Global microRNA profiling using novel miRCURY LNA TM microarrays enables identification of tumors of unknown primary origin Rolf Søkilde * , Poul Erik Høiby * , Boye Schnack Nielsen * , Kim Bundvig Barken * , Alastair Hansen ** , Søren Møller * , Thomas Litman * * Exiqon A/S, Skelstedet 16, DK-2950 Vedbæk, Denmark. ** Department of Pathology, Herlev University Hospital, DK-2730 Herlev, Denmark. Introduction MicroRNAs (miRNAs) constitute a recently discovered class of tissue specific, small, non-coding RNAs, which regulate the expression of genes involved in many biological processes, including development, differentiation, apoptosis and carcinogenesis. Around 5% of all newly diagnosed metastatic cancers are carcinoma of unknown primary (CUP, Figure 1) origin, where the site of the primary tumor cannot be determined, despite the use of advanced immunohistochemical or radiological techniques. Because effective cancer treatment depends on early identification of the primary tumor, CUP patients have poor prognosis with a median survival of 3-6 months and a 1-year survival rate of less than 25%. By applying a novel microarray platform based on locked nucleic acid (LNA TM ) modified detection probes, which enable highly sensitive and specific detection of miRNAs, we are able to identify tumor- specific microRNA signatures for each cancer histology. Our vision is to identify the primary tumor in all cancer patients, without delay, in order to offer the most relevant treatment. As part of this effort we are developing a classifier tool based on miRNA profiles for diagnosis of CUP, enabling individualized treatment of this otherwise intractable group of cancers. Figure 1 Methods Tissue More than 500 tumor and normal adjacent tissue samples were collected from both fresh frozen and formalin fixed, paraffin embedded sections. Table 1 lists 18 tissues that are included in our analysis and which represent the most common tissues of origin for CUP. Table 1 RNA isolation Total RNA was isolated by guanidinum isothiocyanate/phenol: chloroform extraction. Microarray profiling One microgram total RNA was analyzed for miRNA expression on miRCURY LNA™ Discovery arrays containing T m normalized capture probes for 2090 miRNAs, including 110 human miRPlus sequences not yet annotated in miRBase, +200 miRNAs discovered by 454 high throughput sequencing, as well as the corresponding pre- miRNAs (Figure 2). All hybridizations were made against a common reference pool. Data analysis A miRNA expression database was established for identification of miRNAs with high discriminatory power between tumors. Both unsupervised hierarchical clustering and supervised analysis was performed to generate a multiclass classifier. Properties of the LNA™ capture probes and microarrays Figure 3 Using LNA™, the T m is increased significantly and the T m range is narrowed significantly, com-pared to DNA probes. This results in increased stringency and optimal hybridization conditions for the LNA™ capture probes. Figure 4 Figure 5. It is clear that Supplier A and B have difficulties in designing optimal probes spanning the entire region of GC%. In contrast, the miRCURY LNA™ Array offers optimal performance of all probes using T m -normalized probes. The unmatched sensitivity and specificity is partly explained by the optimal design of the LNA™ capture probes. Method: Each of 557 synthetic microRNAs were used at a concentration of 50 amol (~500 copies per cell if using 1 μg total RNA). Table 2 Results Figure 6 Figure 7. Case story: Identification of the origin of a primary cancer, which has metastasized to a regional lymph node. A biopsy from a lymph node metastasis was profiled on the miRCURY LNA™ microRNA Array. Based on a simple cluster analysis, the lymph node metastasis miRNA profile is shown to belong to the colorectal miRNA cluster and is thus, identified as a colorectal cancer (CRC). This patient did have a primary tumor identified in the colon, from which we also received a sample, as well as from the rectum and from a normal lymph node (LN). All the colorectal samples, including the metastasis, cluster together, and clearly define the origin of the metastasis, namely colon. Therefore, in this case, one would – based on the microRNA profile alone – be able to direct the treatment against a colon cancer without having to perform a full body PET scan searching for the primary tumor. A colonoscopy would be sufficient, as was the case for this patient. Validation and Localization of miRNAs With LNA™ technology, tissue specific markers can by validated with in Situ hybridization (ISH). This can potentially increase the strength of CUP classification. Below is shown an ISH experiment, which allows localization of specific miRNAs. Figure 8 Figure 8: MicroRNA in Situ hybridization in formalin fixed, paraffin embedded tissue sections. A shows staining with a nucleus specific probe, while B shows staining of a microRNA with an LNA™ probe. The normal colon tissue has intact villus structure while the cancer tissue is less differentiated. Upregulation of the microRNA in the cancer tissue is seen compared to the normal tissue. The staining pattern suggests that this particular microRNA is located in the fibroblast like cells adjacent to the tumor cells. Little background staining is observed in the normal tissue. Conclusions • We have generated a miRNA based classifier, which can identify the origin of the primary tumor in CUP patients with metastatic disease. • The classifier is based on a limited number of microRNAs and may be improved by including spatial information on miR expression. • To expand the classifier to include histological subtypes, we are currently sampling Exiqon’s tumorbank of +150.000 specimens. Breast Bladder/Ureter Stomach Genitals Bowel Kidney/Adrenals Liver/Bile ducts Other Pancreas Lung Not Found Primary site found at autopsy Figure 1. Primary cancer site found at autopsy. The chart depicts the relative proportion of sites of origin found at autopsy in 884 CUP patients in 12 different follow-up studies (from Pentheroudakis et al. Oncologist 2007; 12: 418-425). Adaptor Total RNA 1 μg miRNA 5’- -3’ Hybridize Scan Enzyme F Acknowledgments: We wish to thank Stine Jørgensen and Gitte Friis for excellent technical assistance. Poster: #5038 - Job: 913866 Figure 2 Figure 6. The bar diagrams illustrate the differential expression of 7 different microRNAs (the miR numbers are arbitrary) in the 18 tissues. The y-axis is the log-ratio between the tissue microRNA expression and the average expression across all tissues Figure 2. Procedure for miRNA profiling with miRCURY LNA TM -based microarrays. 1. Prepare total-RNA 2. Label RNA with Hy5 / Hy3 dyes 3. Hybridize overnight 4. Scan and analyze Differential expression of 7 miRNAs across 18 tissues 6 5 4 3 2 1 0 -1 -2 -3 Adrenal Bladder Breast Cervix Colon Esophagus Liver Kidney Ovary Lung Pancreas Prostate Rectal Sm intestine Spleen Stomach Testis Uterus miR-1 5 4 3 2 1 0 -1 -2 -3 -4 Adrenal Bladder Breast Cervix Colon Esophagus Liver Kidney Ovary Lung Pancreas Prostate Rectal Sm intestine Spleen Stomach Testis Uterus miR-2 4 3 2 1 0 -1 -2 -3 -4 Adrenal Bladder Breast Cervix Colon Esophagus Liver Kidney Ovary Lung Pancreas Prostate Rectal Sm intestine Spleen Stomach Testis Uterus miR-3 3 2 1 0 -1 -2 Adrenal Bladder Breast Cervix Colon Esophagus Liver Kidney Ovary Lung Pancreas Prostate Rectal Sm intestine Spleen Stomach Testis Uterus miR-4 3 2 1 0 -1 -2 Adrenal Bladder Breast Cervix Colon Esophagus Liver Kidney Ovary Lung Pancreas Prostate Rectal Sm intestine Spleen Stomach Testis Uterus miR-5 3 2 1 -1 -2 0 -3 -4 -5 -6 Adrenal Bladder Breast Cervix Colon Esophagus Liver Kidney Ovary Lung Pancreas Prostate Rectal Sm intestine Spleen Stomach Testis Uterus miR-6 6 4 3 5 2 1 0 -1 Adrenal Bladder Breast Cervix Colon Esophagus Liver Kidney Ovary Lung Pancreas Prostate Rectal Sm intestine Spleen Stomach Testis Uterus miR-7 Optimally designed LNA™ capture probes result in unmatched specificity 0 10 20 30 40 50 60 70 50 55 60 65 70 75 80 Number of DNA capture probes 0 50 100 150 200 250 300 350 Number of LNA™ capture probes Temperature LNA™ capture probes DNA capture probes Figure 4. The miRCURY LNA™ enhanced microarray is extremely sensitive 100 90 80 70 60 50 40 30 20 10 0 0,1 1 10 100 1000 1 amol corresponds to ~10 copies per cell if using 1 μg total RNA 50 amol corresponds to ~500 copies per cell if using 1 μg total RNA %Detected Detection limit (attomole) Exiqon Supplier A Supplier B Exiqon’s miCURY LNA™ Arrays have have unmatched sensitivity for all microRNAs 100 90 80 70 60 50 40 30 20 10 0 <30 30-35 35-40 40-45 45-50 50-55 55-60 60-65 65-70 70< GC% % miRs Exiqon % miRs 100 90 80 70 60 50 40 30 20 10 0 <30 30-35 35-40 40-45 45-50 50-55 55-60 60-65 65-70 70< GC% No signal Signal Supplier A 100 90 80 70 60 50 40 30 20 10 0 <30 30-35 35-40 40-45 45-50 50-55 55-60 60-65 65-70 70< GC% % miRs Supplier B Capture probes not working Working Capture probes GC% <30 30-35 35-40 40-45 45-50 50-55 55-60 60-65 65-70 70< # microRNAs 15 40 48 91 92 89 39 28 16 12 Figure 5 Figure 7 Normal colon profile CRC: Colorectal cancer LNM: Lymph node metastasis Placenta Heart Adrenal Lymph node Proximal colon Colon Rectum CRC LNM CRC Kidney Breast Lung Cervix Bladder Pancreas Duodenum Small intestine Metastasis from lymph node Primary tumor profile Step 1 Metastasis identified in the lymph Step 2 Biopsy taken: miRNA profile conducted Step 3 Profile matched normal colon tissue Step 4 Primary tumor identified in colon Step 5 Focused and optimized treatment Normal colon Colon cancer A A’ B B’ miR probe Nuclear probe Exiqon’s miRCURY LNA™ Arrays have unmatched sensitivity Tissue Adrenal Esophagus Ovary Small intestine Bladder Gallbladder Pancreas Stomach Breast Kidney Prostate Testis Cervix Liver Rectum Uterus Colon Lung Experimental set-up: A dilution series of 557 synthetic microRNAs was hybridized, and for each array the percentage of capture probes detecting microRNAs was plotted. At 50 amol only 40% of the probe of Supplier B and 66% of Supplier A detect their targets. Exiqon capture probes detect 96% of their targets. All arrays were processed according to the manufacturers’ protocols. Figure 3. The miRCURY LNA™ enhanced microarray is T m normalized ensuring all miRNA targets hybridize to the array with equal affinity under high stringency hybridization conditions.