Epithelial–Mesenchymal Transition in Human Prostate Cancer Demonstrates Enhanced Immune Evasion Marked by IDO1 Expression Kimberley Kolijn1, Esther I

Published OnlineFirst June 19, 2018; DOI: 10.1158/0008-5472.CAN-17-3752

Cancer Tumor Biology and Immunology Research

Epithelial–Mesenchymal Transition in Human Prostate Cancer Demonstrates Enhanced Immune Evasion Marked by IDO1 Expression Kimberley Kolijn1, Esther I. Verhoef1, Marcel Smid2,ReneB ottcher€ 3, Guido W. Jenster3, Reno Debets2, and Geert J.L.H. van Leenders1

Abstract

Cancer invasion and metastasis are driven by epithelial– N-cadherin–positive areas. N-cadherin–positive areas also mesenchymal transition (EMT), yet the exact mechanisms exhibited a local decrease of intraepithelial cytotoxic þ that account for EMT in clinical prostate cancer are not (CD8 ) T cells and an increase of immunosuppressive þ þ fully understood. Expression of N-cadherin is considered a regulatory T cells (CD4 /FOXP3 ). In conclusion, EMT in hallmark of EMT in clinical prostate cancer. In this study, clinical prostate cancer is accompanied by upregulated we determined the molecular mechanisms associated with expression of IDO1 and an increased number N-cadherin expression in patients with prostate cancer. We of regulatory T cells. These data indicate that EMT, which performed laser capture microdissection of matched N- is an important step in tumor progression, can be pro- cadherin–positive and -negative prostate cancer areas from tected from effective immune control in patients with patient samples (n ¼ 8), followed by RNA sequencing. N- prostate cancer. cadherin expression was significantly associated with an immune-regulatory signature including profound upregu- Significance: These findings demonstrate EMT is linked lation of indoleamine 2,3-dioxygenase (IDO1; log2-fold to an immunosuppressive environment in clinical pro- change ¼ 5.1; P ¼ 2.98E-04). Fluorescent immunostain- state cancer, suggesting that patients with prostate cancer ings of patient samples confirmed expression of IDO1 can potentially benefit from combinatorial drug therapy. protein and also its metabolite kynurenine in primarily Cancer Res; 78(16); 4671–9. 2018 AACR.

Introduction vimentin, ﬁbronectin, N-, and OB-cadherin combined with suppression of epithelial markers such as keratins and E-cadherin (4). Prostate cancer is a heterogeneous disease in terms of patho- N-cadherin has been recognized as a robust marker for logic growth patterns, molecular aberrations, and clinical out- EMT both in vitro and in vivo (5–8). Although expression of come. While prostate cancer is currently graded according to the N-cadherin is rare in localized low-grade prostate cancer, it is Gleason scoring system, analysis of individual growth patterns increased in high-grade progressive disease (9–11). Concom- provides additional information on tumor cell biology and itant N-cadherin upregulation and E-cadherin downregulation, clinical behavior (1). We recently found that Gleason score 7 also referred to as cadherin switching, was found to be the most prostate cancer with "ill-formed" architecture was enriched for reliable marker for EMT in prostate cancer patients' samples tumor cells that express N-cadherin, and represents a morpho- (2). While various mechanisms involved in cadherin switching logic substrate for epithelial–mesenchymal transition (EMT) in and EMT have been identiﬁed in vitro, the underlying regulatory patients with prostate cancer (2). pathways in clinical prostate cancer are unclear. Therefore, the EMT is characterized by the acquisition of a mesenchymal-like objective of this study was to determine the molecular and phenotype by epithelial cells, and mediates tumor invasion and cellular mechanisms associated with N-cadherin expression in metastasis (3). Epithelial cells that undergo EMT in vitro become patients with prostate cancer by using microdissection and RNA spindle shaped and upregulate mesenchymal markers such as sequencing.

1Department of Pathology, Erasmus Medical Center, Rotterdam, the Nether- Materials and Methods 2 lands. Department of Medical Oncology, Erasmus Medical Center, Rotterdam, Patient specimens 3 the Netherlands. Department of Urology, Erasmus Medical Center, Rotterdam, Patients with prostate cancer, who had undergone surgical the Netherlands. removal of the prostate (radical prostatectomy) without receiving Note: Supplementary data for this article are available at Cancer Research prior radiation or hormonal therapy for their disease, were select- Online (http://cancerres.aacrjournals.org/). ed at the Erasmus Medical Center (Rotterdam, the Netherlands; Corresponding Author: Kimberley Kolijn, Department of Pathology, Erasmus Supplementary Table S1). Prostate specimens were transported Medical Center, P.O. Box 2040, 3000 CA, Rotterdam, the Netherlands. Phone: on ice after surgery to the pathology department. A transverse 316-5000-1696; Fax: 31-10-7038340; E-mail: [email protected]. tissue slide was snap frozen in liquid nitrogen for research doi: 10.1158/0008-5472.CAN-17-3752 purposes and stored at 196C until use. The remaining prostate 2018 American Association for Cancer Research. was injected with neutral-buffered formalin (4%) to ensure fast

www.aacrjournals.org 4671

Kolijn et al.

þ þ and equal fixation prior to paraffin embedding. All prostate The number of CD4 and CD8 cells was scored using photo- specimens were evaluated by a urogenital pathologist (GvL) who graphs made using a light microscope, and a small open area recorded Gleason score according to the WHO/ISUP 2014 guide- overlay in Photoshop (Adobe CS6, Adobe Systems Incorporated) lines, as well as pT-stage (WHO 2009), surgical margin status and to randomly select scoring areas and mask other tissue areas. tumor growth pattern (1). The use of residual tissue for scientific purposes was approved by the institutional Medical Research Laser capture microdissection and RNA isolation Ethics Committee (MEC-2011-295 and MEC-2011-296). Sam- We performed hematoxylin/eosin staining and N-cadherin ples were used in accordance with the "World Medical Associa- IHC on frozen tissue samples derived from radical prostatectomy tion-Declaration of Taipei on Ethical Considerations regarding specimens containing >20% tumor cells. The specimens were Health Databases and Biobanks," as well as national and inter- preselected after screening a larger number of frozen prostate national guidelines. Informed consent was obtained through an samples for the presence of ill-formed Gleason grade 4 and N- opt-out procedure, as described in the "Human Tissue and Med- cadherin expression (Supplementary Table S1). Eight samples ical Research: Code of conduct for responsible use" developed by that contained tumor areas with more than a hundred N-cad- the Dutch Federation of Medical Scientific Societies (FMWV, herin–positive cells were selected for laser capture microdissec- version 2002, update 2011). tion (LCM) and subsequent RNA sequencing. N-cadherin–positive and -negative areas were identified by matching both ill-- Immunostainings formed architecture and N-cadherin expression in consecutive Slices of 4- to 5-mm were cut from formalin-fixed, paraffin- reference slices. Every two slices, an additional reference slice was embedded tissues or frozen tissue and mounted on Starfrost cut for IHC staining to identify and confirm N-cadherin–positive (Knittel) silane-coated glass slides for IHC or KP Frost (Klinipath, tissue regions in LCM slices. N-cadherin–positive and -negative the Netherlands) glass slides for fluorescent immunostaining (IF). prostate cancer areas (0.15 mm2) present in the same tissue For IHC, paraffin-embedded tissues were deparaffinized with section were captured separately in LCM tubes (AdhesiveCap500 xylene and rehydrated. Endogenous peroxidase activity was opaque, Carl Zeiss; Supplementary Fig. S1). These areas contained blocked with 0.3% hydrogen peroxide in PBS for 20 minutes. both malignant epithelial cells and adjacent stroma. Heat antigen retrieval was performed with a specialized micro- LCM and RNA isolation were performed as described previ- wave (MicroMed T/T Mega) with tissues merged in citrate buffer ously (12). Cresyl Violet acetate (Sigma-Aldrich) was used to stain (pH 6.0, Sigma-Aldrich) for 15 minutes. Slides were incubated slides and visualize morphology. RNA was isolated using the overnight at 4C with primary antibodies diluted in PBS that miRNeasy Micro Kit (Qiagen) according to the manufacturer's contained 2% BSA. Secondary antibodies were incubated for 30 guidelines and included an on-column DNAse digestion step minutes at room temperature and visualized using the EnVision (RNase-Free DNase set, Qiagen). RNA Integrity Number (RIN) system (Dako). All primary and secondary antibodies are spec- and quantity were measured in duplicate in a bioanalyzer (2100 ified (Supplementary Table S2). Bioanalyzer, Agilent Technologies) using the RNA 6000 Pico For immunostainings on frozen sections, slides were fixated in Kit (Agilent Technologies) according to manufacturer's protocol. formalin (4%) and permeabilized with 0.5% TritonX-100 in PBS The mean quality of total RNA isolated by LCM was 26.1 ng for 10 minutes. Slides were washed with PBS, blocked with 10% (median 20.6; range 8.5–63.3 ng) with RIN values between BSA/PBS and incubated with primary antibody in 2% BSA/PBS for 6.1 and 8.4 (mean 7.3; median 7.4). The RNA quality and quantity 1 hour at room temperature or at 4C overnight. After washing was not significantly different between N-cadherin–positive with PBS, slides were incubated with secondary antibodies and and –negative regions. chromogenically visualized (EnVision Dako Kit, Dako). For fluorescent immunostainings, slides were incubated with secondary RNA preamplification, sequencing, and data workflow antibodies labeled with Alexa488 and Alexa594. Slides were RNA preamplification and sequencing were performed by mounted in Vectashield (Vector Laboratories) containing AROS (AROS Applied Biotechnology A/S). Total RNA was pre- Hoechst 34580 (4,000, Life Technologies) to visualize nuclei. amplified (Ovation RNAseq v2 System NuGen.) and used to Immunofluorescent tissue sections were analyzed using a confo- generate double stranded cDNA from mRNA and nonpolyade- cal laser scanning microscope (Zeiss LSM 700, Carl Zeiss) with nylated transcripts. The cDNA libraries were prepared from all ZEN 2 imaging software (Carl Zeiss). samples and one control sample (high-quality human reference þ þ To quantify regulatory T cells (CD4 /FOXP3 ) in N-cadherin– RNA) according to the KAPA Library Preparation Kits (Illumina positive and -negative regions, we performed immunofluorescent series, KK8200 and KK8201). Samples were fragmented to 300– triple staining (n ¼ 3). Sections were incubated with primary 500 base pair libraries with an ultrasonicator (M22 Focused, antibodies overnight, washed with PBS-T and PBS, followed by Covaris) and DNA quality was measured (High Sensitivity DNA, incubation with secondary antibodies labeled with Alexa647 and Agilent Technologies). High-quality samples were then paired- Alexa594. After washing, an additional incubation step for 1 hour end sequenced with TruSeq SBS V4 chemistry on the HiSeq2500 at room temperature with CD4 conjugated to Alexa 488 was (Illumina). An average of 111.2 million (total) reads was mapped þ þ performed. The number of CD4 /FOXP3 cells was scored per sample (Supplementary Table S3). All RNA sequence data within an N-cadherin–positive or -negative area between 2.2 and were aligned to the preindexed UCSC human reference genome 3.2 mm2. The regulatory T cell count was corrected for area size. To 19 with TopHat version 2.0.4 and expression levels quantified via quantify lymphocytic infiltrates, we performed chromogenic dou- HTSeq-count (version 0.5.4p1; ref. 13). Accuracy was increased by ble staining of CD4 or CD8 with N-cadherin using the Ventana alignment against the indexed transcriptome prior to alignment Benchmark Ultra platform (Ventana Medical Systems; n ¼ 10). to the genome. Subsequently, the edgeR package (version 3.0.4) Ultraview universal DAB and Alkalin phosphatase were used for was used to investigate differentially expressed genes between the detection and chromogenic staining of secondary antibodies. N-cadherin–positive versus -negative prostate cancer areas (14).

4672 Cancer Res; 78(16) August 15, 2018 Cancer Research

EMT and Immune Evasion in Clinical Prostate Cancer

Differentially expressed genes with a P < 0.05 and a log2 fold change (log2FC) smaller or larger than 1 were analyzed (equal to halve or double gene expression), and functional analyses were performed with Ingenuity Pathway Analysis (IPA, Qiagen).

Cell lines and qPCR Prostate cancer cells lines, LNCaP, PC3, and DU145, were obtained from the ATCC (15). All cell lines were maintained at 37 C, 5% CO2 in RPMI medium supplied with 5% FCS and penicillin/streptomycin (Lonza). Authentication was most recently performed using STR analysis in the year 2014 and Mycoplasma testing was done on a yearly basis with the MycoAlert Mycoplasma Detection Kit (Lonza). Cells were pas- saged for a maximum of 6 months for the experiments done in this study after thawing. After 24-hour culture in serum-deplet- ed DMEM medium without phenol-red (11039–021, Lonza), cells were stimulated in 4-fold with 100 mmol/L L-Kynurenine (K8625, Merck) in the same medium. Subsequently, RNA was isolated using the RNeasy Kit (Qiagen, Venlo) according to manufacturer's protocol. Total RNA was reverse transcribed to cDNA using oligo dT12 primers and M-MLV reverse tran- scriptase (Invitrogen). Real-time PCR reactions contained 5 mL 20 diluted cDNA in Taqman Universal PCR Master Mix, AmpliTaq, and 1 FAM primer/probe mix (CDH1,EntrezGene ID 999; CDH2, Gene ID 1000; and IDO1, Gene ID 3620) in a total of 25 mL. Amplified products were quantified relative to the geomean of hydroxymethylbilane synthase (HMBS, HS00609293_G1) and glyceraldehyde-3-phosphate dehydro- genase (GAPDH, HS02758991_G1; Applied Biosystems). All Ct values were quantified using the 7500 Real-Time PCR system (Applied Biosystems) with a threshold at 0.2 within the expo- nential curve.

Statistical analysis All statistics were performed using SPSS 20 (SPSS) or GraphPad Prism (GraphPad Software, Inc). Data were tested for normality with the Shapiro–Wilk test. A Mann–Whitney U test was per- þ þ formed to compare the number of CD4 and CD8 cells in immunohistochemically stained slices and gene expression in prostate cancer cell lines (PC3, LNCaP, and DU145). A two-sided P < 0.05 was considered signiﬁcant.

Results N-cadherin expression is associated with altered immunoregulatory pathways RNA sequencing of paired N-cadherin–positive and negative prostate cancer regions revealed 871 differentially expressed genes (P 0.05; Supplementary Table S4). Ingenuity Pathway Analysis (IPA) of genes with at least 2-fold change (1 log2FC 1) in expression (n ¼ 668) revealed that the majority of genes were Figure 1. associated with cancer (86.5%), organismal injuries/abnormali- Heatmap of differential expression of inflammatory response genes in ties (87.0%), growth and proliferation (40.4%), cell movement N-cadherin–positive clinical prostate cancer. Heatmap diagram of differentially expressed, inflammatory response genes in N-cadherin–positive which were centralized around the median and clustered. Columns represent versus –negative clinical prostate cancer samples. RNA was isolated from hierarchical clustered prostate cancer samples and N-cadherin–positive N-cadherin–positive (Ncad pos) and –negative (Ncad neg) microdissected samples are shown (bold italic). Each row represents an inflammatory response

tumor areas. Gene expression data were obtained using the Illumina HiSeq2500 gene, while expression levels are shown by log2-fold change and colors platform, TopHat, and edgeR. Differentially expressed genes with a P < 0.05 and indicate up- (red) or downregulation (green). The heatmap demonstrates 1 1 were analyzed with IPA. The diagram represents that inflammatory response genes are differentially expressed in 73 differentially expressed inflammatory response genes identified by IPA, N-cadherin–positive versus –negative prostate cancer samples.

www.aacrjournals.org Cancer Res; 78(16) August 15, 2018 4673

Kolijn et al.

Figure 2. IDO1 and N-cadherin expression coincide in clinical prostate cancer. Fluorescent immunostaining in patient samples (n ¼ 8) demonstrates that N-cadherin (CDH2) is associated with IDO1 expression in prostate cancer (A) and high levels of L-Kynurenine (B). A, N-cadherin–negative prostate cancer areas are primarily negative for IDO1 (top). IDO1 is either expressed by N-cadherin–positive cells (arrowhead) or adjacent tumor cells (arrow) in N-cadherin–positive cancer areas (bottom). B, L-Kynurenine (Kyn) was expressed at medium to high levels in N-cadherin–positive prostate cancer cells in patient samples (n ¼ 4). N-cadherin– negative prostate cancer cells generally expressed low levels of L-Kynurenine (asterisk). Hoechst was used to visualize nuclei (blue). Original magniﬁcation, 63. Scale bar, 10 mm.

(24.7%), and morphology (22.6%; Supplementary Table S5). We In total, IPA identified 73 inflammatory response genes that were identified 25 differentially expressed genes associated with EMT differentially expressed in N-cadherin–negative versus -positive (as defined by IPA), including OB-cadherin (CDH11; log2FC ¼ samples (Fig. 1). In addition, the IPA upstream regulator feature 1.1; P ¼ 0.02), TGFB1 (log2FC ¼ 1.3; P ¼ 0.02) and Wnt family identified IFNg, TGFB1 and lipopolysaccharide as potential dri- member 5A (WNT5A; log2FC ¼ 1.8; P ¼ 0.01; Supplementary vers based on the observed gene expression changes in the dataset. Table S6). Overexpression of N-cadherin at mRNA level (CDH2; Immunosuppressive genes such as indoleamine 2,3-dioxygenase 1 log2FC ¼ 2.6; P ¼ 5.82E-04) confirmed the accuracy of LCM. We (IDO1; log2FC ¼ 5.1; P ¼ 2,98E-04), tryptophan-2,3-dioxygenase 2 did not detect significantly different expression of other EMT- (TDO2, log2FC ¼ 6.6, P ¼ 4,68E-03), cytotoxic T-lymphocyte-asso- associated genes, such as fibronectin (FN1) and vimentin (VIM). ciated protein 4 (CTLA4, log2FC ¼ 2.9, P ¼ 0.01), and T-cell In addition, transcription factors SNAIL, SLUG, TWIST, ZEB1, and immunoreceptor with Ig and ITIM domains (TIGIT; log2FC ¼ 4.9; ZEB2 that have all been associated with EMT induction in vitro P ¼ 1,59E-04) were overexpressed in N-cadherin–positive regions were not differentially expressed in N-cadherin–positive prostate (Supplementary Table S4). Both IDO1 and TDO2 are enzymes cancer regions (16–21). involved in the degradation of tryptophan into kynurenines, Various differentially expressed genes were linked to immune which drive T-cell differentiation toward an immunosuppressive cell trafficking (33.4%) and inflammatory response (17.4%), state through local depletion of tryptophan (22). Local immune such as CD28 signaling in T helper cells (15 genes), iCOS suppression can therefore act as a mechanism, whereby N-cad- signaling in T helper cells (14 genes), and IL4 signaling (12 genes). herin escapes an effective immune response. To confirm immune

4674 Cancer Res; 78(16) August 15, 2018 Cancer Research

EMT and Immune Evasion in Clinical Prostate Cancer

CDH2 CDH1 A B P = 0.03 2.5 P = 0.05 2.5 No Kyn 2.0 100 μmol/L Kyn 2.0

P = 0.03 Rao P = 0.05

Rao 1.5 1.5

Figure 3. 1.0 1.0 L-Kynurenine induced N-cadherin CDH1/HKG upregulation and E-cadherin CDH2/HKG 0.5 0.5 downregulation in prostate cancer cell lines. Stimulation (n ¼ 4) of three 0 0 prostate cancer cell lines with 100 PC3 PC3 mol/L L-Kynurenine (Kyn) induced m DU145 LNCaP DU145 LNCaP EMT marked by signiﬁcant upregulation of N-cadherin (CDH2)in PC3 (A) and downregulation of IDO1 2.5 E-cadherin (CDH1) expression in PC3, C DU145, and LNCaP (B). IDO1 was 2.0 downregulated upon L-Kynurenine stimulation (C). Gene expression was corrected for the expression of 1.5 housekeeping genes (HKG). Mean and P < 0.01 SD are shown. 1.0

IDO1/HKG Rao 0.5

PC3 DU145 LNCaP

evasion, we selected IDO1, which was nonlinearly overexpressed N-cadherin–positive prostate cancer areas harbor less cytotoxic in 6 of 8 samples, and investigated its expression and effects on the T cells and more regulatory T cells local immunoresponse. Because IDO1 affects differentiation and proliferation of immune cells, especially regulatory T cells, we quantified the þ þ N-cadherin–positive prostate cancer areas show enhanced number of CD4 T helper cells and CD8 cytotoxic T cells in expression and activity of IDO1 independent prostate cancer samples (n ¼ 10; Supplementary To validate IDO1 upregulation in N-cadherin–positive tumor Table S1; ref. 22). Because intraepithelial lymphocytes are critical regions, we performed fluorescent immunostainings on both to protect the integrity of the epithelial barrier and maintain þ þ sequenced (n ¼ 5) and independent prostate cancer samples immune control, we quantified the number of CD8 and CD4 (n ¼ 3; Fig. 2; Supplementary Table S1). In each sample, IDO1 T cells in the total area of cancer epithelium with adjacent stroma was expressed in N-cadherin–positive tumor regions, while ex- ("overall") as well as in cancer epithelium alone ("intraepithe- pression was rare in tumor areas without N-cadherin (Fig. 2A). lial") through IHC double stainings (Fig. 4A and B; refs. 23, 24). þ IDO1 was either expressed in malignant epithelial cells directly We found that the number of intraepithelial CD8 T cells was adjacent to N-cadherin–positive tumor cells, or coexpressed by significantly lower (Fig. 4C; P ¼ 0.04) in N-cadherin–positive the same cancer cells. In addition, we found that IDO1 metabolite (median 2.0; range 1–33) than negative regions (median 4.0; þ L-Kynurenine was elevated in N-cadherin–positive regions, which range 1–15). In contrast, the overall number of CD8 cells was indicates enzymatic activity of IDO1 (n ¼ 4; Fig. 2B). higher (Fig. 4D; P ¼ 0.02) in N-cadherin–positive regions (medi- To determine whether local accumulation of kynurenine is able an 11.5; range 1–53) than negative regions (median 7.9; range þ to induce N-cadherin expression, we investigated its effects on 1–37). The number of intraepithelial CD4 T cells was not mRNA expression of N-cadherin (CDH2), E-cadherin (CDH1), significantly different between N-cadherin–positive versus nega- þ and IDO (IDO1) in prostate cancer cell lines (Fig. 3). Stimulation tive regions (Fig. 4E; P ¼ 0.21). The overall number of CD4 T with 100 mmol/L L-Kynurenine led to 1.4 increase (P ¼ 0.05) of cells, however, was slightly higher in N-cadherin–positive tumor N-cadherin mRNA levels in the N-cadherin–positive cell line PC3 regions (median 13; range 3–79) as compared with N-cadherin (mean SD; 1.3 0.6 vs. 1.8 0.3), but did not initiate negative tumor areas (median 11; range 3–70; Fig. 4F; P ¼ 0.02). transcription in the N-cadherin–negative cell lines DU145 and Finally, immunofluorescent triple stainings with Hoechst, CD4, þ LNCaP (Fig. 3A). Furthermore, L-Kynurenine stimulation led to FOXP3, and CDH2 (Fig. 5A) indicated that the increase of CD4 þ þ downregulated expression of E-cadherin in PC3 (0.9 0.3 vs. T cells coincided with the presence of a CD4 /FOXP3 T regu- 0.6 0.1; P ¼ 0.05), DU145 (0.3 0.02 vs. 0.1 0.01; P ¼ 0.03), latory phenotype in N-cadherin–positive prostate cancer regions and LNCaP (2.0 0.4 vs. 0.8 0.3; P ¼ 0.03; Fig. 3B). Notably, (Fig. 5B), although the sample size was too small for statistical IDO1 expression was decreased upon L-Kynurenine in PC3 cell analysis (n ¼ 3). Together, these results indicate that expression of þ line (0.6 0.2 vs. 0.3 0.1; P < 0.01) and suggests a negative N-cadherin and IDO1 leads to an increased number of CD4 feedback loop, which might finally result in EMT reversal. IDO1 (regulatory) T cells with a concordant decrease of intraepithelial þ was not expressed in DU145 and LNCaP (Fig. 3C). CD8 cytotoxic T cells, and indicates immune evasion.

www.aacrjournals.org Cancer Res; 78(16) August 15, 2018 4675

Kolijn et al.

Discussion what mechanisms drive EMT in patients with cancer. In this EMT is a biological process involved in tumor invasion, study, we have shown that EMT, characterized by N-cadherin resistance to chemotherapy, and metastasis (25–27). While expression in ill-formed tumor glands, is related to local EMT has extensively been investigated in vitro,itisunclear enrichment of immunosuppressive molecules such as IDO1

Figure 4. N-cadherin–positive tumor areas harbor less intraepithelial cytotoxic CD8þ T cells. The number of CD8þ and CD4þ T cells was scored in N-cadherin–negative (Ncad)and N-cadherin–positive (Ncadþ) areas through IHC double staining. Red staining from alkaline phosphatase indicates CD8 (A)- or CD4 (B)- positive cells (arrows), whereas brown staining from DAB indicates expression of N-cadherin. Eight areas with a total size of 0.5 mm2 were scored per condition and per patient (n ¼ 10) for CD8þ (C and D) and CD4þ (E and F) cells in prostatic cancer epithelium (C and E) or overall (epithelium and adjacent stroma; D and F). Quantiﬁcation of CD8þ and CD4þ cells demonstrated that the number of intraepithelial cytotoxic CD8þ T cells (C), but not CD4þ cells (E), was decreased in N-cadherin– positive cancer epithelium (Ncadþ). The overall number of both CD8þ (D) and CD4þ (F) cells was higher in N-cadherin–positive areas.

4676 Cancer Res; 78(16) August 15, 2018 Cancer Research

EMT and Immune Evasion in Clinical Prostate Cancer

Figure 5. The number of T regulatory cells (CD4þ/FOXP3þ) is increased in N- cadherin–positive prostate cancer. Fluorescent immunostaining was performed to visualize and quantify the number of CD4þ/FoxP3þ regulatory T cells in N-cadherin (CDH2)–negative and –positive regions in patient samples (n ¼ 3). A, DAPI staining was used to identify nuclei (blue) combined with ﬂuorescently labeled CD4 (green), CDH2 (red), and FoxP3 (magenta). The composite image demonstrates a CD4þ and FOXP3þ regulatory T cell (arrow) in close proximity to an N-cadherin–positive prostate cancer cell. B, The number of T regulatory cells was scored. Matching symbols indicate paired samples. The total number of regulatory T cells was increased in N-cadherin–positive (Ncadþ) tumor regions compared with N-cadherin–negative regions (Ncad-) in all three patient samples. Original magniﬁcation, 63.

and immune cells such as regulatory T cells and cytotoxic T cells significantly affected in our study, this does not exclude an altered in clinical prostate cancer. At a mechanistic level, we demon- expression or activation of the protein. strated enhanced enzymatic activity of IDO1 in tissues led to Our findings are in line with previous studies and provide downregulated expression of E-cadherin and upregulated evidence that interaction between EMT and an immunosup- expression of N-cadherin in prostate cancer cell lines. The pressive response exists in patients with clinical prostate cancer þ þ concomitant enrichment of CD4 /FOXP3 regulatory T cells (28–32). While IDO1 activation was present in N-cadherin þ and reduction of intraepithelial CD8 cytotoxic T cells, indi- positive tumor areas and associated with a suppressive immune cates an intimate interaction between EMT as marked by profile, the primary EMT-initiating step in prostate cancer is not N-cadherin expression and local evasion of an effective anti- yet clear. Pathway analysis revealed TGFb and IFNg signaling as tumor CD8 T-cell response in prostate cancer. putative upstream regulatory pathways (see Supplementary Although the exact mechanisms of EMT and immune cell Table S5). TGFb canbesecretedbybothstromalandimmune response in patients with cancer are still largely unknown, recent cells,andisabletoinducebothEMTandIDO1expression studies indicate an interaction between both essential biological through activation of both Smad-dependent and -independent processes (28, 29). Breast cancer cells that undergo EMT are pathways (33, 34). Binding of TGFb to its receptor activates the associated with an immunosuppressive phenotype and were noncanonical nuclear factor-kB pathway and generates a pos- more resistant to immunotherapy than epithelial tumors (28). itive feedback loop that stimulates IDO1 expression in a phos- Inflammatory cytokine-induced EMT significantly upregulated phatidylinositol-3-OH kinase (PI3K) and Src homology region IDO1 in lung, breast, and liver cancer cell lines (30). Chen and 2 domain-containing phosphatase-1 (SHP-1)–dependent man- colleagues demonstrated that IDO-induced kynurenine expres- ner. In our dataset TGFB1, PTPN6 (SHP-1), and IDO1 were all sion led to activation of the aryl hydrocarbon receptor (AhR) significantly upregulated in N-cadherin–positive regions. In complex, resulting in E-cadherin degradation in breast cancer contrast, we did not find differential expression of Smad genes. (31). In hepatocellular carcinoma cell lines, EMT-associated N- This suggests that Smad-independent TGFb signaling plays a cadherin upregulated expression was abolished through silencing role in the accompanied local induction of both EMT and of the AhR (32). Although the expression of AhR mRNA was not immunosuppression in clinical prostate cancer.

www.aacrjournals.org Cancer Res; 78(16) August 15, 2018 4677

Kolijn et al.

The inflammatory cytokine IFNg has been reported to initiate sion. Finally, the role of other enzymes involved in tryptophan the start of negative feedback loops that result in enhanced degradation, such as IDO2 and TDO2, remains to be elucidated. expression of regulators of immune responses, such as IDO1, In conclusion, we found that N-cadherin expression in clinical PD-L1, and regulatory T cells in melanoma (35). Alongside this prostate cancer is enriched for enzymatic activity of IDO1 result- process, cancer cells may demonstrate enhanced activity of Wnt ing in local accumulation kynurenine. L-Kynurenine may induce and PI3K pathways, which in their turn, advocate further loss of or sustain an immunosuppressive microenvironment that is þ þ immune control and generally results in loss of numbers and characterized by increased regulatory FOXP3 /CD4 T cells and þ activity of CD8 effector T cells within tumor tissue (36, 37). The decreased intraepithelial cytotoxic CD8 immune cells. Together reduced number of CD8 effector T cells within cancer epithelium these findings reveal an important link between EMT and immune combined with the overall (epithelium and stroma combined) evasion in patient with prostate cancer. increased number of CD8 effector T cells, suggests an immune evasion mechanism deployed by tumors (5). This acknowledged þ Disclosure of Potential Conflicts of Interest immune evasion mechanism describes the exclusion of CD8 T No potential conflicts of interest were disclosed. cells from the tumor and thereby allows cancer cells to avoid immune destruction. In this study, there was no significant change in expression of Authors' Contributions transcription factors associated with EMT in vitro such as SNAIL, Conception and design: K. Kolijn, G.J.L.H. van Leenders – Development of methodology: K. Kolijn, G.W. Jenster, G.J.L.H. van Leenders SLUG, TWIST, ZEB1, and ZEB2 (16 21). A putative explanation Acquisition of data (provided animals, acquired and managed patients, is that these transcription factors have mostly been studied in cell provided facilities, etc.): K. Kolijn, E.I. Verhoef, G.J.L.H. van Leenders cultures models based on metastasized and androgen-indepen- Analysis and interpretation of data (e.g., statistical analysis, biostati- dent cells lines that represent end-stage prostate cancer. However, stics, computational analysis): K. Kolijn, M. Smid, R. Bottcher,€ R. Debets, in this study we investigated clinical hormone-na€ve prostate G.J.L.H. van Leenders fi Writing, review, and/or revision of the manuscript: K. Kolijn, E.I. Verhoef, adenocarcinoma that in ltrates and interacts with preexistent € fi M. Smid, R. Bottcher, G.W. Jenster, R. Debets, G.J.L.H. van Leenders prostate stromal tissue. Taken together, this paradoxical nding Administrative, technical, or material support (i.e., reporting or organizing could indicate that alternative mechanisms are involved in EMT in data, constructing databases): K. Kolijn, E.I. Verhoef early- and late-stage disease. Study supervision: G.J.L.H. van Leenders The strength of this study is the primary analysis of EMT in clinical prostate cancer specimens. Microdissection of morpho- Acknowledgments fi logically ill-de ned areas that expressed N-cadherin allowed a This research was sponsored by the Dutch Cancer Society (EMCR2011-5006 detailed study of biological events in specific tumor areas. While to K. Kolijn and E.I. Verhoef). use of whole tissue slides for molecular studies is less laborious, We thank A.M. Hoogland for instructions on laser capture microdissection specific local molecular biological events might be diluted and and the Erasmus MC Tissue Bank for storage and supply of frozen prostate tissue remain undiscovered. On the other hand, application of stringent samples. We kindly thank J.A. Stoop for his support on dual chromogenic IHC. morphologic and IHC criteria for LCM selection combined with The costs of publication of this article were defrayed in part by the payment of the relatively rare occurrence of EMT, hampered analysis on a large page charges. This article must therefore be hereby marked advertisement in number of tissue samples. In addition, we used N-cadherin accordance with 18 U.S.C. Section 1734 solely to indicate this fact. expression as IHC marker for EMT, which does not cover the full range of EMT in prostate cancer, namely those EMT-like processes Received December 4, 2017; revised April 16, 2018; accepted June 13, 2018; that downregulate epithelial markers without N-cadherin expres- published first June 19, 2018.

References 1. Epstein JI, Amin MB, Reuter VE, Humphrey PA. Contemporary Gleason 8. Nakajima S, Doi R, Toyoda E, Tsuji S, Wada M, Koizumi M, et al. N-cadherin grading of prostatic carcinoma: an update with discussion on practical expression and epithelial–mesenchymal transition in pancreatic carcino- issues to implement the 2014 international society of urological pathology ma. Clin Cancer Res 2004;10:4125–33. (ISUP) consensus conference on Gleason grading of prostatic carcinoma. 9. Tomita K, van Bokhoven A, van Leenders GJ, Ruijter ET, Jansen CF, Am J Surg Pathol 2017;41:e1–e7. Bussemakers MJ, et al. Cadherin switching in human prostate cancer 2. Kolijn K, Verhoef EI, van Leenders GJ. Morphological and immunohisto- progression. Cancer Res 2000;60:3650–4. chemical identiﬁcation of epithelial-to-mesenchymal transition in clinical 10. Gravdal K, Halvorsen OJ, Haukaas SA, Akslen LA. A switch from E-cadherin prostate cancer. Oncotarget 2015;6:24488–98. to N-cadherin expression indicates epithelial to mesenchymal transition 3. Thiery JP, Acloque H, Huang RY, Nieto MA. Epithelial–mesenchymal and is of strong and independent importance for the progress of prostate transitions in development and disease. Cell 2009;139:871–90. cancer. Clin Cancer Res 2007;13:7003–11. 4. Thiery JP, Sleeman JP. Complex networks orchestrate epithelial– 11. Jaggi M, Nazemi T, Abrahams NA, Baker JJ, Galich A, Smith LM, et al. mesenchymal transitions. Nat Rev Mol Cell Biol 2006;7:131–42. N-cadherin switching occurs in high Gleason grade prostate cancer. 5. Hegde PS, Karanikas V, Evers S. The where, the when, and the how of Prostate 2006;66:193–9. immune monitoring for cancer immunotherapies in the era of checkpoint 12 Kolijn K, van Leenders GJ. Comparison of RNA extraction kits and histo- inhibition. Clin Cancer Res 2016;22:1865–74. logical stains for laser capture microdissected prostate tissue. BMC Res 6. Hazan RB, Phillips GR, Qiao RF, Norton L, Aaronson SA. Exogenous Notes 2016;9:17 expression of N-cadherin in breast cancer cells induces cell migration, 13. Kim D, Pertea G, Trapnell C, Pimentel H, Kelley R, Salzberg SL. TopHat2: invasion, and metastasis. J Cell Biol 2000;148:779–90. accurate alignment of transcriptomes in the presence of insertions, dele- 7. Tanaka H, Kono E, Tran CP, Miyazaki H, Yamashiro J, Shimomura T, tions and gene fusions. Genome Biol 2013;14:R36. et al. Monoclonal antibody targeting of N-cadherin inhibits prostate 14. Robinson MD, McCarthy DJ, Smyth GK. edgeR: a bioconductor package for cancer growth, metastasis and castration resistance. Nat Med 2010;16: differential expression analysis of digital gene expression data. Bioinfor- 1414–20. matics 2010;26:139–40.

4678 Cancer Res; 78(16) August 15, 2018 Cancer Research

EMT and Immune Evasion in Clinical Prostate Cancer

15. Horoszewicz JS, Leong SS, Kawinski E, Karr JP, Rosenthal H, Chu TM, 26. Zhang P, Sun Y, Ma L. ZEB1: at the crossroads of epithelial–mesenchymal et al. LNCaP model of human prostatic carcinoma. Cancer Res 1983; transition, metastasis and therapy resistance. Cell Cycle 2015;14:481–7. 43:1809–18. 27. Nieto MA, Huang RY, Jackson RA, Thiery JP. Emt: 2016. Cell 2016; 16. Kwok WK, Ling MT, Lee TW, Lau TC, Zhou C, Zhang X, et al. Up-regulation 166:21–45. of TWIST in prostate cancer and its implication as a therapeutic target. 28. Dongre A, Rashidian M, Reinhardt F, Bagnato A, Keckesova Z, Ploegh HL, Cancer Res 2005;65:5153–62. et al. Epithelial-to-mesenchymal transition contributes to immunosup- 17. Yuen HF, Chua CW, Chan YP, Wong YC, Wang X, Chan KW. Significance of pression in breast carcinomas. Cancer Res 2017;77:3982–9. TWIST and E-cadherin expression in the metastatic progression of prostatic 29. Chen L, Gibbons DL, Goswami S, Cortez MA, Ahn YH, Byers LA, et al. cancer. Histopathology 2007;50:648–58. Metastasis is regulated via microRNA-200/ZEB1 axis control of tumour cell 18. Graham TR, Zhau HE, Odero-Marah VA, Osunkoya AO, Kimbro KS, PD-L1 expression and intratumoral immunosuppression. Nat Commun Tighiouart M, et al. Insulin-like growth factor-I-dependent up-regulation 2014;5:5241. of ZEB1 drives epithelial-to-mesenchymal transition in human prostate 30. Ricciardi M, Zanotto M, Malpeli G, Bassi G, Perbellini O, Chilosi M, et al. cancer cells. Cancer Res 2008;68:2479–88. Epithelial-to-mesenchymal transition (EMT) induced by inflammatory 19. Emadi Baygi M, Soheili ZS, Schmitz I, Sameie S, Schulz WA. Snail regulates priming elicits mesenchymal stromal cell-like immune-modulatory prop- cell survival and inhibits cellular senescence in human metastatic prostate erties in cancer cells. Br J Cancer 2015;112:1067–75. cancer cell lines. Cell Biol Toxicol 2010;26:553–67. 31. Chen JY, Li CF, Kuo CC, Tsai KK, Hou MF, Hung WC. Cancer/stroma 20. Xie Y, Liu S, Lu W, Yang Q, Williams KD, Binhazim AA, et al. Slug regulates interplay via cyclooxygenase-2 and indoleamine 2,3-dioxygenase pro- E-cadherin repression via p19Arf in prostate tumorigenesis. Mol Oncol motes breast cancer progression. Breast Cancer Res 2014;16:410. 2014;8:1355–64. 32. Song L, Guo L, Li Z. Molecular mechanisms of 3,304,40,5-pentachlorobi- 21. Esposito S, Russo MV, Airoldi I, Tupone MG, Sorrentino C, Barbarito G, phenyl-induced epithelial–mesenchymal transition in human hepatocel- et al. SNAI2/Slug gene is silenced in prostate cancer and regulates neuro- lular carcinoma cells. Toxicol Appl Pharmacol 2017;322:75–88. endocrine differentiation, metastasis-suppressor and pluripotency gene 33. Lamouille S, Xu J, Derynck R. Molecular mechanisms of epithelial– expression. Oncotarget 2015;6:17121–34. mesenchymal transition. Nat Rev Mol Cell Biol 2014;15:178–96. 22. Munn DH, Mellor AL. IDO in the tumor microenvironment: inflam- 34. Mbongue JC, Nicholas DA, Torrez TW, Kim NS, Firek AF, Langridge WH. mation, counter-regulation, and tolerance. Trends Immunol 2016;37: The role of indoleamine 2, 3-dioxygenase in immune suppression and 193–207. autoimmunity. Vaccines 2015;3:703–29. 23. Hayday A, Theodoridis E, Ramsburg E, Shires J. Intraepithelial lympho- 35. Spranger S, Spaapen RM, Zha Y, Williams J, Meng Y, Ha TT, et al. Up- cytes: exploring the third way in immunology. Nat Immunol 2001;2:997– regulation of PD-L1, IDO, and T(regs) in the melanoma tumor microen- 1003. vironment is driven by CD8(þ) T cells. Sci Transl Med 2013;5:200ra116. 24. Cheroutre H, Lambolez F, Mucida D. The light and dark sides of intestinal 36. Spranger S, Bao R, Gajewski TF. Melanoma-intrinsic beta-catenin signalling intraepithelial lymphocytes. Nat Rev Immunol 2011;11:445–56. prevents antitumour immunity. Nature 2015;523:231–5. 25. Zheng X, Carstens JL, Kim J, Scheible M, Kaye J, Sugimoto H, et al. 37. Peng W, McKenzie JA, Hwu P. Complementing T-cell function: an inhib- Epithelial-to-mesenchymal transition is dispensable for metastasis but itory role of the complement system in T-cell-mediated antitumor immu- induces chemoresistance in pancreatic cancer. Nature 2015;527:525–30. nity. Cancer Discov 2016;6:953–5.

www.aacrjournals.org Cancer Res; 78(16) August 15, 2018 4679

Epithelial−Mesenchymal Transition in Human Prostate Cancer Demonstrates Enhanced Immune Evasion Marked by IDO1 Expression

Kimberley Kolijn, Esther I. Verhoef, Marcel Smid, et al.

Cancer Res 2018;78:4671-4679. Published OnlineFirst June 19, 2018.

Updated version Access the most recent version of this article at: doi:10.1158/0008-5472.CAN-17-3752

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2018/06/19/0008-5472.CAN-17-3752.DC1

Cited articles This article cites 37 articles, 11 of which you can access for free at: http://cancerres.aacrjournals.org/content/78/16/4671.full#ref-list-1

Citing articles This article has been cited by 1 HighWire-hosted articles. Access the articles at: http://cancerres.aacrjournals.org/content/78/16/4671.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at Subscriptions [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/78/16/4671. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.