Article Text

Original article
Dual prognostic significance of tumour-associated macrophages in human pancreatic adenocarcinoma treated or untreated with chemotherapy
  1. Giuseppe Di Caro1,
  2. Nina Cortese1,
  3. Giovanni Francesco Castino1,
  4. Fabio Grizzi1,
  5. Francesca Gavazzi2,
  6. Cristina Ridolfi2,
  7. Giovanni Capretti2,
  8. Rossana Mineri3,
  9. Jelena Todoric4,
  10. Alessandro Zerbi2,
  11. Paola Allavena1,
  12. Alberto Mantovani1,
  13. Federica Marchesi1
  1. 1Department of Immunology and Inflammation, Humanitas Clinical and Research Center, Rozzano, Italy
  2. 2Section of Pancreatic Surgery, Department of Surgery, Humanitas Clinical and Research Center, Rozzano, Italy
  3. 3Molecular Biology Section, Clinical Investigation Laboratory, Humanitas Clinical and Research Center, Rozzano, Italy
  4. 4Laboratory of Gene Regulation and Signal Transduction, Department of Pharmacology, University of California San Diego, San Diego, California, USA
  1. Correspondence to Dr Federica Marchesi, Department of Immunology and Inflammation, Humanitas Clinical and Research Center, Via Manzoni 56, Rozzano (MI) 20089, Italy; federica.marchesi{at}humanitasresearch.it

Abstract

Objective Tumour-associated macrophages (TAMs) play key roles in tumour progression. Recent evidence suggests that TAMs critically modulate the efficacy of anticancer therapies, raising the prospect of their targeting in human cancer.

Design In a large retrospective cohort study involving 110 patients with pancreatic ductal adenocarcinoma (PDAC), we assessed the density of CD68-TAM immune reactive area (%IRA) at the tumour–stroma interface and addressed their prognostic relevance in relation to postsurgical adjuvant chemotherapy (CTX). In vitro, we dissected the synergism of CTX and TAMs.

Results In human PDAC, TAMs predominantly exhibited an immunoregulatory profile, characterised by expression of scavenger receptors (CD206, CD163) and production of interleukin 10 (IL-10). Surprisingly, while the density of TAMs associated to worse prognosis and distant metastasis, CTX restrained their protumour prognostic significance. High density of TAMs at the tumour–stroma interface positively dictated prognostic responsiveness to CTX independently of T-cell density. Accordingly, in vitro, gemcitabine-treated macrophages became tumoricidal, activating a cytotoxic gene expression programme, inhibiting their protumoural effect and switching to an antitumour phenotype. In patients with human PDAC, neoadjuvant CTX was associated to a decreased density of CD206+ and IL-10+ TAMs at the tumour–stroma interface.

Conclusions Overall, our data highlight TAMs as critical determinants of prognostic responsiveness to CTX and provide clinical and in vitro evidence that CTX overall directly re-educates TAMs to restrain tumour progression. These results suggest that the quantification of TAMs could be exploited to select patients more likely to respond to CTX and provide the basis for novel strategies aimed at re-educating macrophages in the context of CTX.

  • PANCREATIC CANCER
  • MACROPHAGES
  • CHEMOTHERAPY
  • IMMUNOREGULATION
  • IMMUNOHISTOCHEMISTRY

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Significance of this study

What is already known on this subject?

  • Tumour-associated macrophages (TAMs) are an important component of the microenvironment of solid tumours, in the majority of cancers associated to unfavourable prognosis. However, their peculiar plasticity allows them to acquire either protumour or antitumour functions. Therefore, TAMs emerge as attractive targets of therapeutic strategies aimed at reprogramming their protumour phenotype into an effective antitumour activity.

  • Human pancreatic ductal adenocarcinoma (PDAC) is a very aggressive disease, displaying low responsiveness to conventional therapies and characterised by an immunosuppressive microenvironment.

What are the new findings?

  • In human PDAC, macrophages accumulate at the tumour–stroma interface and display an immunosuppressive phenotype. However, density of macrophages is a critical determinant in identifying patient responsiveness to conventional chemotherapy (CTX).

  • In vitro, CTX synergises with TAM-mediated tumour cytotoxicity, preventing their tumour-protective role and reinstating their antitumour function, by a T-cell independent mechanism.

  • In patients with PDAC, administration of neoadjuvant CTX is associated to a decreased density of protumour TAMs at the tumour–stroma interface, supporting the ability of CTX to modulate the function of TAMs in the tumour microenvironment.

How might it impact on clinical practice in the foreseeable future?

  • The small therapeutic window of CTX in patients with PDAC would greatly benefit by the identification of novel biomarkers with the ability to detect patients who are more likely to benefit from CTX.

  • Despite its clinically aggressive behaviour, in human PDAC, targeting TAMs may be beneficial to tumour prognosis and in some cases be sufficient to ignite an effective antitumour action.

  • New strategies enhancing CTX-driven antitumour activity might open the way for personalised medicine and more tailored therapeutic treatments.

Introduction

Tumour-associated macrophages (TAMs) are a major component of the tumour immune microenvironment, involved in the majority of cancers in the orchestration of key steps required for cancer occurrence and progression.1–4 Even though this simplified view is substantiated by epidemiological and experimental evidence,5–8 macrophages display a peculiar plasticity, which allows them to exert antagonistic functions, as they can acquire either protumour or antitumour functions;9 ,10 this makes TAMs attractive targets of therapeutic strategies aimed at reprogramming their protumour phenotype into an effective antitumour activity. The antitumour role of macrophages has been previously linked to the orchestration of T-cell antitumour immunity;11 however, recent results showed that tumour immune surveillance could be firmly directed by TAMs when educated by specific treatments, in a T-independent fashion.12 In this scenario, macrophages are being rediscovered as key regulators of tumour response to anticancer therapies.3 ,13–18

Pancreatic ductal adenocarcinoma (PDAC) is the fifth cause of cancer worldwide.19 Approximately 80% of cases are not eligible for surgery for either local or metastatic disease and have dismal survival rates at 1 year from diagnosis,20 with surgery being the only treatment with curative intent for patients with early diagnosis. Lately, new multidisciplinary perioperative approaches and the concentration of care in high volume centres have improved the prognosis of patients undergoing surgery.21 ,22 In this new scenario, the characterisation of PDAC cases with long survival could reveal novel biological determinants of progression for this disease, by comparison with patients with dismal prognosis. To date, most of the data generated by global analyses and standardised assessment of immune infiltrating cells have been focused on the prognostic determination of tumour-infiltrating lymphocytes (TILs) and suggested that the role of adaptive immune responses in controlling tumour growth is a common feature to most solid tumours.23 Notably, while TILs are candidate biomarkers in determining the prognosis in several cancers,23 the idea that TILs might be useful in identifying patients more responsive to conventional chemotherapies in humans23 is merely speculative.24 New strategies or new cellular mediators capable of enhancing chemotherapy (CTX)-driven antitumour activity are required to open the way for personalised medicine and more tailored therapeutic treatments. The small therapeutic window of CTX in patients with PDAC25–27 would greatly benefit from the identification of novel biomarkers with the ability to detect patients with PDAC who are more likely to benefit from chemoimmunotherapeutic approaches and would be highly beneficial to the cure of this disease.

In this study, we approached the microenvironment of human pancreatic cancer by systematically evaluating the macrophage immune component at the tumour–host interface. The occurrence and the extent of CD68+ TAMs (CD68-TAMs) at the tumour–stroma interface were evaluated in a large consecutive series of 110 patients with human pancreatic adenocarcinoma who underwent curative surgical resection. The clinical relevance of TAMs and whether their density associates to disease recurrence and metastatic potential in human PDAC were assessed in relation to postsurgical adjuvant CTX. Here, we show that the density of macrophages is a critical determinant in identifying patient responsiveness to conventional CTX and its response rate was highly modified by the presence of macrophages at the tumour–stroma interface. In in vitro analyses we dissected the synergism of CTX and TAMs and found that macrophages exposed to CTX switched to a proinflammatory phenotype and became tumoricidal by a TAM-autonomous and T-cell independent mechanism.

Materials and methods

Patients and study design

We designed a consecutive and retrospective observational cohort study including 110 patients aged older than 18 years, diagnosed with PDAC and who consecutively underwent surgery with curative intent in Humanitas Clinical and Research Center, from February 2010 to December 2012. All patients gave their informed consent prior to inclusion in the study, and the Ethics Committee of the Humanitas Clinical and Research Center approved the study. Gemcitabine (GEM)-based, postsurgical adjuvant CTX was allocated to most patients (see online supplementary table S1) by a non-random assignment according to adjuvant protocols in use at the time of surgery. Sixteen patients received a GEM-based neoadjuvant CTX regimen before surgery. The continuous, median and quartile values of the percentages of CD8-TIL and CD68-TAM IRA (immunoreactive area) were compared with demographics, clinical and histopathological features and all these variables were tested as predictors of patient's outcome (disease-free survival, DFS), patient's disease-specific survival (DSS) and metachronous metastasis. Further information is listed in the online supplementary files.

Immunohistochemistry and image analysis

From each patient enrolled in the study, 2 mm thick tissue slides from formalin-processed and paraffin-embedded tumour sections were retrieved from the archive of the pathology department and processed for immunohistochemistry as previously done.28 Details on immunostaining and image analysis are available in the online supplementary files.

Macrophage cytotoxicity assays

To assess macrophage cytotoxicity, coculture experiments of human macrophages and pancreatic cancer cells were performed in the presence or absence of GEM. Details are listed in the online supplementary files.

Results

Macrophages accumulate at the tumour–stroma interface in human PDAC and display an immunosuppressive phenotype

To evaluate the contribution of TAMs to the immune microenvironment of human PDAC, we systematically and quantitatively assessed by immunohistochemistry the percentage of IRA of CD68+ macrophages (CD68-TAM %IRA) at the tumour–stroma interface (figure 1A and see online supplementary table S1), in 110 tissue specimens from patients with PDAC who underwent surgical resection at our institution. Whole tissue slide quantification evidenced that the density of CD68-TAMs was heterogeneous among patients with PDAC, with some cases showing a high infiltration (figure 1B) and others low infiltration (figure 1C) at the tumour–stroma interface. Figure 1D reports distribution of TAM density across 110 patients with PDAC. Macrophages may have a spectrum of different polarisation states, which dictate their function in the tumour context.10 Accordingly, macrophages at the tumour–stroma interface of pancreatic cancer (figure 1E) predominantly exhibited an immunosuppressive phenotype, as shown by high expression of receptors CD163 (figure 1F), CD206 (figure 1G) and interleukin-10 (IL-10), a prototypical immunosuppressive cytokine (figure 1H). Notably, while expression of CD206 identified most of the CD68+ macrophages present at the tumour–stroma interface (mean%±SD of positive cells 62.4±16.9, n=12, p=0.003), IL-12, which is mainly associated to an M1-like polarisation state in vitro, was expressed only by a minority of macrophages (mean%±SD of positive cells 8.7±10.1, n=6, p=0.0006) (figure 1I), confirming the consolidated evidence that TAMs in human PDAC are prevalently immunosuppressive.11 ,29 CD68-TAMs (figure 1J) frequently localised in the proximity of tumour glands expressing the chemokine CCL17 (figure 1K), a Th2-type chemokine, and low or any amounts of CXCL10 (figure 1L), an exemplar interferon (IFN)-γ-induced chemokine, suggesting that CD68-TAMs in human PDAC localise in regions dominated by an immunosuppressive milieu. Notably, CD68-TAMs (figure 1M) were also a source of CCL17 themselves (figure 1N), which is associated to a Th2-type response.

Figure 1

CD68-TAMs at the tumour–stroma interface display an immunosuppressive phenotype. (A–D) Histological sections of human PDAC, stained for CD68-TAMs. The image exemplifies the methodology used for the assessment of CD68-TAMs at the tumour–stroma interface (dot line); cancer tissue represents approximately 50% of the microscopic field (A). The density of CD68-TAM cells differs greatly, with some highly infiltrated cases (B) and some scarcely infiltrated cases (C). Distribution of TAM density across 110 patients with PDAC (D). (E–I) TAMs display an immunosuppressive phenotype in human PDAC. Immunohistochemical stainings for CD68 (E), CD163 (F) and CD206 (G) on the same PDAC case and double staining with CD68 and IL-10 (H) antibodies show that TAMs in human PDAC predominantly exhibit an immunoregulatory phenotype. Arrowheads indicate CD68-TAMs expressing IL-10. The majority of macrophages express CD206 (62.4±16.9%, n=12, p=0.003), while IL-12 is expressed only by a minority of macrophages (8.7±10.1%, n=6, p=0.0006). (J–N) CD68-TAMs in human PDAC localise in regions dominated by an immunosuppressive milieu. TAMs (J) surrounding a tumour gland expressing the chemokine CCL17 (K), but not CXCL10 (L). CD68-TAMs (M) were also a source of CCL17 (N), associated to a Th2-type response. Arrowheads indicate CD68-TAMs expressing CCL17. Sections in E–G and in J–L and M and N are consecutive sections. Bars: (A) 500 μm, (B and C) 200 μm, (J–L) 100 μm, (H, M and N) 10 μm. **p=0.003; ***p=0.0006, by Student's t test. IL, interleukin; %IRA, percentage immunoreactive area; PDAC, pancreatic ductal adenocarcinoma; TAMs, tumour-associated macrophages.

Density of CD68-TAMs at the tumour–stroma interface predicts response to CTX

We next analysed the distribution at the tumour–stroma interface and the prognostic value of CD68-TAMs in the microenvironment of human PDAC. In online supplementary table S2, univariate and multivariate Cox regression analyses are shown including demographics, clinical and histopathological features and the density of CD68-TAMs. We recorded 72 (65.7%) outcome events of DFS among 110 patients with PDAC. At multivariate Cox analysis, TNM staging and tumour grade were significantly associated with higher risk of disease recurrence (see online supplementary table S2). Notably, by systematically analysing all the variables that may interact in determining patient prognosis, including the densities of immune cell populations, we found a significant interaction between the density of CD68-TAMs and adjuvant CTX (p=0.004) (see online supplementary table S2). To assess whether CTX interacted also with the adaptive immune infiltrate, we also evaluated the density of CD8-TILs (CD8-TIL %IRA) (see online supplementary figure S1A and tables S1 and S2). In contrast to CD68-TAMs, densities of CD8-TILs did not show interaction with CTX in predicting the risk of patient's outcome (p=0.84) (see online supplementary table S2). Pearson's linear regression analysis revealed no association between distributions of CD68-TAM %IRA and CD8-TILs %IRA (r=0.07, p=0.49) (see online supplementary figure S1B), suggesting that these immune populations belong to distinct immune networks in the tumour microenvironment.

To dissect the interaction of CTX and CD68-TAMs, we designed a subgroup analysis and investigated the prognostic value of CD68-TAMs in patients with PDAC according to the administration of postsurgical CTX (table 1). Surprisingly, while among patients who did not receive any adjuvant CTX (n=36), cumulative increasing values of CD68-TAMs were associated with a higher risk of disease progression (HR=1.57; 95% CI (1.04 to 2.37); p=0.03), among patients with PDAC receiving postsurgical adjuvant CTX (n=74), increasing values of CD68-TAM densities were associated with a better prognosis (HR=0.74; 95% CI (0.58 to 0.95); p=0.02)) (see online supplementary table S1 and figure S2).

Table 1

CD68-TAMs as predictors of postsurgical disease-specific recurrences in 110 patients with pancreatic ductal adenocarcinoma, according to CTX treatment

This result indicates that the prognostic value of TAMs in predicting tumour recurrence is critically influenced by postsurgical CTX. This evidence was also confirmed by analysing whether response to adjuvant CTX varied according to the density of TAMs at the tumour–stroma interface (table 2).

Table 2

Adjuvant CTX treatment as predictor of postsurgical disease-specific recurrences in 110 patients with PDAC, according to the extent of CD68-TAMs at the tumour–stroma interface

Interestingly, among patients with high densities of CD68-TAMs (fourth quartile, n=28) CTX associated with lower risk of disease recurrence (n=28; HR: 0.38; 95% CI (0.13 to 1.00); p=0.05). Differently, CTX associated with worst prognosis in patients with PDAC with low CD68-TAMs (first to third quartile, n=82) (n=82; HR: 2.19; 95% CI (1.15 to 4.16); p=0.02). This suggests that the presence of CD68-TAMs at the tumour–stroma interface highly influences responsiveness of patients with PDAC to CTX, and could therefore be required for CTX efficacy. Therefore, CD68-TAM assessment could be exploited to identify a subgroup of patients with PDAC highly responsive to CTX and who could receive prognostic advantage from this therapeutic adjuvant treatment. In light of the immunosuppressive phenotype of macrophages we described in PDAC, we tested in our human cohort whether the amount of TAMs at the tumour–stroma interface modulates the prognostic abilities of CD8-TILs after CTX. We found that among patients treated with CTX, the density of CD8-TILs was not a prognostic variable in both patients with low (<median) and high (>median) CD68-TAMs (p=0.94, p=0.46) (see online supplementary figure S1C). This result suggests that TAMs are independent of TILs in predicting prognosis and thus their antitumour activities seems to be mediated by T-cell independent pathways.

TAMs are known to greatly influence the occurrence of distant metastases.6 ,30–32 The interaction with postsurgical CTX in patients with PDAC suggests that they might have a role in the suppression of micrometastases, occurring after surgical resection. To investigate this hypothesis, we evaluated the association of TAM density to distant metachronous metastases. Significantly, among patients with PDAC who did not receive CTX (n=36), cumulative increasing densities of CD68-TAMs were significantly associated with higher risk of metachronous metastasis (HR=1.77; 95% CI (1.09 to 2.87); p=0.02), while among those who received CTX (n=74) the densities of CD68-TAMs were not associated with prognosis (HR=0.91; 95% CI (0.69 to 1.22); p=0.54) (table 3, see online supplementary figure S2).

Table 3

CD68-TAMs as predictors of postsurgical distant organ metachronous metastasis in 110 patients with PDAC, according to adjuvant chemotherapy treatment

This result suggests that CD68-TAMs could promote distant metastasis occurrence and that CTX could restrain their prometastatic activity, as evidenced also by Kaplan–Meier prognostic curves (figure 2A, B, see online supplementary figure S2). On the contrary, when we investigated patient's DSS, we found that TAMs were significantly associated with better survival only among patients who received adjuvant CTX (p=0.05) (figure 2D), while among those untreated, TAM had a tendency to associate to worst prognosis, although not significant (p=0.12) (figure 2C). These results highlight a synergism between CTX and TAMs that differs according to the prognostic clinical endpoint and this could reflect the different biological roles of TAMs in local versus metastatic distant tumour progression.

Figure 2

Clinical relevance of CD68-TAMs in human pancreatic adenocarcinoma. (A and B) Kaplan–Meier curves showing metachronous metastasis, according to CD68-TAM density and CTX. A high density (≥median) of CD68-TAMs associated to an increased risk of distant tumour metastasis compared with a lower density only in patients who did not receive adjuvant CTX (p=0.02, n=36) (A), but not in those CTX treated (p=0.68, n=74) (B). (C and D) Kaplan–Meier curves showing DSS%, according to CD68-TAM density and CTX. A high density (≥median) of CD68-TAMs was significantly associated with better survival only among patients who received adjuvant CTX (p=0.05) (C), while there was a tendency to associate to worst prognosis, but not significant in untreated patients (p=0.12) (D). CTX, chemotherapy; DSS, disease-specific survival; TAMs, tumour-associated macrophages.

GEM synergises with TAM-mediated tumour cytotoxicity

To provide insight into the interaction between TAMs and CTX, we performed in vitro experiments of tumour cytotoxicity in the presence of differently polarised macrophages and GEM. At an early time point (24 h), treatment of the PDAC cell line (MiaPaCa-2) with 500 mM GEM resulted in a slightly enhanced but not significant tumour cell death (figure 3A, left); however at later time points (48–72 h), the treatment was toxic on tumour cells, confirming that the latter were susceptible to the drug (not shown). M1 macrophages polarised with INF-γ and lipopolysaccharide (LPS) alone did not induce significant increase in PDAC cell cytotoxicity (figure 3A, middle), consistent with the ability of PDAC cells to dampen macrophage antitumour function. By contrast, when GEM was added to the coculture, there was a significant increase in tumour cell death, suggesting that M1 macrophages and CTX have a synergic cytotoxic effect (figure 3A, middle and 3B, left). In contrast, there was a significant decrease in tumour cell death when PDAC tumour cells were cocultured with M2-like macrophages (polarised with IL-4) compared with PDAC cells alone, suggesting that M2-like macrophages exhibit a protective role towards PDAC tumour cells in vitro (p≤0.001) (figure 3A, right). However, surprisingly, GEM treatment was able to significantly revert the protective effect of M2-like macrophages (figure 3A, right and 3B, right panel), reinstating its original cytotoxic effect (p≤0.01). A dose–response curve with increasing concentrations of GEM showed the specificity of the synergic effect and confirmed that this effect was already induced at low drug concentrations (figure 3C), while at the highest concentration it was not further enhanced. Collectively, these findings reveal that GEM modulates the interaction of macrophages with PDAC cells in vitro and reflect the interaction of GEM and macrophages observed in our human study.

Figure 3

CTX synergises with macrophage antitumour activity. (A–C) CTX enhances macrophage cytotoxicity in vitro. FACS plots of PDAC cells (MiaPaCa2) cocultured with M1 macrophages or M2 macrophages with or without CTX. Tumour cell death was evaluated by Annexin V/7-AAD staining. 24 h GEM treatment on PDAC cells alone slightly increased tumour cell death (bottom left). M1 macrophages did not induce detectable tumour cell death (top middle), but treatment with GEM induced an increase in tumour cell lysis (bottom middle). M2-like macrophages showed a protective effect towards tumour cells, decreasing cell death (top right), but when treated with GEM, this effect was hampered (bottom right) (A). GEM treatment significantly increased M1-mediated tumour cell death (p≤0.05) (left) and reverted the protective effect of M2 macrophages (p≤0.001) (right) in two PDAC cell lines (MiaPaCa2 (top) and PT45 (bottom)) (B). Cytotoxicity experiments performed with macrophages in the presence of different concentrations of GEM (C). Increasing concentrations of GEM enhanced macrophage-mediated tumour cell death. *p≤0.05; **p≤0.005; ***p≤0.001 by Student's t test. CTX, chemotherapy; FACS, fluorescence-activated cell sorting; GEM, gemcitabine; PDAC, pancreatic ductal adenocarcinoma.

We then attempted to dissect the synergism of GEM and macrophages. GEM treatment induced neither selective death of immunosuppressive macrophages (figure 4A) nor tumour cell release of molecules able to activate or reprogram macrophages (figure 4B, C). In contrast, tumour cell death was significantly dose-dependently increased when PDAC tumour cells were cocultured with macrophages pretreated with GEM (MfpreGEM), compared with those cocultured with control macrophages (figure 4B, D), thus suggesting that GEM exerted its modulatory effect by acting directly on the macrophages and thwarting the protective function of macrophages towards PDAC cells. Notably, depending on the cell line tested, the interaction between GEM and macrophages was either abrogated (figure 4D, left panel) or maximum (figure 4D, right panel) at the highest GEM dose. We then investigated the mechanism whereby GEM enhanced macrophage cytotoxicity. Inhibition of macrophage death signals, including tumour necrosis factor (TNF)-α and TRAIL, did not affect the synergism between macrophages and GEM in killing PDAC tumour cells (see online supplementary figure S3). In contrast, GEM treatment induced an early and dramatic increase in cellular reactive oxygen species (ROS) in macrophages (figure 4E), thus suggesting that release of cytotoxic ROS could be responsible for the induced macrophage antitumour activity.

Figure 4

Mechanism of macrophage and CTX synergism. (A) GEM did not induce selective apoptosis of M2 macrophages compared with M1 macrophages in 24 h. (B–D) In vitro experiments of tumour cytotoxicity aimed at clarifying the mechanism underlying CTX-mediated modulation of macrophage function. FACS plots of PDAC cells cocultured with macrophages. Macrophages were exposed to supernatant of GEM-treated tumour cells (MfSupGEM) or to GEM (MfpreGEM) and then added to PDAC cells, without GEM. Culture of PDAC cells with MfSupGEM did not detectably increase macrophage cytotoxicity, while coculture with macrophages pretreated with GEM (MfpreGEM) resulted in the reversal of macrophage function, with an increase in tumour cell death. FACS plots depict one representative of three experiments (B). Histograms representative of three independent experiments with two different PDAC cell lines: macrophages significantly decrease PDAC cell death but this effect is not mediated by tumour-derived factors (C). However, this effect is reverted by pretreatment of macrophages with GEM and the effect is dose dependent (D). (E) GEM treatment induces release of ROS in macrophages. Histograms shows fold increase of mean fluorescence intensity of treated macrophages compared with untreated, at two different time points. Mean of three independent experiments is shown. *p≤0.05. **p≤0.005 by Student's t test. Bars represent SE. CTX, chemotherapy; FACS, fluorescence-activated cell sorting; GEM, gemcitabine; PDAC, pancreatic ductal adenocarcinoma; ROS, reactive oxygen species.

GEM directly modifies macrophage polarisation in vitro and in the tumour microenvironment of human PDAC

To reveal molecular actors regulating the CTX-mediated modulation of macrophage function, we evaluated the effect of GEM on the mRNA expression profile of macrophages. GEM-treated macrophages significantly increased expression of several immune-stimulating cytokines, such as IL-12, IL-27, IFNG, as well as cytokines typically involved in inflammatory responses, including prostaglandin-endoperoxide synthase, Fas ligand and TNF-α (figure 5A). In accordance with the evidence that GEM induces ROS production in macrophages (figure 4E), there was also an increase in molecules regulating microbial responses such as inducible nitric oxide synthase and nicotinamide adenine dinucleotide phosphate, although not significant. Remarkably, the increase in proinflammatory mediators was paralleled by a significant decrease in IL-10, also confirmed by decreased protein production (see online supplementary figure S5), suggesting an overall switch in activation status.

Figure 5

GEM directly modifies macrophage polarisation in vitro and in the tumour microenvironment of human PDAC. (A–C) GEM switches on an antitumour expression programme in macrophages in vitro. GEM-treated macrophages show significantly increased expression of typical M1 mediators compared with untreated macrophages: immune activating cytokines, such as IL-12, IL-27, IFNG, IL-23, and inflammatory cytokines, such as PTGS2, FASLG and TNF. However, expression of IL-10 is significantly decreased. Typical tumour-promoting M2 mediators expression is significantly reduced in GEM-treated macrophages: growth factors, such as PDGFA, PDGFB, PGF and matrix proteins, such as OP, and FN (p≤0.05). Data represent mean fold change±SEM in expression of GEM-treated macrophages compared with untreated. *p≤0.05; **p≤0.005 by Student's t test (A). FACS analysis of macrophages exposed to GEM for 24 h (4 h for CCR7). GEM treatment significantly upregulated expression of the M1-markers HLA-DR, CD40-L and the chemokine receptor CCR7 and downregulated the M2-markers CD163 and CD206 (B). Histograms representative of three independent experiments. Data represent mean fold change±SEM in GEM-treated macrophages compared with the untreated ones. The dotted line indicates the unitary value. *p≤0.05; **p≤0.001; ***p<0.001 by Student's t test (C). (D–G) Neoadjuvant CTX modifies macrophage polarisation in human PDAC. CD68 and CD206 immunohistochemical stainings performed on consecutive sections, in human PDAC specimens from patients receiving GEM as neoadjuvant therapy (D). Significantly less macrophages were CD206-positive in PDAC sections from patients receiving neoadjuvant CTX. p=0.0054 by Student's t test (E). CD68 (green) and IL-10 (brown) immunohistochemical staining in human PDAC specimens from patients receiving or not GEM as neoadjuvant therapy. Arrowheads indicate CD68/IL-10+ macrophages (left) and CD68/IL-10 macrophages (right), respectively (F). Among CD68+ macrophages, significantly more macrophages are IL-10+ in PDAC sections from patients not receiving neoadjuvant CTX, since the ratio of CD68/IL-10+ macrophages is decreased in treated patients. *p<0.05 by χ2 test (E). Bar: 200 μm (D), 50 μm (F). FACS, fluorescence-activated cell sorting; FASLG, Fas ligand; FN, fibronectin; GEM, gemcitabine; IFNG, interferon gamma; IL, interleukin; OP, osteopontin; PDAC, pancreatic ductal adenocarcinoma; PDGFA and PDGFB, platelet growth factor A and B; PGF, placental growth factor; PTGS, prostaglandin-endoperoxide synthase; TNF, tumour necrosis factor.

In accordance with these results, the gene expression profile of GEM-treated macrophages also evidenced a significantly decreased expression of some growth factors, such as platelet growth factor A and B, placental growth factor and matrix proteins, such as osteopontin and fibronectin (figure 5A), which are mediators typically upregulated in protumour M2-like TAM. Overall, these findings strongly suggest that GEM treatment acts directly on macrophages by inducing a downregulation of trophic factors, concurrently inducing an increased expression of inflammatory mediators able to revert this protective function and likely to explain the protumour activity of CTX-untreated macrophages. In accordance with gene expression data, GEM treatment significantly modified surface expression of polarising markers, upregulating expression of the M1-markers HLA-DR, CD40 and the chemokine receptor CCR7 and downregulating the M2-markers CD163 and CD206 (figure 5B, C).

To further confirm the effect of CTX on macrophage repolarisation, we took advantage of a cohort of consecutive patients who were administered neoadjuvant CTX and evaluated the phenotype of macrophages at the tumour front, by immunohistochemistry on tumour tissue slides. The density of CD68-TAMs and CD8-TILs at the tumour–stroma interface was not significantly modified after neoadjuvant CTX, although a slight tendency to decreased density of CD68-TAMs was observed (p=0.11, see online supplementary table S1). However, despite a comparable density, the phenotype of macrophages at the tumour–stroma interface was affected by CTX administration: while in untreated patients a considerable amount of CD68-TAMs expressed the M2-marker CD206 (figure 5D, upper panel), significantly less macrophages were CD206-positive in PDAC sections from neoadjuvant-treated patients (figure 5D, lower panel and 5E). Similarly, while the majority of macrophages in human PDAC was IL-10+, in patients treated with CTX the presence of IL-10+ macrophages decreased significantly (p<0.05) (figure 5F, G). Therefore, neoadjuvant CTX, rather than affecting the abundance of TAMs, was efficient in decreasing the amount of CD206+ and IL-10+ macrophages (figure 5D–G), thus suggesting the ability of CTX to modulate the function of TAMs in the tumour microenvironment.

Discussion

Immunohistological analyses of the immune microenvironment of solid tumours have recently been gaining consideration as a source of both prognostic biomarkers33 and predictive biomarkers of novel therapeutic targets.34 Among immune populations, TILs have emerged as the most robust prognostic tool in the identification of postsurgical tumour recurrence, suggesting that they might be valuable biomarkers in refining therapeutic treatments.35–37 However, whether the quantification of immune cells in the tumour microenvironment might be effective in identifying patient response in the context of postsurgical CTX has not yet been elucidated.24

Macrophages hold a prime position in the tumour microenvironment of solid tumours in the majority of studies associated to unfavourable prognosis,5 with some exceptions.38 ,39 Recent evidence suggests that macrophages are also important determinants of the efficacy of anticancer therapies,3 ,4 ,12 opening to the possibility of their targeting as a potentially effective complementary strategy.4 Here we found that the extent of TAMs, quantitatively and systematically assessed in human PDAC microenvironment, modifies the prognostic responsiveness to postsurgical conventional CTX in patients undergoing successful resective surgery. While we show that TILs are prognostically irrelevant, the density of TAMs behaves as a predictive biomarker of response to postsurgical CTX. Accordingly, adjuvant CTX was most efficient in exerting antitumour activity and had a better response rate in tumours with high levels of TAMs. Thus, the density of TAMs in the tumour microenvironment identifies patients with PDAC likely to receive clinical benefit from adjuvant CTX. We acknowledge that the impact of infiltrating immune cells on tumour progression was obtained by their measurement within tumour tissues that had been radically resected at the time of surgery, as previously done in colorectal cancer studies.28 Although this could appear counterintuitive because tumour-infiltrating leucocytes are supposedly eliminated during surgery, it is conceivable to speculate that the ability of PDAC cells to recruit immune cells to their primary site before surgery is likely maintained also at the metastatic site and ultimately might determine the ability of the immune system to contain or enhance micrometastatic growth.

TAMs represented a predominant immune population in the tumour microenvironment of human PDAC, in line with previous observations in preclinical models.1 ,29 ,40–42 Characterisation of human PDAC-infiltrating macrophages revealed a prevalent immunoregulatory phenotype, identified by the expression of phagocytic receptors commonly associated to a remodelling activity (CD163, CD206), IL-10 and of CCL17, a chemokine attractant of regulatory immunosuppressive cells. Plasticity is a hallmark of macrophages, as they can acquire both antitumour and protumour functions. This accounts for their ability to adapt to environmental signals,9 ,10 which, in the case of PDAC microenvironment, are known to be strongly immunosuppressive.12 ,40 Notably, TAMs also distributed in tumour areas enriched in CCL17 and devoid of CXCL10, a prototypical IFN-γ-induced chemokine involved in the recruitment of cytotoxic T and NK cells,11 ,42 strongly supporting the hypothesis that TAM takes part in immunosuppressive networks in the tumour microenvironment of PDAC.

In our analysis, the interaction between CTX and TAMs seemed to have different prognostic effects that vary according to the prognostic clinical endpoint and might reflect the different biological roles of TAMs on local versus metastatic distant tumour progression. On the one hand, when considering distant metastases, CTX was capable of restraining TAM prometastatic activity; on the other hand, it seemed to arouse macrophage antitumour function. Essentially, while in patients with PDAC with a natural history of disease (ie, only in those not receiving CTX) TAMs had a prometastatic significance, CTX restrained their protumour prognostic relevance in mediating distant metastasis occurrence. This data was further confirmed by the prevalently immunosuppressive profile of TAMs at the tumour–stroma interface in human PDAC specimens collected at the time of surgery preceding CTX and allows to hypothesise that CTX in human PDAC might provide a prognostic advantage by inhibiting TAM prometastatic effects. In contrast, when considering overall tumour recurrences, CTX unleashed the antitumour ability of macrophages in human patients. Accordingly, our in vitro data evidenced that GEM treatment impacted both on M2 polarised macrophages by inhibiting their protumour function and synergised with M1-macrophage cytotoxicity. Altogether, both clinical and in vitro data suggest that TAMs interact with CTX in mediating their effect on tumour growth. The variety of different microenvironment signals present in diverse tumour milieus can modify the behaviour of macrophages, and previous evidence has shown that the phenotype of TAMs may differ in the primary tumour compared with that in metastatic tissues.32 Based on our data, it is tempting to hypothesise that CTX and TAMs might synergise in removing residual local disease after surgery (tumour local remnants) while concurrently inhibiting the ability of TAM to support distant metastasis occurrence.

It has been known for a long time that various forms of conventional CTX agents might arouse TAM antitumour activities,13 ,14 and TAM regulation of tumour responses to anticancer therapies has recently been rediscovered.3 ,4 Distinct mechanisms regulate the bidirectional interaction of CTX and macrophages that highly depend on the antitumour agent and the tumour type. Macrophages infiltrating tumour tissues exposed to chemotherapeutic agents might be engaged in a tissue repair response, which may sustain tumour growth and limit CTX efficacy.43 Some antitumour compounds have been shown to directly act on macrophages inducing the functional skewing of TAMs into an M2-like direction.44 ,45 Conversely, our results strongly pointed to the hypothesis that, overall, CTX modulates the polarisation of TAMs to an antitumour phenotype, providing further evidence that TAMs might mediate an antitumour activity if properly educated by therapeutic treatments. In fact, GEM greatly activated macrophages in vitro, inducing them to directly exert tumour cell death, by reinforcing production of proinflammatory mediators and immune-activating molecules and concurrently downregulating M2 mediators, such as growth factors and matrix proteins and scavenger receptors. Our data suggest that in PDAC the modulation of macrophage function could be due to a direct action of the cytostatic drug on the phenotype and biological behaviour of macrophages, rather than an indirect mechanism of GEM treatment mediated by tumour cells.46 In this regard, previous evidence concerning myeloid modulation of CTX responsiveness in PDAC partly contrast with our data,47 ,48 and suggest that TAMs follow different mechanisms of modulation by CTX compared with myeloid-derived suppressor cells. In our model, TAMs were detrimental to PDAC progression, as previously shown,49 but GEM treatment inhibited their protumoural effect. In a novel view, we propose that this effect could be responsible for the prognostic advantage of CTX retained only in patients with PDAC with high density of TAMs. Intriguingly, a series of evidence points to the possibility that re-education of TAMs could mediate their ability to kill tumour cells.12 ,44 ,49 ,50

Our findings add clinical evidence to support the hypothesis that, in pancreatic cancer, in spite of its very aggressive clinical behaviour and of the scarce therapeutic options, the tumour microenvironment could be a critical determinant of responsiveness to CTX. Moreover, our studies provide functional validation and molecular bases supporting the idea that targeting the innate immune system may be beneficial to tumour prognosis and in some cases be sufficient to ignite an effective antitumour action. Based on our results, only tumours with high infiltration of TAMs would obtain prognostic advantage in response to CTX, while this effect is lost in patients with scarcely infiltrated tumours. This hypothesis should be validated in prospective randomised controlled designed clinical trials addressing whether the prognostic effectiveness of postsurgical CTX might vary according to the amount of TAMs at the tumour–stroma interface. The hypothesis that CTX treatment might be detrimental to patients with PDAC with low amounts of TAMs could biologically explain the low therapeutic window of this not well tolerated treatment in patients with already severely compromised general conditions such as cachexia and immunodepression. The introduction of TAM immunohistological quantification in clinical practice as a relevant feature in the decision-making process regarding allocation to CTX could spare patients of unnecessary toxic treatments and provide a step forward in personalised medicine, which is required to maximise CTX efficacy in individual patients.

References

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Footnotes

  • GDC and NC contributed equally.

  • Contributors GDC, PA, AM and FM: conception and design of the study. GDC, NC, GFC and FM: acquisition, analysis and interpretation of data. FG, RM and JT: development of methodology. FG, CR, GC and AZ: surgical specimen collection, clinical database handling and statistical analysis. GDC, NC and FM: writing of the manuscript. AZ, PA and AM: critical review of the manuscript. FM: study supervision. All the authors read and approved the final version of the manuscript.

  • Funding This work was supported by the Humanitas Clinical and Research Center (grant Translation in Medical Oncology 2013 to FM and AZ), the Italian Ministry of University and Research, (FIRB grant RBAP11H2R9 to AM), Associazione Italiana Ricerca sul Cancro (AIRC 5×1000 IG-12182 to PA), Fondazione Italiana Ricerca sul Cancro (FIRC fellowship 15041 to GDC). The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.

  • Competing interests None declared.

  • Ethics approval Humanitas Clinical and Research Ethic Committee.

  • Provenance and peer review Not commissioned; externally peer reviewed.