa 2+  plays a role in the expression of TNFα by Mφ1\n( 48 ). Intracellular Ca 2+ \noscillations are likely to induce permanent changes in Mφ physiology, and a\nsupra-physiologic elevation of Ca 2+  in mitochondria can be cytotoxic and\ninduce apoptosis in the long term ( 48 ).\nWhether TNFα-mediated events play a role in Mφ function awaits further\nstudies.\nOver the past decade, high-throughput sequencing techniques have challenged\nthe dogma of the sterility of the uterine endometrium ( 49 – 54 ), and\nin particular an altered endometrial microbiome in women with endometriosis has been\nproposed ( 55 ). In addition, the endometrial\nmicrobiome also correlates with IVF outcomes ( 52 ), although whether this occurs in women with endometriosis awaits\nfurther study. However, treatment with antibiotics resulted in reduced numbers of\nendometriosis lesions in a mouse model, with concomitant alteration of the gut\nmicrobiome ( 56 ), although the endometrial\nmicrobiome was not reported in this study ( 54 ). Interestingly, a recent systematic review supports the use of\nantibiotics prior to oocyte retrieval in patients with endometriosis, among other\ngynecologic disorders ( 57 ). The presence of\npathogenic, non-commensal bacteria in the endometrium may induce an altered immune\ncell profile and activation (increased numbers and activation of Mφ1 and\nactivation of Mφ2) that could impact the production of cytokines by immune\nresident cells that adversely affect embryo implantation ( 58 ). In addition to effects on reproductive outcomes, the\nobserved greater pro-inflammatory endometrial environment herein could be related to\nthe pathophysiology of the disease. While attractive, we are aware that the sample\nsize of the study is small. Therefore, these results should be taken with caution.\nFinally, whether the pro-inflammatory phenotype of the Mφ2 population\nreported herein is in response to commensal bacteria or pathogens, or if Mφ\npopulations are implicated in reproductive outcomes, is not clear. However, it is\nanticipated that this important area of research could have profound implications\nclinically and diagnostically.\n\nOverall, the results of the current study lead to the conclusion that both\nMφ1 and Mφ2 in eutopic endometrium of women with endometriosis display\na higher pro-inflammatory phenotype compared to controls without disease.\nEndometrial Mφ2 appear to be predisposed to Mφ1 polarization in women\nwith endometriosis, thus increasing their inflammatory phenotype. These findings\nsuggest that eutopic endometrium has different Mφ gene signatures depending\non the presence or absence of disease and that the endometrial environment of women\nwith endometriosis is more pro-inflammatory than control endometrium. Whether\nsubtypes of the disease are associated with different subsets of immune and whether\nthe macrophage pro-inflammatory status is related to bacteria in the endometrium of\nwomen with disease are yet to be determined. Finally, the results herein may have\nimplications regarding the impact of the macrophage phenotypes on reproductive\noutcomes and possible novel therapeutics for microbiome-related symptoms and\nresponse for fertility and pain in women with endometriosis.","source_license":"CC0","license_restricted":false}