High macrophage infiltration into tumours often correlates with poor prognoses; in colorectal, stomach and skin cancers, however, the opposite is observed but the mechanisms behind this phenomenon remain unclear. Here, we sought to understand how tumour-associated macrophages (TAMs) in colorectal cancer execute tumour-suppressive roles. We found that TAMs in a colorectal cancer model were pro-inflammatory and inhibited the proliferation of tumour cells. TAMs also produced chemokines that attract T cells, stimulated proliferation of allogeneic T cells and activated type-1 T cells associated with anti-tumour immune responses. Using colorectal tumour tissues, we verified that TAMs in vivo were indeed pro-inflammatory. Furthermore, the number of tumour-infiltrating T cells correlated with the number of TAMs, suggesting that TAMs could attract T cells; and indeed, type-1 T cells were present in the tumour tissues. Patient clinical data suggested that TAMs exerted tumour-suppressive effects with the help of T cells. Hence, the tumour-suppressive mechanisms of TAMs in colorectal cancer involve the inhibition of tumour cell proliferation alongside the production of pro-inflammatory cytokines, chemokines and promoting type-1 T-cell responses. These new findings would contribute to the development of future cancer immunotherapies based on enhancing the tumour-suppressive properties of TAMs to boost anti-tumour immune responses.
Macrophages are the primary immune cell-type infiltrating solid tumours 1, contributing up to 50% of the tumour cell mass 2. Consequently, these tumour-associated macrophages (TAMs) play important roles in determining the clinical outcome 3, 4. Like tissue macrophages, TAMs exhibit a continuum of phenotypes ranging from pro-inflammatory to anti-inflammatory 1, 5, and these phenotypes vary in their effects on tumour cells. While pro-inflammatory TAMs can suppress tumour growth, TAMs exhibiting an anti-inflammatory phenotype appear to promote tumour growth 2, 6. In human cancers, TAMs are generally associated with promoting tumour growth 7, but in certain cancers such as colorectal, stomach and skin, the presence of TAMs correlates with good prognoses 4, 8. However, it remains unclear how TAMs in these cancers exert their tumour-suppressive effects. Here, we aim to dissect the mechanisms underlying the tumour-suppressive effects of TAMs in colorectal cancer.
To elucidate the roles of TAMs, we first used an in vitro model known as the multi-cellular tumour spheroid (MCTS) model. This model has been proven to exhibit micro-environmental heterogeneity comparable to that of tumours in vivo, in terms of oxygen, nutrient, catabolite and metabolite gradients, resulting in sub-populations of proliferative and necrotic tumour cells typical of non-vascular tumour micro-regions 9, 10. Compared with using animal models, this MCTS model offers the advantages of studying the interactions between tumour cells and TAMs without confounding factors from other cell types, and in a ‘human’ microenvironment. In this study, we used colorectal cancer as a model to study the mechanisms underlying the tumour-suppressive effects of TAMs. We co-cultured primary human monocytes with human colorectal tumour cells for 8 days as MCTSs, during which time the monocytes would differentiate into TAMs. We performed global gene expression profiling to obtain an overview of the biological functions of TAMs, followed by validation with functional assays. Subsequently, we verified the in vitro findings with tumour tissues from colorectal cancer patients.
The TAMs in the colorectal cancer model were pro-inflammatory and inhibited the proliferation of tumour cells. The TAMs also secreted chemokines that attract T cells and expressed surface molecules for antigen presentation and T-cell co-stimulation. In a mixed lymphocyte reaction (MLR) assay, the TAMs stimulated proliferation of allogeneic T cells and activated type-1 T cells, which are associated with anti-tumour immune responses 11. To confirm these findings, we assessed primary tumour tissues from colorectal cancer patients. TAMs in vivo were indeed pro-inflammatory. Also, the number of tumour-infiltrating T cells correlated with the number of TAMs, supporting the finding that TAMs in colorectal cancer secreted chemokines to recruit T cells. In addition, T cells of the type-1 inflammatory phenotype were present. Clinical data of the patients strongly support the findings that TAMs, together with tumour-infiltrating T cells, exert tumour-suppressive effects. For the first time, we demonstrated the tumour-suppressive properties of TAMs and have begun to dissect the underlying processes. These findings will help us understand the potential beneficial actions of TAMs, so that future cancer immunotherapy can be developed based on enhancing these tumour-suppressive effects of TAMs to boost anti-tumour immune responses.
Tumour cells co-cultured with macrophages exhibit reduced proliferation
We co-cultured human primary monocytes with a human colorectal cell line, HT29, as MCTSs for 8 days (this set-up will be referred to as ‘co-culture spheroids’ hereafter). To mimic tumours with no macrophage infiltration, we cultured tumour cells alone as spheroids (hereafter referred to as ‘tumour spheroids’). To determine if monocytes co-cultured with tumour cells differentiated into macrophages, we checked the expression of CD68 and CD14, markers up-regulated and maintained, respectively, during monocyte-to-macrophage differentiation. In contrast, CD68 and CD14 expression are down-regulated in monocyte-to-dendritic cell (DC) differentiation (Supporting Information Fig. 1A–C). All the monocytes (CD45+) co-cultured with tumour cells for 8 days up-regulated the expression of CD68 (Fig. 1A) and maintained the expression of CD14 (Fig. 1B), compared with freshly isolated monocytes (Supporting Information Fig. 1A), indicating that the monocytes have differentiated into macrophages. Monocyte cultured alone for 8 days under the same conditions, in the absence of tumour cells, do not spontaneously differentiate (Supporting Information Fig. 1D). In addition, from day 4 to 8, CD68+ cells in the co-culture spheroids displayed increase in size, number of cytoplasmic granules and heterogeneity of cell shape characteristic of monocyte-to-macrophage differentiation (Fig. 1C). Together, these observations indicated that the monocytes have differentiated into macrophages after 8 days of co-culture with tumour cells.
To study the interaction between tumour cells and macrophages, we carried out global gene expression profiling on three groups of cells: (I) tumour cells from tumour spheroids; (II) tumour cells sorted out from co-culture spheroids and (III) tumour cells and TAMs from co-culture spheroids (Fig. 2A). To assess the changes induced in the tumour cells upon co-culture with macrophages, we compared the gene expression profiles of (I) and (II), which gave 286 differentially expressed genes (DEGs; Supporting Information Table 1). Sorted tumour cells in (II) had a purity of 92.6±4.2%, with only 0.5±0.2% TAMs remaining (Supporting Information Fig. 2), making the comparison valid. Twenty-eight of the 286 DEGs (10%) were associated with proliferation and apoptosis (Fig. 2B). Tumour cells from co-cultures in (II) displayed an overall down-regulation of genes associated with proliferation and up-regulation of genes associated with apoptosis, suggesting that in the presence of TAMs, the growth of tumour cells was suppressed.
To determine whether TAMs could indeed inhibit proliferation and induce apoptosis of colorectal tumour cells, we monitored the proliferation and apoptosis of three colorectal tumour cell lines (HT29, SW620 and LS174T) in co-culture spheroids, compared with tumour spheroids. Tumour cells in the co-cultures were identified by EpCAM expression (Supporting Information Fig. 3A). To monitor proliferation, PI staining was used to visualise the DNA content; single cells within the S to G2 phases were considered proliferating cells (Supporting Information Fig. 3B and C). Throughout the 8-day culture, the percentage of proliferating tumour cells in all the three cell lines was significantly lower in the co-culture spheroids compared with tumour spheroids (Fig. 2C). To identify the apoptotic cells, annexin V staining was used (Supporting Information Fig. 3D). In two of the three colorectal cell lines (HT29 and LS174T), the percentage of apoptotic tumour cells was higher (although not statistically significant) when co-cultured with TAMs (Fig. 2D). These data show that TAMs in colorectal cancer inhibited tumour cell growth by both suppressing their proliferation as well as promoting their apoptosis. The effect of TAMs on suppressing the tumour cell proliferation appeared to be greater. This observation was supported by the gene expression profile whereby 15 out of 19 genes (79%) related to proliferation were down-regulated, whereas only 6 out of 9 genes (67%) related to apoptosis were up-regulated in tumour cells in co-culture (Fig. 2B).
TAMs expressed genes associated with inflammation, chemotaxis and antigen presentation
To obtain the genes expressed by TAMs, we compared the gene expression profiles of (II) tumour cells sorted from co-culture spheroids and (III) tumour cells and TAMs from co-culture spheroids (Fig. 2A). A total of 348 genes were up-regulated in (III) compared with (II) (Supporting Information Table 2 and Supporting Information Fig. 4A), representing the genes expressed by the TAMs (hereafter referred to as ‘TAM genes’). When mapped into biological functions in silico with MetaCore, the immune-related biological functions associated with these TAM genes included inflammation (18%), differentiation (18%), chemotaxis (8%), MHC Class II antigen presentation (3%), and phagocytosis and endocytosis (2%). The remaining (51%) consisted of other basic biological functions, e.g. cellular metabolic processes, protein localisation and cellular transport, with each function making up <2% of all the TAM genes (Fig. 3A). The genes associated with differentiation supported the earlier data (Fig. 1) that the monocytes differentiated into macrophages after co-culture with the tumour cells. In this study, we focused on three main immune-related functions that potentially have associations with tumour suppression: inflammation, chemotaxis and antigen presentation. The expression levels of some genes from these functions were validated by real-time PCR (Supporting Information Fig. 4B), and the results were consistent with the global gene expression profiling.
The co-culture cytokine microenvironment was pro-inflammatory and tumour suppressive
In support of the pro-inflammatory gene expression profile for colorectal TAMs (Fig. 3A), we detected pro-inflammatory cytokines IL-6, IL-8/CXCL8, IFN-γ and CCL2 at high levels in the supernatants of colorectal co-cultures spheroids, whereas they were barely detectable or significantly lower in the supernatants of tumour spheroids (Fig. 3B). Of these four cytokines, IL-6, IL-8 and CCL2 were detected at the gene expression level of TAMs (Fig. 3A). To assess if the production of these pro-inflammatory cytokines were induced upon interaction with tumour cells, we also tested the supernatants of monocyte cultured alone, in the same spheroid culture conditions, for 8 days (hereafter referred to as ‘monocyte culture’). Supernatants from the monocyte culture contained significantly lower levels of IL-6, IL-8 and CCL2 than the co-cultures, indicating that co-culturing with tumour cells stimulated an increase in the production of these pro-inflammatory cytokines by the TAMs. In addition, vascular endothelial growth factor (VEGF), an anti-inflammatory, tumour-promoting, angiogenic factor produced by tumour cells 12, was present at significantly higher levels in tumour-spheroid cultures than the co-cultures, and absent in monocyte culture. This suggested that the pro-inflammatory TAMs suppressed the production of VEGF by the tumour cells.
We also assessed the levels of the pro-inflammatory cytokines in spheroid models of other cancers in which TAMs have been reported to promote tumour growth, such as prostate cancer (using Du145, DuCap and LnCap cell lines), ovarian cancer (using ES2 cell line) and breast cancer (using MCF7 and SKBR3 cell lines; Supporting Information Fig. 5). IL-6, IL-8 and CCL2 levels were significantly lower in the co-culture supernatants of these other cancers compared with co-culture supernatants of colorectal cancers. Notably, IFN-γ production was suppressed in breast and ovarian co-cultures, while VEGF production was increased in ovarian and certain prostate co-cultures. These observations imply that TAMs in colorectal cancer exhibit a more pro-inflammatory phenotype than TAMs in other cancers in which TAMs promote tumour growth.
The co-culture microenvironment contained T-cell-attracting chemokines
The attraction of T cells into tumours is important since T cells are found to be the major effectors in anti-tumour immune responses 11, 13. Since the TAM genes indicated that the TAMs were involved in chemotaxis and antigen presentation (Fig. 3A), we tested the supernatants of colorectal co-culture spheroids, tumour-spheroids and monocyte culture for the presence of chemokines that attract T cells 14, including CCL2, CCL3, CCL4, CCL7, CCL8, CXCL9, CXCL10 and CXCL12 (Fig. 3B). Of these chemokines, CCL2, CCL3, CCL4, CCL7 and CCL8 were detected at the gene expression level of TAMs (Fig. 3A). Five chemokines, CCL2, CCL7, CCL8, CXCL9 and CXCL10, were present at high levels in the supernatants of co-culture spheroids, but almost absent or significantly lower in the supernatants of both tumour spheroids and monocyte cultures, indicating that co-culturing with tumour cells stimulated the production of these chemokines by the TAMs.
When the supernatants of co-culture spheroids of other cancers (prostate, ovarian and breast) were assessed, CCL2, CCL7, CXCL9 and CXCL10 were present at significantly lower levels compared with that of colorectal cancer (Supporting Information Fig. 5). This implies that TAMs in colorectal cancers secrete more chemokines to attract T cells than TAMs in other cancers in which TAMs promote tumour growth.
To ascertain that the chemokines present in the supernatants of the colorectal tumour model were functionally capable of attracting T cells, we performed Transwell assays using two supernatants: supernatants from co-culture spheroids to mimic microenvironment of a tumour with macrophage infiltration, and supernatants from tumour spheroids to mimic microenvironment of a tumour without macrophage infiltration. Indeed, the supernatants of co-culture spheroids attracted significantly more of both CD4+ and CD8+ T cells than the supernatants of tumour spheroids (Fig. 3C), showing that the chemokines in the supernatants of co-culture spheroids were functionally able to attract T cells.
TAMs exhibited antigen presentation and T-cell co-stimulatory capacities
As the TAM genes indicated that the TAMs were involved in antigen presentation (Fig. 3A), and chemokines that attract T cells were present in the co-culture supernatants, we assessed colorectal TAMs for the expression of cell surface molecules involved in interaction with T cells. The TAMs expressed molecules for antigen presentation (HLA-DR, CD74), T-cell co-stimulation (CD40, CD80, CD86) and CD54 (or ICAM-1), an adhesion molecule that stabilises cell contact during T-cell co-stimulation (Fig. 4A, top panel) 15. To obtain an idea of the level of expression of these molecules on colorectal TAMs, we compared them with in vitro differentiated macrophages and freshly isolated monocytes. The median fluorescence intensity (MFI) of the expression of the molecules (Fig. 4A, middle panel) as well as the percentage of cells that expressed the molecules (Fig. 4A, bottom panel) were studied. Colorectal TAMs exhibited higher expression of all the molecules compared with in vitro MCSF-differentiated macrophages, and up-regulated the expression of all molecules except CD74 compared with freshly isolated monocytes. In addition, a significantly larger percentage of TAMs (than macrophages or monocytes) expressed CD74, CD40, CD80 and CD86. This observation indicated that co-culturing with colorectal tumour cells promoted the differentiation of monocytes to TAMs with enhanced expression of antigen presentation and T-cell co-stimulation molecules.
When TAMs in co-culture spheroids of other cancers (prostate cancer, ovarian cancer and breast) were assessed, CD74, CD40, CD80, CD86 and CD54 were expressed at significantly lower levels, and CD74, CD80 and CD86 were expressed on a significantly lower percentage of the TAMs compared with the TAMs in colorectal co-culture spheroids (Supporting Information Fig. 6). This implies that TAMs in colorectal cancer possess a greater capacity to present antigen and co-stimulate T cells than TAMs in other cancers.
To assess the functional capacity of colorectal TAMs in co-stimulating T cells, we performed an MLR assay. TAMs were sorted from colorectal co-culture spheroids and incubated with allogeneic T cells for 4 days, after which T-cell proliferation was measured by tritiated-thymidine incorporation. Indeed, the TAMs were highly competent at stimulating T-cell proliferation (Fig. 4B). Tumour cells sorted from the co-cultures were unable to stimulate T-cell proliferation, indicating that tumour cells per se do not possess T-cell co-stimulatory properties, and in vitro differentiated macrophages were poor stimulators. Together, these observations indicated that TAMs acquired T-cell co-stimulation capabilities during the co-culture with colorectal tumour cells.
Of the T cells that proliferated upon incubation with TAMs, 71% expressed CD25, an activation marker, and 62% produced IFN-γ, a type-1 inflammatory cytokine (Fig. 4C), indicating that TAMs were able to activate type-1 T cells. There was no activation of type-2, type-17 or regulatory-T cells, indicated by the lack of IL-4, IL-17A or FoxP3 (Fig. 4C and D). Together, these results illustrated that TAMs in the colorectal cancer model were capable of stimulating T-cell proliferation and promoting type-1 T-cell responses.
Examination of TAMs and T cells in colorectal tumour tissue sections support in vitro findings
To confirm the in vitro findings on colorectal TAMs, we studied primary tumour tissues from five colorectal cancer patients (Table 1). Pro-inflammatory TAMs were detected in the colorectal tumour sections, as they stained positive for IFN-γ (Fig. 5A, white arrows). The percentage of TAMs that were IFN-γ+ in each tumour sample was quantified using the software TissueQuest, on five images (each ∼350×250 μm) randomly taken from each tumour tissue section. The images were analysed together to give a representative plot for every tumour sample (Supporting Information Fig. 7). This approach takes into account variations from different parts of the tissue section. The percentage of macrophages that were IFN-γ+ in the tumour samples varied from 6.6 to 50% (Fig. 5B and Table 1). To confirm the in vitro findings that TAMs in colorectal cancers could attract T cells, we quantified the numbers of tumour-infiltrating T cells and TAMs. Indeed, the numbers of tumour-infiltrating T cells and TAMs were highly correlated (r2=0.66, Fig. 5C). Furthermore, the TAMs and T cells were often observed to be in close contact (Fig. 5D, black arrows), suggesting direct interaction of the two cell types, such as antigen presentation to and co-stimulation of T cells by TAMs. In addition, T cells of the type-1 phenotype were present in the tumour tissue sections, as determined by positive IFN-γ staining (Fig. 6A, white arrows). The percentage of T cells that were IFN-γ+ in each patient sample varied from 33 to 90% (Fig. 6B and Table 1). Taken together, the presence of pro-inflammatory TAMs and type-1 T cells, and the correlation of numbers of TAMs and T cells in the primary tissue sections support our in vitro findings (Figs. 3 and 4).
Table 1. Clinical data and a summary of the histological analysis of the primary colorectal tumour tissues
No. of TAMs per FOVa) (% IFN-γ+)
No. of T cells per FOV (% IFN-γ+)
a) FOV, field of view.
Lung and liver
We aimed to elucidate the mechanisms underlying the tumour-suppressive effects of TAMs in colorectal cancer, an important but under-studied property of TAMs. We found that TAMs in the colorectal cancer model were pro-inflammatory (Fig. 3B). Pro-inflammatory TAMs have been associated with anti-tumour properties, such as production of cytotoxic products such as reactive oxygen intermediates, serine proteases and lytic factors, and enhanced the ability to process and present tumour antigens to T cells 2, 6, 16. Notably, IFN-γ was amongst the pro-inflammatory cytokines secreted by the colorectal TAMs, in the co-cultures as well as in vivo (Fig. 5A). This is an important observation as the production of IFN-γ has been mainly associated with type-1 T cells or NK cells 17. Activated macrophages can secrete IFN-γ, but at lower levels 18. IFN-γ is a potent anti-tumour cytokine; its production has been highly correlated with tumour regression in immunotherapy 17. Recently, IFN-γ has been shown to switch tumour-promoting, anti-inflammatory (M2-like) TAMs to the tumour-suppressive, pro-inflammatory (M1-like) phenotype 19, 20, supporting the hypothesis that T-cell responses orchestrate TAM polarisation early on during cancer development 8. Here, our data suggest an alternative: TAMs can produce IFN-γ and other pro-inflammatory cytokines to create a pro-inflammatory microenvironment which activates type-1 T cells, which in turn produce more IFN-γ (Figs. 4–6). IFN-γ can elicit other downstream anti-tumour immune responses, such as sensitising tumour cells to apoptosis 17, potentiating monocyte cytotoxicity against tumour cells 21 and anti-angiogenic activities in vivo 22.
Notably, the production of the tumour-promoting angiogenic factor, VEGF, by the tumour cells was suppressed by the colorectal TAMs in co-cultures (Fig. 3B). Besides promoting angiogenesis, VEGF has been associated with increased risk of relapse in colorectal cancer patients 23. VEGF also exerts other undesirable tumour-promoting effects, such as decreasing production of cytotoxic mediators like granzyme B and perforin by T cells, and decreasing TNF-α and IFN-γ secretion by NK cells 24. Additionally, VEGF promotes the infiltration of immune-suppressive cells such as myeloid-derived suppressor cells and regulatory T cells into tumours 25, which facilitate tumour growth. In fact, clinical studies have shown VEGF to be a valid therapeutic target for colorectal cancer 12. Therefore, the lowering of VEGF levels by TAMs could have multiple important tumour-suppressive effects in vivo.
TAMs in the colorectal cancer model were also found to produce chemokines that attract T cells (Fig. 3B and C). The attraction of T cells is particularly important since T cells are known to be the major effectors in anti-tumour immune responses 11, 13. Amongst these chemokines, CXCL9 and CXCL10, both IFN-γ inducible chemokines, are strong chemoattractants for TH1 cells 26. TH1 cells are important for promoting the killing of tumour cells by cytotoxic T cells 27, 28, and the presence of TH1 cells in colorectal tumours has been correlated with good clinical outcome 11. In addition, TAMs isolated from the co-culture spheroids were capable of stimulating allogeneic T-cell proliferation and activating type-1 T cells (Fig. 4). Taken together, the data suggest that TAMs in colorectal cancers create a type-1 inflammatory microenvironment. These new findings establish the link between clinical observations where (i) a high macrophage infiltration and (ii) a type-1 adaptive immunity in human colorectal tumours independently have been correlated with beneficial clinical outcomes 11, 29.
Importantly, the in vitro findings were also observed in primary colorectal tumour tissues (Figs. 5 and 6). TAMs in vivo were pro-inflammatory, the number of tumour-infiltrating T cells correlated well with the number of TAMs and T cells of the type-1 inflammatory phenotype were present. Notably, the two patients with metastasis of the primary colorectal tumour (25271 and 25316) had the lowest TAM (23–35 TAMs per FOV) and T-cell infiltration (37–55 T cells per FOV, Table 1). Amongst these two patients, the one who had more metastasis and did not survive beyond 5 years (25316) had a lower percentage of IFN-γ-positive TAMs (6.6%) and T cells (45%). This supports our hypothesis that the attraction and activation of type-1 T cells into the tumour by pro-inflammatory TAMs play a crucial role in suppressing tumour progression.
For the first time, we have dissected the potential tumour-suppressive roles of TAMs in human colorectal tumours. The data suggest that in vivo, pro-inflammatory TAMs recruit T cells to the tumour site, present antigens and provide co-stimulating signals to activate the T cells, and subsequently promote the type-1 inflammatory response that leads to downstream anti-tumour immune activities. These findings explain the observation that high macrophage infiltration into colorectal cancers correlates with good patient prognoses. Besides helping us to understand how TAMs execute their tumour-suppressive role, these novel findings will contribute towards the rational design of therapeutic strategies to harness the power of TAMs for cancer treatment in future.
It is noteworthy that the tumour types in which TAMs have been observed to exert a tumour-suppressive effect are located in the barrier organs of the body, namely the colon, stomach and skin. This suggests that there may be a qualitative difference in the macrophages operating in these barrier microenvironments compared with other organs, giving rise to a tumour-suppressive phenotype in these macrophages when cancer develops. Detailed studies on the effects of TAMs on tumour cells will further help in understanding the mechanisms of action of TAMs. Together, these would aid in the development of strategies to manipulate and re-educate TAMs to mount anti-tumour responses.
Materials and methods
All blood samples and procedures in this study were approved by the Domain Specific Review Board (DSRB), National Healthcare Group, Singapore (Reference code: 08-352E). Informed consent was given in accordance with the Declaration of Helsinki. Peripheral blood mononuclear cells were isolated from buffy coats (National University Hospital Blood Donation Center, Singapore) by Ficoll-Hypaque density gradient centrifugation; monocytes were positively selected using CD14 Microbeads (Miltenyi). Purity and viability of monocytes obtained were 98.0±1.7 and 98.7±0.8%, respectively, assessed by flow cytometry.
Generation of MCTSs
Human colorectal cancer cell lines (HT29, SW620, LS174T, authenticated by CellBank, Australia), prostate cancer cell lines (Du145, DuCap and LnCap), ovarian cell line (ES2) and breast cancer cell lines (MCF7 and SKBR3) were used to generate MCTSs by the liquid overlay method: 104 tumour cells and 104 monocytes (co-culture spheroids) or only 104 tumour cells (tumour spheroids) or 104 monocyte (monocyte culture) were seeded in 200 μL medium in 96-well coated with 0.8% w/v Agar Noble (Difco, BD). Cells were cultured in IMDM (Hyclone) with 5% human serum (HS; Innovative Research) at 37°C with 5% CO2 for 8 days. Culture medium was changed on day 4, when half the medium was replaced with fresh medium.
Generation of macrophages and DCs in vitro
Monocytes were treated with 100 ng/mL M-CSF for 8 days to generate macrophages, or 100 ng/mL GM-CSF and 25 ng/mL IL-4 for 8 days to generate DCs.
Dead cells were excluded using live/dead fixable dead cell stain (Invitrogen). For intracellular labelling, manufacturer's instructions for the fixation/permeabilisation kit (BD Biosciences) were followed. Antibodies: EpCAM (9C4), CD68 (Y1/82A), CD14 (61D3), HLA-DR (LN3), CD40 (5C3), CD80 (2D10), CD86 (IT2.2), CD54 (HA58), CD3 (III471), CD25 (BC96), IFN-γ (4S.B3), IL-4 (8D48), IL-17A (64DEC17), FoxP3 (PCH101) and their respective isotypes were from eBioscience. CD74 (LN2) was from BioLegend. Data were analysed using FlowJo (Tree Star, Ashland, OR, USA).
Fluorescence-activated cell sorting (FACS)
Co-culture MCTSs were dissociated with Accumex (Innovative Cell Technologies) and labelled with anti-EpCAM-FITC (tumour cells), anti-CD14-PE (macrophages) for sorting (FACSAriaII, BD). The percentage of TAMs in the co-culture spheroids after 8 days of culture was 7±2% (n=4). Tumour cells were sorted directly into Trizol-LS (Invitrogen).
Chloroform (0.2 mL) was added per 1 mL Trizol-LS, mixed and centrifuged (12 000 rpm, 15 min, 4°C). The upper aqueous phase was extracted and an equal volume of 70% ethanol was added. Total RNA was extracted using the RNeasy kit (Qiagen), according to manufacturer's protocol. cRNA preparation, purification and labelling, array hybridisation and scanning were performed according to the manufacturer's protocol (Illumina).
Global gene expression profiling
Gene expression was compared between (I) tumour spheroids, (II) tumour cells only sorted out from co-cultures and (III) co-culture spheroids. Four biological replicates (monocytes from different donors) for co-culture spheroids and two replicates for tumour spheroids were studied using Illumina HumanRef-8 v.2.0 chips. Illumina BeadStudio was used for background correction and generating average signal intensity. R/Bioconductor 30 was used for quantile normalisation 31. Probes showing low variability were discarded by applying interquartile filter (IQR=0.25). A linear model 32 was employed by controlling the number of false positives by false discovery rate (FDR) with adjusted p-value of ≤0.05 33. A log2 fold change signal threshold (log2(FC)≥1.0) was applied for comparison of (I) and (II), and (log2(FC)≥1.5) for comparison of (II) and (III). MultiExperiment Viewer 34, 35 was used for hierarchical clustering, setting a Euclidean distance as the measure of dissimilarity and average linkage as the linkage method. To identify biologically relevant regulatory processes and pathways, MetaCore with a FDR adjusted p-value of ≤0.05 was used.
Real-time polymerase chain reaction (PCR)
Primers were designed using Primer-3 (Supporting Information Table 3). Real-time PCR was performed with Stratagene Mx3000P (Agilent). All gene expressions were normalised to housekeeping gene hypoxanthine phosphoribosyltransferase (HPRT).
Tumour cells were labelled with anti-EpCAM-FITC, stained with propidium iodide (15 μg/mL; Sigma) and analysed by flow cytometry.
Multiplex microbead immunoassay and ELISA
Pre-cleared cell culture supernatants (days 5–8) were used for cytokine measurement (Bio-Plex Pro Human Cytokine Groups I and II, BioRad). Culture medium with 5% HS was used as blank and diluent. Detection was carried out with Luminex 200TM. Results were acquired using IS 2.3 software. CCL8 and CCL13 were detected using Duoset ELISA Development Systems (R&D Systems). Samples were diluted and prepared according to manufacturer's protocol. Culture medium with 5% HS was used as the blank.
Supernatants were incubated at 37°C, 5% CO2, 30 min to allow pH equilibration before assay. Total white blood cells (WBC) were obtained from buffy coats after red blood cell lysis. Totally, 7.5×105 WBC were seeded onto cell culture inserts (BD) with 8.0 μm pore sizes, incubated with supernatants in wells for 1 h. T cells (CD3, CD4, CD8) that trans-migrated through the inserts were distinguished by fluorescence labelling and analysed by flow cytometry (LSRII, BD). Countbright beads (Invitrogen) were used for cell number quantification. WBC from six donors and supernatants from three replicates (spheroid cultures) were used.
Mixed lymphocyte reaction (MLR)
Tumour cells from co-culture spheroids were labelled with anti-EpCAM-FITC followed by anti-FITC microbeads (Miltenyi) for magnetic sorting into tumour cells and TAMs. Sorted tumour cells and TAMs from co-cultures, and in vitro differentiated macrophages (stimulator cells) were treated with mitomycin C (50 μg/mL, 30 min) (Sigma) to inhibit T-cell proliferation. Total T cells were isolated from blood of another donor using CD3 MicroBeads (Miltenyi). 105 T cells (T) per well were incubated with stimulator cells (S) at T/S ratio of 10:1. Cells were incubated for 4 days, pulsed with 0.5 μCi 3H-thymidine (PerkinElmer, Boston, MA, USA) per well for the last 18 h. T-cell proliferation was determined using a TopCount Microplate Scintillation Counter (Packard Instruments). For intracellular cytokine staining, T cells from MLR assay were re-stimulated with 50 ng/mL PMA (Sigma), 1 μM ionomycin (Sigma) and treated with monensin (BioLegend) overnight. Monocytes and allogeneic T cells from three donors each were used.
Human tumour samples
All paraffin-embedded tumour tissue samples and procedures were approved by the Centralised Institutional Review Board (CIRB), Singhealth, Singapore (Reference code: 2009/1001/B).
Paraffin sections were stained with anti-CD68 (PG-M1, Novus Biologicals) and anti-CD3 (polyclonal, Dako), detected using DakoCytomation EnVision+ HRP System and peroxidase substrate AEC Kit (Vector Laboratories).
Paraffin sections were stained with anti-IFN-γ (polyclonal, Abcam), anti-CD3 (F7.2.38, Dako) and anti-CD68 as above, detected using AlexaFluor488 donkey anti-rabbit, AlexaFluor546 donkey anti-mouse secondary antibodies, mounted with Prolong® anti-fade containing DAPI (Invitrogen). Images were acquired with the TissueFAXS platform (TissueGnostics, Austria).
For IHC, manual quantification of CD68+ and CD3+ cells in ten images (each ∼1200×500 μm) randomly taken from each tumour tissue sample was performed. Correlation of the two cell types was assessed using linear regression. For IF, quantification of staining was performed using the software TissueQuest (TissueGnostics) on five images (each ∼350×250 μm) randomly taken from each tumour tissue sample.
Student's t-tests were used: *p<0.05; **p<0.01; ***p<0.001; ns, not significant. All data plotted represent mean±standard deviations (SD).
The authors thank NUH Blood Donation Center for supplying buffy coats; the staff Histology and Microarray Units (Biopolis Shared Facilities), Ms. Poon Lai Fong, Mr. Adrian Lai Tuck-Siong and Dr. Esther Koh for technical assistance; Dr. Shi Xianke (Carl Zeiss, Singapore) for the loan of TissueFAXS and TissueQuest platform; Dr. Lucy Robinson for scientific editing of the manuscript, Dr. Jean-Pierre Abastado and Dr. Subhra K. Biswas for critical reading of the manuscript; Dr Rotzschke's Lab for SW620 and LS174T cell lines; and members of PK Lab for their input. This research is funded by the Biomedical Research Council, A*STAR, Singapore.
Conflict of interest: The authors declare no financial or commercial conflict of interest.