Chemokines modulate the tumour microenvironment in pituitary neuroendocrine tumours

Fresh human PitNET tissues from 24 patients (8 somatotropinomas and 16 clinically non-functioning PitNETs (NF-PitNETs), including 13 gonadotropinomas, 1 silent corticotropinoma and 2 null cell PitNETs) were obtained at the time of transsphenoidal surgery (Table 1), and a fragment was processed for primary culture. This study was approved by the local Ethics Committee (MREC No. 06/Q0104/133) and patients gave written informed consent. Normal pituitary (NP) autopsy samples collected 4–24 h post-mortem from subjects (age 18-81yr) with no known endocrine, immune or malignant disease (n = 5).

Immunostains were performed on 4 μm paraffin-embedded tissue sections using Ventana Discovery DAB Map System (Ventana, Illkirch, France). Slides were deparaffinised and processed for antigen retrieval for 30 min with cell conditioning solution CC1 (Ventana), which is a Tris base buffer (pH~ 9). After blocking with Blocker D solution (Ventana), slides were incubated with primary antibody for 60 min (Additional file 4: Table S1) and then with the universal secondary antibody (Ventana) for 20 min. Slides were counterstained with haematoxylin. A negative control, where primary antibody was omitted, was included per experiment. Immunohistochemical studies assessed macrophages using CD68, CD163 and HLA-DR [42, 44, 71, 81], lymphocytes using CD8 for cytotoxic T cells, CD4 for T helper cells, FOXP3 for T regulatory cells and CD20 for B cells. Neutrophils were assessed with neutrophil elastase and endothelial cells with CD31 as well as by their location and morphology. Slides were scanned and analysed with Pannoramic Scanner and Viewer Software (3DHISTECH, Budapest, Hungary). Full stained slides were inspected at low magnification (5x and 10x magnification) in order to identify “hot spots” areas, i.e. areas with high abundance of immunopositive cells within the tumoural tissue in PitNET slides and within the normal pituitary tissue in NP slides. All selected areas were first verified on the respective haematoxylin and eosin slides, to ensure that cell counting was correctly performed in tumoural or normal pituitary tissue respectively. We excluded intra-vascular and non-adenomatous immunopositive cells. Immunopositive cells were counted in 5 different “hot spots” high power field (HPF) using the ImageJ software (National Institutes of Health, USA). Counterstained nuclei identifying tumour cells were also counted, and the data were expressed as percentage of immunopositive immune cells relatively to the total number of tumour cells per HPF. Vessels (stained for CD31) were manually counted in 3–5 different “hot spot” fields (20x magnification) allowing the estimation of microvessel density (number of vessels per HPF), and the vessels contour was manually traced using the ImageJ software to obtain an estimation of the total microvessel area (μm²), as described previously [78].

Blood samples from each patient were routinely taken within 7–10 days before the pituitary surgery as part of the pre-operative work-up protocol. All the serum full blood counts were performed in a certified National Health Service laboratory in a standardised manner on a Sysmax automated counter.

IL-8 and CCL2 and chemokine receptors CXCR2 and CCR5 were detected using the RNAscope 2.5 HD Duplex Chromogenic Assay (Advanced Cell Diagnostics, ACD, USA), according to the manufacturer’s protocols. Briefly, 4 μm paraffin-embedded PitNET tissue sections were baked at 60 °C for 90 min, deparaffinised, and then boiled with pre-treatment retrieval reagent for 15 min. Protease digestion was performed at 40 °C for 30 min, followed by hybridisation for 2 h at 40 °C with Probe mix according to 1:50 ratio of C2 to C1 probes: mix IL-8 (ACD, cat. no. 310381-C2) and CXCR2 (ACD, cat. no. 468411), and mix CCR5 (ACD, cat. no. 601501-C2) and CCL2 (ACD, cat. no. 423811). Hybridisation signals were amplified and visualised with RNAscope 2.5 HD Duplex Chromogenic Assay reagents. Cell nuclei were counterstained with haematoxylin, and slides were mounted with VectaMount mounting medium (Vector Laboratories, cat. no. H-5000). Probe-DapB (ACD) was used as negative control. Slides were scanned and analysed with Pannoramic Scanner and Viewer Software (3DHISTECH, Budapest, Hungary).

Fresh PitNET tissue was collected in high glucose Dulbecco’s Modified Eagle’s Medium (DMEM, Sigma, Gillingham, UK, cat. no. D6429) supplemented with 10% heat-inactivated foetal bovine serum (FBS, Gibco, Loughborough, UK, cat. no. 16000044) and 0.5% gentamicin (Sigma, cat. no. G1397). Tissue was washed with magnesium and calcium-free Phosphate Buffered Saline (PBS) (Sigma, cat. no. D8537), cut into small pieces and incubated for 45 min at 37 °C in 10 times diluted Trypsin-EDTA 0.05% (1X) Phenol Red (Gibco, cat. no. 25300054) with frequent pipetting allowing effective cell dispersion. Trypsin digestion was stopped by adding complete medium, cells were transferred to a tube and allowed to stand for 10 min for sedimentation of undigested debris. Supernatants containing tumour cells were transferred to a separate tube, centrifuged at 800 g for 5 min, and gently re-suspended in 1 mL complete medium. Viable cells were assessed with Tryptan Blue Solution (Sigma, cat. no. T8154) and 2 × 10⁶ cells were seeded in complete medium in a well coated with Poly-L-lysine (Sigma, cat. no. P4707) from a 6-well plate when viability was > 90%. Cells were incubated overnight at 5% CO₂ at 37 °C. The next day plates were examined under the microscope, complete medium was aspirated, cells were carefully washed 3 times with PBS, and then 1 mL serum-free medium was added. After incubation for 24 h, supernatants were collected to clean tubes and placed immediately on ice to avoid cytokine degradation. Tubes were centrifuged at 10,000 rpm for 10 min at 4 °C to remove debris, and supernatants were collected and stored in − 80 °C for 3–6 months until being assayed in duplicate with the human Millipore MILLIPLEX cytokine 42-plex array.

Rat pituitary somatomammotroph GH3 cell line, obtained from European Collection of Authenticated Cell Cultures, and the murine RAW 264.7 macrophages (kind gift from Dr. Giulia Marelli, Barts Cancer Institute) were incubated at 5% CO₂ at 37 °C and cultured in high glucose DMEM supplemented with 10% FBS and 0.5% gentamycin. Once cells were 70–90% confluent, they were passaged after washes with magnesium- and calcium-free PBS, and mobilisation with Trypsin-EDTA 0.05% (1X) Phenol-Red for GH3 cells or with Accutase® solution (Sigma, cat. no. A6964) for RAW 264.7 macrophages. Once cells were detached (confirmed by light microscopy), cell suspension was put into new flasks or spun (3 min, 1200 g) and re-suspended in medium to be further used in in vitro experiments.

GH3 cell-conditioned medium (CM) was generated by seeding 5 × 10⁶ GH3 cells in T75 culture flasks for 72 h in 10 mL complete medium. Macrophage-CM was generated from 5 × 10⁶ RAW 264.7 macrophages in T75 culture flasks for 24 h in 10 mL complete medium (−PMA_Raw.CM) or stimulated with 5 nM of Phorbol 12-Myristate 13-Acetate (PMA) (Sigma, cat. no. P8139) in 10 mL complete medium (+PMA_Raw.CM).

Cell culture supernatants for cytokine array were generated by seeding 5 × 10⁵ GH3 cells in 12-well culture plates for 24 h in serum-free medium conditions at baseline and after treatment with RAW 264.7 macrophage-CM. Supernatants were carefully transferred to clean tubes, centrifuged at 10,000 rpm for 10 min in 4 °C to remove debris, and then supernatants were collected and stored in − 80 °C for 3–6 months until being assayed.

Cytokine arrays on human PitNET-derived supernatants were performed by Eve Technologies (Calgary, Alberta, Canada), according to their protocol by using Bio-Plex™ 200 system (Bio-Rad Laboratories, Inc., Hercules, CA, USA), and human cytokine/chemokine array with IL-18 (HD42) kit (Millipore, St. Charles, USA). This array measured 42 different cytokines, chemokines and growth factors in the same sample: G-CSF, GM-CSF, IFNα2, IFNγ, IL-1α, IL-1β, IL-1ra, IL-2, IL-3, IL-4, IL-5, IL-6, IL-7, IL-8, IL-9, IL-10, IL-12(p40), IL-12(p70), IL-13, IL-15, IL-17A, IL-18, CXCL1, CXCL10, CCL2, CCL3, CCL4, CCL5, CCL7, CCL11, CCL22, CX3CL1, sCD40L, Flt-3 L, PDGF-AA, PDGF-BB, TGF-α, TNF-α, TNF-β, VEGF-A, EGF and FGF-2.

Millipore MILLIPLEX arrays on supernatants from rat GH3 cells were also performed by Eve Technologies, using a different species-specific kit array: rat cytokine/chemokine array 27-plex (RD27) (Millipore) able to measure 27 different cytokines, chemokines and growth factors in the same sample: G-CSF, GM-CSF, IFNγ, IL-1α, IL-1β, IL-2, IL-4, IL-5, IL-6, IL-10, IL-12(p70), IL-13, IL-17A, IL-18, CXCL1, CXCL2, CXCL10, CCL2, CCL3, CCL5, CCL11, CX3CL1, TNF-α, VEGF, EGF, Leptin and LIX.

Invasion assays were carried out using the BioCoat Matrigel Invasion Chambers with 8 μm pores (24-well insert; BD Biosciences, CA, USA, cat. no. 354480). Invasion chambers were hydrated for 2 h with 500 μl of serum-free medium at 37 °C. After Matrigel rehydration, 750 μl of macrophage-CM or complete medium was added to the lower chamber as chemoattractant and 2.5 × 10⁴ GH3 cells in 500 μl serum-free medium were added to the upper chamber and incubated at 37 °C. After 72 h, invading cells through Matrigel were fixed in 100% methanol and stained with 2% Giemsa blue (Sigma-Aldrich, MO, USA, cat. no. G5637-5G). Total number of invading cells per chamber were counted, and normalised to invading cells towards complete medium. Macrophage chemotaxis were evaluated by Transwell BioCoat Migration insert plates with 8 μm pores (24-well insert; Corning Fisher Scientific, USA, cat. no. 354578) following a similar protocol as described for invasion assay. Migration and invasion studies were run in duplicates and were repeated at least 3 times. Morphological changes were assessed, as previously described [10], by measuring 6 different shape parameters using ImageJ: area (area of selection in calibrated square units, μm²); perimeter (μm); Feret’s diameter (longest distance between any 2 points along selection boundary); roundness (representing shape, 4 × [Area] / π × [Major axis]², with value of 1 for a circle and 0 for very elongated shapes); circularity (representing perimeter smoothness, 4π × [Area][Perimeter]², with value of 1 indicating a perfect circle and value close to 0 indicating elongated shape) and solidity (value of 1 indicating more stiffness and less deformable cells). Per treatment condition, 5 images at 40x were taken and 15 cells were measured per image, thus 75 cells were analysed per experiment (minimum of 3 experiments were done).

GH3 cells (5 × 10⁴) were plated on 15 mm coverslips placed in 12-well plates. After overnight attachment, GH3 cells were treated with -PMA_Raw.CM, +PMA_Raw.CM or complete medium for 24 h. Cells were fixed in 4% paraformaldehyde for 15 min at room temperature, following washes with PBS cells were permeabilised with 0.1% Triton X-100 in PBS for 5 min at 4 °C. Cells were blocked in 1% bovine serum albumin for 30 min at room temperature, and then incubated with primary antibodies (listed in the Additional file 4: Table S1) followed by a 30 min incubation with secondary conjugated antibodies (Alexa Fluor 568-conjugated goat anti-mouse IgG, Alexa Fluor 488-conjugated donkey anti-mouse IgG and Alexa Fluor 488-conjugated donkey anti-rabbit IgG; 1:1000; Molecular Probes, Invitrogen). Coverslips with stained cells were mounted with Fluoroshield with DAPI mounting medium (Sigma, cat. no. F6057). Stained slides were visualised on confocal microscope LSM 880 Zeiss and images taken at 63x magnification. E-cadherin and ZEB1 fluorescent intensities were quantified using the software Carl Zeiss Zen Blue Edition v2.3.

RNA from GH3 cells and RAW 264.7 macrophages was extracted using Qiagen’s RNeasy micro kit (cat. no. 74004) according to manufacturer’s protocol. RNA samples were assessed by NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Rockland, DE, USA). Complementary DNA was synthesised from 1 μg of RNA using High-Capacity cDNA Reverse Transcription Kit (Thermofisher Scientific, cat. no. 4374966) following the manufacturer’s protocol. RT-qPCR reactions were prepared using Brilliant III Ultra-Fast SYBR Green QPCR Master Mix (Agilent Technologies, Palo Alto, CA, USA, cat. no. 600882) and run with Thermal Cycler with MxPro software (Agilent) using a 2-step programme: pre-incubation 3 min at 95 °C, then 40 cycles of 20s at 95 °C and 20s at 60 °C. Cycle threshold (CT) values were analysed with ∆∆CT quantification method. Target gene expression was normalised to GAPDH expression used as internal control. Primers sequence (Sigma-Aldrich) were as follows: CCR5 forward 5′-GTATGTCAGCACCCTGCCAA-3′, reverse 5′-GAGCAGGAAGAGCAGGTCAG-3′; CX3CR1 forward 5′-CCATCTGCTCAGGACCTCAC-3′, reverse 5′-CACCAGACCGAACGTGAAGA-3′; MMP9 forward 5′-CTTGAAGTCTCAGAAGGTGGATC-3′, reverse 5′-CGCCAGAAGTATTTGTCATGG-3′.

Total RNA from a different set of 7 human sporadic PitNET samples (3 somatotropinomas and 4 NF-PitNETs) and 5 NPs was isolated using the Qiagen’s RNeasy micro kit and target labelling and hybridisation were performed using Affymetrix GeneChip 3′ IVT Express Kit (Affymetrix, Santa Clara, CA, USA) according to manufacturer’s instructions (samples described in detail in [10]). Microarray data have been deposited with the National Center for Biotechnology Information Gene Expression Omnibus (www.​ncbi.​nlm.​nih.​gov/​geo, accession number GSE63357). For deconvolution, we used the gene-signature webtool method xCell [2, 56], which inferred different immune and stromal cells from microarray expression data giving a score per cell type. This tool uses the M1 and M2 macrophage nomenclature.

In order to identify the most relevant cytokines derived from human PitNETs, we established primary cultures from 24 PitNETs. All tumours were larger than 1 cm in diameter, 10 had cavernous sinus invasion and 5 had Ki-67 ≥ 3% (Table 1). We assessed 42 different cytokines in fresh tumour culture supernatants (Additional file 5: Table S2). The cytokine array identified IL-8, CCL2, CCL3, CCL4, CXCL10, CCL22, CXCL1 and CX3CL1 as the main PitNET-derived cytokines (Table 2), all chemokines specialised in immune cell recruitment [5]. Ninety percent of PitNETs secreted IL-8, CCL2 and CCL3, while CXCL1 was secreted by 50% of the tumours (Table 2). RNAscope data showed that CCL2 and IL-8 are mainly synthesised by pituitary tumour cells, while these have low expression of chemokine receptors; chemokine receptors were, in turn, strongly expressed in scattered perivascular cells, morphologically distinct from tumour cells, likely corresponding to immune cells (Fig. 1a).

NF-PitNETs secreted higher amounts of cytokines/chemokines than somatotropinomas, especially CCL2 (16x more), IL-8 (25x more) and CCL4 (27x more), except for FGF-2 which was found in higher concentrations in somatotropinoma supernatants (Fig. 1b). NF-PitNETs were more often secreting IL-8, CCL2, CCL4, IL-6 and PDGF-AA than somatotropinomas (Fig. 1c). Secretome differences between NF-PitNETs and somatotropinomas were not explained by pre-operative somatostatin analogue treatment, as there were no secretome differences between pre-treated and non-pre-treated somatotropinomas (Additional file 1: Figure S1a). There were also no secretome differences between sparsely and densely granulated somatotropinomas (Additional file 1: Figure S1b). The PitNET-derived cytokine secretome was not associated per se with cavernous sinus invasion, elevated Ki-67 or presence of hypopituitarism at diagnosis (data not shown).

PitNETs contained more CD68+ macrophages (4.6 ± 0.4 vs 1.2 ± 0.2%, p < 0.001) and CD4+ T cells (1.0 ± 0.1 vs 0.6 ± 0.1%, p = 0.005), but fewer neutrophils (0.7 ± 0.2 vs 1.4 ± 0.1%, p = 0.047) and a trend for fewer CD8+ T cells (1.8 ± 0.2 vs 2.6 ± 0.3%, p = 0.077), with a significant 2-fold decrease in the CD8:CD4 cell ratio compared to NPs (Fig. 2). There were no significant differences in B and FOXP3+ T cells content (Fig. 2).

Macrophages are the most abundant immune cell type in PitNETs, while other immune cells are present in lower amounts (Fig. 2). Macrophages in PitNETs are predominantly CD163+, while HLA-DR+ macrophages predominate in NPs, resulting in a 3-fold increase in the CD163:HLA-DR macrophage ratio in PitNETs (Fig. 3a). These immunohistochemical findings were confirmed on a different set of samples [10] using gene expression analysis with xCell, with PitNETs having a higher score than NPs for macrophages (0.090 ± 0.016 vs 0.031 ± 0.013; p = 0.025) and CD163+ macrophages (0.042 ± 0.007 vs 0; p = 0.001) (Additional file 2: Figure S2). The CD163+ macrophage phenotype in PitNETs may be due, at least in part, to higher concentrations of PitNET-derived M2-polarising cytokines, as IL-4 levels were almost 5x higher than interferon-γ (M1-polarising cytokine) (Fig. 3b).

We observed significant correlations between infiltrating immune cell populations in PitNETs, namely between CD8+ and CD4+ T cells (r = 0.534; p = 0.007), CD8+ and FOXP3+ T cells (r = 0.504; p = 0.012), CD4+ T cells and neutrophils (r = 0.490; p = 0.05).

NF-PitNETs had more neutrophils than somatotropinomas (0.9 ± 0.1 vs 0.1 ± 0.1%, p = 0.002), but there were no differences regarding other immune cells, neither CD163:HLA-DR, CD8:CD4 or CD8:FOXP3 cell ratios (Additional file 6: Table S3). There were no significant differences in infiltrating immune cells among the different NF-PitNET types (Additional file 6: Table S3) or between sparsely and densely granulated somatotropinomas (data not shown).

PitNETs with higher macrophage content were associated to higher levels of IL-8 (p = 0.023), CCL2 (p = 0.216), CCL3 (p = 0.065), CCL4 (p = 0.036) and CXCL1 (p = 0.024) (Fig. 4a), chemokines known to promote macrophage chemotaxis [1, 5, 41, 43]. We observed higher CCL2 (p = 0.036), CCL4 (p = 0.086), CXCL10 (p = 0.134) and VEGF-A (p = 0.025) levels in the supernatants from PitNETs with higher CD8+ T cell contents (Fig. 4b). PitNETs with more neutrophils released higher levels of CCL2 (p = 0.033) and CCL4 (p = 0.044), cytokines known to attract neutrophils [16, 63, 92], and there was a noteworthy trend to release higher levels of chemokines involved in neutrophil chemotaxis (Fig. 4c), namely IL-8 (p = 0.073), CXCL1 (p = 0.097) and CXCL10 (p = 0.098). There were no significant associations between PitNET-derived cytokine secretome and infiltrating CD4+, FOXP3+ and B cells (data not shown).

PitNET-infiltrating immune cells did not correlate with circulating immune cell types, suggesting that immune infiltrates are subject to differential recruitment into the PitNET rather than altered bone marrow production (Additional file 3: Figure S3).

PitNETs with higher proliferative index (Ki-67 ≥ 3%) had lower CD8:CD4 ratio, as well as lower CD8:FOXP3 and CD68:FOXP3 ratios as a result of increased infiltration of FOXP3+ T cells (0.7 ± 0.2 vs 0.3 ± 0.6%; p = 0.013) (Fig. 5a-b). All PitNETs with “deleterious immune infiltrate phenotype”, i.e. higher content of macrophages, T helper lymphocytes, FOXP3+ T regulatory cells and B cells (CD68^hiCD4^hiFOXP3^hiCD20^hi) had a Ki-67 ≥ 3% (Fig. 5c). There were no differences between PitNETs with or without cavernous sinus invasion regarding immune cell content or cell ratios (data not shown). The CD163:HLA-DR macrophage ratio was positively correlated with microvessel density (p = 0.015) and area (p < 0.001) (Fig. 6).

To study the interactions between pituitary tumour cells (GH3 mammosomatotroph tumour cell line) and macrophages (RAW 264.7 macrophage cell line), we established an in vitro model using CM from each of the cell line as a chemoattractant agent for the other. To investigate the role of GH3 cell-derived factors in macrophage chemotaxis, we performed a transwell migration assay, observing a remarkable 36-fold increase in macrophage migration towards GH3-CM in comparison to complete medium or recombinant CX3CL1 (Fig. 7a). CX3CL1 was used as positive control, as this was the chemokine with the highest concentration in GH3 supernatants (Additional file 7: Table S4), and has a recognised chemoattractant effect on RAW 264.7 macrophages [22]. Immune cell chemotaxis depends not only on tissue chemokine gradient, but also on chemokine receptor expression in trafficking cells [5]. GH3-CM increased more than 12x the expression of CX3CR1 (receptor with specific affinity for CX3CL1 and highly expressed in RAW 264.7 macrophages [22]), as well as the expression of CCR5 (p = 0.051) (Fig. 7b). Thus, the GH3-CM macrophage chemoattractant effect can be explained, at least in part, by upregulation of chemokine receptor expression in RAW 264.7 macrophages. After GH3-CM treatment, macrophages showed morphological changes typical of activated macrophages: increase in cell area, perimeter, Feret’s diameter and spindle-shaped morphology [42, 44] and decrease in solidity, roundness and circularity (Fig. 7c), representing cells with enhanced migration phenotype.

RAW 264.7 macrophage-CM induced significant changes in GH3 cells: morphology changes, increased invasion, epithelial-to-mesenchymal transition (EMT) activation (Fig. 8) and alteration in cytokine secretion (Additional file 7: Table S4). Macrophage-CM from untreated (−PMA_Raw.CM) or PMA-treated macrophages (+PMA_Raw.CM) increased GH3 cell area, perimeter and Feret’s diameter, and reduced their solidity, roundness and circularity indicating that GH3 cells acquired an EMT-like phenotype. Macrophage-induced morphology changes in GH3 cells were confirmed with immunocytochemistry for actin: GH3 cells treated with macrophage-CM developed a granular pattern of actin with prominent stress fibres and numerous spikes (Fig. 8a), representing an EMT-like cytoskeletal alteration [50].

GH3 cells showed higher invasion towards +PMA_Raw-CM compared to complete medium (Fig. 8b). As invasion depends on cell ability to secrete proteases to degrade extracellular matrix, we hypothesised that macrophage-CM upregulates matrix metalloproteinases (MMPs) expression in GH3 cells allowing them to invade. MMP-9 is a key protease for type IV collagen degradation (main component of Matrigel [32], and of the pituitary capsule and cavernous sinus wall [14, 31, 36]), as well as MMP-9 overexpression is associated with PitNET invasiveness [36]. Hence, we studied MMP-9 expression in GH3 cells after treatment with macrophage-CM, and we observed a significant MMP-9 upregulation in GH3 cells exposed to macrophage-derived factors (Fig. 8c).

Macrophage-CM induced EMT activation in GH3 cells, decreasing E-cadherin and increasing ZEB1 expression (Fig. 8d), two hallmarks of EMT activation [17]. These findings are in line with the morphological changes and invasion assays results, suggesting that macrophage-derived factors interact with pituitary tumour cells influencing their behaviour and invasiveness.

We also found that PMA-activated macrophage-CM induced cytokine secretion changes in GH3 cells, increasing the release of CX3CL1, CCL3, CXCL1, CXCL10, IL-1β, IL-10, IL-13 and VEGF (Additional file 7: Table S4), peptides that play a role in different tumourigenic mechanisms [5, 7, 40, 41].