Differential Transcriptome of Tolerogenic Versus Inflammatory

ORIGINAL ARTICLE Differential transcriptome of tolerogenic versus inﬂammatory dendritic cells points to modulated T1D genetic risk and enriched immune regulation

T Nikolic1,5, NJC Woittiez1,5, A van der Slik1, S Laban1, A Joosten1, C Gysemans2, C Mathieu2, JJ Zwaginga1, B Koeleman3 and BO Roep1,4

Tolerogenic dendritic cells (tolDCs) are assessed as immunomodulatory adjuvants to regulate autoimmunity. The underlying gene expression endorsing their regulatory features remains ill-deﬁned. Using deep mRNA sequencing, we compared transcriptomes of 1,25-dihydroxyvitaminD3/dexametasone-modulated tolDCs with that of non-modulated mature inﬂammatory DCs (mDCs). Differentially expressed genes controlled cellular interactions, metabolic pathways and endorse tolDCs with the capacity to regulate cell activation through nutrient and signal deprivation, collectively gearing tolDCs into tolerogenic immune regulators. Gene expression differences correlated with protein expression, designating low CD86 and high CD52 on the cell surface as superior discriminators between tolDCs and mDCs. Of 37 candidate genes conferring risk to developing type 1 diabetes (T1D), 11 genes differentially expressed in tolDCs and mDCs regulated immune response and antigen-presenting activity. Differential-expressed transcripts of candidate risk loci for T1D suggest a role of these ‘risk genes’ in immune regulation, which targeting may modulate the genetic contribution to autoimmunity.

Genes and Immunity advance online publication, 10 August 2017; doi:10.1038/gene.2017.18

INTRODUCTION capacity to inhibit effector T-cell responses, while inducing 10 A healthy immune system maintains a delicate balance between antigen-specific regulatory T cells. These combined VD3/Dex- surveillance of harmful entities such as pathogens or transformed modulated tolDCs showed a unique protein expression pattern, 10 cells, and maintenance of self tolerance. Dendritic cells contribute compared with tolDCs generated with either modulator alone. to both operating modes of the immune system by employing However, the mechanisms that control tolDC induction and different functional entities; mature inflammatory dendritic cells immunomodulating functions remain largely unknown. Therefore, (mDC) promote adaptive immunity leading to active clearance of we performed next-generation deep mRNA sequencing of VD3/ pathogens or tumors, whereas tolerogenic dendritic cells (tolDCs) Dex-modulated tolDCs and non-modulated inflammatory mDCs to dampen immune reactions and induce specific tolerance to self- reveal transcriptional networks supporting tolerogenic function. In antigens. addition, we focused on the expression pattern of 37 genes In the pathogenesis of autoimmune diseases such as type 1 conferring genetic risk for T1D in an attempt to shed light on the diabetes (T1D) or rheumatoid arthritis, investigations have been contribution of these genes to autoimmunity versus tolerogenic primarily directed toward defining autoreactive responses leading pathways and tissue protection. to tissue destruction. Therapies suppressing such immune responses nonspecifically show reduced autoreactivity but only temporarily or in subgroups of patients,1,2 while ablating necessary immune defense to pathogens and tumors. Several RESULTS immune intervention strategies are currently explored as a Global gene expression segregates tolDC and mDC therapy to repair an impaired immune tolerance causing To analyse molecular networks that differentiate VD3/Dex- autoimmunity. Recently, promising approaches were introduced modulated tolDCs from non-modulated inflammatory mDCs, we to induce tissue-specific immune tolerance by antigen performed next-generation deep mRNA sequencing of mature administration.3–5 Improved induction of tissue-specific protection tolDCs compared with mDCs that were generated from mono- may be achieved using antigen-loaded tolDCs and this approach cytes of four healthy donors. Detected reads were further mapped is current being clinically tested for a number of autoimmune and to human transcriptome and genes with a reliable mapping score inflammatory diseases.6–8 Several protocols have been reported (430%, Po0.00001) were further transformed into reads per that generate tolDCs.9 Using combined modulation with 1,25- kilobase per million mapped reads (RPKMs; Figure 1).11 Genes dihydroxyvitamin D3, the active form of vitamin D3 (VD3) and were excluded from the analyses if the average gene expression in dexamethasone (Dex), we generated tolDCs with a superior the four donors was lower than 1 RPKM in both tolDCs and mDCs.

1Department of Immunohematology and Blood Transfusion, Leiden University Medical Center (LUMC), Leiden, The Netherlands; 2Department of Clinical and Experimental Endocrinology, University of Leuven, Leuven, Belgium; 3Department of Genetics, University Medical Center Utrecht, Utrecht, The Netherlands and 4Department of Diabetes Immunology, Diabetes and Metabolism Research Institute, City of Hope, Duarte, CA, USA. Correspondence: Dr T Nikolic, Department of Immunohematology and Blood Transfusion, Leiden University Medical Center (LUMC), Albinusdreef 2, 2333 ZA Leiden, The Netherlands. E-mail: [email protected] 5These authors contributed equally to this work. Received 21 January 2017; revised 11 May 2017; accepted 22 June 2017 Differential transcriptome of tolDCs T Nikolic et al 2 The remaining 12 754 genes were used in a principal Total RNA component analysis that clearly segregated tolDC from mDC samples (Figure 2a). Comparisons of individual gene expression Fragmentation between tolDC and mDC identified 4527 genes with ⩾ 2 fold different expression in each donor. These differentially expressed genes clustered into two groups of similar size: 2142 RNA reads genes higher expressed and 2385 genes lower expressed in Quality control tolDC compared with mDC (Figure 2b). Taking into account Reverse transcription the variation in expression between donors, a total of 3121 genes demonstrated significantly different expression between tolDC and mDC (one-way analysis of variance, qo0.01). cDNA reads These genes designated as differential tolDC/mDC transcriptome were considered for subsequent Gene Ontology (GO) analyses. Reads mapping to reference When aligned with GO terms, the differential transcribed genome genes were mainly associated with cell activation and in particular with T-cell activation (Figure 2c). Further GO terms represented in the differential gene set were biological Quantification of gene processes involved in immune regulation and response to external expression and data analysis stimuli. Figure 1. Flowchart showing the steps taken in the mRNA-sequencing analysis. Total mRNA from mature tolDC and mDC samples of four Gene pathways differentiating tolDCs from mDCs donors was isolated, fragmented and mRNA reads quantified. Raw reads Molecular pathways represented by the differential tolDC/mDC were mapped to human genome (Hg19) and normalized to RPKM. fi transcriptome were investigated using Partek Pathway analysis Genes with a mapping ef ciency of 30% and higher and average RPKM software that simultaneously considers qualitative and quantita- of > 1 across all samples were selected for further analyses (n = 4). tive changes in a given data set (Table 1). We identified 22

150

100

0 PC#2 21.3% -50

-100 tolDC mDC -150 -150 -100 -50 0 50 100 150 PC#1 53.7%

regulation of cytokine production - immune effector process granulocyte migration - defense responses - IL-10 production negative regulation of - protein transport cell activation regulation of inflammatory response cell activation response to wounding

response to molecule of bacterial origin

leukocyte activation involved in immune response

regulation of protein kinase activity

hemopoiesis T cell activation

-2.020.00 2.02 Figure 2. Principal component analysis (PCA) and unsupervised clustering segregate tolDCs from mDCs. (a) Genes that fulfilled the mapping and expression threshold were used for PCA analysis. Green circles represent tolDC samples and red circles represent mDC samples. The line connects tolDC and mDC samples of the same donor. (b) Expression data were filtered for genes that show ⩾ 2-fold (log2) differential expression between tolDC and mDC. In total, 4527 genes were selected and used for unsupervised clustering, separating tolDC (green columns) from mDC (red columns). The bar shows the color scaling of the higher (yellow) and lower (blue) expressed genes between compared samples. (c) Genes that show low variance between donors (3121 genes; qo0.01, Partek analysis of variance) were further filtered and used for GO terms analysis using ClueGO software. The pie chart depicts GO terms that were significantly (Po0.001) represented by the selected genes. Colors represent different terms, which are sorted based on the frequency of genes presented in the differential transcriptome.

Genes and Immunity (2017), 1 – 8 © 2017 Macmillan Publishers Limited, part of Springer Nature. Differential transcriptome of tolDCs T Nikolic et al 3 pathways that were ⩾ 2-fold different in tolDCs compared with Table 1. Differentially expressed pathways as determined using mDCs (Po0.001, Figure 3). Three differentially expressed path- Partek pathway-ANOVA analysis software ways related to cellular interactions were twofold lower in tolDCs Pathway name Fold change P-value (Po0.0002) and lysosome degradation was threefold higher in (tolDC/mDC) tolDCs (P = 0.0008). Two of the three pathways in the information processing group (extracellular matrix-receptor interaction and Cellular processes sulfur relay system) were higher in tolDCs (Figure 3a), whereas the Lysosome 3.0 0.0008 hippo-signaling pathway regulating cell size and apoptosis Peroxisome 1.9 0.01 showed lower expression in tolDCs compared with mDCs. Of the Cell growth and death Cell cycle − 1.5 0.001 15 metabolic pathways, oxidative phosphorylation and the Apoptosis − 1.5 0.0007 pathways of lipid and sugar metabolism were twofold higher Cellular community expressed in tolDCs (Po0.001), whereas metabolism of α-linoleic Adherens junction − 2.7 0.0001 and essential amino acids were twofold lower in tolDCs than in Tight junction − 2.4 0.0002 mDCs (Po0.005). Focal adhesion − 2.0 0.0001 Cytoscape programs (that is, ClueGO and CluePedia) were used to speciﬁcally analyze immunologic GO terms and pathways. Environmental information processing Genes reduced in tolDC compared with mDCs control immune Signal transduction κ Hippo signaling pathway − 2.7 0.0001 activation through pathways such as nuclear factor- B, ErB, JAk- Rap1 signaling pathway − 1.9 0.0002 STAT and FoxO signaling (Figure 3b). Genes with increased ErbB signaling pathway − 1.6 0.0005 expression in tolDCs than in mDCs were classiﬁed to the Fc- NF-κB signaling pathway − 1.5 0.009 receptor and complement system activation, antigen processing Signaling molecules and interaction and presentation machinery represented by the phagosome and extracellular matrix-receptor 3.9 0.03 lysosome pathways (Figure 3c). interaction

Metabolism Differential gene expression in relation to previous transcriptome Global and overview maps and proteome analyses Degradation of aromatic compounds 2.6 0.008 To evaluate the extent to which our differential transcriptomes Carbon metabolism 1.5 0.0008 relates to previous analyses, we analyzed the mRNA expression Biosynthesis of amino acids 1.4 0.002 of the previously reported 27 regulators of tolerogenicity in Energy metabolism 12 Nitrogen metabolism 5.8 0.008 tolDCs. Of these 27 reported genes, 11 were in our Sulfur metabolism 2.2 0.0000 differential transcriptome data set: 6 genes were expressed lower Oxidative phosphorylation 2.2 0.02 in tolDCs and 5 were expressed higher in tolDCs than in mDCs Amino acid metabolism (42-fold, qo0.01) (Table 2). Corroborating our pathway D-Glutamine and D-glutamate − 4.2 0.0003 analysis results, genes expressed lower in tolDC are involved in metabolism κ − nuclear factor- B and JAK-STAT pathways, whereas the higher Tryptophan metabolism 2.3 0.005 expressed genes in tolDCs are involved in antigen uptake and Phenylalanine, tyrosine and − 2.2 0.006 tryptophan biosynthesis processing or cell inhibition. Taken together, genes differentially Histidine metabolism − 2.0 0.02 expressed between tolDCs and mDCs in the current genome- Valine, leucine and isoleucine 1.8 0.003 wide RNA-sequencing analyses showed an overlap of 40% with biosynthes the molecules known to be the regulators of tolerogenic DC Glutathione metabolism 1.8 0.01 action. Carbohydrate metabolism Next, we aligned earlier reported expression ratio (tolDC/mDC) Amino sugar and nucleotide sugar 1.9 0.001 of a subset of 84 intracellular proteins from proteomic analyses,10 met. with the corresponding mRNA ratio (supplementary Figure S1). Of Pentose and glucuronate 2.0 0.0000 interconversions 84 genes encoding these proteins, all were expressed in our tolDC Pentose phosphate pathway 2.4 0.0002 and mDC samples, and 16 genes (19%) were found in our Galactose metabolism 2.7 0.0000 differential tolDC/mDC transcriptome. Ten of these 16 molecules Fructose and mannose metabolism 2.7 0.0002 (12% of reported proteins) demonstrated the same qualitative Glycan biosynthesis and metabolism differences at mRNA and protein level (both higher or both lower Other glycan degradation 2.7 0.0000 in tolDC compared with mDC), whereas 6 molecules showed Glycosphingolipid biosynthesis 2.2 0.0005 opposite ratio at mRNA and protein level (higher as mRNA, while Glycosaminoglycan degradation 1.9 0.002 Mucin type O-Glycan biosynthesis − 1.7 0.007 lower as protein in tolDCs vs mDC). Lipid metabolism Ether lipid metabolism 2.2 0.002 Differential gene expression reveals surface proteins as markers of Arachidonic acid metabolism 2.0 0.009 tolerogenic modulation Fatty acid biosynthesis − 2.7 0.02 α-Linolenic acid metabolism − 2.6 0.004 In search for new surface markers of a distinct tolDC phenotype, Metabolism of cofactors and vitamins we performed class prediction analyses using our differential Porphyrin and chlorophyll 4.9 0.003 tolDC/mDC transcriptome data set by applying the leave-one-out metabolism cross-validation method.13 This identified 20 genes (10 higher and Riboflavin metabolism 3.7 0.02 10 lower expressed in tolDC) that with a low-rate error (Po0.001) Vitamin B6 metabolism 2.2 0.02 correctly classified samples as being tolDC or mDC (Table 3). Of Thiamine metabolism 2.0 0.002 − these 20 predictor genes, CD52 and CD16 were molecules One-carbon pool by folate 2.7 0.004 expressed on the cell surface. We did not find CD16 on the Abbreviations: ANOVA, analysis of variance; mDC, mature inflammatory surface of DCs by flow cytometry, we could indeed confirm higher dendritic cell; NF-κB, nuclear factor-κB; tolDC, tolerogenic dendritic cell. expression of CD52 on tolDCs both in the immature and the mature differentiation phase (Figure 4b).

Figure 3. Differential pathways represented by the differential tolDC/mDC transcriptome. (a) Partek Pathway analysis was used to determine signiﬁcant changes in molecular pathways based on the total tolDC and mDC transcriptome. Graph depicts pathways that were ⩾ 2-fold higher (green bars) or lower (red bars) in tolDCs compared with mDC samples. (b and c) The selected data set as described in Figure 1c was further analyzed using ClueGO and CluePedia software to group genes that were on average ⩾ 2-fold lower (b) or higher (c) expressed in tolDCs than in mDCs with respect to the contribution to immunologic pathways. Diamond symbols represent depicted immunological pathways, with connected genes that are present in the analyzed tolDC/mDC dataset. The size of each diamond is determined by the signiﬁcance level (small Po0.05, medium Po0.005 and large Po0.0005) determined by the representation of the genes from the differential transcriptome within the total number of genes per pathway.

To validate the value of CD52 molecule compared to 11 other expression as two high probable predictors for tolDC phenotype surface molecules and three cytokines previously used to (P = 0.00045). discriminate tolDC from mDCs,10,14 the mRNA-sequencing values fl and the corresponding protein expression (measured by ow T1D risk genes in the regulatory transcriptome cytometry or by Luminex) from the same cells of the four donors The functionality of genetic polymorphisms increasing the risk for were log2 transformed and aligned in a heatmap (Figure 4a). T1D has been predominantly sought after in lymphocytes that are We included CD25 in this analyses, the high-affinity interleukin considered as the main contributors to the destructive immune (IL)-2 receptor by which DCs catch and fix the soluble IL-2 to the cell surface to enhance T-cell stimulation.15 Next to response. In view of important functional differences between tolDCs and mDCs with regard to immunoregulation, we evaluated CD52, gene transcripts for eight tested surface proteins 16 (CD14, ILT-3, CD274, CD25, CD80, CD83, CD40 and CD209) the expression of 37 validated risk genes in tolDC and mDCs and cytokine IL-12 were found in the differential tolDC/mDC transcriptomes (Figure 5). Transcripts of 26 out of 37 genes were transriptome (42-fold, qo0.01), showing a better overlap (62.5%) found in dendritic cells, 20 genes showed varying expression than with the intracellular protein fraction from proteomic levels between tolDCs and mDCs, of which 11 risk genes were 4 analyses. included in the differential tolDC/mDC transcriptome ( 2-fold, Subsequently, we assessed the surface expression of these 16 qo0.01). Genes for CCR5 (C–C motif chemokine receptor 5), CTSH molecules on tolDC and mDC generated from independent 35 (Cathepsin H) and RAC2 (RAS-related C3 botulinum substrate 2) donors by flow cytometry. The principal component analysis genes were higher expressed in tolDCs (42-fold, qo0.01). clustering demonstrated that the analysed surface proteins The remaining eight risk genes from the differential tolDC/mDC combined define tolDC and mDC as distinct groups (Figure 4c). transcriptome, classifying as activators of the immune response The class prediction analysis using the leave-one-out cross- were expressed lower in tolDCs than in mDCs (42-fold, qo0.01). validation method allocated high CD52 and low CD86 surface These were as follows: IKZF4 (Eos), IKZF1 (Ikaros), SH2B3 (SH2B

Table 2. Average gene expression ratio (tolDC/mDC) of reported regulators of DC tolerogeneicity

Regulator EntrezID Fold change (tolDC/mDC ) P-value Associated signaling pathways

NF-κB1 p50 4790 − 9.5a 0.0000 Induction of inflammatory cytokines BLIMP-1 639 − 5.9 0.0001 Suppression of IL-6 and Ccl2 transcription PD-L1 29 126 − 5.0 0.0001 PD-1 expressing T-cell inhibition IDO 3620 − 4.1 0.016 Tryptophan degradation SOCS-1 8651 − 3.6 0.0035 Suppressor of cytokine signaling—Th1 induction STAT3 6774 − 2.8 0.0019 Inhibition of NF-κB activation A20/TNFAIP3 7128 − 2.8 0.015 Degradation of intermediate NF-κB molecules Lyn 4067 − 2.5 0.0003 Negative regulation of MyD88 pathway AhR 196 − 2.4 0.063 Induction of IDO DEC-205 6729 − 1.5 0.025 Delivery of endocytosed proteins to MHC molecules ERK 5594 − 1.5 0.085 Decrease of NF-κB DNA binding Wnt5a 7474 − 1.2 0.59 Induction of STAT3 and SOCS3 HO-1 3162 1.0 0.84 Inhibition of NF-κB and IRF3 CD39 953 1.2 0.31 Activation of NLRP3 inflammasome SOCS3 9021 1.3 0.66 Suppressor of cytokine signaling—Th2 induction ILT4 6606 1.3 0.10 Inhibition of cell activation PPAR-g 5468 1.3 0.47 Inhibition of NF-κB nuclear localization RALDH2 8854 2.1 0.082 Repression of proinflammatory cytokines DCIR 50 856 4.6 0.039 Antigen uptake and negative signaling molecule SHP-1 5777 6.5 0.0005 Inhibition of pro-survival signals and cell activation DC-SIGN 30 835 8.9 0.0007 Modulates TLR signals favoring IL-10 production MFG-E8 4240 11.6 0.017 Suppression of NF-κB, activation of STAT3 and A20 TLR2 7097 21.5 0.058 Induction of RALDH2 ILT3 11 006 21.8 0.0001 Inhibition of cell activation FcgRIIB 2213 27.1 0.003 Inhibition of cell activation through SHP-1 GILZ 1831 32.9 0.16 Blockage of NF-κB, induction of Tr-1-like features C1q 712 65.2 0.0054 Complex formation with DC-SIGN, cell inhibition Abbreviations: ANOVA, analysis of variance; DC-SIGN, Dendritic Cell-Specific Intercellular adhesion molecule-3-Grabbing Non-integrin; IDO, Indoleamine 2,3-dioxygenase; IL-6, interleukin-6; IL-10, interleukin-10; mDC, mature inflammatory dendritic cell; MHC, major histocompatibility complex; NF-κB, nuclear factor-κB; TLR, Toll-like receptor; tolDC, tolerogenic dendritic cell. aGenes with negative tolDC/mDC ratio have lower expression in tolDC.

Table 3. Top 10 ranked genes that predict tolDC vs mDC phenotype (class prediction analysis)

Rank Symbol P-value T-value Fold change tolDC/mDC ) EntrezID Gene name

Genes higher in tolDC 1 C1QA 2.8e − 6 11.2 166.8 712 Complement component 1, q, A chain 2 SLC11A1 3.1e − 5 8.1 146.2 6556 Solute carrier family 11, member 1 3 FBP1 o1ev7 29.5 140.1 2203 Fructose-1,6-bisphosphatase 1 4 C1orf162 4.0e − 7 14.5 98.7 128346 Chromosome 1 open reading frame 162 5 CD52 o1e − 7 20.2 92.2 1043 CD52 molecule 6 RGCC o1e − 7 20.1 83.6 28984 Regulator of cell cycle 7 FCGR3A 3.0e − 7 14.6 75.0 2214 Fc fragment low afﬁnity IIIa, receptor (CD16a) 8 MCEMP1 o1e − 7 17.1 72.2 199675 Mast cell-expressed membrane protein 1 9 FN1 o1e − 7 19.7 66.3 2335 Fibronectin 1 10 MATK o1e − 7 19.2 62.1 4145 Megakaryocyte-associated tyrosine kinase

Genes lower in tolDC 1 AOC1 5.6e − 5 − 7.5 0.021 26 Amine oxidase, copper containing 1 2 GPR157 2.6e − 6 − 11.3 0.023 80045 G-protein-coupled receptor 157 3 ADAM19 o1e − 7 − 19.7 0.023 8728 ADAM metallopeptidase domain 19 4 PLEKHA5 o1e − 7 − 18.1 0.033 54477 Pleckstrin homology domain, family A-5 5 FSCN1 2.1e − 6 − 11.6 0.035 6624 Fascin homolog 1, actin-bundling protein 6 ANKRD33B 1.0e − 7 − 16.5 0.037 651746 Ankyrin repeat domain 33B 7 PPP1R16B 2.3e − 6 − 11.5 0.04 26051 Protein phosphatase 1, regulatory subunit 16B 8 ADAM12 2.0e − 7 − 15.6 0.041 8038 ADAM metallopeptidase domain 12 9 EHF 8.0e − 7 − 13.2 0.042 26298 Ets homologous factor 10 PIK3CG 1.0e − 7 − 16.4 0.043 5294 Phosphoinositide-3-kinase, catalytic, gamma Abbreviations: mDC, mature inﬂammatory dendritic cell; tolDC, tolerogenic dendritic cell. adaptor protein 3), ORMDL3 (ORMDL sphingolipid biosynthesis DISCUSSION regulator 3), TYK2 (tyrosine kinase 2), IL2RA (IL-2 receptor-α, CD25), Given the critical role for dendritic cells in immune regulation, a PTPN2 (protein tyrosine phosphatase, non-receptor type 2) and plethora of modulators has been tested aiming to induce effective ICOSLG (inducible T-cell co-stimulator ligand). tolDCs. We found that combined modulation with VD3 and

genes proteins 8 mDC tolDC mDC tolDC

CD52 CD14 ILT-3 IL-12 TNF IL-10 CD274 CD25 CD86 CD83 CD1a

HLA-DR PC#24.94% CD40 CD80

CD209 PC#3 2.49% TLR2 PC#1 87.6%

Figure 4. Comparison of gene and protein expression upon tolDC modulation. (a) Heatmaps depict the mRNA (left) and corresponding protein expression (right) of a set of 16 molecules (rows) determined in the mDCs (red) or tolDCs (green) from four donors arranged in the same order in corresponding columns (n = 4). Gene expression data (RPKM), mean ﬂuorescence intensity (MFI) and Luminex (pg ml − 1) values were log2-transformed, to allow the same scaling (bottom bar) for both mRNA and protein expression. (b) Surface expression of CD52 during tolDC and mDC differentiation in vitro. Empty histogram represents isotype control, blue histogram shows CD52 expression on monocytes used for DC differentiation (day0). Green (tolDCs) and red (mDCs) histograms show CD52 expression by DC at immature (day 6) and mature (day 8) differentiation stage. Modulation of monocytes into tolDCs prevents the loss of CD52 on differentiating DCs. (c) principal component analysis (PCA) clustering of tolDCs and mDCs from 35 donors using the expression of 16 proteins as given in a. Green spheres represent tolDCs and red spheres represent mDCs (n = 35).

Dex generates stable tolDC with potent immunomodulating DCs to implement aerobic metabolism or switch to glycolysis and capacities, including the induction of antigen-specific regulatory determine their survival in tissues. Similar differences in metabolic T cells.9,10,14,17 We next investigated molecular changes in DCs rate exist between regulatory and effector T cells exposed to caused by the combined modulation using deep sequencing by inflammation,19 embracing metabolism as an inevitable contributor Illumina Technology (Illumina, Inc., San Diego, CA, USA). The gene to the regulation of immune processes. expression profile differentiating tolDCs from mDCs revealed Our analyses show that combined VD3/Dex modulation several molecular pathways important for tolerogenic function increases metabolism of carbohydrates while reducing amino that extends previous findings.12 Besides profound metabolic acids and fatty acid synthesis in tolDC compared to mDC. Such changes toward the aerobic energy maintenance, tolDCs show nutrient deprivation supports regulatory T-cell induction,20 and lower expression of immune activation networks and reduced represents an important functional feature of tolDCs. Together ability to establish firm cellular interactions, which contribute to with decreased capacity for cellular interactions, biased produc- their immunomodulating function. Furthermore, combined mod- tion of inhibitory cytokines (that is, transforming growth factor-β ulation with VitD3 and Dex increased antigen uptake and and IL-10) rather than IL-12 and IL-15, and predominant processing pathways in tolDCs. Previously observed changes in expression of inhibiting co-stimulatory molecules of the B7 and the modulated intracellular tolDC proteome did not directly tumor necrosis factor gene families, our tolDCs appear extremely correlate to the differential tolDC/mDC transcriptome. Yet, the well equipped to provide negative signals while ceasing activation combined mRNA and surface protein expression analyses in this and survival signals to keep T cells in control. In contrast, genes study revealed CD52 as a reliable predictor of tolerogenic expressed higher in the immune transcriptome of tolDCs support modulation. Finally, our analyses implicate T1D risk genes as the resistance to bacterial infection (involving Fc receptors and controllers of DC function, revealing potential targets for complement), phagocytosis and lysosome activity. These genes immunomodulation in vivo. also importantly contribute to antigen uptake and processing. The GO terms represented by differential tolDC/mDC transcrip- Although complement factors and Fc receptors have been linked tomes selected cell activation and in particular T-cell activation to the control of inflammation,21–23 the role of increased antigen capacity as differentiating features between our tolerogenic and processing in immune regulation remains to be established. inflammatory DCs. Genes in these immunological GO terms were The differential tolDC/mDC transcriptome included several mainly those expressed lower in tolDCs than in mDCs, whereas the previously described regulators of tolerogenicity (40% overlap) genes higher expressed in tolDCs were found in metabolic rather and many non-overlapping regulator genes participate in path- than immunologic pathways. Our transcriptome data further match ways found significant in our current analyses. This suggests that a the changes in glucose processing and oxidative phosphorylation combined action of genes as pathways rather than single genes or noted in proteomic analyses of VD3/Dex tolDC10 and transcriptome molecules control tolerogenic capacity in tolDCs. Furthermore, the of VD3-only modulated tolDCs.18 Besides controlling differentiation non-overlapping regulators of tolerogenic state may support of tolDC and mDC in vitro, these processes manage the ability of variable phenotype and regulating capacities in tolDCs induced by

Genes and Immunity (2017), 1 – 8 © 2017 Macmillan Publishers Limited, part of Springer Nature. Differential transcriptome of tolDCs T Nikolic et al 7 mDC tolDC procedure to modulate differentiation and maturation of monocytes into tolDCs target proteins that contribute to genetically associated susceptibility or protection to development of T1D. ICOSLG Achieving such ‘regulatory transcriptome’ by therapeutic immune ORMDL3 SH2B3 modulation through small molecules that target these genetic risk TYK2 loci could favorably modulate the genetically inferred risk for T1D IL2RA or other autoimmune diseases. IKZF4 IKZF1 PTPN2 MATERIALS AND METHODS CCR5 fold > 2 CTSH In vitro generation of tolDCs and mDCs q < 0.01 RAC2 Mature tolDCs and control non-modulated DCs (mDCs) were generated as previously described.10 In short, buffy coats of healthy human donors were CD226 purchased from Sanquin Blood Bank (Amsterdam, The Netherlands) and IFIH1 peripheral blood mononuclear cells were isolated by density gradient RASGRP1 centrifugation (Ficoll, Axis-Shield, Oslo, Norway), and CD14+ monocytes CTLA4 fi HLA-DQB were puri ed (MACS, Milteny Biotec, Bergisch Gladbach, Germany). HLA-DRB Isolated monocytes (490% pure) were cultured in RPMI 1640 medium − 1 BCAR1 (Gibco, Paisley, Scotland) supplemented with 500 IU ml penicillin– fold > 2 IL-10 streptomycin (Life Technologies, Bleiswijk, The Netherlands) and 10% q > 0.01 PTPN22 heat-inactivated fetal calf serum (Gibco), in six-well plates (TPP, Transadin- gen, Switzerland). For inducing DC differentiation, 500 U ml − 1 recombi- GSDMB nant human IL-4 (Gentaur, Brussel, Belgium) and 800 U ml − 1 recombinant GPR183 human granulocyte–macrophage colony-stimulating factor (Gentaur) were DEXI added. Medium and cytokines were refreshed on day 3. For generation of CD69 fold < 2 tolDCs, VD3 was added at the beginning of the culture at a final IL7R − 8 q > 0.01 concentration of 10 M and refreshed at day 3. Dex was added on day 3 at HLA-DQA − 6 a final concentration of 10 M. On day 6, immature DCs were collected and maturation was induced using a cytokine mix (IL-6, 500 U ml − 1; IL-1b, − 1 − 1 -2.430 2.43 1600 U ml ; tumor necrosis factor, 335 U ml ; all from Miltenyi Biotec) − 1 and prostaglandin E2 (2 ug ml , Sigma-Aldrich, Zwijndrecht, The Nether- Figure 5. Expression of T1D risk genes in tolDC defines ‘regulatory lands). After 48 h, mature DCs were collected for functional/morphological transcriptome.’ Heatmap depicts the expression of 26 validated T1D analysis or for protein extraction. risk genes present in the tolDC and mDC transcriptome. Clustering of columns is supervised by the donor. Genes (rows) are clustered fl depending on the expression level, segregating genes that belong Antibodies, ow cytometry and cytokine analysis to the differential tolDC/mDC transcriptome (42-fold, qo0.01, top The following antibodies for flow cytometry were used to analyse panel), genes that show different expression in tolDC versus mDC phenotype of tolDC and mDC: fluorescein isothiocyanate-conjugated but have a strong variance between donors (42-fold, q40.01, anti-CD80 (clone L307.4) and anti-HLA-DR (clone G46-6); PE-conjugated middle panel and genes with similar expression in tolDC and mDC anti-CD86 (clone IT2.2), anti-CD1a (clone HI-149), anti-CD14 (clone M5E2), (o2-fold, q40.01, bottom panel). PE-Cy7-conjugated anti-CD25 antibody (clone M-A251) and IgG isotype controls, all were from BD Pharmingen (San Diego, CA, USA). PE- different modulating agents.9,12 We found a low overlap (12%) conjugated anti-TLR-2 (clone TL2.1) was from Santa Cruz Biotec (Heidel- with the differential proteome measured by two-dimensional berg, Germany). APC-conjugated anti-CD40 (clone SC3), anti-CD209 (DC fl sign; clone ebH209) and anti-CD274 (PD-L1; clone MIH1) were from uorescence difference gel electrophoresis. This technique pre- e-Bioscience (Vienna, Austria). fluorescein isothiocyanate-conjugated anti- dominantly detects hydrophilic proteins with substantial CD83 (clone HB15e) and PE-Cy7-conjugated CD85k (ILT3, clone ZM3.8) 24 expression, enforcing the focus of the proteomic study on were from Beckman Coulter (Woerden, The Netherlands). Fluorescein abundantly present intracellular proteins and pathways.10 This isothiocyanate-conjugated anti-CD52 (clone YTH34.5) and corresponding may clarify why many hydrophobic membrane-bound proteins isotype control (IgG2b) was from AbD Serotec (Oxford, UK). including the surface protein CD52 were absent from the For flow cytometry analysis, aliquots of 2 × 105 cells were incubated differential proteome.10 Our flow cytometry analyses much better with a cocktail of monoclonal antibodies, incubated on ice for 30 min, aligned with our differential transcriptome data (62.5%), but washed with phosphate-buffered saline supplemented with 0.5% fl bovine serum albumin. Cells were acquired on a FACS Calibur (Becton remain limited by the availability of uorescent monoclonal Dickinson, San Diego, CA, USA). Mean fluorescence intensity values were antibodies. Temporal differences in gene vs protein expression, extracted and log2 transformed to allow clustering and generation of half-life or protein loss due to secretion may result in different heatmaps. outcomes at RNA and protein level. Such was CD86, showing as Culture supernatants from mDCs and tolDCs after a 48 h maturation protein and not as RNA a strong predictive value. Nonetheless, were analyzed for the presence of cytokines (IL-10, IL-12 and tumor deep RNA-expression analyses proved of significant practical value necrosis factor) using Luminex bead array (Biorad, Veenendaal, The prompting the discovery of CD52 as a new tolDC biomarker, thus Netherlands). Similar to surface expression, the protein concentrations enabling a straight-forward clinical release criterion of functional detected in Luminex were log2 transformed before analyses. tolDCs generated in vitro for clinical immune intervention trials. In the uncharted research area of tolDC-specific pathways, our Preparation of mRNA-sequencing sample and deep sequencing results encourage further exploration and validation of whether Total RNA was isolated using the Nucleospin miRNA kit (Bioké, Leiden, The the induction of regulatory transcriptome in DCs depends on Netherlands). RNA sequencing was performed on 300–500 ng of total RNA ⩾ concomitant therapies that affect monocytes. This may be a (RNA Integrity Number 7.0) at Beijing Genomics Institute-Shenzhen, particular issue shaping the treatment with tolDCs for patients Shenzhen, China (http://www.genomics.cn/index.php). Sequencing was performed using Illumina Truseq RNA-SEQ 160 bp short-insert library with autoimmune diseases (that is, Myasthenia gravis, RA or SLE) preparation and the HiSeq 2000 platform according to the manufacturer’s with immunosuppressive therapies as a standard of care. instructions (Illumina, San Diego, CA, USA). For each sample 4 Gb clean Many candidate risk loci for T1D proved differentially expressed data per sample was provided. For analysis, the received data file with in the transcriptomes of tolDCs vs mDCs. This may reflect that our reads per gene was converted into RPKMs with the following formula:

© 2017 Macmillan Publishers Limited, part of Springer Nature. Genes and Immunity (2017), 1 – 8 Differential transcriptome of tolDCs T Nikolic et al 8 RPKM = 109 × C/(N × L), where C represents the number of reads for a gene, 8 Harry RA, Anderson AE, Isaacs JD, Hilkens CM. Generation and characterisation of N is the number of total mapped reads and L is the gene length in therapeutic tolerogenic dendritic cells for rheumatoid arthritis. Ann Rheum Dis basepairs.11 Genes with o30% coverage of mapped reads onto a gene’s 2010; 69: 2042–2050. exons (hg19 ensembl release 75) or mean RPKM of o1 across all the 9 Nikolic T, Roep BO. Regulatory multitasking of tolerogenic dendritic cells - lessons samples (s.d.o1) were filtered out, selecting only genes adequately taken from vitamin d3-treated tolerogenic dendritic cells. Front Immunol 2013; represented in the RNA-sequencing sample for further analyses. 4: 113. 10 Ferreira GB, Kleijwegt FS, Waelkens E, Lage K, Nikolic T, Hansen DA et al. Statistical analysis Differential protein pathways in 1,25-dihydroxyvitamin d(3) and dexamethasone modulated tolerogenic human dendritic cells. J Proteome Res 2012; 11: For principal component analysis, clustering and pathway analysis, we used – 25 26,27 941 971. the softwares Partek Genomics Suite 6 and ClueGo/CluePedia. 11 Mortazavi A, Williams BA, McCue K, Schaeffer L, Wold B. Mapping and quantifying Differences in gene expression between groups were analysed by one- – o mammalian transcriptomes by RNA-Seq. Nat Methods 2008; 5:621 628. way analysis of variance and pathway analysis of variance tools P 0.05 12 Schinnerling K, Soto L, Garcia-Gonzalez P, Catalan D, Aguillon JC. Skewing den- using Partek Genomics Suite. Where appropriate, the multiple testing dritic cell differentiation towards a tolerogenic state for recovery of tolerance in correction of P-values was performed by positive false discovery rate (q rheumatoid arthritis. Autoimmun Rev 2015; 14: 517–527. value) implemented in Partek Genomics Suite. Class prediction analyses 13 Simon R, Lam A, Li MC, Ngan M, Menenzes S, Zhao Y. Analysis of gene expression were performed using BRB-ArrayTools software.13 Flow cytometry analyses data using BRB-ArrayTools. Cancer Inform 2007; 3:11–17. were performed on FlowJo 7.6 (Tree Star, Ashland, OR, USA). 14 Unger WW, Laban S, Kleijwegt FS, van der Slik AR, Roep BO. Induction of Treg by monocyte-derived DC modulated by vitamin D3 or dexamethasone: differential role for PD-L1. Eur J Immunol 2009; 39: 3147–3159. CONFLICT OF INTEREST 15 Herr F, Lemoine R, Gouilleux F, Meley D, Kazma I, Heraud A et al. IL-2 phos- The authors declare no conflict of interest. phorylates STAT5 to drive IFN-gamma production and activation of human dendritic cells. J Immunol 2014; 192: 5660–5670. 16 Onengut-Gumuscu S, Chen WM, Burren O, Cooper NJ, Quinlan AR, Mychaleckyj JC ACKNOWLEDGEMENTS et al. Fine mapping of type 1 diabetes susceptibility loci and evidence for colo- This work was supported by European Union 7th Framework Programme (FP7/2007– calization of causal variants with lymphoid gene enhancers. Nat Genet 2015; 47: 2013) Grant 241447 (NAIMIT) and the Innovative Medicines Initiative 2 Joint 381–386. Undertaking under Grant Agreement 115797 (INNODIA) (this Joint Undertaking 17 Suwandi JS, Toes RE, Nikolic T, Roep BO. Inducing tissue specific tolerance in receives support from the Union’s Horizon 2020 research and innovation program autoimmune disease with tolerogenic dendritic cells. Clin Exp Rheumatol 2015; 33 and ‘EFPIA,’‘JDRF,’ and ‘The Leona M. and Harry B. Helmsley Charitable Trust.’), (4 Suppl 92): S97–S103. Netherlands Organization for Scientific Research Vici Grant 918.86.611, the Dutch 18 Ferreira GB, Vanherwegen AS, Eelen G, Gutierrez AC, Van Lommel L, Marchal K Arthritis Foundation (LLP16) and by an Expert Center Grant from the Dutch Diabetes et al. Vitamin D3 induces tolerance in human dendritic cells by activation of Research Foundation. intracellular metabolic pathways. Cell Rep 2015; 10: 711–725. 19 Zeng H, Chi H. The interplay between regulatory T cells and metabolism in immune regulation. Oncoimmunology 2013; 2: e26586. REFERENCES 20 Procaccini C, Galgani M, De Rosa V, Matarese G. Intracellular metabolic pathways – 1 Woittiez NJ, Roep BO. Impact of disease heterogeneity on treatment efficacy of control immune tolerance. Trends Immunol 2012; 33:1 7. immunotherapy in type 1 diabetes: different shades of gray. Immunotherapy 21 Elvington M, Scheiber M, Yang X, Lyons K, Jacqmin D, Wadsworth C et al. 2015; 7: 163–174. Complement-dependent modulation of antitumor immunity following radiation 2 Ostrov BE. Immunotherapeutic biologic agents in autoimmune and autoin- therapy. Cell Rep 2014; 8:818–830. flammatory diseases. Immunol Invest 2015; 44:777–802. 22 Skoberne M, Somersan S, Almodovar W, Truong T, Petrova K, Henson PM et al. 3 Kim KS, Surh CD. Induction of immune tolerance to dietary antigens. Adv Exp Med The apoptotic-cell receptor CR3, but not alphavbeta5, is a regulator of human Biol 2015; 850:93–118. dendritic-cell immunostimulatory function. Blood 2006; 108:947–955. 4 Gibson VB, Nikolic T, Pearce VQ, Demengeot J, Roep BO, Peakman M. Proinsulin 23 Bieber T. The pro- and anti-inflammatory properties of human antigen-presenting multi-peptide immunotherapy induces antigen-specific regulatory T cells and cells expressing the high affinity receptor for IgE (Fc epsilon RI). Immunobiology limits autoimmunity in a humanized model. Clin Exp Immunol 2015; 182:251–260. 2007; 212:499–503. 5 Thrower SL, James L, Hall W, Green KM, Arif S, Allen JS et al. Proinsulin peptide 24 Rabilloud T. Membrane proteins and proteomics: love is possible, but so difficult. immunotherapy in type 1 diabetes: report of a first-in-man Phase I safety study. Electrophoresis 2009; 30 (Suppl 1): S174–S180. Clin Exp Immunol 2009; 155:156–165. 25 Kozulin P, Natoli R, O'Brien KM, Madigan MC, Provis JM. Differential expression of 6 Mansilla MJ, Selles-Moreno C, Fabregas-Puig S, Amoedo J, Navarro-Barriuso J, anti-angiogenic factors and guidance genes in the developing macula. Mol Vis Teniente-Serra A et al. Beneficial effect of tolerogenic dendritic cells pulsed with 2009; 15:45–59. MOG autoantigen in experimental autoimmune encephalomyelitis. CNS Neurosci 26 Bindea G, Galon J, Mlecnik B. CluePedia Cytoscape plugin: pathway insights using Ther 2015; 21: 222–230. integrated experimental and in silico data. Bioinformatics 2013; 29:661–663. 7 Benham H, Nel HJ, Law SC, Mehdi AM, Street S, Ramnoruth N et al. Citrullinated 27 Bindea G, Mlecnik B, Hackl H, Charoentong P, Tosolini M, Kirilovsky A et al. peptide dendritic cell immunotherapy in HLA risk genotype-positive rheumatoid ClueGO: a Cytoscape plug-in to decipher functionally grouped gene ontology and arthritis patients. Sci Transl Med 2015; 7: 290ra87. pathway annotation networks. Bioinformatics 2009; 25: 1091–1093.

Supplementary Information accompanies this paper on Genes and Immunity website (http://www.nature.com/gene)