VEGFR1⁺ Metastasis–Associated Macrophages Contribute to Metastatic Angiogenesis and Influence Colorectal Cancer Patient Outcome

Despite early detection and improvements in treatment, colorectal cancer is still the third most common cause of death among all cancer entities (1). Overall survival (OS) is reduced in patients diagnosed with metastases (2), highlighting the need to improve our understanding of the metastatic process. While the principles of the metastatic cascade are of interest, a deeper understanding of the growth of established metastasis is of vital importance for finding new therapeutic targets and improving patients' welfare. As more than two thirds of colorectal cancer metastases are found in the liver (2), we focused our study on understanding how macrophages can influence angiogenesis and growth of liver metastasis and thereby influence patient outcome.

Macrophages can exert pro- or antitumorigenic effects (3, 4) in the tumor microenvironment and play various roles in each step of disease progression. Accordingly, several studies exist with contradictory findings about the correlation of macrophages with patient survival and disease recurrence. While tumor-associated macrophages (TAMs) have been described to act more antitumorigenic in primary colorectal cancer (5, 6), metastasis-associated macrophages (MAMs) have not yet been well characterized. In this study, we hypothesized that TAMs and MAMs have different angiogenic potentials.

Within the multifaceted features of macrophages, one important role is the induction of angiogenesis, a hallmark of cancer (7, 8). Hypoxia drives VEGF upregulation in tumors leading to the development of blood vessels from preexisting ones (9). In primary colorectal cancer tumors, VEGF expression correlates positively with the occurrence of liver metastasis. Likewise, vessel density was found to negatively influence OS in metastatic patients (10). VEGF binds to its tyrosine kinase receptors, the VEGFRs, expressed ubiquitously in the tumor microenvironment. In the monocyte/macrophage lineage, VEGF signals via VEGFR1 which recruits macrophages from bone marrow to the tumor (11, 12) and influences metastatic initiation and growth (13, 14). In the liver, VEGFR1-positive (VEGFR1⁺) macrophages have been shown to infiltrate hypoxia-driven injured areas via the peripheral blood (15). Moreover, in the context of VEGF/VEGFR inhibition, macrophage depletion further suppressed tumor growth and hence, highlights their value on regulating disease outcome (16). In this study we evaluated the role of VEGFR1⁺ macrophages in liver metastasis of colorectal cancer.

As prediction of disease recurrence after metastatic liver resection remains a clinical challenge, we also evaluated VEGFR1⁺ macrophages as liquid biopsy to assess patients' risk profiles after surgical interventions. Several clinical and biological features have been identified as prognostic factors of colorectal cancer relapse, including tumor markers in the blood (17, 18). Also, advances on the genetic characterization of tumors resulted in promising assays, which correlate with OS of patients with colorectal liver metastasis, but still can only be assessed in resected tumor tissue (19). Therefore, a noninvasive method that could be used would be desirable.

In summary, the aim of this study was to elucidate the role of macrophages in liver metastasis angiogenesis and patient outcome. Our results reveal VEGFR1⁺ MAMs as highly angiogenic and predictors of progression-free survival (PFS). Moreover, we found a new prognostic marker of relapse as circulating VEGFR1⁺ monocytes in blood of patients with liver metastasis correlated with PFS and site of disease recurrence.

A total of 210 patients diagnosed with primary colorectal carcinoma (pTU) and 200 patients diagnosed with colorectal liver metastasis were included in this study (Supplementary Table S1). All patients underwent surgery between 2007 and 2017 at the Department of General, Visceral and Transplantation Surgery of the University Hospital Heidelberg (Heidelberg, Germany). From 37 patients, both the pTU and the liver metastasis were analyzed, which had synchronous liver metastasis at the time of surgery (Supplementary Table S1, synchronous). Our study was approved by the Ethics Committee of the Medical Faculty Heidelberg and was conducted in accordance with the declaration of Helsinki. Written informed consent was obtained from all enrolled patients. Patient data were collected prospectively. Tumor tissue samples of 93 patients with pTU and 123 patients with liver metastasis were obtained for IHC purposes (Supplementary Table S1). Venous blood samples were obtained from 80 patients with pTU and 77 patients with liver metastasis (Supplementary Table S1; flow cytometry). Of the patients with liver metastasis, 64.9% received neoadjuvant therapy (Supplementary Table S2). Tumor and nontumor tissue samples of 13 patients with pTU and 17 patients with liver metastasis were obtained for cell isolation by magnetic cell separation in 2016/2017.

Stainings were performed following an in-house protocol (20). Further details are specified in the Supplementary Data.

All experimental protocols of mice were approved and performed in accordance to local authorities and animal welfare officers of Baden-Wuerttemberg, Germany. All mice were cared for in accordance with the “Guide for the Care and Use of Laboratory Animals” by the National Academy of Sciences. A highly metastatic cell line was established by positive selection as described previously (21). Cells were infected with luciferase-expressing lentiviral premade particles (Ambiso). Tumor and metastatic growth were evaluated using bioluminescence imaging. Briefly, 200 μL of Beetle luciferin (15 mg/mL in water; Promega) were intraperitoneally injected in anesthetized mice. After 5 minutes, bioluminescent images were acquired using the IVIS Imaging System (Xenogen).

Cell culture was performed as described previously (20, 22). Fibrin gel sprouting assays were performed as described previously (23). Further details are specified in the Supplementary Data.

Freshly collected tissues, both from mouse and human, were used for isolation of CD11b⁺ cells as described previously (5). For both, murine CD11b⁺ and CD11b⁺/F4/80⁺ isolation prior to MACS separation, cells were applied onto a Percoll Gradient (Sigma-Aldrich). Cells were incubated with F4/80-biotin antibody (clone REA126, 1:10; Miltenyi Biotec) for 15 minutes at 4°C. Then, cells were incubated with streptavidin microbeads for 15 minutes at 4°C (1:10; Miltenyi Biotec). F4/80⁺ cell populations were isolated and then incubated with CD11b MicroBeads (clone M1/70.15.11.5, 1:10; Miltenyi Biotec) for 15 minutes at 4°C. After application into MS Columns, the CD11b⁺/F4/80⁺ cell populations were obtained.

Single-cell suspensions from murine pTU and liver metastasis were prepared as described for magnetic cell separation. Cells were incubated with F4/80-FITC (clone A3-1, 1:20; AbD Serotec) in flow cytometry buffer (eBioscience). Then, for M1 and M2 staining, cells were fixed with IC Fixation Buffer (Thermo Fisher Scientific) for 20 minutes at room temperature, permeabilized with Permeabilization Buffer (Thermo Fisher Scientific) and stained in this buffer with Nos2-PerCP (clone N20, 1:50; Santa Cruz Biotechnology) and CD206-BV421 (clone C068C, 1:50; BioLegend) for 30 minutes at room temperature, respectively. For sorting TAMs and MAMs, fresh single-cell suspensions were also stained with CD11b-PE (clone M1/70.15.11.5, 1:10; Miltenyi Biotec). CD11b⁺/F4/80⁺ macrophages were sorted with BD FACSAria III sorter directly into RLT Buffer (Qiagen), for RNA isolation. Murine blood was also stained with CD45-PerCP (Clone 30-F11, 1:100, BD Bioscience). Cells from human blood samples were incubated with CD11b-PE (clone M1/70.15.11.5, 1:10; Miltenyi Biotec) for 30 minutes at 4°C. Afterward, fixation and permeabilization was performed as before. CD68 (clone 298807, 1:20; R&D Systems) and VEGFR1 (Y103, 1:100; Abcam) were used for 30 minutes at room temperature. Appropriate secondary antibodies were incubated for 30 minutes at room temperature (1:500; Thermo Fisher Scientific). For all samples, negative and corresponding IgG controls were additionally performed. Data were acquired with a BD FACSCanto II Flow Cytometer (BD Biosciences). Analysis was performed with FlowJo 10.4.2 (FlowJo, LLC).

RNA isolation of macrophages was performed with the RNeasy MiniKit (Qiagen) following the manufacturer's instructions with a DNA digestion step using the RNase-Free DNase Set (Qiagen). For sorted cells, to reverse transcribe RNA samples to cDNA, SuperScript VILO cDNA Synthesis Kit (Invitrogen) was used. qRT-PCR was performed using Fast SYBR Green Master Mix (Thermo Fisher Scientific). ActB was used as internal control. All qPCR results were obtained from at least three biological repeats. Primers used are listed in Supplementary Table S3.

Plasmids containing the miRNA control and miRNA against Flt1 were generated using the BLOCK-iT Pol II miR RNAi Expression Vector Kit (Invitrogen) following the manufacturer's instructions. Then, validated miRNAs were cloned in the pLVTHM plasmid vector expressing GFP for generating the lentivirus. Lentiviruses were produced by transfection of HEK293-T cells by the calcium phosphate method with the corresponding vectors. Lentivirus-containing supernatants were collected 72 hours after transfection and concentrated by ultracentrifugation at 22,000 rpm for 90 minutes at 4°C.

Hematopoietic stem cells were obtained with the Lineage Cell Depletion Kit (Miltenyi Biotec) following the manufacturer's instructions. Cells were harvested from Balbc/J mice and cultured in StemSpan SFEM Medium (Stemcell Technologies; supplemented with 50 ng/mL mSCF, mTPO, and Flt3-L; R&D Systems) and 1% penicillin-streptomycin and glutamine for 24 hours. Then, cells were plated in a RetroNectin (TaKaRa, 20 μg/mL) coated plate and transduced with lentiviruses containing either the control miRNA or the miRNA against Flt1 with 4 μg/mL polybrene (Thermo Fisher Scientific). After 48 hours, GFP-positive cells were sorted. For lethal irradiation, receptor mice were irradiated twice with 3.5 Gy and 200,000 cells per mouse were injected in the tail vein. After 4 weeks, tumors were implanted in the mice as before.

Venous peripheral blood samples were collected in EDTA tubes. The white blood cell fraction was isolated by a density gradient centrifugation with Histopaque-1077 (Sigma-Aldrich). Erythrocytes were lysed by Red Blood Cell Lysis Buffer (Roche). The collected cells were frozen at −80°C with DMSO (Sigma-Aldrich).

All values shown are mean ± SEM, if not differently stated. To calculate statistical significance, the two-tailed Student t test (when comparing two groups) or one-way ANOVA with Bonferroni multiple comparisons test (when comparing three or more groups) were used in Prism software. For correlation analysis, D'Agostino–Pearson omnibus normality test was performed to evaluate Gaussian distribution and accordingly, correlation was evaluated either with two-tailed Pearson correlation test or nonparametric Spearman correlation test in Prism software. Statistical significance was taken as P < 0.05. OS was measured from time of surgery until death or loss of follow-up, and PFS was measured from time of surgery until loss of follow-up or disease recurrence. Disease recurrence was assessed by cross-sectional imaging. For grouping of the patients' medians for von Willebrand Factor (vWF; 2.1% in pTU, 1.95% in liver metastasis, and 1.99% and 1.97% in synchronous pTU and liver metastasis, respectively), CD68 staining (114 cells/mm² in pTU, 116 cells/mm² in liver metastasis, and 124 cells/mm² and 80 cells/mm² in synchronous pTU and liver metastasis, respectively) and VEGFR1 (167 cells/mm² in pTU, 117 cells/mm² in liver metastasis, and 173 cells/mm² and 124 cells/mm² in synchronous pTU and liver metastasis, respectively) were calculated and patients were grouped as low <2% or high >2% vessel density or as low <100 cells/mm² or high >100 cells/mm² macrophage infiltration or the median itself for VEGFR1, accordingly. Kaplan–Meier method and log-rank test was used for survival analysis. Results for PFS were confirmed with multivariate Cox regression analysis. Statistical analyses were performed with SPSS Statistics 24.0 (IBM).

Within the tumor stroma, macrophages have the ability to rapidly polarize (3). As we hypothesized that TAMs and MAMs have different proangiogenic properties, we first characterized macrophage polarization. To do so, we used a traceable orthotopic CT26 colon cancer mouse model forming primary tumors in the cecum (pTU) that spontaneously metastasize to the liver in almost 90% of the mice (Fig. 1A and B). A simple but useful classification of M1-like [proinflammatory, nitric oxide synthase (NOS₂)-positive] and M2-like (anti-inflammatory, proangiogenic, mannose receptor (MR)-positive] polarized macrophages has been broadly used (24). We used the cell surface marker F4/80 (murine-specific macrophage marker) together with NOS₂ and MR to characterize the TAM and MAM activation phenotype by flow cytometry (Fig. 1C and D). Our results showed that pTUs had a 2.5-fold higher M1-like (F4/80⁺/Nos2⁺) macrophage infiltration than liver metastasis (P = 0.005; Fig. 1C). Conversely, a 3-fold higher infiltration of M2-like macrophages (F4/80⁺/MR⁺) was observed in liver metastasis (P < 0.001; Fig. 1D).

For further characterization, we used the general myeloid marker CD11b and the F4/80 to FACS TAMs and MAMs from pTU and liver metastasis (Supplementary Fig. S1A). Consistently, qPCR-based expression analyses revealed that TAMs exhibited a M1-like immunostimulatory polarization in contrast to MAMs, while MAMs showed a more M2-like immunosuppressive polarization (Fig. 1E and F).

To address the angiogenic potential of TAMs and MAMs, we isolated CD11b⁺ cells from liver metastasis or pTUs by magnetic separation and generated conditioned medium (CM) to stimulate endothelial cells in sprouting assays. Liver metastasis MAM-derived CM was able to significantly induce sprouting angiogenesis (3.2-fold; P < 0.001), while TAM-derived CM induced angiogenesis to a lower extend (1.6-fold; P = 0.079;Supplementary Fig. S1B and S1C).

As it was recently shown that Flt1 (VEGFR1 in humans) is highly expressed by a subset of macrophages that support breast cancer metastasis (13) and as Flt1 signaling in myeloid cells can support tumor angiogenesis (25), we investigated the expression of Flt1 in MAMs. Indeed, qPCR analyses of sorted TAMs and MAMs showed a significant upregulation of Flt1 in MAMs (78.8-fold change; P < 0.001; Fig. 1G). Moreover, we found that MAMs express higher levels of known proangiogenic genes such as Tek, encoding for the tyrosine kinase receptor Tie2 (26). In contrast, we observed that Cxcl11, an angiostatic chemokine, is highly expressed by TAMs (Fig. 1G). By flow cytometry and immunofluorescence on isolated macrophages, we confirmed that MAMs express more FLT1 than TAMs also at protein levels (Fig. 1H; P < 0.01; Supplementary Fig. S1D; and Supplementary Fig. S1E and S1F; P < 0.001).

Furthermore, to address the functional relevance of Flt1^hi MAMs, we sorted CD11b⁺ F4/80⁺ Flt1^hi or Flt1^lo MAMs and TAMs and performed sprouting assays (Fig. 1I; Supplementary Fig. S1G). Interestingly, our results showed that Flt1^hi MAMs were highly angiogenic (2.4-fold increase, compared with negative control; P < 0.01), while Flt1^hi and Flt1^lo TAMs and Flt1^lo MAMs were not. As Flt1^hi TAMs did not induce sprouting angiogenesis, we analyzed whether they were still more M1-like polarized. Indeed, Flt1^hi TAMs express more Nos2 than Flt1^hi MAMs and significantly less Flt1 (Supplementary Fig. S1H).

In summary, these results indicate that Flt1^hi MAMs have a proangiogenic role in the liver metastatic microenvironment.

To address the relevance of Flt1 expression in macrophages in metastatic angiogenesis, we knocked down Flt1 in hematopoietic stem cells. Lineage-depleted bone marrow cells were transduced with a GFP-expressing lentiviral vector containing a miRNA against Flt1 or control. GFP⁺ control bone marrow cells (miR-Ctrl) or Flt1-knocked down cells (miR-Flt1) were transplanted into irradiated mice. Four weeks after bone marrow transplantation, CT26 colon tumors were injected in the cecum and after 3 weeks, knockdown efficiency, tumor growth, and angiogenesis were evaluated (Supplementary Fig. S2A). Survival analysis showed that three of eight mice in the control group died versus one of nine mice in the knockdown group (Supplementary Fig. S2B).

Mouse and liver weight were not different at the end of the experiment (Supplementary Fig. S2C and S2D). Also, primary tumor weight was not different (Fig. 2A). Interestingly, evaluation of liver metastases revealed a reduction of size in the miR-Flt1 group (1.4-fold reduction; P = 0.038; Fig. 2B; Supplementary Fig. S2E). Reduced FLT1 expression was confirmed at protein level in CD11b⁺ cells from bone marrow (Fig. 2C; Supplementary Fig. S2F), as well as in blood (Fig. 2D; Supplementary Fig. S2G). qPCR analysis of sorted CD11b⁺ cells revealed a significant Flt1 reduction in blood (4.8-fold change; P = 0.029) and bone marrow (3.4-fold change; P < 0.001; Fig. 2E and F). Interestingly, a significant reduction of the proangiogenic metalloprotease 9 (Mmp9) expression in CD11b⁺ bone marrow cells was also observed after Flt1 knockdown (3.0-fold change; P < 0.01), while other genes such Vegfa or Tek were not affected (Fig. 2F).

We further evaluated the vasculature of both primary tumors and liver metastasis by staining for platelet endothelial cell adhesion molecule (CD31) and the lymphatic vessel endothelial hyaluronan receptor 1 (Lyve1), which, besides the lymphatic vasculature, mainly labels hepatic blood sinusoidal endothelial cells in the liver (27). Quantification revealed that depletion of Flt1 (miR-Flt1) did not affect vessel density of the pTU but reduced vessel density by 58.7% ± 80% in liver metastasis (Fig. 2G–I).

These results show that Flt1^hi myeloid cells are able to support metastatic growth and angiogenesis in colorectal cancer.

Because we identified a proangiogenic macrophage subpopulation in mice, we assessed by IHC whether macrophage infiltration (CD68, human macrophage marker) within the tumor microenvironment correlated with vessel density (vWF, human vasculature marker) in patients diagnosed with colorectal cancer (Supplementary Table S1). Macrophage infiltration did not correlate with vascular density in pTU (Fig. 3A and B), while it did in liver metastasis (Fig. 3C and D), supporting that macrophages are more proangiogenic in liver metastasis than in pTU.

Then, to elucidate whether VEGFR1⁺ macrophages in human colorectal cancer could also contribute to metastatic angiogenesis, we correlated the infiltration of VEGFR1⁺ cells with vessel density in pTU and liver metastasis. Interestingly, no correlation of VEGFR1⁺ cells with vessel density was observed in pTU samples (Fig. 3E and F), while a significant positive correlation was found in liver metastasis (Fig. 3G and H).

Analysis of tumor macrophage infiltration and vessel density showed no significant impact on OS (Supplementary Fig. S3A and S3B) or PFS of pTU patients (Fig. 3I and J); while known variables such as the International Union Against Cancer (UICC) staging or levels of circulating carcinoembryonic antigen (CEA) negatively influenced PFS and OS, respectively (UICC stage ≥ 3; HR, 5.280; P = 0.003/CEA ≥2.5 μg/L; HR, 2.609; P = 0.043).

While known parameters predicted reduced OS in patients with liver metastasis [Memorial Sloan Kettering Cancer Center (MSKCC) score ≥3; HR, 1.760; P = 0.008/CEA ≥ 200 μg/L; HR, 2.140; P = 0.027], neither macrophage infiltration nor vessel density had an impact on OS (Supplementary Fig. S3C and S3D). However, both high macrophage infiltration and high vessel density predicted reduced PFS in patients with liver metastasis (Fig. 3K and L).

In line, we observed that VEGFR1 staining did not predict OS in pTU (Supplementary Fig. S3E), nor in liver metastasis (Supplementary Fig. S3F). VEGFR1 also did not impact PFS in the pTU (Fig. 3M), but high infiltration in liver metastasis was associated to a reduced PFS (Fig. 3N).

We confirmed all the above-mentioned results in pTU and liver metastasis from patients with synchronous liver metastasis where matched pairs of the pTUs and liver metastasis exist in the same patient (Supplementary Fig. S4A–S4P; Supplementary Table S1).

In summary, these results show that in human colorectal cancer liver metastases, vessel density and macrophage infiltration are prognostic factors of PFS and suggest that, VEGFR1⁺ MAMs may contribute positively to angiogenesis.

To elucidate the angiogenic potential of human macrophages, CM of freshly isolated human CD11b⁺ cells after surgical resection of primary tumors or liver metastasis was used for sprouting angiogenesis assays. By immunofluorescence, we confirmed that VEGFR1⁺ macrophages were present within the isolated CD11b⁺ cells from the liver metastasis and the surrounding “no tumor” liver (ntL), and from the pTU and the surrounding “no tumor” mucosa (Mu). Indeed, there were 20% more VEGFR1⁺ MAMs within the metastasis than in the surrounding tissue and 50% more than in the pTU (Fig. 4A and B). CD11b⁺ MAMs were able to induce angiogenesis to a higher extend with more and longer endothelial cell sprouts than those in the surrounding liver and more than TAMs (Fig. 4C–E; 3-fold increase compared with negative control; P < 0.001).

In summary, we identified a subset of MAMs in human colorectal cancer that express VEGFR1 and are proangiogenic.

As Flt1⁺ macrophages have been shown to be recruited to the metastatic site from the circulation (14, 25), we investigated whether VEGFR1⁺ monocytes were present in the blood of patients with colorectal cancer liver metastasis (Supplementary Table S1).

We performed flow cytometry analysis of human blood from patients with pTU or liver metastasis. We stained for CD11b (myeloid cell marker), CD68 (macrophage marker), and VEGFR1 on 77 liver metastasis and 80 pTU patients (Supplementary Fig. S5A–S5D). These patients were either diagnosed with liver metastasis or had a pTU, which was resected at the day of blood collection. Results showed no differences in percentage of circulating CD11b⁺/CD68⁺ monocytes (P = 0.34) or VEGFR1⁺ (P = 0.65) cells between the different cohorts (Supplementary Fig. S5E–S5G).

Then, we investigated whether VEGFR1⁺ monocytes in blood of patients could predict PFS or OS. In pTU patients, neither CD11b⁺/CD68⁺ nor CD11b⁺/CD68⁺/VEGFR1⁺ cells predicted PFS (Fig. 5A and B) or OS (Fig. 5C and D). Also, we observed no association with numbers of circulating cells and the different clinical and pathologic variables. Conversely, high percentage of CD11b⁺/CD68⁺ (Fig. 5E) and CD11b⁺/CD68⁺/VEGFR1⁺ (Fig. 5F) circulating monocytes predicted reduced PFS in patients with liver metastasis, however not OS (Fig. 5G and H). Interestingly, high percentage of circulating VEGFR1⁺ cells associated with clinical and pathologic variables in patients with liver metastasis (Supplementary Table S4). This group included more patients with high risk MSKCC clinical score (P = 0.03), with more than one liver metastasis (P < 0.001) and that had received neoadjuvant therapy (P = 0.01; Supplementary Table S4).

Thus, to determine the relevance of circulating VEGFR1⁺ monocytes for PFS, we first investigated the general clinical and pathologic variables associated with PFS in liver metastasis (Supplementary Table S5). These analyses revealed that high MSKCC clinical risk score [HR, 1.682; 95% confidence interval (CI), 1.010–2.802; P = 0.046] and high number of VEGFR1⁺-circulating macrophages (HR, 2.542; 95% CI, 1.472–4.390; P = 0.001) were associated with an unfavorable prognosis. To address the dependency of our variable, we next performed a multivariate Cox proportional hazard analysis with the variables associated to VEGFR1⁺ macrophages and PFS (Supplementary Table S6). A high number of circulating VEGFR1⁺ macrophages was an independent prognostic factor for early disease recurrence in patients with liver metastasis (HR, 2.129; 95% CI, 1.153–3.932; P = 0.016; Supplementary Table S6).

Taken together, our results suggest that detection of VEGFR1⁺ monocytes in blood of patients can be a promising marker to predict disease recurrence on liquid biopsies of patients with liver metastasis.

Before metastasis, a favorable soil for cancer cells to seed and grow is formed at distant sites, as a premetastatic niche. VEGFR1⁺ macrophages have been suggested to immunomodulate this niche to support growth of the upcoming tumor cells (14). We observed that high levels of VEGFR1⁺-circulating monocytes predicted independently PFS as well as recurrence (Fig. 5I). Interestingly, patients in the high VEGFR1⁺ group had more frequently a recurrence in the liver (Fig. 5I).

Considering this, the number of VEGFR1⁺-circulating macrophages may be considered as a novel biomarker for future recurrence of metastasis in the liver.

In this study we elucidated the role of MAMs. We found macrophage infiltration to correlate with vessel density only in liver metastases and not in pTU of patients with colorectal cancer. Also, we identified TAMs to be more M1-like and MAMs more M2-like and more proangiogenic. Both in mice and humans, we found VEGFR1 expression in a subset of MAMs, which in vivo support angiogenesis and metastatic growth. High number of VEGFR1⁺ cells in liver metastasis predicted reduced PFS in tissue and blood samples. Notably, circulating VEGFR1⁺ monocytes also predicted intrahepatic recurrence in patients with liver metastasis.

Even though macrophages are involved in almost every step of tumor and metastasis progression (3, 8), little is known about their specific role and influence within metastases. Distinctive tumor microenvironments can differently activate macrophages (4) and in colorectal cancer, different microenvironments in the pTU and liver metastasis have been described previously (5, 28). Therefore, we hypothesized that TAMs and MAMs are differently polarized. In line with the literature (5, 29), in a clinically relevant mouse model of liver metastasis we observed higher infiltration of M1-like TAMs in the pTU, while, more M2-like MAMs were found in liver metastasis. The more M1-like polarization in the pTU can be explained by the inflammatory environment induced by commensal- and pathogen-derived molecules in the gut (5). On the contrary, when metastasis grows in the liver, the immunosuppressive environment supports the observed shift toward an M2-like phenotype (28).

Apart from being more M2-like polarized, in mice, we were able to identify Flt1^hi MAMs as a subset of macrophages, which strongly induces angiogenesis and supports metastasis growth in liver metastasis. In line with our results, Flt1⁺ signaling in bone marrow–derived myeloid cells is necessary to induce the angiogenic response in gliomas (25). Moreover, in lung, pancreatic, and breast cancer models it was shown that Flt1 signaling in TAMs was necessary to support tumor angiogenesis (30). It has also been reported that knockout of VEGFR1 might alter the bone marrow–derived cell infiltration into lung metastasis but not their growth, while others have shown that VEGFR1 is required for metastasis progression (31). In addition, CD11b myeloid cells expressing VEGFR1 were shown to accumulate at the premetastatic niche before the arrival of lung cancer and melanoma cells (14). On the other hand, in a mouse model of intrasplenic injection of CMT93 colorectal cancer cells, Flt1⁺ MAMs could not be detected in the liver metastasis (32). In this tumor model, no pTU is implanted and thus, there are no metastases per se. Cells rapidly form intrahepatic tumor lesions within 7 days, while in our mouse model the first liver metastasis was observed after 15 days of pTU implantation. Our results support the idea that in colorectal cancer, VEGFR1⁺ cells support metastasis and highlight the importance of adequate mouse models to study metastasis.

In human colorectal cancer primary tumors, discrepancies exist about whether or not macrophages play a role in predicting OS and/or PFS (33). In our study, we found no value for OS or PFS prediction in pTU, while others observed that the density of macrophage infiltration inversely correlates with patients' survival (34). Our results from the murine model support the notion that in pTU, M1-like TAMs can account for the benefit in OS observed in other studies (35, 36), and that discriminating TAM subpopulations is needed to better understand the overall impact of macrophage infiltration on patient outcome. Not only the plasticity of macrophages within the microenvironment could explain these differences (4, 24) but also, analysis of different tumor areas like the invasive front (35) or the whole tumor, as we did in this study, could contribute to these discrepancies. In line with the literature (37), we found that the bulk of CD68⁺ macrophages predicted reduced PFS in liver metastasis.

As angiogenesis is a hallmark of cancer (8) and macrophages can support tumor angiogenesis (7), their role in metastatic angiogenesis was of high interest. In colorectal cancer it has been described that high vascular density, as a marker for tumor angiogenesis, negatively influences patient outcome both in the pTU (38) and liver metastasis (39). In our cohort of patients, this hold true for the patients with liver metastasis regarding PFS.

Here, for the first time, we observed a positive relationship between macrophages and vessel density in samples from patients with liver metastasis. Albeit TAMs of colorectal cancer are known to secrete VEGF and thereby promote angiogenesis (40), no association was found in the pTU. In line with our results, Lackner and colleagues (41) also observed no correlation between CD68⁺ macrophages and microvessel density in colorectal cancer stage II and III pTU. This altogether indicates that macrophages in the pTU and the liver metastasis are different, in terms of their angiogenic potential.

Interestingly, our results show that also in human colorectal cancer, MAMs are more angiogenic than TAMs and that VEGFR1⁺ MAMs are a specific angiogenic subgroup. Importantly, here we show that VEGFR1-expressing macrophages not only induce angiogenesis in liver metastasis, but also predict poor prognosis while not affecting prognosis in pTU. High protein levels of VEGFR1 in the pTU were described to either have no association with patient survival (42), supporting our data, or to be associated with worse patient outcome (43). Also, presence of more VEGFR1⁺ cells in human liver metastasis was reported compared with colon mucosa and pTU (44). Moreover, in other tumor entities, VEGFR1 has been reported as a prognostic marker: for example, in glioma, overexpression of VEGFR1 was associated with poor prognosis (45) and in breast cancer, high VEGFR1 in the stroma correlated with worse OS (46); but to the best of our knowledge, the prognostic value of VEGFR1⁺ cells in liver metastasis was so far not assessed.

Many efforts have been made to establish noninvasive methods to assess patient risk profiles before surgical tumor resection. In this regard, circulating tumor cells are promising candidates, but measuring them in patient-derived blood has not yet proven its practicability in clinical use (47). In line with the literature (48), our results suggest that detection of VEGFR1⁺-circulating monocytes in the blood of patients with liver metastasis could be a promising new biomarker for recurrence of liver metastasis. Further validation will be necessary to determine its specificity. We believe that this could be a good addition to the so-called liquid biopsies for colorectal cancer.

As we observed that circulating VEGFR1⁺ monocytes associate with neoadjuvant therapy, it will be important to find out how chemotherapy affects VEGFR1 expression. As plasma soluble VEGFR1 levels were proposed to be good biomarkers for response to neoadjuvant therapy in patients with localized rectal cancer (49), it would also be interesting to understand whether VEGFR1-circulating cells can predict response to therapy in patients with colorectal cancer. In line, low VEGFR1 gene expression levels in metastatic colorectal cancer were predictive for prolonged OS after antiangiogenic therapy (50). Therefore, it would be noteworthy to study the possibility of VEGFR1-circulating monocytes being biomarkers for antiangiogenic therapy response.

Certainly, the importance and validity of all clinical observations are limited by the retrospective character of the analysis. However, the underlying biological mechanism was conserved both in the murine model, as well as in primary isolated cells from patients. In the future, the results should be confirmed in a prospective study.

In summary, our findings show a correlation of macrophages and vessel density in patients with colorectal cancer liver metastasis, identify a new subset of proangiogenic MAMs that express high levels of VEGFR1, and suggest circulating VEGFR1⁺ monocytes as a new biomarker to predict disease recurrence in the liver.

Thomas Schmidt reports receiving speakers bureau honoraria from Servier. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A. Freire Valls, K. Knipper, C. Ruiz de Almodovar, T. Schmidt

Development of methodology: A. Freire Valls, K. Knipper, E. Giannakouri, V. Sarachaga, Y. Shen, C. Ruiz de Almodovar, T. Schmidt

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Freire Valls, K. Knipper, E. Giannakouri, M. Wuehrl, Y. Shen, P. Radhakrishnan, J. Klose, T. Schmidt

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Freire Valls, K. Knipper, V. Sarachaga, P. Radhakrishnan, M. Schneider, T. Schmidt

Writing, review, and/or revision of the manuscript: A. Freire Valls, K. Knipper, E. Giannakouri, V. Sarachaga, P. Radhakrishnan, A. Ulrich, M. Schneider, H.G. Augustin, C. Ruiz de Almodovar, T. Schmidt

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Knipper, E. Giannakouri, S. Hinterkopf, Y. Shen, P. Radhakrishnan, J. Klose, M. Schneider, H.G. Augustin, T. Schmidt

Study supervision: K. Knipper, T. Schmidt

We thank Heike Adler, Melanie Richter, and Benjamin Schieb for great technical support and all members of the laboratory for helpful discussions. We also thank the Flow Cytometry and FACS Core Facility at ZMBH (Heidelberg, Germany) and the Nikon Imaging Center at the University of Heidelberg. Tissues of 13 synchronous patients were provided by the National Center for Tumor Diseases (Heidelberg, Germany) in accordance with the regulations of the tissue bank and the approval of the ethics committee of Heidelberg University. This work was supported by grants from the Deutsche Forschungsgemeinschaft (SCHM 2560/3-1 to T. Schmidt and projects A2 and C5 within CRC1366 “Vascular control of organ function” - project number 39404578 to H.G. Augustin and C. Ruiz de Almodovar). T. Schmidt received funding from the Jung Foundation for Science and Research and grant support by the Heidelberg Foundation of Surgery.

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.