15 kDa Granulysin versus GM-CSF for monocytes differentiation: analogies and differences at the transcriptome level

A detailed description of the procedure has been previously described by Finn et al, 2011[19]. Briefly, a cDNA clone of the 15 kDa Granulysin gene was generated from human peripheral blood cells and cloned into a pet28A E. coli expression vector. After being engineered for insect expression and secretion, the vector was transfected in Hi5 insect cells and after 2 days of culture at 21 C the supernatant was filtered using a 0.45 μM filter and applied to a 5 ml HiTrap Heparin HP (GE Health Care, Uppsala, Sweden). Fractions containing the 15 kDa Granulysin were pooled, purified on 1 ml Resource S column (GE Health Care), concentrated and stored at -80°C.

Human peripheral blood from three healthy donors was collected by apheresis in the Department of Transfusion Medicine of the Clinical Center (NIH) using Amicus Separator (Baxter Healthcare Corp., Fenwal Division, Deerfield, IL). The monocyte fraction was immediately separated by elutriation (Elutra^®, Gambro BCT, Lakewood, CO, USA) according to the manufacturer's instructions and the purity achieved was greater than 80%. Fresh monocytes were cultured in 6-well plates (Corning Costar, Corning Incorporated, Corning, NY, USA) at a concentration of 2 ×10 ⁶ cell/ml in 90% RPMI-1640 media, 10% AB heat inactivated plasma, 10 mcg/ml gentamicin in the presence of 15 kDa Granulysin (10 nM) or GM-CSF (Leukine Sagramostin, 10 ng/ml, 56 IU/ml, Genzyme, Cambridge, MA, USA) and harvested at 4, 12, 24 and 48 hours.

At times 0, 4 h, 12 h, 24 h and 48 h 20 ×10 ⁶ cells from each culture condition were used for total RNA extraction using miRNA Easy Kits (Qiagen, Valencia, CA, USA). RNA quantity and quality were assessed by ND-1000 Spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA) and Agilent 2100 Bioanalyser (Agilent Technologies, Waldbronn, Germany), respectively.

Samples and universal Human Reference RNA (Stratagene, Santa Clara, CA, USA) were amplified and labeled using Agilent kit according to the manufacturer's instructions and hybridized on Agilent Chip (Whole Human genome, 4 × 44 k, Agilent Technologies, Santa Clara, CA, USA). The arrays were scanned with Agilent Microarray Scanner and the images were analyzed using Agilent Feature Extraction Software 9.5.1.1. Resulting data were uploaded onto mAdb Gateway http://​madb.​nci.​nih.​gov, retrieved and analyzed with BRB Array Tools http://​linus.​nci.​nih.​gov/​BRB-ArrayTools.​html. The raw data set was filtered according to a standard procedure to exclude spots below a minimum intensity of 20 in both fluorescence channels. If the fluorescence intensity of one channel was higher than 20, but the other was below 20, the fluorescence of the low intensity channel was arbitrarily set to 20. Flagged spots were also excluded from the analysis. A total of 33757 genes passed the filter and were used for the analysis.

A total of 0.5 μg of purified RNA was used to synthesize cDNA using Random Hexamers (Qiagen, Valencia, CA, USA) and Superscript II RT (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's instruction. The expression of CCL2, CCR7, CD209 and PIM1 were tested using specific TaqMan Gene Expression Assays (Applied Biosystems, Carlsbad, CA, USA). HPRT1 was selected as the housekeeping gene, due to the fact that it has been described as a housekeeping gene in monocytes [20] and it showed low variability in our microarray dataset. RT-PCR reactions were setup with TaqMan Universal PCR Master Mix (Applied Biosystems) in 384-well plates in a final reaction volume of 10 μl. PCR was conducted using a 7900 HT Sequence Detection System (Applied Biosystems) and data were analyzed using SDS 2.3 software package (Applied Biosystems).

Class comparison was conducted with BRB Array Tools using a random variance model. Significant genes were defined as p-value < 0.001 and FDR < 0.1. Hierarchical cluster analysis and TreeView software were used for data visualization (Eisen Lab, http://​rana.​lbl.​gov)[21]. Partek Genomic Suite 6.4 (Partek Inc., St. Louis, MO, USA) was used for the Principal Component Analysis. Database for Annotation, Visualization and Integrated Discovery (DAVID) 2008 software[22, 23] was used for Gene Ontology (GO) enrichment analysis. For the analysis of specific pathways related to DC functions all the genes that, according to Biocarta (http://​www.​biocarta.​com), are part of a specific pathway were selected. For each pathway, similarly to Chaussabel et al 2008[24] a less stringent p- value (0.05) and FDR (0.15) filter was applied and the remaining number of genes was arithmetically computed according to their up/down-regulation.

Elutriated monocytes were cultured in presence of GM-CSF (10 ng/ml, 56 IU/ml) or 15 kDa Granulysin (10 nM). At 4, 12, 24 and 48 hours RNA was isolated and used for global gene expression analysis. Principal component analysis of the entire dataset (Figure 1a) revealed that GM-CSF and 15 kDa Granulysin induced a response in monocytes that was similar at early time points (4 hours) but strongly differed at later time points (12, 24 and 48 hours). In particular, principal component (PC) #1 which accounted for 31.5% of the variability of the dataset did not separate the samples cultured with Granulysin from those cultured with GM-CSF, but clearly placed the 4 and 48 hour samples at the extremes with the other samples in between and closer to the 48 hour samples. This indicated that the two agents induced one group of genes at 4 hours and a second set at later times. PC #2, which accounted for 14.8% of the variability, split the GM-CSF and Granulysin samples into two distinct groups at later time points, indicating that the differences between the GM-CSF- and Granulysin-cultured monocytes became more evident at later time points. The third PC (14.1% of the variability) segregated time 0 samples, the untreated monocytes, from the other samples indicating that both agents induced major changes at the transcriptome level when compared to time 0 samples.

In order to stratify changed transcripts associated with treatment and time in an unbiased fashion, the complete gene set was further filtered to include genes with expression levels ≥ 1.75-fold from the median in at least 20% of the samples[25]. 9951 out of 33757 genes were obtained and used for an unsupervised hierarchical cluster analysis which clearly separated early time point samples (T0 and T4) from the late time point samples (Figure 1b). Moreover, within the cluster of the late time point samples, three subclusters emerged: all 12-hour samples, the late 15 KDa Granulysin and late GM-CSF samples. This analysis revealed that GM-CSF and Granulysin induce in monocytes similar changes at the transcriptome level at early time points, but differences become more evident at later time points.

To analyze genes significantly induced by both GM-CSF and Granulysin compared to time 0 monocytes, we selected only the genes that at each time point were commonly induced following treatment by both agents compared to time 0 monocytes (t-test with p-value < 0.001 and FDR < 0.1). A total of 3191, 2416, 1534 and 1738 genes were induced by both GM-CSF and Granulysin at 4, 12, 24 and 48 hours respectively. We then evaluated gene ontology (GO) families that were statistically overrepresented among up- and down-regulated genes at each time point (Figure 2a, b, c, d). Genes related to apoptosis and cell differentiation were significantly enriched at almost all time points. In particular, genes that negatively regulate apoptosis were up-regulated at 4 hours, whereas at later time points those involved with positive induction of apoptosis were mainly down-regulated, suggesting a general down-regulation of apoptosis at each time point. The opposite was observed regarding proliferation related genes, with proliferation related genes mainly up-regulated at 4 hours and the negative regulation of proliferation related genes down-regulated at later time points, pointing to a general induction of cell proliferation. Moreover, at later time points, genes encoding zinc finger proteins were up-regulated and those encoding ribosomal proteins were down-regulated. Interestingly, at 12 hours both GM-CSF and Granulysin induced genes related to the regulation of the adaptive immune response, including CD40, CD80, PVR, PVRL2 and IDO1. This initial activation of the immune system was followed at 24 and 48 hours by the down-regulation of genes involved in leukocyte activation and proliferation, such as IL-8, IL-15, RAB27A, BCL11, FYN and CLCF1.

To identify genes specifically induced by GM-CSF at each time point we selected only the genes that were differentially expressed (p-value < 0.001 and FDR < 0.1) by GM-CSF-treated monocytes versus both time 0 monocytes and cells treated with Granulysin at the same time points. A total of 98, 768, 756 and 467 genes were specifically induced in GM-CSF-treated monocytes at 4, 12, 24 and 48 hours, respectively (Figure 3). Gene functional categories defined by Gene Ontology (GO) families at each time point were analyzed and only those overrepresented in both up- and down-regulated genes were illustrated (Figure 3). Interestingly, GM-CSF-treated monocytes specifically down-regulated immune related genes at each time point, among which were IL-10, CXCL1, CXCL2, CXCR4, CXCR5, and the co-stimulatory molecules CD27, CD28, FYB (ADAP) and TNFSF4 (OX40L). In particular, cytokine production, lymphocyte mediated immunity and humoral immune response GO families were overrepresented among the down-regulated genes at late time points. In contrast, at 48 hours, antigen processing and presentation were specifically up-regulated, including the overexpression of the genes CD1A, CD1B, CD1E, and HLA-DQA1. Moreover, at 12 hours, GM-CSF specifically up-regulated genes involved in myeloid cell differentiation, including IRF4, CSF1 (GCSF), RUNX1, CBFB and PPARG. In addition, at 12 hours, GM-CSF specifically induced the down-regulation of genes related to cell cycle arrest (among which were the cyclin-dependent kinase inhibitors CDKN1B, CDKN2B, CDNK1C), and thus favored cell proliferation. However, at 48 hours, anti-apoptotic genes, such as PIM3, THBS1, HGF and SERPINB2, were mainly down-regulated. Additionally, among the up-regulated genes specifically induced by GM-CSF at early time points were angiogenesis genes, while at late time points lipid biosynthetic process genes were up-regulated and several histone genes were down-regulated.

A total of 152, 498, 429 and 598 genes were specifically induced in Granulysin treated monocytes at 4, 12, 24 and 48 hours, respectively, versus both time 0 monocytes and cells treated with GM-CSF at the same time points (p-value < 0.001 and FDR < 0.1, Figure 4). GO analysis showed immune response genes were up-regulated by Granulysin at each time point. In particular, a coordinated and time-dependent induction of immune related genes could be detected. At 12 hours, innate immunity related genes were up-regulated, but were later down-regulated. At 48 hours humoral and lymphocyte proliferative genes were mostly up-regulated, including CCL2, TNFRSF4, CD38, EBI3, C2, and C3. In addition, cell adhesion genes, including 4 integrins (ITGB8, ITGA9, ITGAV, ITGB5) and the chemokine C-C motif receptor 7 (CCR7), were specifically up-regulated especially at late time points. In addition, chemotaxis related genes were up-regulated at almost all time points, although the involved genes changed markedly between 4 and 48 hours. In fact, CXCL1, CXCL11, CCL20 and IL-6 were up-regulated at 4 hours, whereas CXCL3, CXCL12, CCL2, CCRL2, NRP2 and SEMA3A were induced at 48 hours. Of special note was the induction of cell proliferation genes: after the negative regulation of cell proliferation genes at 4 hours, a positive regulation of cell proliferation genes was most prominent at 48 hours.

Since one of the main in vitro therapeutic uses of GM-CSF is the differentiation, in combination with IL-4, of monocytes into DCs and considering that our results suggest a partially similar response of monocytes when cultured with GM-CSF or 15 kDa Granulysin, we focused on 6 specific Biocarta pathways primarily involved in DC function (Figure 5). To evaluate the level of activation of each pathway we used gene lists with a less stringent cut off (p-value < 0.05 and FDR < 0.15)[24] and calculated the percentage of genes in each pathway induced by each treatment versus T0 monocytes. GM-CSF- and Granulysin-treated monocytes showed a similar number of genes in the Antigen Processing and Presentation, and Monocyte and Surface Molecules Pathways, although the former pathway revealed a constant up-regulation of genes, whereas for the latter pathway a down-regulation at late time points. In contrast, differences were observed regarding the other four pathways, reinforcing the observations described above. GM-CSF-treated monocytes clearly showed a unique down-regulation of the IL-10 Anti-Inflammatory Signaling and the Co-stimulatory Signal during T-cell Activation Pathways, whereas Granulysin-treated monocytes showed an up-regulation of genes in the latter pathway as well as those in the IL-12 and Stat4 Dependent Signaling in Th1 Development and Dendritic Cells in Regulating Th1 and Th2 Development Pathways. Almost the same conclusions could be outlined by focusing on the fold change of the genes in each pathway instead of the percentage of genes (data not shown).

To validate the microarray data, we performed real-time PCR on CCL2, CCR7, PIM1 and CD209 genes. The selection of CCL2 and CCR7 was based on their up-regulation in Granulysin-treated, but not GM-CSF-treated monocytes in array data. PIM1 was selected because it has been described as being induced by GM-CSF[26], and CD209 was selected since it is a marker of DC differentiation. The analysis was performed only on untreated T0 monocytes and hour 4 and 48 GM-CSF- and Granulysin-treated monocytes. Both CCL2 and CCR7 were statistically up-regulated by both agents at 4 hours, however, at 48 hours they were only up-regulated by Granulysin (p-value < 0.01) with a fold change greater than 70 for CCR7 and greater than 800 for CCL2 compared to time 0 monocytes, confirming the finding by microarray analysis (Additional file 1). At 4 hours the expression of both CCL2 and CCR7 was greater in Granulysin-treated monocytes than in monocytes treated with GM-CSF, with a fold change in Granulysin samples more than doubled for CCR7 and more than quadrupled for CCL2 compared to GM-CSF. Although PIM1 was filtered out in our analysis, RT-PCR showed a statistically significant induction of PIM1 at 48 hours (p- value < 0.05) by both GM-CSF and Granulysin, but its expression was greater in GM-CSF treated cells. This difference can be easily ascribed to the high stringency we used for statistical analysis of the microarray data (p-value < 0.001) where we preferred to select and analyze only those genes showing strong induction compared time 0 monocytes. In addition, we observed that CD209 was up-regulated by both agents at 48 hours (p-value < 0.01, with fold changes between 5 and 20 versus time 0 monocytes), which is similar to what we observed in the microarray dataset (both agents increased the expression CD209 genes with p- values < 0.0001).