Donor monocyte-derived macrophages promote human acute graft versus host disease

Myelopoiesis is invariably present, and contributes to pathology, in animal models of graft versus host disease (GVHD). In humans, a rich inflammatory infiltrate bearing macrophage markers has also been described in histological studies. In order to determine the origin, functional properties and role in pathogenesis of these cells, we isolated single cell suspensions from acute cutaneous GVHD and subjected them to genotype, transcriptome and in vitro functional analysis. A donor-derived population of CD11c+CD14+ cells was the dominant population of all leukocytes in GVHD. Surface phenotype and nanostring gene expression profiling indicated the closest steady-state counterpart of these cells to be monocyte-derived macrophages. In GVHD, however, there was upregulation of monocyte antigens SIRPα and S100A8/9, and transcripts associated with leukocyte trafficking, pattern recognition, antigen presentation, and co-stimulation. Isolated GVHD macrophages stimulated greater proliferation and activation of allogeneic T cells, and secreted higher levels of inflammatory cytokines than their steady-state counterparts. In HLA-matched mixed leukocyte reactions, we also observed differentiation of activated macrophages with a similar phenotype. These exhibited cytopathicity to a cell line and mediated pathological damage to skin explants, independently of T cells. Together, these results define the origin, functional properties and potential pathogenic roles of human GVHD macrophages.


Introduction
Acute graft versus host disease (GVHD) affects up to 50% of patients receiving allogeneic bone marrow transplantation (BMT) and remains a leading cause of morbidity and mortality (1,2). GVHD most commonly affects the skin, gut, liver and may also contribute to idiopathic pneumonia syndrome (3). In animal models, donor T lymphocytes play an essential role in immune-mediated damage to host epithelium (4). In human GVHD, mononuclear infiltrates have been observed that include CD8+ T cells, CD4+ Th1, Th2, Th17 and Tregs although no pathognomic effector subset has been observed in all patient cohorts (5)(6)(7)(8). Despite the obvious importance of effector T cells, they may not be sufficient to mediate GVHD pathology (4). In almost all GVHD models, pathology occurs in the presence of neutrophils, monocytes and other myeloid components that may infiltrate tissues and amplify local immune responses (9).
Animal models previously demonstrated that immunocompetent donor myeloid cells enhance GVHD, without specifying a particular cell type (10, 11). Macrophages have been implicated through observations that GVHD may be modulated by manipulation of the CSF-1 axis, although opposing effects have been reported, depending on the timing of interventions (12)(13)(14)(15). Glucocorticoids also appear to reduce GVHD at least partly through attenuation of macrophage responses (16) and in humanized mice, donor monocytes or DC are absolutely required for xeno-GVHD (17). Knock-out of the ATP receptor P2Y2 on recipient monocytes reduces GVHD lethality (18). Most recently, a specific role of T cell-derived GM-CSF was described in promoting the differentiation of effector macrophages (19).
In humans, a number of reports highlight an increase in myeloid cells bearing macrophage markers, showing that the level of infiltration correlates with clinical severity and outcome (7,20,21). However, as shown by high resolution analysis, the myeloid cell compartment of human skin is highly complex with discrete populations of classical dendritic cells (DC), monocyte-derived cells and resident macrophages, (22)(23)(24)(25)(26). These observations suggest that the nature of myeloid infiltrates cannot be adequately resolved using in situ microscopy; hence their origin and immune functions in GVHD remain undefined.
The role of (recipient) myeloid cells in responding to danger signals is integral to most models of GVHD, but is not known whether human GVHD infiltrates bearing macrophage markers are recipient, donor, immunogenic or anti-inflammatory.
Although donor myelopoiesis usually dominates the peripheral blood compartment during GVHD, recipient dermal macrophages have very slow kinetics of turnover in humans (22) and potentially expand during inflammation (27). Macrophages are capable of mediating a wide spectrum of tolerogenic or pathogenic responses (28).
By extrapolation from mouse models, macrophages are likely to promote GVHD.
However, their ability to stimulate local effector T cells and mediate direct epithelial damage remain untested in humans.
Here we employ direct methods of isolation and testing to show that acute GVHD lesions in human skin are dominated by CD11c+CD14+ myeloid cells with the genotype, phenotype and transcriptional profile of donor monocyte-derived macrophages. These cells have potent immunological functions that are likely to contribute to the pathogenesis of GVHD and may offer opportunities for therapeutic intervention.

Results
In order to investigate the properties of myeloid cells in human cutaneous acute GVHD, mononuclear cells of the human dermis were defined by immunohistochemistry, immunofluorescence microscopy and flow cytometry in healthy controls, BMT recipients without acute GVHD, and BMT recipients with GVHD (Table S1, S2). BMT control patients without GVHD were biopsied on median day 83 post-transplantation (range 28-148 days). GVHD skin biopsies were taken at the onset of an acute onset erythematous rash, prior to initiation of therapy.
Classical acute GVHD, immunosuppression withdrawal acute GVHD, and acute GVHD following donor lymphocyte infusion were all included. A pathological diagnosis of acute-type GVHD was confirmed by standard histological criteria in all cases and patients with clinical or histological features of chronic GVHD were excluded. The median day of biopsy was day 53 (range 13-304; Mann-Whitney p = 0.27 compared with BMT controls). In situ analysis showed an increase in CD3+ T cells and CD11c+ myeloid cells in a perivascular and epidermal interface distribution in GVHD (Figure 1a). The nature of the leukocytic infiltrate was also documented using 4-colour immunofluorescence of whole-mount specimens. There was marked infiltration of perivascular spaces by CD11c+ cells that usually remained distinct from FXIIIA-expressing resident macrophages (22) (Figure 1b). Further comparison of CD11c, FXIIIA and CD163 antigen expression by this approach is shown in Figure   S1a-c.
The infiltrates of acute GVHD infiltrate were characterized by flow cytometry of single cell suspensions. Gating on live singlets expressing CD45 and HLA-DR revealed SSC low lymphocytes and HLA-DR+ SSC high myeloid cells as previously described (22,25). Surprisingly, the proportion of cells falling in the lymphoid gate was not 7 significantly increased in GVHD relative to BMT controls or healthy donors ( Figure   S1e) although a relative expansion of IFNγ-secreting CD4+ T cells was observed in relative to both controls (Figure S1d-f). Myeloid cells were further divided on the bivariate plot of CD14 versus CD11c allowing identification of subsets previously described in healthy control skin, without relying upon autofluorescence to capture resident macrophages (22)(23)(24)26). The linkage between this new gating strategy and previously identified myeloid cell populations is explained in Figure S2a-d.
In contrast to the modest changes in overall cellularity and T cell populations, CD11c+CD14+ myeloid cells were expanded more than 10-fold compared with healthy control skin or BMT skin without GVHD (Figure 1c, d, red gates; Figure   S1a-c). This GVHD-related subset lacked CD1c expression and mapped to parameter space containing monocyte-macrophages in the steady-state (25). Cells in the CD14+ CD11c-gate contained FXIIIA+ CD163+ macrophages with high melanin content and auto-fluorescence, representing 'fixed' or resident macrophages (22,29,30). These were relatively depleted in GVHD as were classical DC2 (cDC2; CD11c+ CD1c+ CD14-) and cDC1 (CD141+ cells in the CD14-CD11c-gate; Figure   1c, d). A ratio of >0.55 between CD11c+ CD14+ GVHD cells and CD1c+ cDC2 was 84% sensitive and 81% specific for the histological diagnosis of GVHD in skin biopsies post BMT (Figure 1f). Sequential biopsies showed resolution of the GVHD infiltrate in parallel with clinical improvement (Figure S2e, f).

15
The preceding data suggest that GVHD macrophages are derived from blood monocytes and achieve a higher state of functional activation than their steady state counterparts. Further evidence of their likely function in GVHD was sought by deriving allo-stimulated macrophages from monocytes and testing their functional properties. HLA-matched donor and recipient blood was taken prior to transplantation and PBMC stored in order to prepare mixed leukocyte reactions. The cytokine milieu of an HLA-matched MLR was similar to that observed when GVHD skin was cultured (Figure 5a), and a prominent population of macrophages appeared, with similar phenotype to GVHD macrophages (Figure 5b). MLRactivated macrophages also expressed cytotoxic molecules perforin, granzyme A, granulolysin and TRAIL, similarly to GVHD macrophages (Figure 5b-d). Many of these products were already upregulated in CD14+ monocytes isolated from the blood of patients with GVHD, compared with healthy control monocytes (Figure 5d).

17
The expression of cytotoxic molecules prompted us to test the possibility that MLRactivated macrophages might mediate cytotoxicity to epidermal cells. We observed that MLR-activated macrophages were directly cytotoxic to a keratinocyte cell line in vitro, in a dose-dependent manner (Figure 6a, b). In order to test a setting more relevant to GVHD, we adapted the in vitro skin explant model. When a small explant of intact skin is exposed to a clone of minor-histocompatibility antigen-specific T cells, GVHD-like epidermal pathology is observed in an HLA-restricted and antigenspecific manner (31). GVHD pathology is also observed, in proportion to HLAmatching and sex differences, when recipient skin is exposed to donor leukocytes pre-sensitized to recipient antigens in a mixed leukocyte reaction (MLR) (32).
Although it has been assumed that GVHD pathology in vitro is exclusively mediated by T cells in the MLR, we were surprised to observe nearly equivalent cytopathic effects when the 'donor' MLR was sorted into macrophage and T cell components (Figure 6c, d). c. MLR outputs were sorted to yield macrophages and lymphocytes and co-cultured with shave biopsies of recipient skin for 3 days. Explants were fixed and stained with Haematoxylin and eosin. Representative images from explants cultured for 3 days in medium alone or medium + MLR macrophages. d. Summary of histological damage to the dermo-epidermal junction graded on the Lerner scale from 6 independent experiments. p values from one-way nonparametric ANOVA and Dunn's multiple comparison tests are shown.

Discussion
In this study we have defined the role of myeloid cells in human cutaneous acute GVHD by characterizing mononuclear infiltrates from primary tissue, isolating the dominant myeloid cell, and defining its origin, transcriptional profile and functional properties. The data indicate that human GVHD lesions are highly infiltrated with donor monocyte-derived macrophages with enhanced allo-stimulatory activity and the potential to mediate epidermal pathology.  (25). Enhanced monocyte priming and recruitment to tissues is also suggested by the phenotypic activation of peripheral blood monocytes, previously been reported in patients with active GVHD (37)(38)(39). We observed a similar phenomenon in the expression of cytotoxic genes by GVHD monocytes compared with those from healthy controls. In corroboration, we also observed that monocytes differentiating into macrophages in HLA-matched MLRs had a similar phenotype and functional properties to GVHD macrophages.

Myeloid cells found in
The data indicate that single surface markers previously used to define GVHD macrophages by histology often have variable expression under more detailed scrutiny. In keeping with previous reports (7,(20)(21)(22), CD163 was detectable by flow cytometry, but was less consistent than CD11c+ in identifying GVHD infiltrates by immunohistochemistry. CD163 is expressed by resident macrophages (40) and chronic inflammatory macrophages found in psoriasis (41), coeliac disease (42) and Crohn's disease (43). Higher CD163 expression was associated with longer intervals post-transplant, consistent with the description of abundant CD163 expression in advanced GVHD lesions (20,21). These findings are also in keeping with the original characterization of CD163 (clone RM3/1) as a 'late phase' macrophage antigen with more delayed kinetics of expression (44).

GVHD macrophages demonstrated enhanced T cell stimulatory functions, compared
with steady-state CD11c+CD14+ cells, including greater expression of pattern recognition, leukocyte adhesion and trafficking, antigen presentation and T cell costimulation genes, production of chemokines and capacity to stimulate allogeneic T cell proliferation. The potential of macrophages to mediate direct cytotoxic effects is described in classical studies but, until recently, has received little attention in the field of GVHD. Early studies showed that GVHD induced priming of macrophages resulting in direct cytotoxicity following LPS challenge (45) and subsequent work elaborated on the secretory properties of activated macrophages (46). However, contemporary models of GVHD present the contribution of macrophages to GVHD almost exclusively in terms of sensing danger and enhancing accessory cell function (9-11). It was not possible to harvest sufficient GVHD macrophages directly from biopsies to test their effector function so we generated allo-stimulated macrophages from monocytes in HLA-matched MLRs, as a surrogate. MLR-stimulated macrophages were capable mediating direct cytopathicity with a cell line and surprisingly, caused a similar degree of immunopathology as T cells in the skin explant model of GVHD. Although the latter lacks all the complexity of GVHD in vivo, it is the only human system amenable to manipulation. Furthermore, the degree of pathological damage consistently reflects levels of major and minor histocompatibility antigen matching and has been used previously to dissect HLArestricted antigen-specific GVHD responses (31,32). This result revises the assumption that T cells are the only relevant effectors when the an MLR product is added to explanted skin.
In classical animal models of GVHD, myelopoiesis is invariable present during the effector phase of GVHD, even when the sole instigator of GVHD is a T cell clone directly targeted to epithelium (47,48). Investigators have now revealed nonredundant functions for myeloid cells in GVHD pathogenesis (12-15, 17, 19). The conceptual advance that T cells are necessary but may not be sufficient for GVHD has important therapeutic implications. Unlike T cells in the human adult, which may take more than 2 years to recover after transplantation, myeloid cells are continuously generated, and rapid immune reconstitution is possible following myeloid-targeted interventions. The ability to isolate discrete mechanisms that govern the infiltration of tissues by myeloid cells, such as GM-CSF dependence, may also offer a means of minimising GVHD without compromising GVL (19).
Inflammatory macrophages with a similar phenotype to GVHD macrophage have been characterized in other pathological states such as psoriasis (41)  initiate GVHD by stimulating an initial host versus graft (HVG) reaction by recipient T cells. Consistent with this hypothesis, it has been observed that DP mismatches in the HVG direction are associated with higher levels of GVHD than fully DP-matched transplants (Reid et al, manuscript in preparation). The possibility that a tissuebased HVG reaction can cause pathology in distinguishable from GVHD is also suggested by the observation that lethal GVHD may be mediated by a 'bystander' 24 MLR effect in which recipient epithelium is not directly recognizable by T cells, in animal models (50).
In summary, the results presented here demonstrate that GVHD lesions contain abundant donor macrophages, likely to be derived from activated circulating classical monocytes. GVHD macrophages secrete chemokines, are capable of stimulating T cells and can mediate direct cytotoxicity. Together, these results shed new light on human macrophage functions and suggest that they may be a universal therapeutic target for the prevention and treatment GVHD.

Human subjects
Sequential patients undergoing allogeneic BMT were recruited from the Northern Centre for Bone Marrow Transplantation at Newcastle upon Tyne Hospitals NHS Foundation Trust over a 3 year period between 2013 and 2016. 5-15mm 2 skin shaves were obtained using 1% lidocaine local anesthesia and a Dermablade TM from BMT recipient controls with no evidence of GVHD and BMT recipients with GVHD.
Biopsies were transported in serum-free medium (X-VIVO, Lonza) and analysed within 24 hours. An independent clinical pathologist provided diagnosis and histological grading of GVHD in controls and GVHD biopsies. Healthy control skin was obtained from patients undergoing mammoplasty or abdominoplasty as previously described (22). Specimens comparable to clinical biopsies were obtained by immobilizing skin strips on a cork block covered with sterile silicon and performing skin shave biopsies.

Cell lines and mononuclear cells
Unless stated otherwise all cells were cultured in RPMI 1640 (Lonza) with 100IU/ml penicillin, 10μ/ml streptomycin, 2mM L-glutamine (Invitrogen), 10% heat-inactivated fetal bovine serum (Sera Lab). MLR macrophages were generated from co-culture of peripheral blood mononuclear cells (PBMCs) from HLA-matched BMT donor and recipient pairs. Recipient PBMCs were irradiated (20Gy) and used as stimulators for donor PBMCs at a 1:1 ratio. Cultures containing 1-5x10 7 donor PBMCs were maintained in RPMI 1640 with 10% human AB serum (Sigma) for 7 days.

Dermal cell isolation
Skin shave biopsies were used whole or split into dermis and epidermis by treatment with dispase 0.5-1.0mg/ml in RPMI for 60-90 minutes at 37°C (Gibco) before digestion in medium with 1.6mg/ml type IV collagenase (Worthington) for 12-16 hours at 37°C in RPMI with 10% heat-inactivated fetal calf serum. Gentle dissociation and passage through a 100μm filter generated single cell suspensions.
For some experiments, an enzyme-free preparation of leukocytes was required. Skin was cultured as above but without collagenase. After 48 hours, migratory leukocytes were harvested from the supernatant. Supernatants were stored at -80°C and later used for cytokine analysis by Luminex assay (see below)

Immunohistochemistry
Formalin-fixed and paraffin-embedded skin shave biopsies from GVHD diagnostic material were used. 4μm sections were made. Antigen retrieval and staining were performed using the BenchMark autostainer (Ventana). CD3, CD163, CD11c, factor XIII and Ki67 primary antibodies and the 'ultraview' detection kit were used (X).

Flow cytometry analysis and sorting
Antibodies used for flow cytometry are listed in Table S3. DAPI was added for dead cell discrimination (X). Flow cytometry analysis and sorting was performed using BD FACS Canto II, BD Fortessa X20, BD FACS Aria II and BD FACS Fusion running FACSDiva version 7 and analyzed with FlowJo (TreeStar version X).

X/Y fluorescence in situ hybridization
Sorted cells were spun onto slides, fixed in methanol and acetic acid and prepared with dual labeled XY FISH probes on a ThermoBrite system (Abbott) in accordance with manufacturer's instructions.

Lymphocyte proliferation assays
Healthy volunteer T cells were prepared from healthy blood donors by immunodensity negative selection (Human T cell enrichment cocktail, Stemcell 15021) and labeled with 1μM CSFE (Invitrogen). T cells were co-cultured with sorted macrophages or DCs at a 25:1 ratio for 7 days , incubated with antibodies to CD3, CD4, CD8, HLA-DR and CD69 (Table S3), and analyzed by flow cytometry.

Cytokine and chemokine quantification
Quantification was performed on medium from isolated macrophage populations stimulated ex-vivo and medium conditioned by explanted GVHD skin and by donorrecipient MLR. For macrophage stimulation, FACS sorted skin macrophages were stimulated with 100ng/ml LPS from E. coli (Sigma) and harvested at 10 hours.
Supernatants were cryopreserved at -80°C and batch-analyzed by Luminex assay (ProcartaPlexTM 34-plex, EBioscience) using a Qiagen Liquichip 200 running Luminex 100 integrated system software version 2.3. Procartaplex Analyst version 1.0 was used to define standard curves.

Skin explant assay
Blood from BMT donor and recipient pairs was prospectively collected into EDTA, prior to transplnatation. Recipient PBMCs were irradiated (20Gy) and used as stimulators for donor PBMCs at a 1:1 ratio. Cultures containing 1-5x10 7 donor PBMCs were maintained in RPMI with 10% human AB serum (Sigma) for 7 days.
Macrophages and T cells were sorted from MLR outputs as HLA-DR+CD14+CD11c+ cells and SSC lo CD3+ cells respectively. Shave biopsies of recipient skin were taken prior to BMT conditioning. Biopsies were divided and incubated with medium alone (negative control), 2x10 5 MLR lymphocytes or 2x10 5 MLR macrophages for 3 days in RPMI 1640 with 20% heat inactivated autologous 29 serum. Explants were fixed in 10% buffered formalin, sectioned and stained with haematoxylin and eosin. Severity of histological damage was graded using the Lerner scale by an experienced assessor blinded to experimental conditions (XN Wang).

Quantification and statistical analysis
FlowJo version 9.6.7 was used for analysis of flow cytometry data. Principal