CCR1-mediated accumulation of myeloid cells in the liver microenvironment promoting mouse colon cancer metastasis

We demonstrated recently that CCR1 plays critical roles in liver metastasis of Smad4-deficient colon cancer in a mouse model [16]. In the BM, CD11b⁺ Gr-1⁺ myeloid cells expressed 10 times higher level of Ccr1 mRNA than CD11b^– Gr-1^– non-myeloid cells (Supplementary Fig. 1a), suggesting that CCR1 expressing cells were highly enriched in the myeloid cells. To isolate and characterize the CCR1-expressing cells by cell sorting, we tested antibodies from various sources, but were unable to find one that bound to mouse CCR1 specifically and reliably. Accordingly, we resorted to the construction of a reporter transgenic mouse model whose marker gene (membrane-targeted Venus; mVenus) was placed under the control of the Ccr1 promoter. As the source of regulatory elements to reconstitute the endogenous CCR1 expression, we used a BAC clone spanning 8 kb upstream and 34 kb downstream of the mouse Ccr1 gene (Fig. 1a). Thus, we established four independent transgenic lines (Fig. 1b, and Supplementary Fig. 1b, c; see "Materials and methods" section).

A flow cytometric analysis of adult transgenic mouse BM cells showed that expression of the reporter mVenus fluorescence was restricted essentially to the CD11b⁺ Gr-1⁺ myeloid cells (Fig. 1b and Supplementary Fig. 1c). When the CD11b⁺ myeloid cells were divided into four populations based on the intensity of mVenus fluorescence, the level of endogenous Ccr1 mRNA correlated well with that of mVenus fluorescence, suggesting that expression of the reporter precisely reflected that of the endogenous CCR1 (Fig. 1c). Notably, CD11b⁺ Gr-1⁺ myeloid cells could be divided into Ccr1-mVenus^– and Ccr1-mVenus⁺ cells (Fig. 1d, left panel). Giemsa staining of the sorted cells in these subpopulations revealed that the Ccr1-mVenus^– cells were immature neutrophils with ringed nuclei, whereas the Ccr1-mVenus⁺ cells were mature neutrophils with lobulated nuclei, indicating that expression of CCR1 correlated with the maturation of BM neutrophils (Fig. 1d, center and right).

As we reported previously [16], intrasplenic injection of CCL9-expressing CMT93 colon cancer cells can cause their dissemination into the liver, which is followed by metastatic colonization. Employing this model, we monitored the behavior of the Ccr1-mVenus⁺ cells in the cancer metastasis microenvironment first by histochemistry at 14 days post-injection. We found that the density of Ccr1-mVenus⁺ cells among CD45⁺ hematopoietic cells was relatively high (~50/0.01 mm²) in small foci (i.e., Φ < 0.2 mm; Fig. 2a, top) whereas it was very low (<5/0.01 mm²) in large foci (Φ > 0.2 mm; Fig. 2a, bottom). These results suggest that accumulation of CCR1⁺ cells at the metastatic foci takes place at earlier stages around micrometastases rather than at later stages of colonization. At a higher magnification, 50–80 % of the Ccr1-mVenus⁺ cells were Gr-1⁺ neutrophils, whereas the remaining Gr-1⁻ cells contained both mononuclear and non-mononuclear cells (Fig. 2b).

To characterize these myeloid cells further, we then dissociated the liver tissue through the course of dissemination-metastasis and analyzed recovered cells by flow cytometry. Because not all myeloid cells of the liver foci could be collected into cell suspensions even after multiple enzymatic digestions, the remaining CD45⁺ hematopoietic cells in the remnants were also analyzed histochemically (Supplementary Fig. 1d; see below). Before injection of the CRC cells into the spleen, we found only ~0.2 % of the recovered liver resident cells to be propidium iodide (PI)⁻ and CD45⁺ live hematopoietic cells, and only ~5 % of them were Ccr1-mVenus⁺ with their majority being Gr-1⁺ (Fig. 2c, Day 0). On day 1 post-injection, CD45⁺ hematopoietic cells began to accumulate in the liver and peaked between days 7 and 14, comprising up to ~3 % of recovered cells with their 20–30 % being Ccr1-mVenus⁺ (Fig. 2c). These Ccr1-mVenus⁺ cells were mostly neutrophils as identified by their expression of CD11b and Gr-1, and lobulated nuclei (Fig. 2d), which was consistent with the above observation of BM cells (Fig. 1d). Namely, CCR1⁺ myeloid cells in the liver foci were mostly neutrophils when traced after Ccr1-mVenus expression.

We then analyzed the myeloid cells in the liver metastatic foci in more detail. In large metastatic foci (Fig. 2a), most CD45⁺ hematopoietic cells lacked expression of Ccr1-mVenus, although a few expressing cells were also found. To further characterize the hematopoietic cells accumulated at the site of metastasis, we recovered cells from livers of control mice, and from livers with or without metastatic foci, and analyzed them by flow cytometry (Fig. 3a). The live hematopoietic (PI⁻ CD45⁺) fraction was significantly larger in the metastatic foci (~11 vs. ~1 % in non-transplanted control, Fig. 3a top), and contained more CD11b⁺ myeloid cells (~70 vs. ~15 % in the control, Fig. 3a bottom). Then, CD11b⁺ myeloid cells were divided into two subpopulations in terms of Gr-1 expression (Gr-1^high and Gr-1^low; Fig. 3b, top left). The Gr-1^high cells clustered as a distinct population in the scatter analysis (FSC vs. SSC; Fig. 3b, top right), and showed the neutrophil morphology in the sorted specimens [Fig. 3b, Photo (i)]. On the other hand, the Gr-1^low cells were constituted of three subpopulations in the scatter analysis and re-analysis of the data sets for CD11b vs. Gr-1 (Fig. 3b, bottom). When sorted, two subpopulations with lower scatters (FSC^low SSC^low) showed morphologies of eosinophils and monocytes (Fig. 3b, right photos (ii) and (iii), respectively). In addition, we found cells with a profile of high FSC and SSC with the expression level of CD11b higher than that of eosinophils or monocytes (Fig. 3b, bottom left). The Giemsa staining of the sorted specimens revealed large cells with expanded cytoplasm containing vacuoles, suggesting that they are of the monocyte-macrophage lineage (Fig. 3b, right photo (iv)). We found that all these four subsets of CD11b⁺ myeloid cells were of the BM origin because liver metastasis experiments following BM transplantation using congenic donor cells (CD45.1 as the marker [21]) proved their donor origin (Supplementary Fig. 2).

It is known that a monocyte-derived lineage identified as ‘fibrocytes’ expresses a series of mesenchymal molecules as vimentin and collagens together with markers of hematopoietic cells [22]. We found that these monocytes and macrophage-like cells obtained from the metastatic foci expressed such non-hematopoietic marker molecules at substantial levels whereas neutrophils or eosinophils did not (Fig. 4a). Namely, immunofluorescence microscopy of the sorted cells confirmed expression of collagen type I by monocytes and macrophage-like cells derived from the metastatic foci (Fig. 4b). Moreover, most CD45⁺ hematopoietic cells that formed the ‘cap’ [15, 16] around the liver metastatic foci expressed abundant collagen type I (Fig. 4c). Therefore, we identify these macrophage-like cells as fibrocytes. As we reported earlier, these ‘cap’ cells were stained with a particular anti-CD34 antibody (clone: RAM34, Supplementary Fig. 3). However, we recently found that they were not stained with either of two other anti-CD34 antibodies from different sources (MEC14.7 or 3H1240), suggesting that the staining with RAM34 was caused by a non-specific binding of the antibody, conceivably to collagens. In fact, a qRT-PCR analysis of each sorted myeloid population showed that they contained little CD34 mRNA (data not shown). As mentioned above, it was difficult to dissociate and collect fibrocytes from the liver foci due to abundant collagen fibers in the ECM (Supplementary Fig. 1d). Accordingly, fibrocytes did not constitute the major population when isolated cells were analyzed by flow cytometry (see also below).

Collectively, these results indicate that the CD45⁺ hematopoietic cells accumulated around the metastatic foci consist of neutrophils, eosinophils, monocytes and fibrocytes, and that the major types are monocytes and fibrocytes, with the latter producing abundant collagen.

Having identified four distinct myeloid cell types accumulating in the metastatic foci, we investigated the chronological changes in myeloid cells found in the liver metastasis lesions (Fig. 5a, b). The top panels in Fig. 5a show expression of CD11b and Gr-1 among the CD45⁺ hematopoietic cells whereas the bottom panels do FSC and SSC profiles of the CD11b⁺ Gr-1^low population. After splenic injection of CMT93 cells, the number of CD45⁺ hematopoietic cells started to increase on day 1 and peaked on day 14, decreasing thereafter. Neutrophils and monocytes rapidly appeared on day 1 at high numbers before the appearance of eosinophils and fibrocytes. Accumulation of eosinophils and fibrocytes began on day 7 and peaked on day 14. These data suggest that the recruitment of neutrophils and monocytes may have caused that of eosinophils and fibrocytes.

To determine the possible role of neutrophils in the emergence of fibrocytes, we performed antibody-mediated neutrophil depletion experiments in vivo. Administration of anti–Ly-6G antibody significantly decreased the number of neutrophils in the BM and Ccr1-mVenus⁺ neutrophils in the metastatic foci on Day 7 (Supplementary Figs. 4, 5c). At the same time, this treatment caused a significant decrease in the fibrocyte number as well (on Day 7, Supplementary Fig. 5, 5d), suggesting that the preceding accumulation of neutrophils at around day 1 was necessary for that of fibrocytes later at 7–10 days.

To investigate the roles of CCR1 in the accumulation of these myeloid cell types, we introduced the Ccr1-mVenus reporter transgene into Ccr1 ^−/− mice by successive crosses and subjected them to liver dissemination-metastasis experiments (Fig. 6). On day 1 post-splenic injection, we observed rapid accumulation of Ccr1-mVenus⁺ neutrophils (CD11b⁺ Gr-1^high) in Ccr1 ^+/+ control mice. In the Ccr1 ^−/− mutant mice, however, accumulation of Ccr1-mVenus⁺ CD11b⁺ Gr-1^high neutrophils was reduced significantly (Fig. 6a, b). On day 7, the relative abundance of neutrophils (Ccr1-mVenus⁺ CD11b⁺ Gr-1^high cells) between wild type and knockout mice did not differ significantly (P = 0.32, Fig. 6c, d), whereas the proportion of eosinophils was increased in Ccr1 ^−/− mice, with that of monocytes and fibrocytes rather decreased (Fig. 6c, d). These results suggest that CCR1 played key roles not only in the earlier mobilization and accumulation of neutrophils, but also in later accumulation of monocytes and fibrocytes in liver lesions of CRC metastasis, although the molecular mechanisms by which neutrophils recruit fibrocytes remain unclear.

Both MMP9 and MMP2 are required for promotion of liver metastasis of colon cancer in a mouse model [16] To identify the cell type(s) that produce these matrix metalloproteinases, we performed quantitative RT-PCR analyses on the sorted myeloid cells isolated from the liver metastatic foci (Fig. 7a). As anticipated from the results of the Ccr1-mVenus reporter transgenic mice, CCR1 was abundantly expressed by neutrophils, with much lower levels in other myeloid cells. On the other hand, metalloproteinases MMP9 and MMP2 were expressed in neutrophils and monocytes/fibrocytes, respectively, at significant levels. Notably, MMP9 and MMP2 were not expressed simultaneously in any of these myeloid cell types (Fig. 7b, c). To further characterize the cell types that express MMP9 or MMP2, we performed immunohistochemistry for MMP9, and in situ hybridization for MMP2 on serial sections of the metastatic mouse livers (Fig. 7d). As the result, MMP9 was expressed only by a limited number of cells with lobulated nuclei (polymorphonuclear cells). In contrast, MMP2 mRNA (that hybridized to the anti-sense oligonucleotide, but not to the sense oligonucleotide) was found in the majority of tumor-associated stromal cells. These results suggest that, in the liver metastatic lesions, MMP9 and MMP2 were produced by neutrophils and monocyte/fibrocytes, respectively; by distinctly different myeloid cell types.

In order to seek the clinical relevance of these findings in mice, we investigated expression of CCR1 in human leukocytes. In sharp contrast to mouse neutrophils, flow cytometric analysis revealed that CCR1 was not expressed in human neutrophils, but expressed only by monocytes in the peripheral blood (n = 7, Supplementary Fig. 6a). These findings were confirmed by immunofluorescence staining of peripheral blood cytospin specimens. All the cells expressing CCR1 were CD14⁺ monocytes, not CD15⁺ neutrophils (Supplementary Fig. 6b). We further analyzed human peripheral blood for the so-called ‘myeloid-derived suppressor cells’ (MDSC; CD33⁺ HLA-DR⁻ population [23] ), and found that their monocyte (CD14⁺ cells) subpopulation expressed CCR1 (data not shown). In the liver metastatic lesions of human colon cancer, on the other hand, only a small fraction (~10 %) of CD14⁺ monocytes expressed CCR1 by immunohistochemistry, whereas most of the CD15⁺ myeloid cells with mononuclear morphology expressed CCR1 (Supplementary Fig. 6c). These results suggest the possibility that recruitment of MDSC and/or up-regulation of CCR1 in CD15⁺ myeloid cells contribute to liver metastasis of CRCs in humans as well.