Origins of tumor-associated macrophages and neutrophils

Spleen Contributes TAMs and TANs.

The lung contains different populations of macrophages, which can express CD11b at variable levels (20) and include CD11b⁻ CD11c⁺ F4/80⁺ alveolar macrophages. To test whether the spleen participates in macrophage responses during tumor progression, we splenectomized KP mice at week 0 (at tumor initiation) (Fig. 1A) via a procedure that preserves bone marrow and blood cell pools (17). The procedure reduced both CD11b⁺ and CD11b⁻ lung macrophages (TAMs) by week 11 in a large fraction of animals (Fig. 1B and Fig. S1). Splenectomy also significantly reduced the numbers of CD11b⁺ Gr1⁺ Ly-6G⁺ lung tissue neutrophils (TANs) (Fig. 1B).

The experiment was repeated with mice that were splenectomized at week 8 instead of week 0 (Fig. 1A) (i.e., mice with preestablished cancer). These mice likewise showed impaired TAM and TAN responses by week 11 (Fig. 1B). This finding supports the previous notion that TAMs and TANs are continuously replaced during tumor progression (14, 15) and suggests that the spleen acts to contribute these cells at least after week 8. The spleen was also found to be required for full accumulation of TAMs and TANs in two grafted tumor models (Fig. S2).

TAMs produce proteases that degrade ECM proteins and enhance tumor cell invasion and intravasation (21). Ex vivo flow cytometry analysis of KP mice that received a cysteine Cathepsin B-targeted optical sensor revealed high cathepsin activity in TAMs, as expected, but also in splenic monocytes (Fig. S3). We next compared proteolytic activity in vivo by combining the optical sensor with fluorescent-mediated tomography/computed tomography (CT) fusion imaging (2). The lungs of spleen-deficient animals showed significantly reduced proteolytic activity (Fig. 1C), in line with the above findings that splenectomy reduced the TAM and TAN responses. Other tumor-promoting functions have been identified for TAMs (5) and may also be relevant in KP mice.

Spleen Contributes to Tumor Growth.

Given that removal of the spleen prevented full-fledged TAM accumulation and TAM-associated tumor-promoting activity, we investigated whether it also affected tumor progression. To do this, spleen-sufficient and spleen-deficient animals were monitored by CT at weeks 7, 9, and 11 after tumor initiation. The majority of KP mice from which the spleen was removed showed decreased numbers of detectable tumor nodules and reduced total tumor volumes over time (Fig. 1 D and E). Histopathological analysis at week 11 (Fig. S4) confirmed reduced tumor progression in the majority of spleen-deficient mice. This same phenotype was observed regardless of whether mice were splenectomized at week 0 or at week 8. In both cases, splenectomized mice often presented with fewer and smaller tumor nodules, which were either grade 1 adenomatous hyperplasias or grade 2 adenomas compared with grade 2 or grade 3 adenocarcinomas in age-matched control animals.

Splenectomy decreased the TAM response before changes in tumor progression could occur (Fig. S5A), which is in accordance with a causal association between TAMs and tumor progression. Also, in support of the notion that the spleen contributed cells to tumors directly, the number of blood CD11b⁺ Ly-6C^hi monocytic cells decreased shortly after splenectomy in tumor-bearing KP animals (Fig. S5 B and C). Splenectomy likely did not affect bone marrow hematopoiesis because bone marrow leukocyte counts remain normal in steady state (17) and in the presence of growing tumors (Fig. S5D).

Histopathological analysis indicated that the tumor burden at week 11 in spleen-deficient KP mice was similar (i.e., not lower) to that at week 8 in spleen-sufficient mice (Fig. S4); thus, splenectomy delayed tumor progression but did not induce tumor rejection. Antitumor cytotoxic T lymphocytes (CTLs) seemed uninvolved because they remained rare in the tumor stroma of KP mice even in spleen-deficient animals (Fig. S5E). Likewise, the CTLs that did accumulate in tumors did not exhibit enhanced effector functions (Fig. S5F). Splenectomy also did not affect the number of tumor-infiltrating FoxP3⁺ regulatory T cells (Fig. S5G). In contrast, impaired tumor progression in spleen-deficient KP mice was associated with decreased numbers of TAMs and TANs (Fig. S5 H and I).

Inflammatory Ly-6C^hi monocytes express the chemokine receptor CCR2, which is involved in the amplification of TAM responses (15, 22). To test whether CCR2 controlled spleen-derived TAM accumulation, we used a lipid nanoparticle that contains short interfering (si) CCR2 RNA sequences and that selectively silences CCR2 expression in splenic Ly-6C^hi (CCR2⁺) monocytes (23). KP mice received injections of CCR2-silencing siRNA (siCCR2) or control-silencing (siCON) nanoparticles starting on week 12 after tumor initiation and over 12 d (Fig. 2A). The siCCR2-treated mice showed an approximately twofold reduction of TAMs (Fig. 2B). Splenic granulocytic cells do not express CCR2 and, accordingly, siCCR2 treatment did not alter TAN responses (Fig. 2B). Further evaluation of lung tissue showed reduced tumor progression in mice that received siCCR2 nanoparticles (Fig. 2C and Fig. S6). These data indicate that splenic-derived TAMs were recruited via CCR2 and were causally associated with tumor progression. The apparently unaltered TAN response in siCCR2-treated animals was insufficient to promote tumor growth; thus, further investigation will be needed to define the precise impact of spleen-derived TANs.

Splenic Monocytic and Granulocytic Cells Relocate to Cancer Sites.

Based on these results, we attempted to define whether reservoir splenocytes physically relocate to tumor sites. We used a method involving transplantation of an anastomosed spleen and tracking of the cells originating from the transplanted spleen (17) (Fig. 3A). Perfusion of the transplant (Fig. 3B) allowed donor splenocytes to enter the recipient's bloodstream directly, and thus to relocate elsewhere. Spleens transplanted into tumor-free animals did not trigger a measurable cell release (17) (Fig. 3C). However, spleens transplanted into tumor-bearing KP mice (11 wk after tumor initiation) triggered relocation of numerous cells to tumors. Namely, 1.9 ± 1.2 × 10⁵ TANs and 5.1 ± 3.2 × 10⁵ TAMs redistributed within 24 h from the donor spleen to the recipient's tumor stroma. A total of 66 ± 4% and 61 ± 17% of splenic-derived TAMs down-regulated Ly-6C and up-regulated F4/80, respectively, in line with the initiation of a differentiation program of these cells on recruitment to tissue. This approach did not permit us to assess the relative contribution of spleen- vs. bone marrow-derived cells; however, it showed that within 24 h, the spleen replaced as many as 4% and 8–10% of the preexisting endogenous TAM and TAN repertoires, respectively (Fig. 3D). Because these repertoires increased, on average, only 2% daily, it is likely that a small fraction of older cells either died locally or exited the tumor stroma, in part, counteracting the contribution of newly arrived splenic cells. Future studies will have to determine whether TAMs and TANs are short-lived or exit the tumor stroma. Either way, these experiments positioned the spleen as a significant contributor of TAM and TAN precursors.

Spleen Produces CD11b⁺ Ly-6C^hi Monocytic and CD11b⁺ Ly-6G^hi Granulocytic Cells Locally During Cancer Progression.

The spleen may enhance TAM and TAN responses by controlling the quality and/or quantity of immature myeloid precursor/progenitor cells released to tumors. For example, splenic monocytic and granulocytic cells may acquire unique functions locally by receiving microenvironmental instruction signals, which differ from those in bone marrow. Although this mechanism cannot be formally excluded, a comparative analysis of (CD11b⁺ Ly-6C^hi) monocytic and (CD11b⁺ Ly-6G^hi) granulocytic cells in bone marrow vs. spleen of tumor-bearing hosts (Fig. S7A) and of splenic monocytic and granulocytic cells in tumor-bearing vs. control mice (Fig. S7 B and C) did not reveal notable differences. Phenotypic changes were detected mostly only on cell recruitment to the tumor stroma (Fig. S7C and Table S1).

The spleen may also control TAM and TAN responses quantitatively by supplying new monocytic and granulocytic cells continuously during tumor progression. Continuous support of the spleen, however, requires its constant replenishment with new reservoir cells to compensate for the loss of those that are mobilized. Interestingly, the spleen of tumor-bearing mice contained a significantly larger number of cells in the S/G2 phase of the cell cycle, which indicated active cell proliferation within the organ. The proliferating splenocytes exhibited a CD11b⁻ Lin⁻ phenotype (Fig. 4A). Thus, cancer induced the accumulation of proliferating splenocytes that were neither monocytic nor granulocytic but could include their lineage progenitors. Accordingly, in vitro cultures of splenocytes from tumor-bearing mice produced more granulocyte/macrophage colonies than from controls (Fig. 4B).

The established paradigm states the following paths of lineage differentiation for monocytes and neutrophils: HSC→CMP (common myeloid progenitor)→GMP→MDP (macrophage/dendritic cell progenitor)→monocyte (24) and HSC→CMP→GMP→neutrophil (10). The presence of these populations was tested by flow cytometry with a combination of markers used for cell phenotyping in bone marrow (10–12). HSCs, CMPs, GMPs, and MDPs increased largely in the spleen (Fig. 4 C and D and Fig. S8A) in comparison to bone marrow (Fig. S8 B and C) in KP mice at 11 wk after tumor initiation. Similar splenic responses were detected in two other grafted tumor models (Fig. S8D).

Subsequent experiments were restricted to the analysis of HSCs and GMPs. These studies indicated that their expansion in the spleen of tumor-bearing mice was likely the consequence of both their continuous recruitment and their local proliferation. To analyze recruitment, parabiosis was established between two tumor-bearing mice for 16 d until HSC and progenitor cell equilibration. The spleens contained a significant fraction of HSCs (28.5 ± 2%) and GMPs (24 ± 4%) that originated from the other parabiont; thus, circulating HSCs and progenitor cells contributed to the splenic repertoire. These observations are in line with previous evidence that a small fraction of HSCs and progenitor cells exit the bone marrow and traffic through different extramedullary tissues, including the spleen (25, 26), and that in response to inflammatory stimuli, these cells produce tissue-resident cell populations (27–29). To analyze local proliferation, splenic HSCs and GMPs were labeled with a cell cycle marker. Flow cytometry analysis showed that a high fraction of these cells were in the S/G2 phase of the cell cycle, similar to their bone marrow counterparts (Fig. S8E). Splenic HSCs and GMPs were functionally active in vitro, because they formed colonies as efficiently as their bone marrow counterparts.

We performed additional experiments to evaluate the fate of GMPs in vivo. Intravital microscopy 5 d after injection of fluorescently tagged GMPs indicated that the cells gave rise to large numbers of colonies in the splenic red pulp parenchyma next to, but outside, collecting venous sinuses (Fig. 5A). Characterization of the cells by flow cytometry identified that GMP descendants were virtually all monocytic or granulocytic (Fig. 5B). The ability of transferred GMPs to produce myeloid progeny was highest within the spleen of KP animals with cancer. The activity remained relatively low within the bone marrow as well as the liver (another extramedullary site that can exhibit myelopoietic activity) even in splenectomized mice (Fig. 5B and Fig. S8F). We made similar observations on adoptive transfer of HSCs instead of GMPs, although, as expected, the appearance of HSC-derived monocytic and granulocytic cells was delayed by several days. Finally, GMPs had a similar fate in a grafted mouse tumor model (Fig. S8G). The data indicate that splenic HSCs and progenitor cells actively produce Ly-6C^hi monocytic and Ly-6G^hi granulocytic cells locally and participate, at least in part, in replenishing the splenic reservoir in tumor-bearing mice.

Splenic GMPs Contribute to the TAM and TAN Responses.

Tracking of the GMP progeny in the lungs of KP mice also identified that these cells produced tissue macrophages (TAMs) and neutrophils (TANs) (Fig. 5C). The capacity of the transferred GMPs to generate these cells markedly increased in mice with cancer (11 wk after tumor initiation). To test whether the GMP-derived TAMs and TANs required the spleen, we repeated the GMP adoptive experiments (11 wk after tumor initiation); however, we used spleen-sufficient or spleen-deficient mice as recipients this time. Five days after transfer, GMPs failed to produce full TAM and TAN responses in the spleen-deficient mice (Fig. 5C). This result supports the role of splenic GMPs as TAM and TAN progenitors and positions the spleen as an important site for amplification of these cells.

Human Patients with Cancer Expand Splenic Myeloid Progenitor Cells.

Mouse and human spleens are physiologically very similar (30), and the spleen of patients with chronic inflammatory disorders is capable of producing leukocytes (31–33). However, whether the extramedullary myeloid response seen in animal models also occurs in human patients has not been addressed.

Fresh splenic tissue was collected from seven control patients and seven patients with pathological evidence of invasive cancer. Splenic GMP-like cells were sorted based on the expression of cell surface markers used previously to identify human bone marrow GMPs (11) (i.e., CD117⁺ CD34⁺ CD38⁺ IL-3Ra⁺ CD45RA⁺ Lin^−/int; Fig. 6A).

PCR analysis identified that the splenic GMP-like cells produced the transcription factor PU.1 and CCAAT/enhancer binding protein ε (C/EBPε), the granulocyte colony-stimulating factor (CSF), granulocyte-macrophage (GM)-CSF receptors, and the enzyme myeloperoxidase, in line with the study of Manz et al. (11). The cells did not express GATA binding protein 1, GATA binding protein 3, von Willebrand factor, or IL-7 receptor α (IL-7Ra), all of which are involved in developmental lineages disparate from neutrophils and monocytes (Fig. 6B). In vitro, purified splenic GMP-like cells produced colonies and generated cells that were mostly CD14⁺ CD115⁺ CD68⁺ CD11b^+/− (Fig. S9A). Because we could not detect granulocytes, it is possible that the culture conditions skewed maturation toward the mononuclear phagocyte lineage or that the splenic population was already committed to this lineage (and thus may also qualify as human MDPs).

Patients with invasive cancer had significantly higher numbers of splenic GMPs ex vivo compared with controls (P < 0.05; Fig. 6C), and in vitro cultures of splenocytes of patients with cancer (n = 2) produced higher numbers of granulocyte/macrophage colonies than splenocytes of a control patient (Fig. S9B). Also, a flow cytometry protocol established to detect endogenous human splenic monocytes and neutrophils ex vivo (Fig. S9C) showed increased numbers of these cells in patients with cancer (Fig. 6D). These findings suggest that patients with cancer can mount active GMP responses and produce monocytic and granulocytic cells locally.

To address directly whether human splenic GMPs produce monocytic and granulocytic cells in vivo, we purified human splenic progenitor cells obtained ex vivo from a patient with invasive cancer and injected the cells into a nonobese diabetic-SCID mouse bearing an NCI-H226 human non-small cell lung carcinoma xenograft. Within 5 d of the transfer, the GMP progeny accumulated and differentiated mainly into CD11b⁺ CD68⁻ CD11c⁻ monocyte-like cells in the spleen. Human cells also accumulated at the tumor site, where they acquired a CD68⁺ macrophage-like phenotype (Fig. 6E).

Together, the human data appear to mirror our observations in the mouse, and they are consistent with the notion that the spleen can support production of monocytic and granulocytic cells, at least in patients with cancer. Aside from the fact that splenic GMP responses develop in both species, it is noteworthy that the magnitude of these responses was comparable, if not more pronounced, in humans (taking scale differences between species into account; Fig. 6F).