High CD206 levels in Hodgkin lymphoma‐educated macrophages are linked to matrix‐remodeling and lymphoma dissemination

Macrophages (Mφ) are abundantly present in the tumor microenvironment and may predict outcome in solid tumors and defined lymphoma subtypes. Mφ heterogeneity, the mechanisms of their recruitment, and their differentiation into lymphoma‐promoting, alternatively activated M2‐like phenotypes are still not fully understood. Therefore, further functional studies are required to understand biological mechanisms associated with human tumor‐associated Mφ (TAM). Here, we show that the global mRNA expression and protein abundance of human Mφ differentiated in Hodgkin lymphoma (HL)‐conditioned medium (CM) differ from those of Mφ educated by conditioned media from diffuse large B‐cell lymphoma (DLBCL) cells or, classically, by macrophage colony‐stimulating factor (M‐CSF). Conditioned media from HL cells support TAM differentiation through upregulation of surface antigens such as CD40, CD163, CD206, and PD‐L1. In particular, RNA and cell surface protein expression of mannose receptor 1 (MRC1)/CD206 significantly exceed the levels induced by classical M‐CSF stimulation in M2‐like Mφ; this is regulated by interleukin 13 to a large extent. Functionally, high CD206 enhances mannose‐dependent endocytosis and uptake of type I collagen. Together with high matrix metalloprotease9 secretion, HL‐TAMs appear to be active modulators of the tumor matrix. Preclinical in ovo models show that co‐cultures of HL cells with monocytes or Mφ support dissemination of lymphoma cells via lymphatic vessels, while tumor size and vessel destruction are decreased in comparison with lymphoma‐only tumors. Immunohistology of human HL tissues reveals a fraction of cases feature large numbers of CD206‐positive cells, with high MRC1 expression being characteristic of HL‐stage IV. In summary, the lymphoma‐TAM interaction contributes to matrix‐remodeling and lymphoma cell dissemination.

Macrophages (Mφ) are abundantly present in the tumor microenvironment and may predict outcome in solid tumors and defined lymphoma subtypes. Mφ heterogeneity, the mechanisms of their recruitment, and their differentiation into lymphoma-promoting, alternatively activated M2-like phenotypes are still not fully understood. Therefore, further functional studies are required to understand biological mechanisms associated with human tumor-associated Mφ (TAM). Here, we show that the global mRNA expression and protein abundance of human Mφ differentiated in Hodgkin lymphoma (HL)-conditioned medium (CM) differ from those of Mφ educated by conditioned media from diffuse large B-cell lymphoma (DLBCL) cells or, classically, by macrophage colony-stimulating factor (M-CSF). Conditioned media from HL cells support TAM differentiation through upregulation of surface antigens such as CD40, CD163, CD206, and PD-L1. In particular, RNA and cell surface protein expression of mannose receptor 1 (MRC1)/CD206 significantly exceed the levels induced by classical M-CSF stimulation in M2-like Mφ; this is regulated by interleukin 13 to a large extent. Functionally, high CD206 enhances mannose-dependent endocytosis and uptake of type I collagen. Together with high matrix met-alloprotease9 secretion, HL-TAMs appear to be active modulators of the tumor matrix. Preclinical in ovo models show that co-cultures of HL cells with monocytes or Mφ support dissemination of lymphoma cells via lymphatic vessels, while tumor size and vessel destruction are decreased in

Introduction
The tumor microenvironment (TME) plays an important role in numerous malignancies and may even provide a tumor supportive milieu. The TME contains a wide variety of cytokines and growth factors and supports crosstalk between neighboring cells. Among the stromal cells found in the TME, tumor-attracted and re-educated macrophages (Mφ), also designated as tumor-associated Mφ (TAM), have received increasing interest (Mantovani and Sica, 2010;Twum et al., 2017). They are involved in various aspects of tumor progression like aberrant cytokine secretion, immune check point modulation, tumor matrix remodeling, angiogenesis, and resistance to treatment (Mantovani et al., 2017(Mantovani et al., , 2002Mantovani and Sica, 2010). TAMs have been implicated in the pathogenesis of various lymphoma subtypes such as Hodgkin lymphoma (HL), chronic lymphatic leukemias, follicular lymphoma, and aggressive non-HLs such as Burkitt lymphoma and DLBCL (Kridel et al., 2015b(Kridel et al., , 2015aLenz et al., 2008;Pham et al., 2018;Steidl et al., 2011;Steidl et al., 2010;Verdi e et al., 2018).
Hodgkin lymphoma is a paradigm of a TMEdominated tumor that is characterized by a very small portion (< 1%) of malignant cells surrounded by abundant supportive cells, among which T cells often dominate (Aldinucci et al., 2010;K€ uppers, 2009;Scott and Steidl, 2014;Steidl et al., 2011). Additionally, Mφ are an integral part of the TME, and the number of CD68-positive Mφ predicts the patient's outcome (Scott and Steidl, 2014). Therefore, targeting the interaction between HL cells and Mφ presents a promising strategy in the treatment of HL (Locatelli et al., 2018;Ruella et al., 2017). However, our knowledge of the functions and interactions of Mφ in HL is still limited (Mantovani et al., 2002;Martinez and Gordon, 2014;Steidl et al., 2010). The present study sheds new light on the molecular interaction between HL cells and TAMs and shows that HL-educated TAMs are capable of supporting tissue remodeling and lymphoma dissemination involving the mannose receptor 1 (MRC1/CD206).

Cell lines, conditioned media, reagents, and functional assays
The HL cell lines L428, KM-H2, L1236, L540, and HDLM-2 were provided by V. Diehl (Cologne). The DLBCL cell lines HBL-1 and OCI-LY3 were obtained from D. Krappmann (Munich) and the Deutsche Sammlung f€ ur Mikroorganismen und Zelllinien (DSMZ; Braunschweig, Germany), respectively. All cell lines were maintained in Roswell Park Memorial Institute (RPMI) 1640 supplemented with 2 mgÁmL À1 glutamine and 10% FCS. At regular intervals, we used a highly specific Ig PCR to test for cell identity. In addition, L428, L1236, KM-H2, OCI-Ly3, and HBL1 cells had been subjected recently to STR profiling by the DSMZ (Braunschweig, Germany).
For the production of lymphoma-conditioned media (CM), cells were seeded at a density of 5 9 10 5 cells per mL and incubated in complete RPMI 1640 medium for 2 days. Cell supernatants were centrifuged at 300 g for 10 min at 4°C, sterile-filtered, and stored at 4°C for a maximum of 2 weeks.

Monocyte isolation
Peripheral blood mononuclear cells (PBMCs) of healthy donors were isolated from fresh buffy coats by density-gradient centrifugation over Biocoll separating solution (Biochrom, Berlin, Germany). CD14 + monocytes were obtained from PBMCs by magnetic cell separation using CD14 microbeads (Miltenyi Biotec, Bergisch Gladbach, Germany) according to the manufacturer's instructions. Purity of CD14 + cells after magnetic cell separation was determined by staining with specific markers and quantification by flow cytometry using a FACSCanto II (BD Biosciences, Franklin Lakes, NJ, USA).

Macrophage differentiation
Monocyte isolation and macrophage differentiation were performed as described previously (Menck et al., 2014). Monocytes were differentiated either in the presence of RPMI 1640 containing 10% (v/v) FCS, Pen/Strep, and 2.5 ngÁmL À1 recombinant M-CSF (Immunotools, Friesoythe, Germany) or in lymphoma-CM mixed in equal parts with complete RPMI1640. Briefly, M-CSF or CM of the tested lymphoma cell lines was used to incubate primary human monocytes for 7 days in Teflon-coated cell culture equipment. After 7 days, cells were counted, harvested, and transferred into cell culture dishes for additional functional analysis. Freshly isolated monocytes or Mφ were stained with cell-specific antibodies or isotype controls. Expression was quantified by flow cytometry using a fluorescence-activated cell sorting (FACS) Canto II and calculated by dividing the mean fluorescence intensity (MFI) by the MFI of the corresponding isotype control. For intracellular CD68 staining, cells were fixed and permeabilized using the Cytofix/Cytoperm Kit (BD BioSciences) following the manufacturer's instructions.
For migration assays, a Boyden chamber with a 5µm porous membrane was used (Neuro Probe Inc., Gaithersburg, MD, USA). 5 9 10 4 monocytes were seeded per well in 50 µL of RPMI 1640 medium and allowed to migrate for 2 h toward RPMI 1640 supplemented with 1% or 10% (v/v) FCS or lymphoma-CM. The cells that had migrated into the lower chamber were counted.
To analyze endocytosis, Mφ were allowed to adhere overnight on tissue culture dishes. Cells were washed twice with PBS and maintained in RPMI 1640 containing 10% (v/v) FCS and pen/strep. 10-and 70-kDa FITC dextran (at 1 mgÁmL À1 for 2 h; Sigma Aldrich, Munich, Germany), respectively, or Gelatin Oregon Green TM 488 Conjugate (at 5 µgÁmL À1 for 30 min; Thermo Scientific, Waltham, MA, USA) were added to each plate and incubated at 37°C or on ice. After incubation, cells were washed twice with PBS, harvested using trypsin/EDTA, and fixed in 2% formaldehyde solution in PBS. Uptake was analyzed by means of a FACS Canto II and calculated by dividing the MFI of the 37°C sample by the corresponding MFI of the sample incubated on ice.
The activity of matrix metalloproteinase (MMP)-9 was determined in cell culture supernatants by gelatin zymography. First, proteins were separated by SDS/ PAGE. Gel mixes were used to prepare 8% acrylamide gels containing 1% gelatin and 5% acrylamide stacking gels. Fifteen microliters of cell culture supernatant has mixed with an equal volume of loading buffer (62.5 mM Tris/HCl, pH 6.8, 4% SDS, 25% glycerol, 0.01% bromophenol blue) and loaded onto the gel. Electrophoresis was performed under cooling using freezer packs. Afterward gels were incubated in wash buffer (50 mM Tris/HCl, pH 7.5, 5 mM CaCl 2 , 2.5% Triton X-100) for 1 h. Gels were transferred to renaturation buffer (2.5% Triton X-100) and incubated 1 h. Afterward gels were covered in development buffer (50 mM Tris base, 150 mM NaCl, 10 mM CaCl 2 ) and incubated overnight at 37°C with soft agitation. To visualize the gel degradation by MMP activity, the gels were then stained in staining buffer (0.5% Coomassie blue, 40% methanol, 10% acetic acid) for 1 h, followed by destaining in destaining buffer (40% methanol, 10% acetic acid) for 1.5 h. Gels were then fixed for 30 min in fixation buffer (5% glycerol, 30% methanol), placed between two cellophane membranes and dried overnight. Fixed and dried gels were scanned for image processing.

Fluorescence-labeled antibodies
ELISA kits for the detection of secreted M-CSF and IL-13 in cell culture supernatants were purchased from R&D Systems (Minneapolis, MN, USA). The sensitivity was 11.2 and 13.2 pgÁmL À1 , respectively.
Slides were scanned by a Hamamatsu Nanozoomer 2.0 RS slide scanner (Hamamatsu Photonics, Ammersee, Germany) with 409 magnification for the TMAs and a 209 magnification for the fluorescent-labeled slides of HL and tonsils. Raw image data were saved in.ndpi format and handled by the software kit NDP.VIEW 2 (Hamamatsu Photonics) to save details of the whole image in.jpeg format. Figures were made with Adobe Photoshop CS4 Extended (Adobe Systems Incorporated, San Jose, CA, USA). Chick chorioallantoic membrane (CAM) tissue slides were scanned with a 209 objective (UPlanSApo, NA 0.75) using the virtual slide microscope VS120-L100 (Olympus, Hamburg, Germany) equipped with a VC-50 camera.
The experiments and analysis were undertaken with the understanding and written consent of each subject. The study methodologies conformed to the standards set by the Declaration of Helsinki. The study methodologies were approved by the local ethics committee [Medical Faculty of the Georg-August-University G€ ottingen (No. 16/5/18An); Medical Faculty of the Christian-Albrechts-University Kiel, Germany (No. 447/10)].
In order to access tumor volume and blood content, specimens were stained with phosphotungstic acid (PTA) according to the protocol described recently (Dullin et al., 2017). Then, they were embedded in 1% agarose gel in 2-mL plastic tubes and scanned using the in vivo microcomputed tomography (micro-CT) QuantumFX (Perkin Elmer Health Sciences, Hopkinton, MA, USA) and the following acquisition parameters: 90-kV tube voltage, 200-µA tube current, FOV 20 9 20 mm 2 , 2-min total acquisition time resulting in 3D datasets with a voxel size of 40 9 40 9 40 µm 3 . The software SCRY v6.0 (Kuchel & Sautter GbR, R€ otenbach, Germany) was used for 3D rendering and volume measurement. For this purpose, the CAM around the tumor was manually removed using a virtual scalpel and the tumor mass was segmented based on a brightness threshold.

Transcriptomics
The samples were analyzed by RNA-Seq. Read quality was assessed with fastQC (Andrews (2010): FastQC: a quality control tool for high-throughput sequence data. Available online at: https://www.bioinformatic s.babraham.ac.uk/projects/fastqc/), FastqPuri and QoRTs (Hartley and Mullikin, 2015;Perez-Rubio et al., 2018). First, we removed reads perfectly matching the human ribosomal repeating unit (GenBank accession U13369.1) with FastqPuri. Sequences were then filtered and trimmed based on quality scores (bases with quality scores below 28 were removed from the ends and the remaining read accepted if it had fewer than 5% base qualities below 28). Then, undetermined bases (Ns) were removed from the reads and the largest N-free sequence was kept as long as it was at least 25 nt long. Transcript counts were generated with Kallisto using the Ensembl homo sapiens genome (release 87). The mean pseudo-alignment rate was 87.42%. In order to mitigate the donor effect, the combat function of the R-package sva was applied (Leek J.T. (2018): Available online at: http://biocond uctor.org/packages/release/bioc/html/sva.html). DESeq2 was used for differential gene expression (GE) analysis and P-values were adjusted for multiple testing using the false discovery rate (Hartley and Mullikin, 2015). KEGG pathway analysis was performed using the Rpackage gage (Luo et al., 2009). Variance stabilizing transformation implemented in the R-package DESeq2 was applied to normalize the data for t-SNE plots. Subsequently, t-SNE dimension reduction was performed using the R-package Rtsne (Maaten and Hinton, 2008;Krijthe (2015): Available online at: https://cran.r-projec t.org/web/packages/Rtsne/citation.html). GE data are available at NCBI project number PRJNA592308.

Proteomics
The samples (25 µg per replicate) were subjected to tryptic digestion using the GASP protocol (Fischer and Kessler, 2015). Five micrograms of the resulting peptide mixtures were spiked with 100 fmol of the retention time standard RePLiCal (Polyquant GmbH, Bad Abbach, Germany) and analyzed using an Eksigent ekspert TM nanoLC 400 system coupled to a Triple-TOF 5600 + mass spectrometer (SCIEX, Framingham, MA, USA).
The sequential window acquisition of all theoretical fragment-ion spectra (SWATH) runs was accomplished using a 50 ms full-MS scan from 400-1000 m/z and 60 subsequent SWATH windows of variable size for 35 ms each (mass range, 230-1500 m/z).
The respective libraries were generated from pooled samples measured in data-dependent acquisitions mode (DDA) using the TOP20 method with a full-MS scan for 250 ms and MS/MS scans for 50 ms each. MS/MS spectra from the DDA runs were searched against the respective UniProt database (swissprot-human09-2017) using ProteinPilot 5.0 and imported in PeakView 2.1 using the SWATH MicroApp 2.0 allowing six peptides per protein and five transitions per peptide. Raw values were normalized to total intensity.
The MS proteomics data and search output file have been deposited to the ProteomeXchange Consortium (Perez-Riverol et al., 2019) via the PRIDE partner repository with the dataset identifier PXD01612.

Combining transcriptome and proteome data
Filtering for matching proteins and genes in the combined transcriptome and proteome data yielded 1938 gene-protein pairs. Further filtering for genes and proteins that both differed (P adj < 0.01) in expression between DLBCL and HL stimulated Mφ brought down the number to 13 gene-protein pairs (Table 1, top). Analogously, comparing HL and M-CSF-stimulated Mφ revealed 10 gene-protein pairs (Table 1, bottom). In both comparisons, CD206 was the most differential protein/gene with a log 2 fold change (LFC) of~3.

Statistical and regression analyses
Results are shown as mean or as mean AE SD of the indicated number of samples. The statistical significance of the values was determined using the Student's t-test. If applicable, group results were compared by ANOVA (one-way or two-way analysis of variance) with subsequent Bonferroni's post hoc test to correct for multiple comparisons as indicated. Normal distribution and homogeneity of variance were tested using the Kolmogorov-Smirnov test and the F-test, respectively. The Kruskal-Wallis test with Dunn's post hoc test was performed for nonparametric testing. Significance levels are indicated as *P < 0.05, **P < 0.01 and ***P < 0.001. All statistical analyses and plots were done using GRAPHPAD PRISM 6.04 (GraphPad Software Inc., La Jolla, USA).

Hodgkin lymphoma cell supernatants attract monocytes and initiate their differentiation into Mφ characterized by high mannose receptor 1 expression
To gain insight into the capacity of lymphoma cells to attract and differentiate monocytes into Mφ, the migration of monocytes toward conditioned media (CM) from five HL and two DLBCL cell lines was first investigated by the Boyden chamber assay and compared to unconditioned medium containing 1% and 10% FCS, respectively. Only CM from HL cell lines, with the exception of HDLM-2, exhibited significant chemo-attractive activity (Fig. S1A). However, as exemplified by KMH2-CM, strong chemo-attractive activity does not necessarily translate into strong Mφ differentiation capacity (Fig. S1B). This suggests, in contrast to previous findings (Locatelli et al., 2018;Ruella et al., 2017), that attraction and differentiation are distinct processes regulated by different cytokines or chemokines.
To gain deeper insight into the functions of the L428-educated Mφ, label-free data-independent liquid chromatography-tandem mass spectrometry was used to determine protein expression levels in Mφ. Protein levels were then correlated with the corresponding transcript levels. This analysis confirmed, among others, the striking difference in (MRC1)/CD206 expression, which was significantly higher in L428-educated Mφ than in both, M-CSF and DLBCL-CMderived Mφ (Table 1). Gene and protein expression data can be found in the Tables S1-S5.
In summary, the stimulation of monocytes with L428-CM or M-CSF changes the abundance of surface markers commonly used to monitor Mφ differentiation. This shows that after 7 days, mature Mφ are generated, which belong to the M2 subtype based on high CD163 and CD206 expression. Compared to M-CSFeducated Mφ, L428-CM-differentiated Mφ express more CD1a, CD80, CD40, and PD-L1, which are involved in T-cell interactions. The increased expression of CD11c, CD206, CD33, CD44, and CD54, on the other hand, points toward enhanced cell-matrix or cell-cell interactions. The other markers analyzed were not differentially expressed between L428-CM and M-CSF-treated Mφ. Importantly, CD206 upregulation in lymphoma-educated Mφ was also observed for L1236 and L540-CM-derived Mφ when compared with HBL1 and OCI-Ly3-CM (Fig. 3A). This further emphasizes that the ability of HL cells to differentiate monocytes into M2-like Mφ, which might play important roles in T-cell interactions as well as cell-matrix or cell-cell interactions involving the MRC1 (CD206; Taylor et al., 2005).

HL-educated macrophages are characterized by mannose-dependent endocytosis and increased collagen ingestion
The strong upregulation of CD206 expression in L428-CM vs. M-CSF differentiated Mφ suggested that the former might differ in their endocytic capacity. Therefore, various endocytosis assays were performed. First carboxylate-modified fluorescence-labeled latex beads were used to monitor phagocytic activity (Fig. 3B,C). We observed no differences in the percentage of cells that took up latex beads or in the number of beads taken up, indicating that L428-CM and M-CSF-differentiated Mφ do not differ in the uptake of large particles.
Since MRC1/CD206 can bind and take up polysaccharides, glycosylated proteins, and collagen (Kato et al., 2000;Taylor et al., 2005), we tested whether HL-CM-treated Mφ differed from M-CSF-differentiated Mφ in the uptake of polysaccharides and collagen. We incubated the respective Mφ with 10-kDa FITC dextran in the presence or absence of mannose and measured the uptake of FITC dextran by flow cytometry (Fig. 3D). Mφ differentiated with L428-CM took up significantly more dextran than M-CSFderived Mφ. More importantly, addition of mannose led to a significant inhibition of the dextran uptake in L428 CM-educated Mφ, whereas M-CSF-induced Mφ showed only mannose-unrelated dextran uptake. This also applied to 70-kDa FITC dextran indicating that M Fig. 2. Surface expression of selected proteins on M-CSF or L428-CM-educated Mφ and monocytes. Flow cytometry analysis of surface protein expression on monocytes and Mφ differentiated with either 2.5 ngÁmL À1 M-CSF or L428-CM. MFI was calculated by dividing the MFI of the specific antibody by isotype MFI (mean AE SD, n = 12, paired t-test, two-tailed; *P < 0.05, **P < 0.01 and ***P < 0.001).  higher dextran uptake by HL-CM-educated Mφ is mediated by CD206 (data not shown). Next, we tested the uptake of type I collagen by adding fluorescencelabeled gelatin (gelatin OG-488) to Mφ (Fig. 3E). Interestingly, as observed for dextran, L428-CM-educated Mφ also showed a higher capacity of collagen uptake, further supporting the notion that CD206 on HL-associated M2-like Mφ may participate in the remodeling of the lymphoma microenvironment (Madsen et al., 2013). We also tested whether HL-educated Mφ secrete the matrix metalloproteinases MMP2 and MMP9, which are involved in extracellular matrix (ECM) degradation. Zymography, using gelatin-containing gels, failed to detect any activity of MMP2 in supernatants of HL-and M-CSF-induced Mφ, whereas both secreted equally large amounts of MMP9 (Fig. 3F). L428 cells, in comparison, secreted far less MMP9.
In conclusion, HL-CM-differentiated Mφ differ from M-CSF-induced Mφ in their ability to take up CD206-specific targets. Further, MMP9 secretion hints at an active role of TAMs in remodeling the ECM.

IL13 is present in HL-CM and upregulates CD206 in monocytes
A distinguishing feature of HL cells is their ability to express and secrete IL13 and M-CSF (Lamprecht et al., 2010;Scott and Steidl, 2014;Skinnider and Mak, 2002;Tudor et al., 2014). Additionally, Mφ are known to be polarized toward an M2 phenotype by IL4 or IL13, which includes upregulation of CD206. However, whether either or both IL13 and M-CSF drive CD206 expression remained to be elucidated (Mantovani et al., 2002). Therefore, we isolated CD14 + PBMCs and stimulated them with M-CSF, IL-13, or a combination of both in direct comparison with L428-CM (Fig. 4A). In parallel, the amount of IL13 and M-CSF secreted by HL cells used in this study was determined by ELISA. Both L428 and L1236 secreted large amounts of IL13 and M-CSF, whereas DLBCL cells secreted neither (Fig. S5). We monitored MRC1 GE after 6 and 24 h of stimulation. There was no detectable expression of MRC1 in unstimulated and M-CSF-stimulated monocytes after 6 h. However, upon stimulation with either IL13 or L428-CM, MRC1 was expressed at 6 h, and even more so after 24 h. Addition of M-CSF did not further increase IL13-induced MRC1 expression. Compared to IL13 only, L428-CM-induced expression of CD206/ MRC1 was about tenfold higher after 6 h and remained twofold higher after 24 h. In addition, the surface expression of CD206 was analyzed after 24 h and 7 days (Fig. 4B). In concordance with MRC1 GE levels, surface expression was increased in IL13 and L428-CM-treated cells after 24 h but not in untreated controls or M-CSF-stimulated cells. Again, L428 CMtreated cells expressed about four times more CD206 than IL13-treated cells. Its expression further increased after 7 days and was then also detectable on unstimulated and M-CSF-treated cells. This suggests the activation of an intrinsic program, which needs more time to support MRC1 expression, and that HL-derived IL13 promotes this process.
Having shown that IL13 was involved in the regulation of MRC1 GE and CD206 protein abundance on the cell surface of Mφ, we tested the importance of Jak/STAT signaling. We monitored MRC1 GE in monocytes after 24 h of stimulation with or without ruxolitinib, which targets Jak1 and 2, or pyridone 6, which inhibits all four Janus kinases ( Fig. 4C; Thompson et al., 2002;Verstovsek, 2013). Both IL13 and L428-CM induced MRC1 expression as shown above. The co-incubation with either ruxolitinib or pyridone 6 completely abrogated MRC1 gene activation by IL13 or L428-CM. This argues for a Jak/STAT-dependent regulation of MRC1 expression. It also supports the presence of other Jak/STAT regulating factors in L428-CM, in addition to IL13, that are responsible for an even higher upregulation of MRC1 in L428-CM-educated Mφ. In parallel, an inhibitor screen was performed to estimate not only the impact of ruxolitinib or pyridone 6 on MRC1 expression but also to test their impact on the differentiation of Mφ. We performed Mφ differentiation as described above for 7 days in Teflon-coated cell culture bags either with M-CSF or L428-CM with and without ruxolitinib or pyridone 6, in comparison to Ibrutinib and BKM120 (Fig. 4D,E). The strongest effect was observed for Ibrutinib in L428-CM-educated Mφ. The pan-Jak inhibitor pyridone 6 reduced Mφ differentiation to nearly 50%, whereas ruxolitinib seemed to affect only the outcome of M-CSF-differentiated Mφ. Both M-CSF and L428-CM-induced Mφ differentiation was inhibited by the PI3K inhibitor BKM120. Our observation argues for the involvement of the tyrosine kinase BTK, PI3K and the Jak/STAT pathway in Mφ differentiation. It also indicates that IL13 and IL13-activated Jak/STAT signaling, albeit important for MRC1 expression and CD206 upregulation, is only partially involved in L428-CM-mediated TAM differentiation. Furthermore, our data imply that M-CSF also activates the Jak/STAT and PI3K pathways, with differentiation being dependent to some extent on BTK. These similarities and differences between M-CSF and HL-educated Mφ need to be further dissected in the future.

Lymphoma-Mφ interactions in a preclinical model
To test the interaction of lymphoma cells with monocytes or Mφ in an in vivo setting, we used the CAM assay as it recapitulates important aspects of human lymphoma as shown recently (Klingenberg et al., 2014a(Klingenberg et al., , 2014b. One million Mφ derived by incubation of monocytes with HL-CM were mixed with 2 9 10 6 lymphoma cells (L428, L1236) in Matrigel and applied on the CAM. Four days after application, lymphomas were harvested and characterized macroscopically and Fig. 4. CD206/MRC1 is upregulated by IL13 and L428-CM. (A) CD14 + PBMCs were stimulated with 2.5 ngÁmL À1 M-CSF, 10 ngÁmL À1 IL-13, or both, or with L-428-CM for the indicated time. Gene expression of MRC1 was analyzed by qRT-PCR, relative to GAPDH (mean AE SD, n = 6). (B) Monocytes were seeded in Teflon-coated cell culture bags with 2.5 ngÁmL À1 M-CSF, 10 ngÁmL À1 IL-13, or both, or with L-428-CM mixed with an equal volume of fresh medium. Aliquots of Mφ were taken at the indicated time points and stained with anti-CD206 antibody. MFIRs were calculated as described above (mean AE SD, n = 6; paired one-way ANOVA with Bonferroni's post-test; C) CD14 + PBMCs were preincubated with DMSO or inhibitors for 1 h before 10 ngÁmL À1 IL-13 or L-428-CM were added. Monocytes were cultured for six additional hours. Gene expression of MRC1 was analyzed by qRT-PCR, relative to GAPDH and MRC1 expression in L428-CM-treated monocytes (mean AE SD, n = 6). (D/E) Mφ were differentiated in Teflon-coated cell culture bags with either 2.5 ngÁmL À1 M-CSF (D) or L428 CM (E) for 7 days in the presence of DMSO or indicated inhibitors (1 µM), and cells were counted (mean AE SD, n = 4/6; paired one-way ANOVA with Bonferroni's post-test; *P < 0.05, and ***P < 0.001). histologically by defining their volume and investigating the topography of lymphoma cells and Mφ in relation to lymphatics and blood vessels (Fig. 5).
The first observation concerned the reduced size of hemorrhages in lymphomas with Mφ (Figs 5A,B and  S6A). This was confirmed by micro-CT using PTA to enhance the contrast of blood (Cunningham and Crane, 1966;Saccomano et al., 2018). This may also explain in part the smaller size of lymphomas co-inoculated with Mφ compared to lymphomas without Mφ (Figs 5A,B and S6B). Moreover, immunostaining with a-CD30 revealed that HL-derived CAM lymphomas have a more compact morphology, with lymphoma cells invading the CAM and occupying the whole area between the upper chorionic and lower allantoic epithelium (Figs 5C and S6C). Staining CAM-Mφlymphoma specimens with antibodies against CD30, CD68, and CD206 revealed a compartmentalization of the tumor (Figs 5D-F and S6D). CD30-positive HL cells were located in the upper and lower parts of the tumor and were almost absent in the center. CD68À and CD206-positive Mφ dominated the upper and central parts and were present both in the remaining Matrigel and intermingled with tumor cells that had invaded the CAM. Comparable results were observed when applying CD14 + PBMCs together with L428 cells on the CAM, further supporting the observation that HL cells support the differentiation of monocytes to M2-like Mφ with high CD206 expression (Fig. S6E).
A detailed immunofluorescence analysis of consecutive sections using costaining against CD30 or CD68 with Prox1, a marker for lymphatic endothelial cells, revealed the absence of lymphatic vessels in L428 CAM lymphomas (Fig. 5G). In CAM-Mφ-lymphomas, lymphatic vessels were regularly observed at the invasive front (Figs 5H and S7). Dissemination of lymphoma cells in lymphatics was regularly visible in CAM-Mφ-lymphomas (Figs 5I and S7).

CD206 and primary Hodgkin lymphoma cases
Immunohistochemical analysis of a tissue microarray of 16 HL specimens showed that a few cases to feature a large count of CD206-positive cells in the lymphoma tissue ( Fig. 6A-H). Additional immunofluorescence analysis revealed that while CD163 and CD206-positive Mφ are rare in the germinal center of reactive tonsils, they are readily detected in HL tissues, including CD163/CD206 double-positive Mφ (Fig. S8).
Analysis of publicly available patient microarray data revealed a significant and linear trend for high CD206 expression in HL patients with stage IV (Fig. 6I). This supports the view that Mφ with high CD206 and, thus, probably increased matrix-remodeling capacity are characteristic for advanced stages of classical HL.

Discussion
The lymphoma microenvironment plays an important role in the pathogenesis of HL. TAMs are a regular element of this microenvironment. Hence, it is essential to understand the mechanisms that enable HL cells to recruit and reprogram monocytes into TAMs (Aldinucci et al., 2010;Steidl et al., 2011). However, despite the negative impact of increased counts of M2-like Mφ on outcome and treatment response in HL, there is still more to be learnt about the mechanisms driving the development of HL-associated Mφ (Scott and Steidl, 2014). Our data provide a mechanistic insight into the processes that are involved in the interaction of HL cells with Mφ. We present a combined analysis of transcriptome and proteome data to compare lymphoma-educated to M-CSF-differentiated Mφ. We extend recent data and models, which have proposed CXCL13, CCL5, and M-CSF as central players in the interaction of HL cells with Mφ (Carey et al., 2017;Casagrande et al., 2019;Ford et al., 2015;Locatelli et al., 2018;Ruella et al., 2017;Scott and Steidl, 2014;Tudor et al., 2014). We demonstrate that Jak/STAT signaling and, especially, IL13 are involved in the regulation of the expression of MRC1 and the abundance of CD206 on the surface of HL-educated Mφ. Thus, our study expands the role of IL13 from an autocrine growth factor to a lymphoma-secreted factor that mediates ECM remodeling and immunosuppression by HL-educated Mφ (Skinnider and Mak, 2002), which in turn facilitates the dissemination of lymphoma cells.
We observed significantly higher counts of Mφ upon differentiation with HL-CM as compared to DLBCL-CM or M-CSF. Although M-CSF is expressed by the majority of HL cell lines, and M-CSF is known to promote Mφ differentiation into M2-like subtypes, the observed phenotype and the strongly enhanced CD206 surface expression cannot be attributed to M-CSF in HL-CM. We observed that IL13 is involved in early MRC1 mRNA upregulation and enhanced CD206 incorporation into the cell membrane. Enhanced IL13 expression is characteristic of HL, and it has been shown that IL13 can promote Mφ differentiation (McKenzie et al., 1993;Skinnider et al., 2001). Conditioned media of L428 and L1236 cells, which featured the highest capacity to attract monocytes and Mφ, were also characterized by a high potential to activate MRC1 GE in Mφ. Both HL cell lines secreted large amounts of M-CSF and IL13. The same was true for HDLM2 cells (Fig. S5), but in contrast to L428 and L1236 cells, they failed to induce MRC1 GE in Mφ (Fig. 3A) for reasons yet to be elucidated. However, the observation hints at other not yet identified Jak/ STAT activating factors that complement IL13. The latter binds to an IL4Ra-IL13Ra heterodimer, which intracellularly can interact with JAK1, JAK2, and TYK2 (Bhattacharjee et al., 2013). The absence of MRC1 expression in Mφ after ruxolitinib treatment indicates that either JAK1 or JAK2 rather than TYK2 induces M2-like GE. Thus, IL13 seems to be not only involved in the autocrine support of HL cells but also in rebuilding the lymphoma microenvironment through Mφ differentiation and upregulation of CD206. This offers novel opportunities for targeting the lymphoma microenvironment by Jak/STAT inhibition (Kim et al., 2014). Whether this approach can be used in combination or extension of immune checkpoint inhibitor therapies needs to be evaluated in future studies.
Preliminary data show the additional involvement of BTK-and PI3K-regulated pathways, as respective inhibitors affect both M-CSF-and L428-CM-mediated Mφ differentiation. This supports the recent observation of a preclinical model that used RP6530 to target the interaction between HL cells and Mφ (Locatelli et al., 2018). We propose that M-CSF induces a network of signaling pathways to activate autocrine Mφ differentiation factors, whereas IL13 directly promotes an early differentiation program. Although elevated M-CSF serum levels have been reported in HL patients, and HL cells were found to be M-CSF-positive by immunohistochemistry, the measured M-CSF amounts in the lymphoma-CM in vitro and the differences in Mφ counts and phenotype indicate the involvement of additional factors during HL-specific Mφ differentiation (Casagrande et al., 2019;Kowalska et al., 2012;Zheng et al., 1999). Recently, an autoregulatory loop between TAMs and breast cancer cells has been reported (Cassetta et al., 2019). It might be interesting to test whether described factors such as tumor necrosis factor alpha, SIGLEC1 and CCL8, which is self-reinforcing through the production of M-CSF, also participate in the interaction of HL cells with Mφ.
CD206 is a carbohydrate binding receptor expressed by Mφ, dendritic cells, and lymphatic endothelial cells. It belongs to a family of endocytic receptors with conserved structural elements consisting of an N-terminal cysteine-rich (CR) domain, a fibronectin type II (FNII) domain, and several C-type lectin-like domains (CTLDs; Taylor et al., 2005;Wilting et al., 2009). Through the CTLDs, CD206 can bind to glycoconjugates carrying terminal mannose, N-acetylglucosamine, and fucose residues, while the FNII and CR domain bind to collagens and sulfated carbohydrates, respectively. CD206 is a highly effective endocytic receptor that constantly recycles between the plasma membrane and the early endosomal compartment. Various exogenous and endogenous ligands have been described for C-type lectins, implicating several functions in homeostasis and inflammation. We observed that HL-CMeducated Mφ were able to take up collagen, which concords with previous reports on M2-like Mφ (Madsen et al., 2013). This suggests that these M2-like Mφ play a specific CD206-mediated role in lymphoma tissue reorganization. Matrix remodeling is a common process in tumor progression, and the tumor stroma is characterized by profound proteolytic degradation (Cox and Erler, 2011;Nakayama et al., 2016;Tataroglu et al., 2007). The most striking feature of the nodular sclerosis subtype of classical HL (NSCHL) is the presence of ECM deposits. These are collagen-rich fibrotic bands surrounding nodules of inflammatory and neoplastic cells. Of note, L428 cells are derived from NSCHL. The observed capacity of L428-CM to induce a specific Mφ subtype with high CD206 expression and collagen uptake supports the hypothesis of tissue remodeling by the interaction of lymphoma cells with stromal cells, and adds an additional element to the previously observed mast cell infiltration and fibrosis in HL (Nakayama et al., 2016). It has to be mentioned that C-type lectins facilitate the endocytosis of particular antigen types for cross-presentation, but whether antigen uptake by the C-type lectins will result in immunogenic or tolerogenic cytotoxic T-cell priming remains to be studied (Allen and R€ uckerl, 2017). Taking into account the capacity of HL-CM to enhance CD40 and PD-L1 cell surface expression on Mφ, and the recently described close association of HL cells with PD-L1positive Mφ, it is reasonable to assume that HL-educated Mφ modify T-cell functions at different levels (Carey et al., 2017).
In summary, we hypothesize that the observed in vitro and in ovo regulation of CD206 in Mφ might also occur in vivo and contribute to CD206 abundance in advanced stages of HL. Our data identify CD206 as a potential target for the modulation of Mφ activation in addition to the recently discussed targeting of the CSF1R (Luo et al., 2006;Mart ın-Moreno et al., 2015;Pham et al., 2018;Pyonteck et al., 2013;Ries et al., 2014;Wang et al., 2018;Zhu et al., 2014). Inhibition of the Jak/STAT pathway targets not only lymphoma cells directly, but may also affect lymphoma-educated Mφ development and/or activity. This supports the simultaneous targeting of both lymphoma cells and Mφ with PI3K inhibitors, and the use of Maraviroc for the blocking of CCR5 in these cells (Casagrande et al., 2019;Locatelli et al., 2018). In line with this, targeting the pattern recognition receptor MARCO can alter Mφ polarization and result in reduced tumor growth (Georgoudaki et al., 2016). This type of reprogramming may now be tested in HL, as HL-educated Mφ show specific high MARCO expression. Furthermore, imaging of TAMs seems to be a valuable approach and mannose-coated liposomes or nanobodies are tested for their use in monitoring CD206 TAMs using PET (Azad and Schlesinger, 2015;Locke et al., 2012;Movahedi et al., 2012).

Conclusion
The present study further corroborates the shown capacity of HL cells to attract and differentiate monocytes into alternatively activated M2-like Mφ. The latter are characterized by high CD206 expression, which indicates that HL-TAMs act as immunosuppressive agents. Further, CD206 is associated with a high capacity to bind and take up glycoconjugates with terminal mannose residues as well as type I collagen. Thus, it is likely that CD206 contributes to the remodeling of the tumor microenvironment. Our study highlights the role of Jak/STAT signaling and, particularly, the capacity of IL13 to upregulate CD206 expression on Mφ, thus expanding its role from an autocrine growth factor in HL to a lymphoma-secreted factor capable of modulating the tumor microenvironment. Additionally, the increased expression of CD40 and PD-L1 argues for a role of HL-secreted factors in the reprogramming of monocytes/ Mφ to further support the capacity of HL cells to suppress, for instance, Tcell-mediated anti-lymphoma responses. The CAM assay presents a simple preclinical model to study HL cell-Mφ interactions and the dissemination of HL cells. Finally, MRC1/CD206 expression data show a positive correlation with the stage of HL patients.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. Hodgkin lymphoma cells secret factors that attract and differentiate monocytes into macrophages.      Fig. S6. Mφ affect lymphoma growth in the chick chorioallantoic membrane (CAM), blood vessel destruction and tissue remodeling. Fig. S7. Dissemination of lymphoma cells in lymphatics in CAM-Mφ-lymphomas. Fig. S8. CD163 and CD206 double staining in a representative tonsil and HL tissue sections. Table S1. Differential gene expression of macrophages comparing HL-CM and M-CSF differentiated macrophages. Table S2. Differential gene expression of macrophages comparing HL-CM and DLBCL-CM differentiated macrophages. Table S3. Differential protein abundance of macrophages comparing HL-CM and M-CSF differentiated macrophages. Table S4. Differential protein abundance of macrophages comparing HL-CM and DLBCL-CM differentiated macrophages. Table S5. Differential protein abundance of macrophages comparing DLBCL-CM and M-CSF differentiated macrophages.