Macrophages can either inhibit or enhance endometriosis depending on their origin

Macrophages are intimately involved in the pathophysiology of endometriosis, a chronic inflammatory disorder characterized by the growth of endometrial-like tissue (lesions) outside the uterus. By combining genetic and pharmacological macrophage depletion strategies we determined the ontogeny and function of macrophages in a mouse model of induced endometriosis. We demonstrate that lesion-resident macrophages are derived from eutopic endometrial tissue, infiltrating large peritoneal macrophages (LpM) and monocytes. Endometriosis triggers continuous recruitment of monocytes that contribute to LpM and small peritoneal macrophage (SpM) pools. Depletion of eutopic endometrial macrophages results in smaller endometriosis lesions. Constitutive inhibition of monocyte recruitment significantly reduces peritoneal macrophage populations and increases the number of lesions. We propose a putative model whereby endometrial macrophages and embryo-derived LpM are ‘pro-endometriosis’, whilst monocyte-derived LpM are ‘anti-endometriosis’. These observations highlight the importance of monocyte-derived macrophages in limiting disease progression.


Introduction
Macrophages are exceptionally diverse cells present in all tissues of the body that perform functions vital for immunity, development, tissue homeostasis and repair following 35 injury. They modify their role depending on signals received from their local microenvironment and accordingly exhibit high degrees of transcriptional and phenotypic heterogeneity and tissue-specific function 1, 2 . Macrophages differ in their ontogeny. Whilst early studies suggested macrophages were continually replaced by circulating blood monocytes, more recent lineage-tracing experiments demonstrated that most tissue-40 resident macrophages (exceptions include gut, dermis and heart) are derived from embryonic precursors that seed tissues prior to birth and are maintained by self-renewal or longevity 3,4,5 . Tissue-resident and monocyte-derived macrophages play distinct roles both in health and disease 6 . Usually, tissue-resident macrophages play a homeostatic role, whilst monocyte-derived macrophages that are recruited to tissues during inflammation secrete 45 pro-inflammatory cytokines and stimulate the immune system 6 . Thus, in pathological situations, macrophages are a heterogenous population. For example, hepatic macrophages (Kupffer cells) that usually maintain homeostasis become activated during acute injury and recruit monocytes that differentiate into disease-modified inflammatory macrophages that promote the progression of liver injury 7 . In pancreatic cancer, both tissue-resident and 50 monocyte-derived macrophages populate the tumor and increase in density as the cancer progresses. The two populations are transcriptionally diverse and depletion studies revealed that tissue-resident macrophages are responsible for driving tumor progression whereas monocyte-derived macrophages act as innocent bystanders 8 . This highlights the paradigm that under disease-modified conditions tissue-resident macrophages can become adapted 55 such that they promote disease.
The peritoneal cavity hosts two main macrophage populations: a predominant population expressing high levels of EGF-like module-containing mucin-like hormone receptor-like 1 (EMR1 / F4/80) and low levels of Major histocompatibility class II (MHCII) known as large peritoneal macrophages (LpM), and a less abundant population that are 60 F4/80 lo , MHCII hi (small peritoneal macrophages; SpM). LpM are considered to be tissue macrophages and are largely embryonically derived, however it is now understood that they are gradually replaced by monocytes over time in a sexually dimorphic manner, that occurs much quicker in males 9, 10 . The SpM population are considered pro-inflammatory and are a heterogeneous population 9 . Inflammatory challenge in the peritoneal cavity can result in an 65 increase in SpM and a loss in LpM (the so-called macrophage disappearance reaction (MDR)), with the exception of helminth infections which are characterized by elevated levels of Interleukin 4 (IL-4) and a type 2 inflammatory response 11 12, 13, 14 .
Endometriosis is a chronic inflammatory condition where tissue similar to the endometrium, grows ectopically usually in the peritoneal cavity as 'lesions' 15 . The condition 70 impacts an estimated 190 million women worldwide and is associated with debilitating pelvic pain and infertility 16 . Currently, therapeutic options are very limited, with the goldstandard treatments being surgical removal of lesions or suppression of ovarian hormones.
Surgery is associated with high recurrence rates and ovarian suppression is contraceptive and often has unwanted side effects. A high abundance of macrophages is reported both in 75 the peritoneal cavity and in lesions of women with endometriosis 17 . It is clear that macrophages are intrinsically linked with the pathophysiology of endometriosis where they enhance establishment, proliferation and vascularisation of lesions 18,19 . They are also critical in promoting innervation of lesions and concomitant sensitization of nerve fibres, thus contributing to pain in the condition 20,21 . Evidence from a syngeneic mouse model of 80 induced endometriosis indicates that donor endometrial macrophages as well as hostderived macrophages can be identified in endometriosis lesions 22 . However, the exact origins of the host-derived macrophages and specific functions of the different populations remains to be determined.
Since macrophages play such a key role in many aspects of the pathophysiology of 85 endometriosis they represent an attractive therapeutic target. However, the development of a viable immune-therapy targeting 'disease-promoting' macrophages or enhancing the function of 'protective' macrophages requires a comprehensive understanding of the origin and function of lesion-resident and associated peritoneal macrophages. In this study we have characterized the origin of endometriosis lesion-resident macrophages and examined 90 the dynamics of peritoneal cavity macrophage populations. Finally, we have used a combination of transgenic and pharmacological approaches to selectively deplete different populations of macrophages to assess their impact on development of endometriosis lesions.

Discussion
The pathophysiology of endometriosis remains enigmatic 15 . Although immune cell dysfunction is intrinsically linked with the disorder, our understanding of macrophage origins and respective function remains limited compared with other diseases, such as cancer 24 . Macrophages are critical for endometriotic lesion establishment and survival, as 215 indicated by early depletion experiments 18 . They play a key role in promoting growth and vascularization of ectopic tissue 19 and in promoting nerve growth, activation of nerves and pain generation in endometriosis 20,21 . As the ontogeny of macrophages in diseased tissue is a key determinant of how they respond and contribute to pathogenesis, it is necessary to understand how macrophages derived from different sources impact lesion development. 220 This will provide a greater appreciation of how 'disease-promoting' or 'protective' macrophages could be targeted in the future as a potential therapy. The use of a syngeneic mouse model of induced endometriosis 22 has allowed intricate exploration of macrophage ontogeny and functional consequences of depleting specific populations that is impossible to perform in women. In this study, we have used a model that aims to recapitulate the 225 process of 'retrograde menstruation' by generating 'menses'-like endometrium in donor mice and injecting the tissue into the peritoneal cavity of recipient mice. We have demonstrated that lesion-resident macrophages are derived from the donor endometrium, from infiltrating LpMs and from monocytes. We used different depletion strategies to explore the impact of macrophages from different origins on lesion development. The most 230 striking result was revealed by constitutively limiting monocyte recruitment (using both macrophages that are GFP-, F4/80+. The cells were localized to areas of breakdown, repair and remodelling, respectively 25 . The 'menses-like' endometrium that we recover from donor mice for transfer into recipient mice is collected at the initiation of the 'break-down' phase and is most likely to consist of monocytes and monocyte-derived macrophages. Moreover, the tissue collected is the 'decidual' mass only and does not include the compartment of the 250 uterus where putative 'tissue-resident' macrophages are located. Using an inducible Csf1rknockout to generate donor endometrium we achieved depletion of macrophages (F4/80 hi , Ly6C lo ) and limited the number of monocytes (F4/80 lo , Ly6C hi ) in the tissue transferred to recipient mice. We did not observe any difference in the number of lesions formed between recipient mice that received wild type or macrophage depleted endometrium, however we 255 did find that the lesions recovered were significantly smaller in mice receiving macrophage-depleted endometrium. The breakdown phase of the (donor) endometrium is analogous to the initial inflammatory phase of the wound healing process where pro-inflammatory macrophages play a vital role prior to wound repair 26 . If we consider endometriosis lesions as chronic wounds that do not fully resolve their inflammation, we may presume that the 260 initial inflammatory phase begins as the endometrium breaks down during menstruation and the repair of the translocated endometrium occurs in the peritoneal cavity resulting in the formation of lesions. Importantly, reduced wound closure and delayed repair response is observed when monocytes and macrophages are depleted during the initial inflammatory phase of skin repair 27  We determined that LpM but not SpM enter endometriosis lesions using adoptive transfers; we suggest that these cells change phenotype in lesions very rapidly because only a few cells positive for both F4/80 and GATA6 were identified using dual staining. Peritoneal macrophages perform important functions in other visceral organs in response to 300 inflammation; following sterile liver injury, mature F4/80 hi , GATA6+ LpM traffic directly to the damaged tissue via direct recruitment across the mesothelium. Once in the liver, the macrophages undergo local proliferation, and rapid phenotype switching such that they exhibit a pro-repair profile and activate pathways to restore the injury site 30 . Our data suggest that in endometriosis LpM trafficking to lesions may play a similar role, perceiving 305 the ectopic tissue as a wound and activating repair processes. This is supported by the finding that depletion of LpM using an anti-F4/80 antibody prior to and following endometrial tissue injection results in the recovery of significantly lighter lesions 18 .
Interestingly, our data suggest that SpM may have a neutral role in the pathophysiology of endometriosis since they never exhibit an increase in number and appear to serve as a 310 differentiation intermediate for monocytes transitioning into LpM.
Our data indicates a key role for monocyte recruitment to the peritoneal cavity and ectopic tissue in endometriosis. We detected lesion-resident monocytes as well as cells that were double positive for both Ly6C and F4/80 indicating that monocytes rapidly differentiate in lesions. Moreover, monocyte numbers are elevated in the peritoneal cavity 315 of mice with induced endometriosis and when monocytes are constitutively depleted this impacts significantly on LpM and SpM pools in the peritoneal cavity and results in increased lesions. Thus, we suggest that monocyte-derived macrophages act to protect the peritoneal cavity and can limit the development of lesions (Fig.7). Previous studies have demonstrated infiltration and accumulation of bone marrow derived CD11b+ cells and TEK receptor 320 tyrosine kinase positive (TIE2+) macrophages in lesions 19, 31 supporting our findings that monocytes are recruited to lesions, not just from the periphery but perhaps also directly from the bone marrow. Our findings show that, although monocytes were depleted in Ccr2-/mice with induced endometriosis, they were not completely absent, as we observed in Immunofluorescence. Immunofluorescence was performed as previously described 20,21,41 .
In brief, sections were antigen retrieved with heat and pressure (buffers pH 6.0 or pH 9.0) or 440 trypsin tablets dissolved in dH2O (for F4/80 antibody; Sigma) and incubated with sections for 20 min at 37°C. Sections were blocked for endogenous peroxidase (6% H 2 O 2 in methanol) and nonspecific epitopes (species-specific serum diluted 1:5 in Tris-buffered saline and 5% bovine serum albumin, or blocking serum from species specific ImmPRESS® kit; Vector Laboratories) and incubated with primary antibody (Table 2)  Values were expressed as % of total DAPI+ cells. The area of H&E stained lesions captured using 2.5x magnification was measured in Fiji by setting the scale to a known size (198 pixels = 500µM) and then drawing around boundary of the lesion (excluding peritoneal and 460 adipose tissue) and using the 'measure' function.
Definiens analysis. Ly6C, F4/80 dual immunofluorescence was automatically quantified using slide scanning and machine learning. Stained tissue sections were imaged on a Zeiss Axioscan.Z1 (Carl Zeiss AG, Oberkochen, Germany) at 20x using fluorescence filters configured for DAPI, FITC, and Cy3. Whole-slide .czi files were imported into TissueStudio 465 2.4 (Definiens AG, Munich, Germany) for automated tissue detection followed by manual correction of ROIs to delineate endometriosis lesion, peritoneum, haemorrhage, and adipose tissue. Tissue studio's built-in nuclear segmentation, using the DAPI channel, was applied within these regions to identify cell objects and these objects were then classified as positive or negative for each channel based on intensity thresholds which were used across 470 all samples.

Statistical analysis.
Statistical analysis was carried out in GraphPad Prism 7.02. Data was first analysed for normality using an Anderson Darling normality test. If data were normally distributed, either an ANOVA with a Tukey's post-hoc test (more than 2 samples) or a t-test (2 samples) was performed. If data were not normally distributed, non-parametric tests 475 were used; either Kruskal-Wallis with a Dunn's post hoc test (more than 2 samples) or a Mann-Whitney U test (2 samples). Statistical significance was reported at p<0.05.

Acknowledgements
We thank the QMRI flow cytometry and cell sorting facility technicians (University of 480 Edinburgh) for advice on panel design, and Fiona Ballantyne for technical assistance. We   Data are presented as mean ± SEM. Statistical significance was determined using a student's t-test or a Mann-Whitney test. *;p<0.05. 690