Monocyte-Derived Leukemia-Associated Macrophages Facilitate Extramedullary Distribution of T-cell Acute Lymphoblastic Leukemia Cells

Macrophages take part in a broad range of physiologic and pathologic processes. They arise from successive waves of embryonic and adult hematopoiesis. Despite the fact that tissue-resident macrophages are established prenatally and self-maintained throughout adulthood, monocyte-derived macrophages (MOMF) are indispensable in tissues (1). A proportion of resident macrophages are constantly replaced by MOMF during adulthood, although the kinetics and extent of replacement vary among tissues (2). Furthermore, massive infiltration of MOMF occurs in most tissues upon inflammation (3, 4).

Macrophages exhibit great plasticity and diversity in differentiation and activation. The well-accepted M1 (classically activated) and M2 (alternatively activated) macrophages represent only two extremes of activation phenotypes. In fact, most macrophages are shaped by a specific microenvironment and are signal-dependently activated into a continuum of states between the M1 and M2 extremes in vivo, although they are always classified as M1/M2 or M1/M2-like macrophages (5). Normally, the phenotype of MOMF eventually becomes close to that of local resident macrophages (6). However, it was proposed that MOMF and resident macrophages might play different roles, especially under pathologic conditions (7). MOMF is highly phagocytic and inflammatory, whereas resident microglia show a suppressed cellular metabolism in the inflamed central nervous system (4, 8). MOMF promotes, whereas resident alveolar macrophages dampen, acute allergic lung inflammation (9). Nevertheless, the different pathological roles of MOMF and resident macrophages have not been fully revealed due to the complexity and an inadequate understanding.

Macrophages play important pathologic roles in malignancies and tumor-associated macrophages (TAM) are widely studied. In many cases, TAM participates in forming a tumor microenvironment favorable for tumor progression via multiple mechanisms (10). In some cases, TAM suppresses tumor progression by promoting immune responses against tumor cells (11). Although they are normally regarded as M2 macrophages, TAM from different tumors or different regions of tumors exhibit heterogeneous activation phenotypes and different activation subpopulations have been identified (12). Currently, it is widely accepted that TAM subpopulations with more M1-like characteristics have antitumor effects whereas those with more M2-like characteristics have protumor effects. It is worth noting that most TAM is MOMF. MOMF has a close relationship with inflammation, which is proposed as the seventh hallmark of cancer (13). Tumor-induced inflammation, a kind of sterile inflammation, plays important roles in tumor progression. Although a phenotypic shift in TAM (from M1-like to M2-like) has been observed, the evolution details and pathologic roles of TAM subpopulations have not been well elucidated.

Compared with solid tumors, leukemia has unique pathologic and clinical features. Leukemia-associated macrophages (LAM) are different from TAM in gene expression and activation phenotype (14). It seems that LAM is M2-like macrophages with pro-leukemic effects because repolarizing LAM with more M1-associated characteristics (15) or depletion of patrolling monocytes and macrophages delays leukemia progression (16). However, the recruitment of different immune cells into the leukemic microenvironment is observed (14, 17–20), and LAM may also participate in the leukemia-induced inflammation and exhibit M1-like characteristics. In fact, LAM from different tissues is heterogeneous and subpopulations of LAM with different activation phenotypes have been classified in both mouse and human studies (14, 18, 19, 21). Does MOMF contribute to LAM and how it evolves? Which LAM subpopulation contributes to hepatosplenomegaly, which is frequently observed in T-cell acute lymphoblastic leukemia (T-ALL) patients? As severe hepatosplenomegaly is associated with inferior prognosis and is considered as a valuable marker that has a negative effect on the quality of patient lives (22, 23), what is the pathological role of a particular LAM subpopulation in leukemia-associated inflammation in extramedullary tissues? We hypothesized that a specific LAM subpopulation might contribute to the leukemia-induced sterile inflammation and participate in the pathologic process of hepatosplenomegaly in T-ALL.

Here, we used Notch1-induced T-ALL models in wild-type (WT) and CCR2^−/− mice to answer these questions. MOMF contributed to macrophage subsets in resting and leukemic liver and spleen. The Ly6C⁺ monocytes differentiate into the Ly6C⁺ subset in the spleen. More monocyte-derived LAM, especially the Ly6C⁺ subset, accumulated in extramedullary tissues of leukemic mice. The Ly6C⁺ LAM exhibited increased M1-related characteristics and contributed to sterile inflammation. In CCR2^−/− leukemic mice, reduced Ly6C⁺ LAM, relieved sterile inflammation, and reduced distribution of leukemia cells were detected. In addition, Ly6C⁺ LAM expressed high levels of CCL8 and CCL9/10. Blocking CCR1 and CCR2 relieved the hepatosplenomegaly and inhibited extramedullary distribution of leukemia cells in T-ALL mice.

C57BL/6J (CD45.2) and B6.SJL (CD45.1) mice were provided by the Animal Centre of the Institute of Hematology and Blood Diseases Hospital, CAMS & PUMC. The CCL2/CCR2 axis is essential for BM Ly6C⁺ monocyte trafficking and B6.129S4-Ccr2^tm1Ifc/J (CCR2^−/−) mice are widely used to study physiologic and pathologic processes lacking inflammatory monocytes and macrophages (24, 25). CCR2^−/− mice were introduced from the Jackson Laboratory and bred in the same animal center. Six- to eight-week-old mice were used and randomly assigned to different groups. In most experiments, female mice were used. In the survival experiments, five male and five female mice were used each group. All mice were maintained in an SPF-certified facility and all procedures for animal experiments were approved by the Animal Care and Use Committees at the Institution.

The Notch1-induced nonirradiated T-ALL model has been described previously (14, 26). Briefly, GFP⁺ T-ALL cells were isolated from leukemic mice and transplanted into C57BL/6J (CD45.2) and B6.SJL (CD45.1) mice (5 × 10⁵ cells/mouse) without irradiation. The T-ALL model in CCR2^−/− mice was established by transplantation of the same T-ALL cells. All mice suffered and died from T-ALL. No gender difference in the disease progression and survival was observed in C57BL/6J and B6.SJL mice. The distribution of leukemia cells in peripheral blood (PB), bone marrow (BM), and spleen was monitored. Days 10, 15, and 20 were selected as typical time points for the early, middle, and late stages of leukemia, respectively. Mice were sacrificed by cervical dislocation. Tissue samples were collected for further analyses.

All cells were cultured in a humidified atmosphere of 5% CO₂ at 37°C. All culture supplies were endotoxin free. A fluorescence-conjugated antibody against F4/80 (BM8, APC) was purchased from BD Biosciences. Fluorescence-conjugated antibodies against CD45.1 (A20, FITC), CD45.2 (104, PerCP-Cy5.5), CD45.2 (104, PE), CD3 (17A2, PerCP-Cy5.5), Gr-1 (RB6-8C5, Pacific blue), CD115 (AFS98, PE), CD11b (M1/70, APC-Cy7), Ly6C (HK1.4, PE-Cy7), Ki67 (16A8, PE), CCR1 (S15040E, APC), and CCR2 (SA203G11, Brilliant Violet 421) were purchased from Biolegend.

Spleen and liver cells were harvested as single-cell suspensions in PBS with EDTA and filtered through graded nylon filters. After removal of red cells by ammonium chloride lysing buffer, cells were resuspended in PBS containing 1% FBS, stained with a combination of antibodies, washed, and subjected to FACS. The Canto II and LSR II flow cytometers (BD Biosciences) were used for FACS analysis and the FACS Aria III (BD Biosciences) was used for cell sorting. FlowJo software (version 7.6.1) was also used for data analysis. Standard protocols were followed for all experiments. The splenic macrophages were gated as the Gr-1^loF4/80⁺CD115^intSSC^int/lo subpopulation from the GFP⁻CD3⁻ population. An antibody against Ly6C was used to identify macrophage subsets. The hepatic macrophages were gated as GFP⁻CD45⁺F4/80⁺CD11b⁺Ly6C^int/−.

Macrophage subsets from the spleens of normal and leukemic mice were sorted and spun to slides. Wright-Giemsa staining was performed according to manufacturer's instructions. The slides were examined under a light microscope (AXIO Observer A1; ZEISS).

Peripheral blood mononuclear cells (PBMC) were prepared from normal CD45.2⁺ donors following standard Ficoll separation protocols. Ly6C⁺ and Ly6C⁻ monocytes were sorted from CD45.2⁺CD115⁺CD11b⁺ monocytes. PBMCs or monocyte subsets were intravenously injected into normal or T-ALL (day 10) recipients (CD45.1⁺). PB and spleen samples were collected at the indicated time points for further analysis of donor-derived cells by flow cytometry.

Cells were stained with propidium iodide (PI) according to standard protocol. Briefly, cells were fixed and permeabilized by cold 70% ethanol on ice for 1 hours. Then RNA was digested by RNase at final concentration of 0.2 to 0.5 mg/mL for 1 hours at 37°C. At last, PI was added at the final concentration of 10 μg/μL prior to analysis on FACS. Leukemia cells were fixed and permeabilized by Cytofix/Cytoperm TM Fixation/Permeabilization Solution Kit (BD Biosciences). Then cells were stained with anti-Ki67 before analyzed by flow cytometry.

Latex bead uptake experiments were used to measure the phagocytic activity of macrophage subsets. FACS-sorted macrophage subsets from the spleen of healthy and leukemic mice on day 15 were incubated for 15 minutes with FITC-labeled 2-μm latex beads (Sigma-Aldrich) before flow cytometry analysis.

Coculture assays were used to test the effects of LAM subsets on leukemia cells. spleen macrophage subsets from healthy and leukemic mice (day 15) were sorted by flow cytometry. They were respectively cocultured with sorted GFP⁺ T-ALL cells at the ratio of 1:2 in 24-well plates in α-MEM supplemented with 10% FBS and antibiotics (HyClone) for 8 hours under a Live Cell Imaging System (Nikon Ti). Leukemia cells cultured alone were used as a blank control. Randomly selected fields were photographed. GFP⁺ cells at 0 hours (control) and 8 hours (test) from each well were counted by using ImageJ software (version 7.6.1). The phagocytotic rate was calculated as follows: ((control − test)/control − (blank control − blank test)/blank control) × 100%.

Transwell experiments were used to measure the in vitro migration activity of cells according to the manufacturer's instructions (Millipore). To test the migration of macrophage subsets, macrophages from normal and leukemic mice (n = 5 each group) were sorted and pooled, separately. Then, 5 × 10⁴ spleen macrophages were seeded into the upper compartments (8.0 μm) in 200 μL of RPMI1640 whereas 1.5 × 10⁵ normal mouse spleen cells were placed in the lower compartments in 600 μL of RPMI1640 with 0.5% BSA. After 24 hours, membrane inserts were collected and noninvading cells on the upper surface were removed. Macrophages on the lower surface were stained with 0.1% crystal violet for 30 minutes and counted in at least six random fields under a microscope (ZEISS) for each membrane.

To study the in vitro migration of leukemia cells to the spleen leukemic microenvironment, 8 × 10⁵ GFP⁺ leukemia cells were seeded in the upper compartment (5.0 μm) in 200 μL of RPMI1640 whereas 600 μL of supernatant (2 × 10⁸ spleen cells from normal and WT or CCR2^−/− leukemic mice were suspended in PBS with 0.5% BSA for 0.5 hours at 4°C) was placed in the lower compartment. After 6 hours, T-ALL cells that migrated into the lower compartment were collected, and cells from each well were counted by hemocytometer.

To identify important chemokines mediating the migration of leukemia cells, sorted leukemia cells were first preincubated with or without the CCR1 antagonist BX471 (1 μmol/L, Sigma-Aldrich) or the CCR2 antagonist INCB3344 (15 nmol/L, Sigma-Aldrich) for 30 minutes, and then seeded in the upper compartments (5.0 μm). Supernatant with or without CCL8 or CCL9/10 (PeproTech) at different concentrations was placed in the lower compartments. After 6 hours, T-ALL cells that migrated into the lower compartments were collected, and cells from each well were counted by hemocytometer.

WT and CCR2^−/− leukemic mice were daily intraperitoneally administered with BX471 (20 mg/kg) or INCB3344 (30 mg/kg) in 200 μL of vehicle for 8 days from day 10 after transplantation of leukemia cells (n = 5). Mice were sacrificed on day 17 to check the effects of the CCR1 and the CCR2 antagonists on the distribution of leukemia cells in vivo.

A single cell suspension of spleen (3 × 10⁸) or liver (5 × 10⁸) cells was obtained in 500 μL of PBS and transferred to filters pre-rinsed with another 500 μL of PBS. The supernatant was harvested by centrifugation. Chemokines were screened by using a Proteome Profiler Mouse Chemokine Array Kit (R&D) according to the manufacturer's protocols. The concentrations of cytokines (TNFα, IL1β, IFNγ, eBioscience; CCL8, Abcam; CCL9/10, Thermo Scientific) were measured by ELISA according to manufacturers' protocols. At least three independent mouse samples were analyzed.

Cells were lysed and total RNA was isolated using RNeasy Mini Kit (Qiagen). Reverse transcription was achieved using SuperScript III (Invitrogen). Real-time PCR was performed on the StepOne real-time PCR system (Applied Biosystems) and ABI-Prism 7900 Sequence Detector (Applied Biosystems). The expression level of target genes was analyzed by the RQ value calculated through ΔΔC_t method [ΔΔC_t = (Ct_TARGET − Ct_GAPDH)_sample − (Ct_TARGET − Ct_GAPDH)_calibrator]. The sequences for all primers are listed in Supplementary Table S1.

Spleen macrophage subsets were sorted by flow cytometry from healthy and leukemic mice (day 15). RNA sequencing (RNA-seq) was carried out by the Beijing Genomics Institute (BGI) following standard protocols. The library products were sequenced using an Illumina HiSeq 4000. Standard bioinformatics analyses were performed by BGI. K-Mean cluster, gene ontology (GO), and pathway analyses were performed by MeV 4.8.1, FunNet, and Kyoto Encyclopedia of Genes and Genomes (KEGG), respectively. The RNA-seq data are available in the National Center for Biotechnology Information Gene Expression Omnibus database under accession number GSE101621.

A two-dimensional illustration was established to describe the activation phenotype of macrophages (15). For each gene, the value of the Ly6C⁺F4/80⁺ macrophage subset from normal spleen or liver was designated as 0, respectively. The M1 and M2 values of a specific macrophage subset at a specific time point were calculated as follows and the mean values of relative expression of genes were used for calculation:

GraphPad Prism 6.0 software and the spleenSS17.0 software package (SPSS) were used. All results are expressed as the mean ± SEM. Comparisons between two groups were analyzed by an unpaired Student t test, whereas comparisons of multiple groups were analyzed by one-way ANOVA. Survival time was compared by Kaplan–Meier analysis. Statistically significance was accepted when the P value was less than 0.05.

Spleen and liver are the major extramedullary hematopoietic organs, which are most frequently infiltrated by T-ALL cells. Two macrophage subsets have been classified in liver (27). CD45⁺F4/80⁺CD11b⁺Ly6C^int/− hepatic macrophages (Supplementary Fig. S1A) were classified into Ly6C⁺F4/80⁺ MOMF, which accounted for 35% of total macrophages, and Ly6C⁻F4/80⁺ Kupffer cells (KC), which are fetal origin (Fig. 1A). CD3⁻CD115^intGr1^loF4/80⁺SSC^lo spleen macrophages (Supplementary Fig. S1B) were subdivided into three distinct subsets: Ly6C⁺F4/80⁺, Ly6C⁻F4/80⁺, and Ly6C⁻F4/80^hi (abbreviated as Ly6C⁺, Ly6C⁻, and F4/80^hi, respectively). The majority were Ly6C⁻ macrophages. All of them had kidney- or horseshoe-shaped nuclei, but Ly6C⁺ and Ly6C⁻ macrophages were smaller than F4/80^hi macrophages (Fig. 1B). RNA-seq analysis revealed that different subsets had different expression profiles of monocyte/macrophage-related genes (Fig. 1C).

To investigate the contribution of MOMF to macrophage subsets in the resting spleen, CD45.2⁺ PBMCs, Ly6C⁺, or Ly6C⁻ monocytes (Supplementary Fig. S1C) were intravenously transplanted into CD45.1⁺ recipient mice (Fig. 1D). Donor-derived macrophages were detected in all subsets 24 hours after transplantation, although the main subset was Ly6C⁻ (Fig. 1E; Supplementary Fig. S1D). Although Ly6C⁺ monocytes can differentiate into Ly6C⁻ monocytes in peripheral blood (PB; ref. 28), Ly6C⁻ monocytes appeared after 18 hours, and the majority of monocytes were Ly6C⁺ (Fig. 1F; Supplementary Fig. S1E). By contrast, Ly6C⁺ monocytes gave rise to Ly6C⁺, Ly6C⁻, and F4/80^hi macrophages within 18 hours, and the main subset was Ly6C⁻. Furthermore, Ly6C⁻ monocytes could not differentiate into Ly6C⁺ macrophages, and the main subset was Ly6C⁻ (Fig. 1G; Supplementary Fig. S1D). Collectively, MOMF contributed to macrophage subsets in the resting liver and spleen. In spleen, only Ly6C⁺ monocytes gave rise to Ly6C⁺ macrophages, which further differentiated into Ly6C⁻ and F4/80^hi macrophages.

A mouse T-ALL model was used to study the role of MOMF in leukemia. During leukemia progression, the percentage of total LAM in the liver and spleen increased at the early stage and then decreased. Furthermore, there was a significant increase in LAM subsets, that is liver MOMF, spleen Ly6C⁺ macrophages, and Ly6C⁻ macrophages (Fig. 2A and B). To determine whether PB monocytes contribute to increased spleen LAM and LAM subsets, CD45.2⁺ PBMCs were transplanted into CD45.1⁺ leukemic mice at the early stage (Fig. 2C). Increased numbers of donor-derived LAM (about 2.5-folds) was detected (Fig. 2D). Furthermore, the increased percentage of the Ly6C⁺ cells (Fig. 2E) corresponding to 4.8-folds increase in cell number was also detected. Cell proliferation assays showed that Ly6C⁻ and F4/80^hi LAM was more proliferative than their normal counterparts, whereas no difference was found between Ly6C⁺ LAM and their normal counterparts (Supplementary Fig. S2A). Hence, the increase in Ly6C⁺ LAM was mainly resulted from their differentiation from Ly6C⁺ monocytes.

To study the activation phenotype of spleen LAM subsets, the dynamic expression of phenotype-associated genes (6 genes for M1 and 6 genes for M2) was monitored. The LAM subsets showed considerable differences (Fig. 2F). To obtain an intuitive views of the phenotypic evolution of LAM subsets, a two-dimensional illustration of macrophage phenotypes was used (15). Interestingly, Ly6C⁺ LAM was in M1 area at the early and middle stages of leukemia, whereas both splenic Ly6C⁻ and F4/80^hi LAM was in M2 area during leukemia progression. The phenotype of MOMF in the liver was also in the M1 area at the early and middle stages of leukemia (Supplementary Figs. S2B and S2C). These results suggested that the liver and spleen leukemic microenvironments recruited an increased number of monocyte-derived Ly6C⁺ LAM, which had more M1-related characteristics.

RNA-seq was performed to study gene expression profiles in spleen LAM subsets. The expression levels of M1- and M2-related genes are shown in Supplementary Fig. S3A. Hierarchical cluster (HCL) of differentially expressed genes (DEG) between LAM subsets and their normal counterparts, which reflected the modulation effects of leukemic microenvironment on the gene expression in LAM subsets, demonstrated that the response of Ly6C⁺ subset was different from the other two subsets (Fig. 3A). The numbers of DEGs between LAM subsets and their respective normal counterparts are shown (Supplementary Fig. S3B). These DEGs were subjected to GO enrichment (Supplementary Fig. S3C) and KEGG analysis (Fig. 3B). The KEGG results showed that the response of Ly6C⁺ LAM to leukemic microenvironment was different from Ly6C⁻ or F4/80^hi LAM, which shared eight identical annotations. It is interesting that the top two annotations of the Ly6C⁻ or F4/80^hi LAM were associated with cell proliferation, whereas those of the Ly6C⁺ LAM were associated with active protein expression. Notably, the annotations “cytokine-cytokine receptor signaling pathway” and “phagosome” were enriched in Ly6C⁺ LAM (Fig. 3B). Some of the related genes were selected and verified by quantitative RT-PCR (Fig. 3C). These results suggested that Ly6C⁺ LAM showed distinct transcriptional modulations in response to the leukemic environment from Ly6C⁻ or F4/80^hi LAM, which might result in different biological activities.

Migration and phagocytosis are important functional characteristics of macrophages. In the steady state, the migration of Ly6C⁺ and Ly6C⁻ macrophages was higher than that of F4/80^hi macrophages. In the leukemic microenvironment, increased migration was detected in Ly6C⁺ LAM, whereas decreased migration was detected in Ly6C⁻ and F4/80^hi LAM (Fig. 3D). Notably, the migration of Ly6C⁺ LAM was higher than that of Ly6C⁻ or F4/80^hi LAM. Both non-specific and specific phagocytosis of LAM was assessed. Non-specific phagocytosis was assessed by latex bead uptake experiments. No difference was detected in Ly6C⁺ LAM. Upregulation and downregulation of phagocytosis were detected in Ly6C⁻ and F4/80^hi LAM, respectively (Supplementary Figs. S4A and S4B). Specific phagocytosis against T-ALL cells was assessed by a live cell imaging system. No difference in phagocytosis was detected in three subsets in the steady state. Ly6C⁺ LAM maintained a high phagocytotic ability, whereas Ly6C⁻ and F4/80^hi LAM had a decreased phagocytotic ability (Fig. 3E). Most importantly, the phagocytotic ability of Ly6C⁺ LAM was much higher than that of Ly6C⁻ and F4/80^hi LAM. The above results indicated that spleen Ly6C⁺ LAM was distinct from the other two subsets in response to the leukemic microenvironment. The recruited spleen Ly6C⁺ LAM not only had an increased migration potential but also maintained the high phagocytotic potential.

The CCL2/CCR2 axis is essential for BM Ly6C⁺ monocyte trafficking (24, 25), and CCR2^−/− mice are widely used to study physiologic and pathologic processes lacking inflammatory monocytes and macrophages. In CCR2^−/− mice, the PB monocyte count was normal (Supplementary Fig. S4C), and tissue sections (liver, spleen, lung, kidney, and BM) showed no significant difference in the steady state (Supplementary Fig. S4D). However, the increase in monocytes, which was observed in WT leukemic mice, was absent in CCR2^−/− leukemic mice, but the ratio between monocyte subsets remained unchanged (Fig. 4A). For total liver LAM in CCR2^−/− leukemic mice, a significant decrease was detected when compared with WT counterparts, although a mild increase was observed when compared with healthy mice. Furthermore, a decrease in MOMF was observed in CCR2^−/− leukemic mice when compared with WT counterparts (Fig. 4B). In spleen, the increase in total LAM mainly resulted from Ly6C⁻ and F4/80^hi LAM in CCR2^−/− leukemic mice. By contrast, the number of Ly6C⁺ LAM in CCR2^−/− leukemic mice was lower than WT leukemic mice (Fig. 4C). These results demonstrated that the increase in Ly6C⁺ LAM was absent in the liver and spleen in CCR2^−/− leukemic mice.

Furthermore, ELISA results demonstrated that the concentrations of typical proinflammatory cytokines (TNFα, IL1β, and IFNγ) in both the hepatic and splenic microenvironments were lower in CCR2^−/− leukemic mice than WT counterparts (Fig. 4D and E). These results suggested that the sterile inflammation relieved in CCR2^−/− leukemic mice. As we previously demonstrated that increased expression of TNFα and IL1β was detected in Ly6C⁺ LAM in both liver and spleen (Fig. 2F; Supplementary Fig. S2B). Hence, monocyte-derived LAM, especially the Ly6C⁺ subset, might contribute to the formation of a sterile inflammatory microenvironment in the extramedullary tissues at the early stage of leukemia, although other cell types also participate the process.

We investigated the pathologic role of monocyte-derived LAM in T-ALL. The CCR2^−/− and WT mouse groups showed similar survival times when mice were transplanted with either 1 × 10⁵ or 5 × 10³ leukemia cells (Fig. 5A). However, alleviation of hepatosplenomegaly was observed in the CCR2^−/− leukemic mice (Fig. 5B–C). The PB leukemia cells were dynamically monitored, showing a significant decrease on day 17 in CCR2^−/− mice (Fig. 5D). The distribution of leukemia cells was studied on day 17. No difference was detected in BM (Supplementary Fig. S4E). However, less leukemia cells were detected in the liver and spleen in CCR2^−/− mice than WT mice (Fig. 5E). Extramedullary tissues from leukemic mice on day 17 were subjected to HE-staining (Fig. 5F). The leukemia cell-infiltrating area in liver and spleen of CCR2^−/− leukemic mice was significantly smaller than that in WT leukemic mice. The infiltration of leukemic cells in lung and their aggregation around bronchial were more severe in WT leukemic mice than CCR2^−/− leukemic mice. Massive aggregation of leukemia cells was detected in renal capsule and medulla in WT leukemic mice, whereas less leukemia cells were detected in CCR2^−/− leukemic mice. Hence, decreased distribution of T-ALL cells in the extramedullary tissues was detected in CCR2^−/− mice. Although multiple mechanisms are involved in the process, monocyte-derived LAM might facilitate the distribution of T-ALL cells in the extramedullary tissues.

The proliferation and apoptosis of leukemia cells were studied to investigate the mechanism leading to decreased distribution of T-ALL cells in the extramedullary tissues. However, no significant difference was observed (Supplementary Figs. S4F and S4G). Transwell experiments were used to study the migration of leukemia cells to spleen leukemic microenvironments. Less leukemia cells were detected in CCR2^−/− leukemic mice than in WT counterparts (Fig. 6A). Hence, the microenvironment in CCR2^−/− leukemic mice may account for the decreased migration of T-ALL cells to extramedullary tissues.

Chemokine profiles of spleen (Supplementary Fig. S5A) and liver (Supplementary Fig. S5B) microenvironments were screened to unravel chemokines vital for the altered migration of T-ALL cells. Attention was paid to the chemokines that were upregulated in WT leukemic mice. Furthermore, the expression of these chemokines was lower in CCR2^−/− leukemic mice than in WT leukemic mice. CCL8 and CCL9/10 showed these patterns in both spleen (Fig. 6B) and liver (Supplementary Fig. S5B) microenvironments. Their levels in spleen (Fig. 6C) and liver (Supplementary Fig. S5C) were further validated by ELISA. As they are mainly expressed in blood cells (29, 30), their expressions in major blood cells in the spleen microenvironment were screened. High levels of CCL8 and CCL9/10 expression were only observed in LAM among the tested cell types (Supplementary Fig. S5D). Furthermore, the Ly6C⁺ subset expressed the highest levels of CCL8 and CCL9/10 (Fig. 6D). These results suggested that high levels of CCL8 and CCL9/10 in the leukemic microenvironments of extramedullary tissues were mainly a result of LAM, especially Ly6C⁺ LAM.

The main receptors for CCL8 are CCR1 and CCR2, whereas the main receptor for CCL9/10 is CCR1 (31, 32). T-ALL cells expressed both receptors (Supplementary Fig. S6A), suggesting that the CCL8–CCR1/CCR2 and CCL9/10–CCR1 axes might contribute to the extramedullary distribution of leukemia cells. Transwell experiments were performed to test whether antagonists of CCR1 (BX471) and CCR2 (INCB3344) blocked the migration of T-ALL cells. More T-ALL cells migrated to the leukemic spleen microenvironment in WT mice than to the healthy spleen microenvironment, and both antagonists decreased the migration of T-ALL cells to the leukemic spleen microenvironment in WT mice (Fig. 6E). Furthermore, exogenous CCL8 and CCL9/10 dose-dependently enhanced the migration of leukemia cells to the leukemic spleen microenvironment in CCR2^−/− leukemic mice (Supplementary Figs. S6B and S6C). Moreover, BX471 and INCB3344 decreased the effect of exogenous CCL8 (100 ng/mL) whereas BX471 decreased the effect of exogenous CCL9/10 (100 ng/mL; Fig. 6F).

These in vitro results suggested that monocyte-derived LAM, especially Ly6C⁺ LAM, expressed high levels of CCL8 and CCL9/10, which promoted the extramedullary distribution of T-ALL cells.

To further investigate the role of these axes on the extramedullary distribution of leukemia cells in vivo, leukemic mice were intraperitoneally administered INCB3344 or BX471 daily for 8 days (Fig. 7A). Alleviation of hepatosplenomegaly was observed upon INCB3344 treatment, whereas alleviation of splenomegaly was observed upon BX471 treatment (Fig. 7B and C). FACS analysis revealed that administration of INCB3344 or BX471 decreased the proportion of leukemia cells in both liver and spleen, but had no effect on BM leukemia cells (Fig. 7D). Different extramedullary tissues from leukemic mice on day 17 were subjected to HE-staining, showing that the leukemia cell-infiltrating area in liver, spleen, lung, and kidney in leukemic mice was significantly smaller in the INCB3344 and BX471 group than in the control group (Fig. 7E), whereas no significant difference was detected in BM samples (Supplementary Fig. S6D). These results demonstrated that blocking CCR1 and CCR2 considerably relieved hepatosplenomegaly and inhibited the extramedullary distribution of leukemia cells in T-ALL mice.

Heterogeneity is a distinct characteristic of macrophages. Diversity in their origin, activation phenotype, function, and role has been reported under both physiologic and pathologic conditions (28). Evidence shows that MOMF exerts distinct functions in various pathologic microenvironments (7). Heterogeneity of LAM in activation phenotype and function has been reported (14–16, 18, 19, 33). This study aims to elucidate the origin of LAM and the pathologic role of monocyte-derived LAM in T-ALL by using mouse models in both WT and CCR2^−/− mice.

Although it is well accepted, the M1/M2 classification of macrophage activation is far from perfect. Strictly, M1 and M2 macrophages only represent the states polarized by IFNγ/LPS and IL4/IL13, respectively (34). Most macrophages are signal-dependently activated to a continuum of states between the M1 and M2 extremes. However, all macrophages are simply classified into M1/M2, or M1/M2-like phenotypes in the literature (35). Currently, M2 macrophages refer to all non-M1 macrophages, and are further divided into M2a to M2d subpopulations (36). However, such classification causes much confusion. For example, TAM with the M1 phenotype has been reported, although TAM is widely accepted as M2 macrophages (37). We classified LAM into M2 macrophages, but we managed to distinguish heterogeneous phenotypes and phenotypic shifts in LAM and LAM subsets based on the expression of M1- and M2-related genes. Thereafter, LAM was described as macrophages with more M1 or M2 characteristics.

MOMF is an important source of macrophages in many tissues in the steady state (2). Here we showed that they were an unneglectable component of the macrophage pool in liver and spleen, two extramedullary tissues. Macrophages show great heterogeneity and can be subdivided into subsets due to the complexity of origin, structural organization, and disease content. Red pulp, metallophilic and marginal zone macrophages have been classified and characterized by F4/80⁺VCAM1⁺, CD169⁺MOMA1⁺, and SIGN-R1⁺MARCO⁺ in the spleen based on the structural organization, respectively (38, 39). However, these classifications are not suitable to elucidate the pathologic role of monocyte-derived LAM in T-ALL. The monocytic myeloid-derived suppressor cells are F4/80⁻ (40) and have little overlap with the subsets described in this paper. The Ly6C⁺ subset was detected in both tissues, and three subsets were detected in the spleen macrophage pool. Circulating monocytes are also heterogeneous, and two major subsets, that is, CCR2⁺Ly6C⁺CX3CR1⁻ inflammatory monocytes and CCR2⁻Ly6C⁻CX3CR1⁺ patrolling monocytes, have been identified (41). We showed that both monocyte subsets differentiated into spleen macrophages. Nevertheless, inflammatory monocytes rather than patrolling monocytes, gave rise to the Ly6C⁺ macrophage subset in spleen. Hence, the presence of Ly6C⁺ macrophages largely depends on the continuous entry of inflammatory monocytes into spleen, because they will easily differentiate into Ly6C⁻ and F4/80^hi macrophages with distinct transcriptional profiles and phenotypical characteristics.

In leukemia, macrophages were educated to LAM with the infiltration of leukemia cells into extramedullary tissues, such as spleen and liver. LAM in extramedullary tissues comprises both resident macrophages and MOMF, which are different from TAM. It is worth noting that the increased Ly6C⁺ LAM was observed at the early stage. Furthermore, the spleen leukemic microenvironment recruited more monocytes, and had more Ly6C⁺ LAM cells. Moreover, the increase in Ly6C⁺ LAM was suppressed in CCR2^−/− leukemic mice. These results suggest that more inflammatory monocytes enter the extramedullary tissues at the early stage of leukemia, and these monocytes account for the increased number of Ly6C⁺ LAM.

Hepatic LAM exhibits a proinflammatory phenotype (18), and the recruitment of other immune cells is also observed in extramedullary leukemic microenvironments (17, 42). Hence, sterile inflammation, which plays pathologic roles in the initiation and malignant evolution of malignancies, occurs in extramedullary tissues at the early stage of leukemia. Many cell types, especially immune cells, participate local sterile inflammation. The fact that such inflammation is considerably reduced in CCR2^−/− leukemic mice suggests that monocyte-derived LAM is important players in this process. As monocyte-derived LAM can differentiate into three subsets in spleen, whether they play similar roles is of interest. In general, M1-like macrophages have proinflammatory effects, whereas M2-like macrophages have anti-inflammatory effects. As Ly6C⁺ LAM presented more M1-related phenotypic and expressed higher level of proinflammatory cytokines than the other two subsets, the Ly6C⁺ LAM may be the main subset among monocyte-derived LAM contributing to sterile inflammation in the extramedullary tissues. Nevertheless, the Ly6C⁺ subset undergoes a phenotypic shift and evolves into the other two subsets, which have more M2-related characteristics, within 18 hours in spleen. Such a shift has also been observed in other tissues and pathologic processes. Monocytes not only give rise to MOMF but also further differentiate into KCs in liver (43). Ly6C⁺ MOMF gradually loses M1-related characteristics and obtains more M2-related characteristics, such as MRC1, MMP9, and CCL17 gene expression, in liver injury (44). Hence, the maintenance of the Ly6C⁺ subset and proinflammatory microenvironment depends on the continuous entry of Ly6C⁺ monocytes.

Significant extramedullary infiltration of leukemia cells is frequently observed in patients with T-ALL, and more than 65% of patients have activating mutations in the Notch1 gene (45). In addition, patients with severe hepatosplenomegaly have an inferior prognosis, which also has a negative effect on the quality of patient lives (22, 23). Hepato-splenomegaly may be caused by inflammation, which raised the speculation that monocyte-derived Ly6C⁺ LAM-related sterile inflammation may contribute to this pathologic process. Our results demonstrated that CCR2^−/− leukemic mice, which had reduced Ly6C⁺ LAM and sterile inflammation, showed relieved hepatosplenomegaly and, most importantly, attenuated extramedullary infiltration of leukemia cells. The relationship between inflammation and malignancies is complex. On the one hand, inflammation is associated with the accumulation of immune cells against tumors. However, continued inflammation with high levels of serum proinflammatory cytokines contributes to tumor pathomechanism (46). TAM promotes tumorigenesis and development through inappropriate expression of proinflammatory cytokines, such as IL1β, IL6, and TNFα, via anomalous inflammatory signaling pathways (47). A chronic inflammatory state also contributes to chronic lymphocytic leukemia (CLL) progression as it initiates prosurvival and proliferation signaling in CLL cells (48). Our results suggested that sterile inflammation caused by monocyte-derived LAM might facilitate extramedullary the distribution of T-ALL cells.

Chemokines mediate the migration of cells. Here we found that CCL8 and CCL9/10, mainly released by LAM, especially Ly6C⁺ LAM, were two key chemokines in the mouse T-ALL model. Blocking either CCR2 or CCR1 suppressed the extramedullary distribution of T-ALL cells in not only liver and spleen, but also lung and kidney. Although CCL3 and CCL5 are the main ligands of CCR1, whereas CCL2 is the main ligand of CCR2, their expressions were almost undetectable in spleen and liver microenvironments (Supplementary Figs. S5A and S5B). Hence, Ly6C⁺ LAM might facilitate the distribution of T-ALL cells in extramedullary tissues via the CCL8–CCR1/CCR2 and CCL9/10–CCR1 axes in this mouse model. It is worth noting that sterile inflammation and extramedullary distribution of leukemia cells is also frequent in other types of leukemia. M1 and M1-like macrophages in different pathologic microenvironments express various elevated levels of chemokines (49, 50), which are summarized in Supplementary Table S2. At the same time, different types of leukemia cells from acute myeloid leukemia, B-cell ALL (B-ALL), and CLL have been shown to express the higher levels of chemokine receptors, such as CXCR7, CCR5, and CCR7 (51–53). These results imply that Ly6C⁺ LAM might facilitate the distribution of leukemia cells in extramedullary tissues via corresponding axes in other types of leukemia.

Extramedullary distribution of leukemia cells has clinical significance. It has long been observed that extramedullary distribution of leukemia cells is a bad marker for leukemia patients. Studies on isolated extramedullary relapse, which arise from the central nervous system, testis, ovary, liver, lymph nodes, kidney, and skin (54), demonstrated that minimal residual leukemia cells or leukemia stem cells in extramedullary tissues are important sources for relapse. Isolated extramedullary relapse accounts for approximately 5.8% of leukemia and is more frequently observed in T-ALL than in acute myeloid leukemia or chronic myeloid leukemia (55, 56). As a poor predictor of survival, patients with isolated extramedullary relapse rarely achieve long-term survival (57). Furthermore, patients suffer isolated extramedullary relapse (with or without BM relapse) has worse prognosis than those with BM relapse. Thus, Ly6C⁺ LAM might play adverse pathologic roles in leukemia progression and clinical outcome upon treatment.

Taken together, our findings reveal the multifaceted pathologic roles of monocyte-derived Ly6C⁺ LAM in T-ALL progression. Ly6C⁺ LAM in extramedullary tissues originates from circulating Ly6C⁺ inflammatory monocytes and possesses more M1-related characteristics. Although M1 macrophages are believed to have proinflammatory and antitumor effects, Ly6C⁺ LAM not only contributes to sterile inflammation, but also facilitates the extramedullary distribution of leukemia cells, which at least partly contributes to the hepatosplenomegaly in T-ALL. Repolarizing macrophages to obtain more M1-related characteristics is a widely accepted strategy for macrophage-based immunotherapy against malignancies. Our results suggest that a comprehensive interpretation of the pathologic roles of LAM is necessary. We also suggest that other pathologic roles of LAM and/or TAM be considered, and the method to repolarize macrophages should be optimized to maximize the antitumor effects and minimize the side effects.

No potential conflicts of interest were disclosed.

F. Yang: Conceptualization, resources, data curation, software, formal analysis, validation, investigation, writing-original draft. W. Feng: Conceptualization, data curation, formal analysis, validation, investigation. H. Wang: Resources, software, validation, investigation. L. Wang: Resources, data curation, formal analysis. X. Liu: Data curation, software, validation, investigation. R. Wang: Resources, data curation, software, formal analysis. C. Chen: Resources, software, investigation. X. Yang: Data curation, software, formal analysis. D. Zhang: Resources, software. Q. Ren: Resources, supervision. G. Zheng: Conceptualization, supervision, funding acquisition, writing-review and editing.

This work was supported by grants 81770183, 81970155, and 81570153 from the National Natural Science Foundation of China (NSFC); programs 2016-I2M-2-006 and 2017-I2M-1-015 from the CAMS Innovation Fund for Medical Sciences (CIFMS); and Graduate Student Innovation Fund (2016-0710-10) from Peking Union Medical College. G. Zheng is a recipient of the New Century Excellent Talents in University (NCET-08-0329).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.