Hets 2021

Scientific Publications of Arrow Project Participants to mark 15 years since its establishment 2007-2021



Scientific Publications of Arrow Project Participants to mark 15 years since its establishment 2007-2021 ‫פרויקט ח״ץ‬



Dear Mentors and Students The Arrow Project for Young Researchers at Sheba Medical Center was established 15 years ago by Prof. Anat Achiron as a groundbreaking program for integrating medical students’ studies with research conducted at Sheba Medical Center. Over the years, the program has expanded its activities to medical students from all faculties and programs in the country, to nursing students and outstanding high school students studying health related science and the Magshimim program (Cyber unit-8200 Israel Defence Force). Today, more than fifty medical and nursing students participate in the program, about twenty high school students with dozens of researchers from Sheba Medical Center. The research carried out as part of the Arrow Project is groundbreaking and deals with a very wide range of topics. These include clinical, epidemiological research, basic sciences, health policy, medical technologies, data science and artificial intelligence. The participants in the project receive training and skills that help them continue on their professional path and enable them to be at the forefront of researchers and doctors in Israel and around the world. The project is a combination of an experienced generation of Sheba doctors and researchers who lead the research and are mentors to a young and talented generation of students. This multidisciplinary approach of intergenerational integration of researchers, and many diverse research topics has enormous scientific potential. The Arrow Project is a scientific-educational project but also an important social project that involves all groups and sectors in the State of Israel. This booklet, consists of 96 studies published as part of the activities of the Arrow Project, expresses the important scientific contribution of the program to the world of medicine and science in the State of Israel and around the world. I would like to thank the management of Sheba Medical Center for its unprecedented support, and all the participants in the project during its 15 years of activity for their work, efforts and contribution to science and society in Israel. Prof. Eldad Katorza MD, MSc, MBA Director, Arrow Project Director, Gertner Institute Sheba Medical Center, Tel-Hashomer

contents‎ No. Published Written by Article Name Published in page 1 2007 Anat Achiron, Michael Gurevich, Y Snir, Zinc-ion binding and cytokine activity Clinical and experimental 11 E Segal, M Mandel regulation pathways predicts outcome in immunology relapsing-remitting multiple sclerosis 2 2009 Michael Gurevich, Tamir Tuller, Prediction of acute multiple sclerosis BMC Medical genomics 21 Udi Rubinstein, Rotem Or-Bach, Anat Achiron relapses by transcription levels of peripheral blood cells 3 2010 Michael Gurevich, T Gritzman, Rotem Orbach, Laquinimod suppress antigen Journal of 37 T Tuller, A Feldman, Anat Achiron presentation in relapsing-remitting neuroimmunology multiple sclerosis: In-vitro high- throughput gene expression study. 4 2011 ‫ ענת אחירון‬,‫גלעד וינדר‬ ‫טרשת נפוצה טבה‬ ‫הרפואה‬ 47 Merav Lidar, Hagith Yonath, Naama Shechter, Incomplete response to colchicine in Autoimmunity reviews 51 5 2012 Fabienne Sikron, Siegal Sadetzki, M694V homozygote FMF patients Pnina Langevitz, Avi Livneh, Elon Pras Alon Skaat; Ifat Sher; Andrew Kolker; Pupillometer-based objective chromatic Investigative perimetry in normal eyes and patients Ophthalmology & Visual 6 2013 Sivan Elyasiv; Elkana Rosenfeld; with retinal photoreceptor dystrophies Science 57 Mohamad Mhajna; Shlomo Melamed; Michael Belkin; Ygal Rotenstreich 7 2013 Shaye Kivity, Aviva Katzav, Maria Teresa Arango, 16/6-idiotype expressing antibodies BMC Medicine 69 Moran Landau-Rabi, Yaron Zafrir, induce brain inflammation and cognitive Nancy Agmon-Levin, Miri Blank, Juan-Manuel impairment in mice: the mosaic of central Anaya, Edna Mozes, Joab Chapman, nervous system involvement in lupus Yehuda Shoenfeld 8 2013 Asaf Achiron, Joab Chapman, Sigal Tal, Superior temporal gyrus thickness Brain Structure & 79 Eran Bercovich, Harari Gil, Anat Achiron correlates with cognitive performance in Function multiple sclerosis Nicola Maggio, Zeev Itsekson, Dan Dominissini, Thrombin regulation of synaptic 9 2013 Ilan Blatt, Ninette Amariglio, Gideon Rechavi, plasticity: implications for physiology and Experimental Neurology 89 David Tanne, Joab Chapman pathology Yoav Yinon, Eldad Katorza, D I Nassie, Late diagnosis of fetal central nervous Prenatal Diagnosis 101 10 2013 Elad Ben-Meir, L Gindes, Chen Hoffmann, system anomalies following a normal second trimester anatomy scan Shlomo Lipitz, Reuven Achiron, Batia Weisz 11 2013 Nir Peled, Leor Zach, Ori Liran, Maya Ilouze, Effective crizotinib schedule for brain Case Reports 109 Paul A Bunn Jr, Fred R Hirsch metastases in ALK rearrangement metastatic non-small-cell lung cancer 12 2014 Rotem Orbach, Michael Gurevich, Anat Achiron Interleukin-12p40 in the spinal fluid as a Multiple Sclerosis Journal 113 biomarker for clinically isolated syndrome Yoav Yinon, Elad Ben Meir, Circulating angiogenic factors in American Journal of 123 13 2014 Alexandra Berezowsky, Boaz Weisz, Eyal Schiff, monochorionic twin pregnancies Obstetrics & Gynecology complicated by twin-to-twin transfusion Shali Mazaki-Tovi, Shlomo Lipitz syndrome and selective intrauterine growth restriction 14 2014 Yael Levy-Shraga, Inbal Gazit, Pituitary function in children following 131 Dalit Modan-Moses, Orit Pinhas-Hamiel infectious diseases of the central nervous Pituitary system Chen Hoffmann, Boaz Weisz, Shlomo Lipitz, Regional apparent diffusion coefficient Neuroradiology 139 15 2014 Gal Yaniv, Eldad Katorza, Dafi Bergman, values in 3rd trimester fetal brain Anat Biegon Yinon Gilboa, Zvi Kivilevitch, Meri Oren, Anogenital distance in male and female Prenatal Diagnosis 147 16 2014 Yisca Pazit Cohen, Eldad Katorza, fetuses at 20 to 35 weeks of gestation: centile charts and reference ranges Reuven Achiron, Shmuel Stienlauf, Eyal Meltzer, Daniel Kurnik, Potential drug interactions in travelers Travel Medicine and 155 17 2014 Eyal Leshem, Eran Kopel, Bianca Streltsin, with chronic illnesses: a large Infectious disease retrospective cohort study Eli Schwartz

No. Published Written by Article Name Published in page Placenta 163 Yoav Yinon, Elad Ben Meir, L. Margolis, Low molecular weight heparin therapy 18 2014 Shlomo Lipitz, E. Schiff, S. Mazaki-Tovi, during pregnancy is associated with elevated circulatory levels of placental M.J. Simchen growth factor 19 2014 ‫ ענת אחירון‬,‫סער אניס‬ :‫טרשת נפוצה במהלך התקפי‬ ‫הרפואה‬ 169 ‫תרופות הדור החדש‬ 175 Dana Ben-Ami S hor, Tomer Bashi, 20 2015 Jordan Lachnish, Mati Fridkin,Giorgia Bizzaro, Phosphorylcholine-tuftsin compound Journal of Autoimmunity prevents development of dextransulfate- Iris Barshak, Miri Blank, Yehuda Shoenfeld sodium-salt induced murine colitis: implications for the treatment of human Yael Levy-Shraga, Keren Dallalzadeh, inflammatory bowel disease 21 2015 Keren Stern, Gideon Paret and Orit The many etiologies of neonatal Pediatric Emergency Care 183 Pinhas-Hamiel hypocalcemic seizures 22 2015 ‫ ענת אחירון‬,‫ גד סגל‬,‫ רותם אורבך‬,‫אסף אחירון‬ ‫ חינוך למצוינות רפואית‬- ‫פרויקט ח\"ץ‬ ‫הרפואה‬ 189 ‫ תל השומר‬,‫במרכז הרפואי שיבא‬ 23 2015 Michael Gurevich, Gadi Miron, Anat Achiron Optimizing multiple sclerosis diagnosis: Annuals of clinical and 193 gene expression and genomic association translational neurology 24 2015 Achiron Asaf, Stone Evan, Achiron Anat Injury to white matter tracts in relapsing- NeuroImage Clinical 201 remitting multiple sclerosis: A possible therapeutic window within the first 5 years from onset using diffusion-tensor imaging tract-based spatial statistics Roee Ber, Omer Bar-Yosef, Normal fetal posterior fossa in MR American Journal of 209 25 2015 Chen Hoffmann, David Shashar, imaging: new biometric data and possible Neuroradiology clinical significance Reuven Achiron and Eldad Katorza 26 2015 Ronen Shavit, Maya Ilouze, Tali Feinberg, Mitochondrial induction as a potential Cellular oncology 219 Yaacov Richard Lawrence, Yossi Tzur, Nir Peled radio-sensitizer in lung cancer cells - a (Dordrecht) short report Irina Lojkin, Tami Rubinek, Sandra Orsulic, Reduced expression and growth 227 27 2015 Omer Schwarzmann, Beth Y Karlan, inhibitory activity of the aging suppressor Cancer Letters klotho in epithelial ovarian cancer Shikha Bose, Ido Wolf Michael Gurevich, Gadi Miron, Transcriptional response to interferon BMC Neurology 237 28 2015 Rina Zilkha Falb, David Magalashvili, beta-1a treatment in patients with secondary progressive multiple sclerosis Mark Dolev, Yael Stern, Anat Achiron Dorit Gamus, Saralee Glasser, Elisheva Langner, Journal of Back and Musculoskeletal 29 2016 Aliza Beth-Hakimian, Israel Caspi, Narin Carmel, Psychometric properties of the Hebrew Rehabilitation 247 Itzhak Siev-Ner, Hagai Amir, A Ziv, M Papa, version of the Oswestry Disability Index Liat Lerner-Geva Nisreen Shehada, John C Cancilla, Jose S Torrecilla, Enrique S Pariente, 30 2016 Gerald Brönstrup, Silke Christiansen, Silicon Nanowire Sensors Enable American Chemical 257 Douglas W Johnson, Marcis Leja, Diagnosis of Patients via Exhaled Breath Society - Nano Michael P A Davies, Ori Liran, Nir Peled, Hossam Haick Ron Chibel, Ifat Sher, Daniel Ben Ner, Chromatic Multifocal Pupillometer for American Academy of 269 Mohamad O Mhajna, Asaf Achiron, Objective Perimetry and Diagnosis of Ophthalmology 31 2016 Soad Hajyahia, Alon Skaat, Yakir Berchenko, Patients with Retinitis Pigmentosa Bernice Oberman, Ofra Kalter-Leibovici, Laurence Freedman, Ygal Rotenstreich Anat Achiron, Gadi Miron, Rina Zilkha-Falb, Host cell virus entry mechanisms Journal of NeuroVirology 285 32 2016 David Magalashvili, Mark Dolev, Yael Stern, enhance anti-JCV-antibody switch in natalizumab-treated multiple sclerosis Michael Gurevich patients 33 2017 Asaf Polat, Stewart Barlow, Roee Ber, Volumetric MRI study of the intrauterine European Radiology 297 Reuven Achiron, Eldad Katorza growth restriction fetal brain

No. Published Written by Article Name Published in page Smadar Gertel, Gidi Karmon, Sivan Vainer, Immunomodulation of RA Patients' Mediators of 307 34 2017 Ora Shovman, Martin Cornillet, Guy Serre, PBMC with a Multiepitope Peptide inflammation Derived from Citrullinated Autoantigens Yehuda Shoenfeld, Howard Amital Dekel Shlomi, Manal Abud, Ori Liran, Jair Bar, Detection of Lung Cancer and EGFR Journal of Thoracic 317 35 2017 Naomi Gai-Mor, Maya Ilouze, Amir Onn, Mutation by Electronic Nose System Oncology Alon Ben-Nun, Hossam Haick, Nir Peled Merav Lidar, Yael Brantz, Yael Shinar, A high and equal prevalence of the Clinical and Experimental 327 Haike Reznik-Wolf, Avi Livneh, Ilan Ben Zvi, Q703K variant in NLRP3 patients Rheumatology 36 2017 Rinat Cohen, Yaakov Berkun, Philip J Hashkes, with autoinflammatory symptoms and Hagit Peleg, Aharon Kessel, Gleb Slobodin, ethnically matched controls Michael Rozenbaum, Ofra Goldzweig, Elon Pras 37 2017 Roee Ber, D Hoffman, Chen Hoffman, Volume of Structures in the Fetal Brain American Journal of 333 Asaf Polat, E Derazne, A Mayer, Eldad Katorza Measured with a New Semiautomated Neuroradiology Method 38 2017 Smadar Gertel, Hussein Mahagna, Gidi Karmon, Tofacitinib attenuates arthritis Clinical Immunology 341 Abdulla Watad, Howard Amital manifestations and reduces the pathogenic CD4 T cells in adjuvant arthritis rats 39 2017 Gadi Miron, Michael Gurevich, S Baum, Psoriasis comorbidity affects multiple Journal of European Anat Achiron, A Barzilai sclerosis neurological progression: a Academy of Dermatology 347 retrospective case - control analysis and Venereology Ricardo Tarrasch, Narin N Carmel-Neiderman, The Effect of Reflexology on the Pain- Journal of Alternative 355 40 2018 Sarah Ben-Ami, Bella Kaufman, Raphi Pfeffer, Insomnia-Fatigue Disturbance Cluster of and Complementary Breast Cancer Patients During Adjuvant Medicine Merav Ben-David, Dorit Gamus Radiation Therapy Rachel Shapira, Nadia Ilyayev, Ruben Attali, Early Detection and Staging of Colorectal Journal of Cancer Science 363 Gal Westrich, David Halle, Chen Speter, Cancer Using a Panel of Micro RNAs and Therapy Amalia V. Stavropoulos, Marina Roistacher, 41 2018 Vera Pavlov, Ronit Grinbaum, Mladjan Protic, Ali O. Gure, Anton J. Bilchik, Alexander Stojadinovic Stella Mitrani-Rosenbaum, Aviram Nissan 42 2018 Lior Orbach, Shay Menascu, Chen Hoffmann, Focal cortical thinning in patients with Neuroradiology 371 Shmuel Miron, Anat Achiron stable relapsing-remitting multiple sclerosis: cross-sectional-based novel estimation of gray matter kinetics Smadar Gertel, Gidi Karmon, Eszter Szarka, Anticitrullinated Protein Antibodies The Journal of 381 43 2018 Ora Shovman, Esther Houri-Levi, Edna Mozes, Induce Inflammatory Gene Expression rheumatology Profile in Peripheral Blood Cells from Yehuda Shoenfeld, Howard Amital CCP-positive Patients with RA Iris Morag, Orly Stern Levkovitz, Postnatal Growth Disadvantage of the Nutrients 393 44 2018 Maya Siman-Tov, Mor Frisch, Small for Gestational Age Preterm Twins Orit Pinhas-Hamiel, Tzipi Strauss Fingolimod-improved axonal and myelin 45 2018 Michael Gurevich, Roy Waknin, Evan Stone, Anat integrity of white matter tracts associated CNS Neuroscince & 405 Achiron with multiple sclerosis-related functional Therapeutics impairments 46 2018 Ran Harel, Maya Nulman, Zvi R Cohen, Nachshon Anterior cervical approach for the British Journal of 415 Neurosurgery Knoller treatment of axial or high thoracic levels Brain Lesion Load and Anatomic 47 2018 Shay Menascu, Carolina Legarda, Shmuel Miron, Distribution in Patients With Juvenile Journal of Child 421 Anat Achiron Clinically Isolated Syndrome Predicts Neurology Rapidly Advanced to Multiple Sclerosis Vered Yahalom, Nir Pillar, Yingying Zhao, Shirley Modan, Mingyan Fang, Lydia Yosephi, 48 2018 Orna Asher, Eilat Shinar, Gershon Celniker, Haike SMYD1 is the underlying gene for the European Journal of 429 Haematology Resnik-Wolf, Yael Brantz, AnWj-negative blood group phenotype Hagit Hauschner, Nurit Rosenberg, Le Cheng, Noam Shomron, Elon Pras Anat Achiron, Rina Zilkha-Falb, Anna Feldman, 49 2018 Maria Bovim, Onn Rosenblum, Polymerase-1 pathway activation in Autoimmunity Reviews 437 Ida Sarova-Pinhas, David Magalashvili, acute multiple sclerosis relapse Mark Dolev, Shay Menascu, Michael Gurevich

No. Published Written by Article Name Published in page 443 Soad Haj Yahia, Amit Hamburg, Ifat Sher, Effect of Stimulus Intensity and Visual Investigative Field Location on Rod- and Cone- Ophthalmology & Visual 453 50 2018 Daniel Ben Ner, Saeed Yassin, Ron Chibel, Mediated Pupil Response to Focal Light Science Michael Mimouni, Estela Derazne, Stimuli Pediatric Cardiology Michael Belkin, Ygal Rotenstreich MiRNA-208a as a Sensitive Early Biomarker for the Postoperative Course Keren Zloto, Tal Tirosh-Wagner, Yoav Bolkier, Following Congenital Heart Defect 51 2018 Omer Bar-Yosef, Amir Vardi, David Mishali, Surgery Yael Nevo-Caspi, Gidi Paret Daniel Ben Ner, Ifat Sher, Amit Hamburg, Chromatic pupilloperimetry for objective Clinical Ophthalmology 461 Mohamad O Mhajna, Ron Chibel, diagnosis of Best vitelliform macular 52 2019 Estela Derazne, Inbal Sharvit-Ginon, Eran Pras, dystrophy Hadas Newman, Jaime Levy, Samer Khateb, Dror Sharon, Ygal Rotenstreich Prenatal diagnosis for de novo mutations: Molecular Genetics & 473 Experience from a tertiary center over a Genomic Medicine 481 Ori Eyal, Michal Berkenstadt, Haike 10-year period 497 53 2019 Reznik-Wolf, Hana Poran, Tomer Ziv-Baran, 503 Diurnality, Type 2 Diabetes, and Journal of Biological Lior Greenbaum, Hagit Yonath, Elon Pras Depressive-Like Behavior Rhythms 513 Carmel Bilu, Paul Zimmet, Vicktoria 54 2019 Vishnevskia-Dai, Haim Einat, Galila Agam, Ehud Grossman, Noga Kronfeld-Schor 55 2019 Ran Harel, Maya Nulman, Nachshon Knoller Intraoperative imaging and navigation for Surgical Neurology C1-C2 posterior fusion International Carmel Bilu, Haim Einat, Katy Tal-Krivisky, Red white and blue - bright light effects Chronobiology 56 2019 Joseph Mizrahi, Vicktoria Vishnevskia-Dai, in a diurnal rodent model for seasonal international affective disorder Galila Agam, Noga Kronfeld-Schor Carmel Bilu, Haim Einat, Orly Barak, Linking type 2 diabetes mellitus, cardiac hypertrophy and depression in a diurnal 57 2019 Paul Zimmet, Vicktoria Vishnevskia-Dai, animal model Scientific Reports Amanda Govrin, Galila Agam, Noga Kronfeld-Schor Shalev Fried, Abraham Avigdor, Bella Bielorai, 58 2019 Amilia Meir, Michal J Besser, Jacob Schachter, Early and late hematologic toxicity Bone marrow 521 Avichai Shimoni, Arnon Nagler, Amos Toren, following CD19 CAR-T cells transplantation Elad Jacoby Doron Amsalem, Doron Gothelf, Reducing Stigma Toward Psychiatry The Primary Care 531 Alexandra Dorman, Yaron Goren, Oren Tene, Among Medical Students: Companion for CNS Assaf Shelef, Itai Horowitz, Liora Libman A Multicenter Controlled Reducing Disorders 59 2020 Dunsky, Stigma Toward Psychiatry Among Eldor Rogev, Efrat Hirsh Klein, Medical Students A Multicenter Ehud Mekori-Domachevsky, Tsvi Fischel, Controlled Trial Yechiel Levkovitz, Andres Martin, Raz Gross 60 2020 Doron Amsalem, Doron Gothelf, Omer Soul, Single-Day Simulation-Based Training 539 Alexandra Dorman, Amitai Ziv, Raz Gross Improves Communication and Psychiatric Frontiers in Psychiatry 547 Skills of Medical Students 553 561 Imri Amiel, Roi Anteby, Moti Cordoba, Experienced surgeons versus novice 61 2020 Shlomi Laufer, Chaya Shwaartz, Danny Rosin, surgery residents: Validating a novel knot Surgery tying simulator for vessel ligation Mordechai Gutman, Amitai Ziv, Roy Mashiach 62 2020 Ofir Zmira, Alex I Halpern, Lital Abraham, Efficacy and safety of alemtuzumab Acta Neurologica Belgica Anat Achiron treatment in a real-world cohort of patients with multiple sclerosis Ilan Volkov, Luciana Seguro, Elaine P Leon, Profiles of criteria and non-criteria 63 2020 László Kovács, Dirk Roggenbuck, anti-phospholipid autoantibodies are Auto - immunity Peter Schierack, Boris Gilburd, Andrea Doria, associated with clinical phenotypes of the Highlights Maria G Tektonidou, Nancy Agmon-Levin antiphospholipid syndrome Ariella Bar-Gil Shitrit, Ami Ben-Ya'acov, Matan Siterman, Matti Waterman, Ayal Hirsh, Doron Schwartz, Eran Zittan, Yehonatan Adler, Safety and effectiveness of ustekinumab for induction of remission in patients with 64 2020 Benjamin Koslowsky, Irit Avni-Biron, Crohn's disease: A multicenter Israeli United European 571 Yehuda Chowers, Yulia Ron, Eran Israeli, study Gastroenterology Journal Bella Ungar, Henit Yanai, Nitsan Maharshak, Shomron Ben-Horin, Rami Eliakim, Iris Dotan, Eran Goldin, Uri Kopylov

No. Published Written by Article Name Published in page Ifat Sher, Yisroel Tucker, Maya Gurevich, Chromatic Pupilloperimetry Measures 65 2020 Amit Hamburg, Ettel Bubis, Jonathan Kfir, Shlomit Correlate With Visual Acuity and Visual Transtational Vision 579 Science & Technology Zorani, Estela Derazne, Alon Skaat, Field Defects in Retinitis Pigmentosa Ygal Rotenstreich Patients Imri Amiel, Roi Anteby, Moti Cordoba, Feedback based simulator training American Journal of 593 66 2020 Shlomi Laufer, Chaya Shwaartz, Danny Rosin, reduces superfluous forces exerted by Surgery novice residents practicing knot tying for Mordechai Gutman, Amitai Ziv, Roy Mashiach vessel ligation Ramit Maoz-Segal, Tanya Levy, Soad Haj-Yahia, Combination therapy with omalizumab The World Allergy 599 67 2020 Irena Offengenden, Mona Iancovich-Kidon, and an immune-suppressive agent for Organization Journal resistant chronic spontaneous rrticaria - Nancy Agmon-Levin A real-life experience 68 2020 Leora Allen, Odelia Leon-Attia, Meirav Shaham, Autism risk linked to prematurity is more PloS One 609 Shahar Shefer, Lidia V Gabis accentuated in girls 69 2020 Raoul Orvieto, Baruch Feldman, Marine Wiesel, Is Day-4 morula biopsy a feasible PloS One 623 Hagit Shani, Adva Aizer alternative for preimplantation genetic testing? Oran Ben-Gal, Amit Benady, Sean Zadik, Using the loading response peak for Journal of Biomechanics 631 70 2020 Glen M Doniger, Michal Schnaider Beeri, defining gait cycle timing: A novel solution for the double-belt problem Meir Plotnik Baruch Feldman, Raoul Orvieto, Marine Weisel, Obstetric and Perinatal Outcomes Obstetrics & Gynecology 639 71 2020 Adva Aizer, Raanan Meyer, Jigal Haas, in Pregnancies Conceived After Preimplantation Genetic Testing for Michal Kirshenbaum Monogenetic Diseases 72 2020 Rina Zilkha-Falb, Tatyana Rachutin-Zalogin, RAM-589.555 favors neuroprotective Journal of 651 Lakota Cleaver, Michael Gurevich, Anat Achiron and anti-inflammatory profile of CNS- neuroinflammation resident glial cells in acute relapse EAE affected mice 73 2020 Ori Eyal, Yael Shinar, Mordechai Pras, Elon Pras Familial Mediterranean fever: Penetrance 669 of the p.[Met694Val];[Glu148Gln] and Human Mutation p.[Met694Val];[=] genotypes Elad Zvi, A. Shemer, S. toussia-Cohen, D. Zvi, 74 2020 Y. Bashan, L. Hirschfeld-dicker, Fetal Exposure to MR Imaging: Long- American Journal of 675 N. Oselka, M.-M. Amitai, O. Ezra, Term Neurodevelopmental Outcome Neuroradiology Omer Bar-Yosef and Eldad Katorza Binyamin B Neeman, Elad Maor, 75 2020 Israel M Barbash, Ilan Hai, Ori Vaturi, Outcomes of Patients Turned Down for The Journal of Invasive 681 Sagit Ben Zekry, Amit Segev, Micha Feinberg, Percutaneous Mitral Valve Repair Cardiology Victor Guetta, Paul Fefer 76 2020 Carmel Bilu, Haim Einat, Paul Zimmet, Beneficial effects of daytime high- 695 Vicktoria Vishnevskia-Dai, Noga Kronfeld-Schor intensity light exposure on daily rhythms, Scientific Reports metabolic state and affect Lee Cohen Ben-Meir, David Soriano, The Association Between Ultrachall in der Medizin 713 77 2020 Michal Zajicek, Vered Yulzari, Jerome Bouaziz, Gastrointestinal Symptoms and Transvaginal Ultrasound Findings in Marc Beer-Gabel, Vered H Eisenberg Women Referred for Endometriosis Evaluation: A Prospective Pilot Study Michal Gafner, Shalev Fried, Noa Gosher, Fetal Brain Biometry: Is there an European Journal of 723 78 2020 Danielle Jeddah, Eliel Kedar Sade, Eran Barzilay, Agreement among Ultrasound, MRI and Radiology the Measurements at Birth? Arnaldo Mayer, Eldad Katorza 79 Or Bercovich, Tal Tirosh-Wagner, Lior Goldberg, Immunomodulatory miRNAs as Potential Congenital Heart Disease 729 2020 Amir Vardi, David Mishali, Gideon Paret, Biomarkers for the Postoperative Course Following Surgery for the Repair of Yael Nevo-Caspi Congenital Heart Defects in Children 80 2021 Rachel Shemesh, Eedy Mezer, Publication modifiers of abstracts Eye 741 Tamara Wygnanski-Jaffe submitted to the American Association of Pediatric Ophthalmology and Strabismus Annual Meeting

No. Published Written by Article Name Published in page Yael Levy-Shraga, Ophir Megnazi, Trabecular Bone Score in Children and Journal of Clinical Adolescents With Inflammatory Bowel Densitometry: the 81 2021 Dalit Modan-Moses, Liana Tripto-Shkolnik, Diseases official journal of the 745 Noah Gruber, Yael Haberman, Dror S Shouval, International Society of Clinical Densitometry Batia Weiss Michael Peled, Tali H Bar-Lev, Efrosiniia Talalai, Mesencephalic astrocyte-derived neurotrophic factor is secreted from 82 2021 Haggar Zoë Aspitz, Inbal Daniel-Meshulam, interferon-γ-activated tumor cells PloS One 755 Jair Bar, Iris Kamer, Efrat Ofek, Adam Mor, through ER calcium depletion Amir Onn Noga C Minsky, Dafna Pachter, Galia Zacay, Managing Obesity in Lockdown: Survey Nutrients 767 83 2021 Naama Chishlevitz, Miriam Ben-Hamo, of Health Behaviors and Telemedicine Dana Weiner, Gabriella Segal-Lieberman Amit Benady, Sean Zadik, Oran Ben-Gal, Vision Affects Gait Speed but not Frontiers in 777 84 2021 Desiderio Cano Porras, Atalia Wenkert, Patterns of Muscle Activation During Bioengineering and Inclined Walking-A Virtual Reality Study Biotechnology Sharon Gilaie-Dotan, Meir Plotnik Keren Zloto, Liat Mor, Omer Bar-Yosef, MiRNA-124a: a Potential Biomarker for Journal of Cardiovascular 791 85 2021 Tal Tirosh-Wagner, Amir Vardi, David Mishali, Neurological Deficits Following Cardiac Translational Research Surgery in Pediatric Patients Gideon Paret, Yael Nevo-Caspi Liana Tripto-Shkolnik, Iris Vered, Bone Mineral Density of the 1/3 Radius Endocrine Practice 801 86 2021 Naama Peltz-Sinvani, David Kowal, Refines Osteoporosis Diagnosis, Correlates With Prevalent Fractures, and Inbal Goldshtein Enhances Fracture Risk Estimates Anat Achiron, Sapir Dreyer-Alster, Definitions of primary-progressive multiple sclerosis trajectories by rate of 87 2021 Michael Gurevich, Shay Menascu, clinical disability progression Multiple Sclerosis and 807 David Magalashvili, Mark Dolev, Yael Stern, related disorders Tomer Ziv-Baran Limor Marko, Asaf Shemer, Merav Lidar, Anakinra for colchicine refractory familial 88 2021 Chagai Grossman, Amit Druyan, Avi Livneh, Mediterranean fever: a cohort of 44 Rheumatology 815 Shaye Kivity patients 89 2021 Lee Rima Madi, Naama Fisch Shvalb, Central precocious puberty after MNJ Case Reports 823 Chen Sade Zaltz, Yael Levy-Shraga resection of a virilising adrenocortical oncocytic tumour Yael Levy-Shraga, Lee Rima Madi, Mor Shalev, Effectiveness of Metformin for Weight Journal of Child 829 90 2021 Kineret Mazor-Aronovitch, Reduction in Children and Adolescents and Adolescent Treated with Mixed Dopamine and Psychopharmacology Maya Schwartz-Lifshitz, Doron Gothelf Serotonin Receptor Antagonists: A Naturalistic Cohort Study 91 2021 Sapir Dreyer-Alster, Aviva Gal, Anat Achiron Optical Coherence Tomography Is Journal of Neuro- 835 Associated With Cognitive Impairment in Ophthalmology Multiple Sclerosis Ayelet Rishpon, Eli Sprecher, Stephen W Dusza, Morphological features of benign International Journal of 845 92 2021 Elana Kleinman, Sara Haupt, Harold Rabinovitz, pigmented ear lesions: a cross-sectional Dermatology study Alon Scope Amit Benady, Sean Zadik, Gabriel Zeilig, Gait Speed Modulations Are Proportional Sharon Gilaie-Dotan, Meir Plotnik 93 2021 to Grades of Virtual Visual Slopes-A Frontiers in Neurology 855 Virtual Reality Study Correlation between 2D and 3D 94 2021 Shalev Fried, Michal Gafner, D Jeddah, N Gosher, Fetal Brain MRI Biometry and American Journal of 869 D Hoffman, Roee Ber, A Mayer, Eldad Katorza Neurodevelopmental Outcomes in Neuroradiology Fetuses with Suspected Microcephaly and Macrocephaly Anat Achiron, Michael Gurevich, SARS-CoV-2 antibody dynamics and Clinical Microbiology and 877 95 2021 Rina Falb, Sapir Dreyer-Alster, Polina Sonis, B-cell memory response over time in Infection COVID-19 convalescent subjects Mathilda Mandel 96 2021 Tom Halevy, Meirav Nezer, Jorden Halevy, Twin discordance: a study of volumetric European Radiology 885 Tomer Ziv-Baran, Eran Barzilay, Eldad Katorza fetal brain MRI and neurodevelopmental outcome



‫‪Zinc-ion binding and cytokine activity regulation pathways‬‬ ‫‪predicts outcome in relapsing-remitting multiple sclerosis‬‬ ‫‪Clinical and experimental immunology | 2007‬‬ ‫הייתה לי זכות גדולה להקים ולהוביל את פרויקט חוקר צעיר(ח״ץ) המרכז‬ ‫מנחה‪ :‬פרופ' ענת אחירון‬ ‫הרפואי שיבא ובמסגרת זו עברתי חוויה מלמדת ומיוחדת של עשייה מדעית‬ ‫משותפת עם הסטודנטים והמנחים בפרויקט ‪.‬הקמתי את הפרויקט ב‪2007 -‬‬ ‫מייסדת פרויקט ח\"ץ‪ ,‬מנהלת המרכז‬ ‫בהשראתו של בני אסף‪ ,‬שהחל את לימודי הרפואה בפקולטה לרפואה‬ ‫לטרשת נפוצה ואחראית הקתדרה‬ ‫בתל אביב‪ .‬בסיום השנה הראשונה ללימודיו ביקש להשתלב במחקר‬ ‫למחלות אוטואימוניות אוניברסיטת ת\"א‬ ‫באחת המעבדות המובילות בפקולטה‪ ,‬ובמסגרת זו פעילותו המחקרית‬ ‫הייתה שטיפת צלחות פטרי‪ ...‬באותו זמן נעזרתי בו לצורי חד עם רופאי‬ ‫‪[email protected]‬‬ ‫וחוקרי המרכז בשיבא‪ .‬בהתאמה‪ ,‬פירסמתי בפקולטה פניה לסטודנטים‬ ‫המעוניינים להשתלב במחקר קליני בשיבא תוך כדי הלימודים‪ ,‬והסטודנטית‬ ‫הראשונה שהגיעה אלי ‪ -‬כיום דר׳ רותם אורבך רופאת ילדים בכירה‬ ‫בביח איכילוב‪ ,‬הייתה הסנונית הראשונה שהחלה את אביב המחקר של‬ ‫פרויקט ח״ץ ‪ .‬העבודה מול הסטודנטים הביאה לשדרוג של המחקרים שלי‬ ‫ועד היום הרעיונות החדשים שלהם‪ ,‬החשיבה היצירתית לעיתים קרובות‬ ‫מחוץ לקופסא‪ ,‬הסקרנות‪ ,‬והרצון לדעת‪ ,‬להבין ולעשות יותר טוב‪ ,‬גורמים‬ ‫לי להמשיך ולחנך עוד סטודנטים ולהיות מלאת גאווה כשאני רואה את‬ ‫הח\"צים שלי גדלים ופוגעים במטרה!‬ ‫יאיר שניר‬ ‫מנחה‪ :‬דר' מיכאל גורביץ‬ ‫אונ' תל אביב‬ ‫מנהל המעבדה הנוירואימונולוגית‪,‬‬ ‫השתתף כסטודנט בפרויקט ח״ץ‬ ‫המרכז לטרשת נפוצה‬ ‫בין השנים ‪2007-2009‬‬ ‫‪[email protected]‬‬ ‫‪[email protected]‬‬ ‫‪11‬‬

Clinical and Experimental Immunology ORIGINAL ARTICLE doi:10.1111/j.1365-2249.2007.03405.x Zinc-ion binding and cytokine activity regulation pathways predicts outcome in relapsing–remitting multiple sclerosis A. Achiron*†, M. Gurevich*†, Y. Snir*†, Summary E. Segal‡ and M. Mandel§ Multiple sclerosis (MS) is a demyelinating disease characterized by an unpre- *Multiple Sclerosis Center, †Neurogenomics Unit, dictable clinical course with intermittent relapses that lead over time to sig- and §Blood Bank Center, Sheba Medical Center, nificant neurological disability. Clinical and radiological variables are limited Tel-Hashomer, Sackler School of Medicine, in the ability to predict disease course. Peripheral blood genome scale analy- Tel-Aviv University, Israel, and ‡Department of ses were used to characterize MS patients with different disease types, but not Computer Science and Applied Mathematics, for prediction of outcome. Using complementary-DNA microarrays we Weizmann Institute of Science, Rehovot, Israel studied peripheral-blood gene expression patterns in 53 relapsing–remitting MS patients. Patients were classified into good, intermediate and poor clinical Accepted for publication 28 March 2003 outcome established after 2-year follow-up. A training set of 26 samples was Correspondence: A. Achiron MD, PhD, Multiple used to identify clinical outcome differentiating gene-expression signature. Sclerosis Center, Sheba Medical Center, Supervised learning and feature selection algorithms were applied to identify Tel-Hashomer, 52621, Israel. a predictive signature that was validated in an independent group of 27 E-mail: [email protected] patients. Key genes within the predictive signature were confirmed by quan- titative reverse transcription–polymerase chain reaction in an additional 10 patients. The analysis identified 431 differentiating genes between patients with good and poor clinical outcome (change in neurological disability by the expanded disability status scale was -0·33 Ϯ 0·24 and 1·6 Ϯ 0·35, P = 0·0002, total number of relapses were 0 and 1·80 Ϯ 0·35, P = 0·00009, respectively). An optimal set of 29 genes was depicted as a clinical outcome predictive gene expression signature and classified appropriately 88·9% of patients. This pre- dictive signature was enriched by genes related biologically to zinc-ion binding and cytokine activity regulation pathways involved in inflammation and apoptosis. Our findings provide a basis for monitoring patients by pre- diction of disease outcome and can be incorporated into clinical decision- making in relapsing–remitting MS. Keywords: gene expression, multiple sclerosis, pathways, prediction, regulation Introduction reported to account for more variance than lesion burden in predicting cognitive impairment [4]. However, all these Multiple sclerosis (MS) is a central nervous system disease clinical and radiological variables are limited in their ability affecting young adults in which 85% of patients experience a to predict disease outcome, especially during the early stages relapsing–remitting (RR) clinical course [1]. Clinical of the disease. Gene microarray technology, that analyses outcome differs between patients, as the rate of disease pro- simultaneously the expression of thousands of genes [5], can gression and frequency of relapses vary along the disease be used as a comprehensive analysis method to correlate course [2]. It has been suggested that age of disease onset gene expression patterns with numerous clinical parameters below 35 years, rapid development and regression of initial related to patients’ outcome. Attempts to correlate MS gene symptoms, a single symptom at onset and visual loss as the expression with disease activity disclosed that activity corre- initial symptom indicate a good prognosis. Brain magnetic lated with the frequency of CX3CR1-positive natural killer resonance imaging (MRI) parameters have also been impli- (NK) cells [6], and that MS expression profiling identified cated as important in the evaluation of MS course by mea- responder and non-responder phenotypes to interferon suring disease load over time [3]. Brain atrophy has been (IFN)-b treatment [7]. © 2007 British Society for Immunology, Clinical and Experimental Immunology, 149: 235–242 235 12

A. Achiron et al. Table 1. Demographic and clinical variables of the study relapsing–remitting multiple sclerosis (RRMS) population. Validation Variable Differentiating clinical Clinical outcome Biological function outcome group classifier group by qRT–PCR group n = 26 n = 27 n = 10 Age (years) 40·2 Ϯ 5·8 40·5 Ϯ 1·6 36·1 Ϯ 2·1 F (M) 21 (5) 16 (11) 8 (2) Disease duration (years) Relapse rate 9·9 Ϯ 4·2 10·3 Ϯ 1·6 8·5 Ϯ 1·4 EDSS 1·3 Ϯ 0·7 0·9 Ϯ 0·2 0·9 Ϯ 0·2 Treated 2·0 Ϯ 1·0 2·5 Ϯ 0·2 1·9 Ϯ 0·3 13 11 None EDSS: Expanded Disability Status Scale; qRT–PCR: quantitative reverse transcription–polymerase chain reaction. We have demonstrated previously that gene expression tory treatments for at least 3 months prior to the gene signature of peripheral blood mononuclear cells (PBMC) expression study, and 13 patients were naive to immuno- significantly differentiates RRMS patients from healthy modulatory treatment. In the validation group, 11 patients subjects. Having also demonstrated that different gene sig- were receiving immunomodulatory treatments for at least nature characterizes MS disease stage (relapse versus remis- 3 months prior to the gene expression study, and 16 patients sion) [8], in the current study we sought to evaluate whether were naive to immunomodulatory treatments. Within up to gene expression profiling can differentiate RRMS patients 1 month from withdrawal of blood, all patients were treated according to their clinical course - either favourable or poor. with IFN-b1a. None of the patients had ever received cyto- Our idea was to use an informative subset of original train- toxic treatments and all were free of steroid treatment for at ing samples. This subset consists of only good-outcome least 30 days before blood was withdrawn. All patients had RRMS patients who did not deteriorate neurologically peripheral blood counts within the normal range. The study within a 2-year period, and patients with poor outcome who was approved by the Sheba Medical Center Institutional increased their disability and demonstrated clinical disease Review Board, and all patients gave written informed activity within the same follow-up period. These extreme consent for participation. training samples yielded a clear platform from which to identify genes whose expression is related to clinical Clinical follow-up outcome. The discriminating genes were then integrated by a support vector machine (SVM) to build a prediction model, Patients were followed-up prospectively for a period of by which each validation sample was assigned a good or poor 2 years. Neurological examination was performed once every risk score for MS progression. We found that RRMS patients 3 months and at the time of a suspected relapse, and EDSS in high- and low-risk groups are clearly distinguishable. Our assessment was completed accordingly. Relapse was defined results indicate that gene expression profiles combined with as the onset of new objective neurological symptoms/signs carefully chosen learning algorithms can predict patient or worsening of existing neurological disability, not accom- outcome and may be incorporated in individualized, tailored panied by metabolic changes, fever or other signs of infec- management of RRMS. tion, lasting for a period of at least 48 h accompanied by objective change of at least 0·5 points in the EDSS score. For Methods EDSS evaluations we used only stable EDSS scores that were confirmed at 3-month follow-up examinations. Confirmed Patients relapses and EDSS scores were recorded consecutively. Sixty-three patients with definite RRMS (45 females, Definition of clinical outcome 18 males), aged 38·2 Ϯ 3·9 years, disease duration 9·2 Ϯ 2·8 years, annual relapse rate 1·1 Ϯ 0·5 and neurologi- Clinical outcome was defined according to neurological dis- cal disability evaluated by the Expanded Disability Status ability as the primary criterion and total number of relapses Scale (EDSS) [9], 1·9 Ϯ 0·6, were included in the study; 26 as the secondary criterion. patients participated in the differentiating clinical outcome analysis, 27 patients in the validation process of prediction Good outcome and 10 patients in the functional biological validation. Good outcome comprised patients who had not deteriorated The clinical and demographic variables were similar in their neurological disability and had not experienced any between groups (Table 1). In the differentiating clinical relapse during the 24 months of follow-up. outcome group, 13 patients were receiving immunomodula- 236 © 2007 British Society for Immunology, Clinical and Experimental Immunology, 149: 235–242 13

Poor outcome Predictive regulatory pathways in multiple sclerosis Poor outcome comprised patients who deteriorated in their mean Ϯ standard deviation (s.d.). Student’s t-test was used neurological disability (DEDSS increased by at least 0·5 to compare the difference in clinical variables between points) within the 24 months of follow-up, either with or groups, and P < 0·05 was considered statistically significant. without relapses. Predictive genes analysis and validation Intermediate outcome To depict the predictive genes from the differentiating clini- Intermediate outcome comprised patients who did not dete- cal outcome signature, the SVM in combination with riorate in their neurological disability yet experienced at least Forward feature selection algorithm were applied (http://ro. one relapse during the 24 months of follow-up. utia.cz/fs/fs_algorithms.html) [14–16]. SVM generates a classifier based on a known labelled training set (19 of 26 RNA isolation and microarray expression profiling RRMS patients with good or poor clinical outcome from the differentiating clinical outcome group). Then, the classifica- PBMC were separated on Ficol-Hypaque gradient, total RNA tion power of the generated classifier is evaluated by apply- was purified, labelled, hybridized to a Genechip array ing it to an independent test set (nine of 27 RRMS patients (U95Av2 and HU-133A) and scanned (Hewlett Packard, from the validation group). The feature selection algorithm GeneArray-TM scanner G2500A) according to the manufac- finds a subset of predictive genes that enables the generated turer’s protocol (Affymetrix Inc, Santa Clara, CA, USA), as classifier to achieve the highest classification rate [14]. To described previously [8]. validate the power of the predictive genes, the classifier was applied to an additional independent set (18 of 27 RRMS Clinical outcome differentiating genes analysis patients from the validation group). Additionally, to confirm independently the obtained expression profiling data, we RMAExpress software was used to analyse the scanned arrays performed quantitative reverse transcription–polymerase [10]. In order to be consistent with the ontology and array chain reaction (qRT–PCR) on PBMC samples from 10 MS type, all the transcripts in U95Av2 microarray were con- patients for three key genes of the predictive signature. verted to the corresponding transcripts in HU-133A using NetAffex comparison table. Probe sets that did not have a The study design is depicted in Fig. 1. signal present in at least 90% of the samples were filtered. Noise effect was reduced by fitting a multiple effect model Biological functional analysis for each gene modelling the log-ratio measurement as a sum of contributions for age, gender, batch, subject state (naive or Functional annotation of the clinical outcome differentiat- treated) and time from last steroid treatment. ing and predictive gene signatures was performed using functional classification tools (FCT; David Bioinformatics Statistical methods Resources: http://david.abcc.ncifcrf.gov/home.jsp). Gene enrichment was defined as a group of genes associated highly Statistical analysis was performed using the ScoreGenes soft- with a specific biological function and measured statistically ware tools (http://compbio.cs.huji.ac.il/scoregenes/). Data by one-tailed Fisher’s exact probability value using the David were analysed by t-test, the threshold number of misclassifi- system. Biological regulatory pathway reconstruction for the cations (TNoM) method and the Info-test score. Differenti- predictive gene signature was performed using: (1) Pathwa- ating genes were defined as genes whose expression was yArchitect software (http://www.stratagene.com) based on significantly higher or lower, with P < 0·05 in all three statis- published data in the literature, and (2) Genomica soft- tical tests. Overabundance analysis was used to compare ware (http://genomica.weizmann.ac.il), based on Bayesian between the number of observed and expected genes that network methods taken from the field of machine learning, differentiated between the good and poor clinical outcome and was applied to our results of the differentiating gene under the null hypothesis that the classification of the microarray expression signature [17]. This evaluation was samples was random [11,12]. To verify further the accuracy aimed to identify potentially target genes that share a of the classification we used the leave-one-out cross- common regulatory mechanism. validation (LOOCV) statistical method [13]. LOOCV simu- lates removal of a single sample for every trial and trains on Results the rest. The procedure is repeated until each sample is left out once and the number of correct and incorrect predic- Clinical classification of study patients tions is counted. The demographic variables are presented as Patients were classified into three groups based on their clinical disease outcome: (1) patients with good outcome (n = 9, mean age 39·3 Ϯ 3·3 years, disease duration 10·7 Ϯ 3·4 years), (2) patients with intermediate outcome © 2007 British Society for Immunology, Clinical and Experimental Immunology, 149: 235–242 237 14

A. Achiron et al. Fig. 1. Flowchart of the study design. Overview Classification rate was 70·4% using only one gene (RRN3) of the strategy used for the identification and and reached a rate of 85·2% using six genes (RRN3, KLF4, validation of predictive clinical outcome HAB1, TPSB2, IGLJ3, COL11A2). The addition of one or all gene-expression signature in of the remaining predictive genes resulted in maximal clas- relapsing–remitting multiple sclerosis. sification rate of 89·0%. This suggests that maximal predic- tive ability could be achieved using only seven genes. As there (n = 7, mean age 35·8 Ϯ 5·4 years, disease duration is no preference to any of the genes beyond the six predictive 8·6 Ϯ 4·7 years) and (3) patients with poor outcome (n = 10, genes, we analysed the biological relevance of all 34 predic- mean age 46·3 Ϯ 4·2 years, disease duration 10·3 Ϯ 0·9). As tive genes. the aim of our study was to evaluate the differences within the extremes, the analysis was performed between the good Independent validation of the predictive clinical and the poor clinical outcome groups. The comparison dem- outcome gene expression signature onstrated significant differences between patients with good and poor clinical outcomes. Change in neurological disabil- Applying the resulting SVM-generated classifier, based on ity assessed by the EDSS was -0·33 Ϯ 0·24 and 1·6 Ϯ 0·35, the 34 predictive genes to an additional data set of 18 of 27 P = 0·0002, and total number of relapses was 0 and patients from the validation group, maintained the high clas- 1·80 Ϯ 0·35, P = 0·00009, respectively. sification rate of 88·9%, P < 0·00001. Differentiating clinical outcome gene qRT–PCR performed in PBMC samples from 10 MS expression signature patients for three key genes of the predictive signature (S100B, KLF4 and RRN3) showed a perfect correlation of the The distinctive clinical outcome gene expression pattern expression levels of the candidate genes analysed with the between patients with good and poor clinical outcomes mean expression levels obtained from the microarray included 431 differentiating genes which passed the three experiments. S100B was lower by 1·47, P = 0·003; KLF4 was statistical tests, with P < 0·05 (Fig. 2a). Functional analysis higher by 1·87, P = 0·044; and RRN3 was higher by 2·29, disclosed genes associated with signal transduction, catalytic P = 0·003, in MS patients with poor outcome. activity, adhesion and inflammation (Fig. 2b). Overabun- dance analysis of the observed compared with the expected Biological regulation of the predictive clinical outcome number of genes that distinguished significantly between gene expression signature patients with good or poor clinical outcome was higher than expected (431 versus 200 genes at P = 0·03) (Fig. 2c). Functional annotation of the 34 predictive genes demon- LOOCV resulted in a high classification rate of 90%, strated that this group of genes was enriched significantly by P < 0·0001 (Fig. 2d), suggesting that the differentiating genes zinc-ion binding protein genes (S100B, KLF4, CAII) and by signature is reliable and not related to spurious differences genes with cytokine activity (CCL17, MUC4, PTN VEGFB), due to multiple testing. P = 0·02 and P = 0·005, respectively (Fig. 3b). The Genomica software confirmed enrichment by the zinc-ion binding gene Predictive clinical outcome gene expression signature family and by cytokine activity genes using all the 431 dif- ferentiating gene expression signature data (Fig. 3c). Using Application of the SVM on data from 19 of 26 patients with these enriched gene-families, regulatory pathways were good (nine patients) or poor (10 patients) outcome as a reconstructed (Fig. 3d,e). These pathways suggest that training set, and nine of 27 additional patients from the validation group as test set, resulted in a high classification rate of 89%. This high classification was achieved by the Forward feature selection algorithm using 34 gene tran- scripts (29 genes), defined accordingly as predictive (Fig. 3a). 238 © 2007 British Society for Immunology, Clinical and Experimental Immunology, 149: 235–242 15

Predictive regulatory pathways in multiple sclerosis (a) Clinical outcome (b) Good Poor Molecular function unknown 22 Cellular component 18 Structural molecular activity 22 Adhesion 23 Transcription activity Inflammation 36 Development 37 Signal transducer Catalytic activity 49 Binding 86 100 207 (c) Observed (TNoM) 400 Number of genes Expected 200 0 0·01 0·02 0·03 0 P-value (d) 40 Error rate % 30 t-test Info EDSS 0 20 Deta EDSS neg Deta EDSS pos 10 TNoM Relapse 0 0 0·0002 0·0004 0·0006 0·0008 0·001 P-value Fig. 2. (a) Heatmap of clinical outcome differentiating genes. Heatmap of the 431 differentiating genes that distinguishes between patients with good and poor clinical outcome. Each row of the heatmap represents a gene and each column represents a patient sample. Genes with increased expression are shown in progressively brighter shades of red, and genes with decreased expression are shown in progressively darker shades of green. The bottom matrix shows corresponding clinical outcome attributes marked in black when positive. (b) Functional annotation histogram. Distribution of differentiating gene expression signature according to biologically relevant functional groups. (c) Overabundance analysis. Actual number of genes (blue line) is significantly more abundant than expected (red line) for threshold number of misclassifications (TNoM) statistical test. x-Axis denotes P-value; y-axis denotes number of genes. (d) Leave-one-out cross-validation (LOOCV) classification. Division of errors between patients with good and poor clinical outcome using TNoM, Info and t-test demonstrated high classification rate of 90% at P < 0·0001. x-Axis denotes P-value; y-axis denotes error rate in percentage. apoptosis regulation through zinc-ion binding and cytokine adhesion and cell migration such as CD44 and COL11A2, activity is responsible for Th1/Th2 shift and may play a role and T cell receptor genes such as TCRVB; all play an impor- in the clinical outcome of RRMS. tant role in MS pathogenesis. Genomica reconstruction of regulatory gene expression Discussion networks based on all 431 differentiating genes resulted in a regulation pathway in which the predictive zinc-ion binding Prognostic modelling of patients with RRMS is a challenge gene KLF4, in association with CLPP and RRLP, mediate in view of the unpredictable course of the disease. To the best downstream genes including S100B (Fig. 3f). Other interest- of our knowledge, our study is the first that correlates gene ing functional groups in the 29 predictive genes include © 2007 British Society for Immunology, Clinical and Experimental Immunology, 149: 235–242 239 16

A. Achiron et al. 18 Goodkin DE, Hertsgaard D, Rudick RA. Exacerbation rates and adherence to disease type in a prospectively followed-up popula- delineation of patients at high risk who may benefit from tion with multiple sclerosis. Implications for clinical trials. Arch early therapy. Neurol 1989; 46:1107–12. References 19 Kurtzke JF, Beebe GW, Nagler B, Kurland LT, Auth TL. Studies on the natural history of multiple sclerosis - 8. Early prognostic fea- 1 Confavreux C, Vukusic S. Natural history of multiple sclerosis: tures of the later course of the illness. J Chronic Dis 1977; 30:819–30. implications for counselling and therapy. Curr Opin Neurol 2002; 15:257–66. 20 Weinshenker BG. The natural history of multiple sclerosis: update 1998. Semin Neurol 1998; 18:301–7. 2 Trojano M, Paolicelli D, Bellacosa A, Cataldo S. The transition from relapsing–remitting MS to irreversible disability: clinical 21 Weinshenker BG, Rice GPA, Noseworthy JH, Carriere W, Basker- evaluation. Neurol Sci 2003; 24 (Suppl. 5):S268–70. ville J, Ebers GC. The natural history of multiple sclerosis: a geo- graphically bases study: 3. Multivariate analysis of predictive factors 3 Simon JH. Contrast-enhanced MR imaging in the evaluation of and models of outcome. Brain 1991; 114:1045–56. treatment response and prediction of outcome in multiple sclerosis. J Magn Reson Imaging 1997; 7:29–37. 22 Runmarker B, Andersen O. Prognostic factors in a multiple sclero- sis incidence cohort with 25 years of follow-up. Brain 1993; 4 Benedict RH, Weinstock-Guttman B, Fishman I, Sharma J, Tjoa 116:117–34. CW, Bakshi R. Prediction of neuropsychological impairment in multiple sclerosis: comparison of conventional magnetic resonance 23 Kantarci OH, Weinshenker BG. Prognostic factors in multiple imaging measures of atrophy and lesion burden. Arch Neurol 2004; sclerosis. In: Cook DS, ed. Handbook of multiple sclerosis, 3rd edn. 61:226–30. New York: Marcel Dekker, 2001:449–63. 5 Mantripragada KK, Buckley PG, de Stahl TD, Dumanski JP. Ge- 24 Tremlett H, Paty D, Devonshire V. Disability progression in mul- nomic microarrays in the spotlight. Trends Genet 2004; 20:87–94. tiple sclerosis is slower than previously reported. Neurology 2006; 66:172–7. 6 Infante-Duarte C, Weber A, Kratzschmar J et al. Frequency of blood CX3CR1-positive natural killer cells correlates with disease 25 Zhang W, Geiman DE, Shields JM et al. The gut-enriched Kruppel- activity in multiple sclerosis patients. FASEB J 2005; 19:1902–4. like factor (Kruppel-like factor 4) mediates the transactivating effect of p53 on the p21WAF1/Cip1 promoter. J Biol Chem 2000; 7 Sturzebecher S, Wandinger KP, Rosenwald A et al. Expression pro- 275:18391–8. filing identifies responder and non-responder phenotypes to interferon-beta in multiple sclerosis. Brain 2003; 126:1419–29. 26 Chen ZY, Shie JL, Tseng CC. STAT1 is required for IFN-gamma- mediated gut-enriched Kruppel-like factor expression. Exp Cell Res 8 Achiron A, Gurevich M, Friedman N, Kaminski N, Mandel M. 2002; 281:19–27. Blood transcriptional signatures of multiple sclerosis: unique gene expression of disease activity. Ann Neurol 2004; 55:410–17. 27 Feinberg MW, Cao Z, Wara AK, Lebedeva MA, Senbanerjee S, Jain MK. Kruppel-like factor 4 is a mediator of proinflammatory sig- 9 Kurtzke JF. Rating neurologic impairment in multiple sclerosis: an naling in macrophages. J Biol Chem 2005; 280:38247–58. expanded disability status scale (EDSS). Neurology 1983; 33:1444– 52. 28 Petzold A, Eikelenboom MJ, Gveric D et al. Markers for different glial cell responses in multiple sclerosis: clinical and pathological 10 Li C, Wong WH. Model-based analysis of oligonucleotide arrays: correlations. Brain 2002; 125:1462–73. expression index computation and outlier detection. Proc Natl Acad Sci USA 2001; 98:31–6. 29 Petzold A, Brassat D, Mas P et al. Treatment response in relation to inflammatory and axonal surrogate marker in multiple sclerosis. 11 Eisen MB, Spellman PT, Brown PO, Botstein D. Cluster analysis Mult Scler 2004; 10:281–3. and display of genome-wide expression patterns. Proc Natl Acad Sci USA 1998; 95:14863–8. 30 Faffe DS, Whitehead T, Moore PE et al. IL-13 and IL-4 promote TARC release in human airway smooth muscle cells: role of IL-4 12 Kaminski N, Friedman N. Practical approaches to analyzing results receptor genotype. Am J Physiol Lung Cell Mol Physiol 2003; of microarray experiments. Am J Respir Cell Mol Biol 2002; 285:L907–14. 27:125–32. 31 Dabbagh K, Takeyama K, Lee HM, Ueki IF, Lausier JA, Nadel JA. 13 Ben-Dor A, Bruhn L, Friedman N, Nachman I, Schummer M, IL-4 induces mucin gene expression and goblet cell metaplasia in Yakhini Z. Tissue classification with gene expression profiles. vitro and in vivo. J Immunol 1999; 162:6233–7. J Comput Biol 2000; 7:559–83. 32 Zhu Z, Lee CG, Zheng T et al. Airway inflammation and remodel- 14 Furey TS, Cristianini N, Duffy N, Bednarski DW, Schummer M, ing in asthma. Lessons from interleukin 11 and interleukin 13 Haussler D. Support vector machine classification and validation of transgenic mice. Am J Respir Crit Care Med 2001; 164:S67–70. cancer tissue samples using microarray expression data. Bioinfor- matics 2000; 16:906–14. 33 Soto P, Price-Schiavi SA, Carraway KL. SMAD2 and SMAD7 involvement in the post-translational regulation of Muc4 via the 15 Statnikov A, Aliferis CF, Tsamardinos I, Hardin D, Levy S. A com- transforming growth factor-beta and interferon-gamma pathways prehensive evaluation of multicategory classification methods for in rat mammary epithelial cells. J Biol Chem 2003; 278:20338–44. microarray gene expression cancer diagnosis. Bioinformatics 2005; 21:631–43. 34 Achour A, Laaroubi D, Caruelle D, Barritault D, Courty J. The angiogenic factor heparin affin regulatory peptide (HARP) induces 16 Aha DW, Bankert RL. A comparative evaluation of sequential proliferation of human peripheral blood mononuclear cells. Cell feature selection algorithms. In: Fisher D, Lenx JH, eds. Proceed- Mol Biol 2001; 47:OL73–7. ings of the 5th International Workshop on Artificial Intelligence and Statistics. New York: Springer-Verlag, 1995:1–7. 35 Heroult M, Bernard-Pierrot I, Delbe J et al. Heparin affin regula- tory peptide binds to vascular endothelial growth factor (VEGF) 17 Segal E, Shapira M, Regev A et al. Module networks: identifying and inhibits VEGF-induced angiogenesis. Oncogene 2004; regulatory modules and their condition-specific regulators from 23:1745–53. gene expression data. Nat Genet 2003; 34:166–76. 242 © 2007 British Society for Immunology, Clinical and Experimental Immunology, 149: 235–242 17

expression profiles segregating RRMS patients into classes Predictive regulatory pathways in multiple sclerosis associated with clinical outcome. As MS has a winding course and the rate of disease progression differs between cells in MS involves proinflammatory cytokines such as patients [18–20], it is of great importance to look for surro- IFN-g and tumour necrosis factor (TNF)-a that promote gate markers that will enable future envisaging of the disease disease activity. Conversely, anti-inflammatory cytokines process in an individual patient. Prediction of outcome in such as TGF-b, interleukin (IL)-4 and IL-10 decrease proin- MS was reported to relate to different clinical variables such flammatory activation. The molecular transcripts we identi- as age at disease onset, gender and the type of neurological fied regulate the balance of these opposing effectors and are symptomatology presented at onset. The major clinical thus associated with clinical outcome prediction. The zinc- determinants of more severe disease are male sex, relatively ion binding genes in the predictive signature include KLF4, older age at onset, motor or cerebellar symptoms at onset known to be activated by (signal transucers and activators of and high annual relapse rate [21–23]. However, the ability of transcription (STAT1), which is activated by S100B [25], and these variables to predict the clinical course is imperfect. This increases IFN-g expression [26], a well-known proinflamma- uncertainty in forecasting disease outcome means that some tory cytokine involved in MS disease activity. MS patients who need aggressive treatment do not receive it, while others are treated unnecessarily and as a result are KLF4 is markedly induced in response to IFN-g, exposed to the risk of side effects without a sound rationale lipolpolysaccharide (LPS) or TNF-a. Over-expression of [24]. In the current study we used a comprehensive approach KLF4 is associated with macrophage activation marker to gain detailed understanding of the evolution of MS by inducible nitric-oxide synthase and with TGF-b1 inhibition. analysing disease process-relevant gene signatures. Our KLF4 interacts with the NF-kB family member p65 (RelA), rationale was to look for differentiating gene transcripts and has an important role as a regulator of key signalling that are related to and predict clinical outcome and pathways that control macrophage activation [27]. not - although it might be - are the causes for the change in clinical outcome. It is evident that the patients included in The S100B gene protein is known to be involved in intra- the study had a priori different clinical outcomes, but the cellular and extracellular regulatory events within the central ability to take a snapshot in time and identify a specific gene nervous system. S100B was found to be elevated in acute signature that characterizes good or poor clinical outcome is brain lesions of RRMS patients [28], and its plasma levels the major significance of our findings. Classifying patients were reported elevated in RRMS patients responding to into homogeneous groups based on the progression of neu- IFN-b treatment [29]. This is in accordance with our find- rological disability and number of relapses, and then sorting ings, demonstrating decreased S100B expression in patients between different disease outcome groups according to gene with poor outcome. Additionally, a novel association of the expression anchors, enabled us to expand the knowledge of zinc-ion binding gene CA11 was identified in the network the disease phenotypes. The clinical outcome expression sig- reconstruction. nature includes genes involved in proliferation, stimulation of T cell and T cell receptors, inflammation and adhesion. The balance between T helper 1 (Th1) and Th2 immune The reliability of the differentiating 431 clinical outcome responses plays an important role in the pathogenesis of MS. gene-expression pattern was validated further by the low In addition to the recognition of encephalitogenic epitopes, error estimates using LOOCV as well as by the overabun- the ability to produce Th1 cytokines is an important func- dance analysis, showing a significant difference between the tional requirement by which myelin-reactive T cells mediate expected and observed data. These results provided further the disease, while Th2 cells secreting IL-10 suppress the evidence that the identified genes were indeed representative ongoing inflammation. The cytokine activity-enriched gene of true biological processes that lead to different clinical family identified in the prediction signature included phenomena. We assumed that not all the genes that differ- CCL17, MUC4, PTN and VEGFB. CCL17 displays chemot- entiated between good and poor clinical outcome were effec- actic activity for Th2 lymphocytes [30], and its activity is well tive predictive genes. Accordingly, we applied the SVM known to be enhanced by the Th2-related cytokines IL-4 and method combined with the Forward feature selection algo- IL-13, leading to inhibition of inflammation. MUC4 expres- rithm to evaluate outcome predictive genes and validated sion is dependent upon IL-4 and IL-13 levels [31–33]. PTN is further the predictive pattern in an additional group of involved in regulation of cell-mediated immunity [34], and RRMS patients. Classification rate was 70·4% using only one negatively regulates VEGF activity [35]. These findings dem- gene (RRN3), and reached a rate of 85·2% using six genes onstrate that the poor clinical outcome predictive signature (RRN3, KLF4, HAB1, TPSB2, IGLJ3, COL11A2). In larger in RRMS is affected mainly by decreased Th2 cytokine activ- groups, the applicable methods would use much smaller ity and aberrant regulation of inflammation. groups of genes (six or 34). The clinical outcome predictive gene expression signature was enriched by zinc-ion binding In conclusion, the predictive outcome gene expression sig- and cytokine activity pathways. Activation of mononuclear nature is sensitive to RRMS evolution and as such provides a new perspective on disease progression. Moreover, our find- ings suggest that the co-stimulatory regulatory pathways of zinc-ion binding and cytokine activity-related genes within the predictive signature may serve as new targets for thera- peutic interventions. Finally, the predictive signature may enable planning of tailored therapeutic strategies, and allow © 2007 British Society for Immunology, Clinical and Experimental Immunology, 149: 235–242 241 18

A. Achiron et al. Fig. 3. (a) Predictive classification chart. The (a) Classification rate classification rate of 29 predictive genes is 1 demonstrated. Highest classification rate is 5 10 15 20 25 30 35 40 achieved using only seven genes, yet according 0·9 Number of genes to the feature selection algorithm, genes are 0·8 added to the subset as long as the classification 0·7 rate is not decreased. y-Axis denotes 0·6 classification rate; x-axis denotes the number of 0·5 genes. (b) Gene enrichment. Direction of an 0·4 over-expressed (1) or down-expressed (- 1) 0·3 gene is demonstrated in the enriched groups 0·2 within the poor versus good outcome signature. 0·1 (c) Heatmap of module analysis using differentiating clinical outcome genes. 0 Enrichment of zinc-ion binding gene set for 0 patients with relapses and cytokine activity gene set for patients with stable disease [no change (b) in neurological disability, Expanded Disability Status Scale (EDSS) = 0] are demonstrated. The Genes upper left graphical panel is a matrix of gene ADD1 sets versus arrays, where a coloured entry CA11 indicates that the genes in the gene set had CCL17 changed significantly in a co-ordinated fashion CD44 in the respective array (red, increased; green, COL11A2 decreased). The centre graphical panel shows CRYGD individual clinical outcome attributes to which DNM1 each array belongs. The bottom graphical panel DR1 demonstrates overall clinical outcome attributes GNMT in which gene sets were significantly enriched. GPP3 (d) Reconstructed zinc-ion binding pathway. GSTA1 Pathway analysis performed using genes from HAB1 the predictive signature (yellow circles) and HSPA8 genes brought into the pathway based on IGLJ3 literature-known relationships according to IGLVJ PathwayArchitect software (green circles). IL3RA Arrows indicate regulatory interactions KIAA0980 confirmed by literature database, dashed arrows KLF4 indicate suggested gene interactions. (e) KLK1 Reconstructed cytokine activity pathway. MUC4 Pathway analysis performed using genes from NY-REN-24 the predictive signature (grey circles) and genes ODZ2 brought into the pathway based on PTN literature-known relationships according to RRN3 PathwayArchitect software (blue circles). Arrows S100B indicate regulatory interactions confirmed by TCRBV literature database, dashed arrows indicate TOP3B suggested gene interactions. (f) Gene expression TPSB2 regulatory network module. The single gene VEGFB expression module from the gene expression p value regulatory network of 431 differentiating genes is demonstrated. Each node in the regulation Direction 1 –1–1 1 –1–1 1 1 –1–1–1–1–1 1 1 –1 1 1 –1–1–1 1 –1 1 –1–1–1–1–1 tree represents a regulating gene. The expression of the regulating genes themselves is Cytokine activity 0·005 shown below their node. Cluster of gene expression profiles (rows represent genes, Zinc ion binding 0·02 columns represent patients’ arrays) arranged according to the regulation tree. Note that (c) zinc-ion binding related genes KLF4 (regulating gene) and S100B (regulated gene) belong to the (d) (e) same regulatory module (black asterisks). S100B IL4 IFNY MUC4 CCL17 KLF4 STAT1 IL13 VEGF-β CA11 PTN (f) 240 © 2007 British Society for Immunology, Clinical and Experimental Immunology, 149: 235–242 19



‫‪Prediction of acute multiple sclerosis relapses‬‬ ‫‪by transcription levels of peripheral blood cells‬‬ ‫‪BMC Medical genomics | 2009‬‬ ‫מנחה‪ :‬דר' מיכאל גורביץ‬ ‫מנחה‪ :‬פרופ' ענת אחירון‬ ‫מנהל המעבדה הנוירואימונולוגית‪,‬‬ ‫מייסדת פרויקט ח\"ץ‪ ,‬מנהלת המרכז‬ ‫המרכז לטרשת נפוצה‬ ‫לטרשת נפוצה ואחראית הקתדרה‬ ‫למחלות אוטואימוניות אוניברסיטת ת\"א‬ ‫‪[email protected]‬‬ ‫‪[email protected]‬‬ ‫בשנת ‪ 2006‬הצטרפתי לפרופ' ענת אחירון ודר' מיכאל גורביץ כשפרויקט‬ ‫רותם אורבך‬ ‫ח\"צ החל להפוך מחלום למציאות‪ ,‬נפלה בחלקי הזכות להיות \"הח\"צ‬ ‫הראשונה\"‪ .‬במהרה המרכז לטרשת נפוצה התמלא בסטודנטים לרפואה‬ ‫אונ' תל אביב‬ ‫שרק חיכו להזדמנות מהסוג שפרופ' אחירון שמחה להציע‪ .‬יחד ולחוד‬ ‫השתתפה כסטודנטית בפרויקט ח״ץ‬ ‫העשרנו את עולמנו בהרצאות‪ ,‬שיעורים בסטטיסטיקה והזמנות אמיתית‬ ‫ללמוד מהו מחקר‪ ,‬והחשוב מכל – לעשות בעצמנו‪ ,‬משלב ההרהורים ועד‬ ‫בין השנים ‪2007-2009‬‬ ‫ללחיצה המיוחלת על הכפתור ‪ .submit -‬עוד לפני שהספקתי לסיים את‬ ‫בית הספר לרפואה‪ ,‬התמלא‪ ‬אולם ההרצאות במספר מרשים של סטודנטים‪,‬‬ ‫‪[email protected]‬‬ ‫שכבר עבדו בשלל מחלקות ומעבדות בית החולים‪ .‬כמה מעורר השראה היה‬ ‫לראות את המפעל החשוב‪ ‬הזה מקבל את כוחו ומהדהד בקרב הסטודנטים‬ ‫החדשים‪.‬‬ ‫שמי רותם‪ ,‬אני רופאת ילדים‪ ,‬מומחית בנוירולוגית ילדים והתפתחות‬ ‫הילד‪ .‬כעת אני נמצאת במרילנד להתמחות בתוכנית משולבת של‬ ‫בית חולים‪ Johns Hopkins ‬וה‪ ,NIH-‬ב‪National Institute of,-‬‬ ‫‪ ,Neurological Disorders and Stroke‬במחלות‪ ‬גנטיות ומחלות‪ ‬עצב‬ ‫שריר בילדים‪ .‬תמיד ידעתי שאני רוצה להיות רופאת ילדים‪ ,‬אך אין לי‬ ‫ספק שההחלטה להתמחות בנוירולוגית ילדים היא תוצאה של ההשפעה‬ ‫העצומה שהיתה לפרופ' אחירון עלי‪ .‬אותו‪ ‬פטיש רפלקסים שהעניקה לי‬ ‫לאחר שהצגתי את עבודת המחקר שלי בכנס של ה‪ AAN-‬עדיין יקר לי מכל‬ ‫ומסמל עבורי בחירה בנתיב קריירה‪ .‬לאורך דרכי כמתמחה‪ ,‬הכלים שרכשתי‬ ‫בתוכנית נתנו לי בסיס טוב וחזק ליזום ולבצע מחקר בעצמי‪ .‬אין לי ספק‬ ‫שאותן חוויות ייחודיות ומלמדות בתחילת דרכי כסטודנטית הובילו אותי‬ ‫להתמחות הנוכחית שלי‪.‬‬ ‫גם היום במרחק השנים‪ ,‬אני מלאת הוקרה לפרויקט ובפרט לפרופ' אחירון‪,‬‬ ‫דר' גורביץ ואנשי המעבדה במרכז לטרשת נפוצה‪ ,‬ומודה ממעמקי ליבי על‬ ‫ההזדמנויות הייחודיות שזכיתי להן ועל האופן שבו עצבו את ה\"דנ\"א\" שלי‬ ‫כרופאה‪.‬‬ ‫‪21‬‬

BMC Medical Genomics BioMed Central Research article Open Access Prediction of acute multiple sclerosis relapses by transcription levels of peripheral blood cells Michael Gurevich†1, Tamir Tuller*†1, Udi Rubinstein2, Rotem Or-Bach1 and Anat Achiron*1 Address: 1Multiple Sclerosis Center, Sheba Medical Center, Tel-Hashomer, Sackler School of Medicine, Tel-Aviv University, Tel Aviv, Israel and 2School of Computer Science, Tel Aviv University, Tel Aviv, Israel Email: Michael Gurevich - [email protected]; Tamir Tuller* - [email protected]; Udi Rubinstein - [email protected]; Rotem Or-Bach - [email protected]; Anat Achiron* - [email protected] * Corresponding authors †Equal contributors Published: 22 July 2009 Received: 10 December 2008 BMC Medical Genomics 2009, 2:46 doi:10.1186/1755-8794-2-46 Accepted: 22 July 2009 This article is available from: http://www.biomedcentral.com/1755-8794/2/46 © 2009 Gurevich et al; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Abstract Background: The ability to predict the spatial frequency of relapses in multiple sclerosis (MS) would enable physicians to decide when to intervene more aggressively and to plan clinical trials more accurately. Methods: In the current study our objective was to determine if subsets of genes can predict the time to the next acute relapse in patients with MS. Data-mining and predictive modeling tools were utilized to analyze a gene-expression dataset of 94 non-treated patients; 62 patients with definite MS and 32 patients with clinically isolated syndrome (CIS). The dataset included the expression levels of 10,594 genes and annotated sequences corresponding to 22,215 gene-transcripts that appear in the microarray. Results: We designed a two stage predictor. The first stage predictor was based on the expression level of 10 genes, and predicted the time to next relapse with a resolution of 500 days (error rate 0.079, p < 0.001). If the predicted relapse was to occur in less than 500 days, a second stage predictor based on an additional different set of 9 genes was used to give a more accurate estimation of the time till the next relapse (in resolution of 50 days). The error rate of the second stage predictor was 2.3 fold lower than the error rate of random predictions (error rate = 0.35, p < 0.001). The predictors were further evaluated and found effective both for untreated MS patients and for MS patients that subsequently received immunomodulatory treatments after the initial testing (the error rate of the first level predictor was < 0.18 with p < 0.001 for all the patient groups). Conclusion: We conclude that gene expression analysis is a valuable tool that can be used in clinical practice to predict future MS disease activity. Similar approach can be also useful for dealing with other autoimmune diseases that characterized by relapsing-remitting nature. Page 1 of 14 (page number not for citation purposes) 22

BMC Medical Genomics 2009, 2:46 http://www.biomedcentral.com/1755-8794/2/46 Background Imaging is considered as a more sensitive tool for predict- ing MS progression; it was reported that various parame- Multiple sclerosis (MS) is an autoimmune demyelinating ters like brain MRI lesion load including the number, central nervous system (CNS) disease characterized by an volume and location of lesions, as well as the presence of unpredictable relapsing-remitting course. In MS and other enhancing lesions and brain atrophy can predict disease autoimmune diseases, a relapse is defined as the new outcome [24]. In CIS patients, T2 lesion volume at onset onset or worsening of clinical neurological symptoms, correlates with disability over the next 10 years, and with and is followed by periods of remissions with no disease the time to progress to definite MS [25]. activity. Relapses are the basic feature of MS and other autoimmune diseases such as myasthenia gravis [1], sys- The possibility to use peripheral blood gene expression temic lupus erythemathosus [2], rheumatoid arthritis [3], analysis for prediction of clinical outcome in MS patients and Crohn's disease [4]. In MS, relapses are the conse- was demonstrated in our previous work [26] where we quence of complex immunological and neurodegenera- showed that peripheral blood mononuclear cells (PBMC) tive processes. Relapses in MS are associated with myelin gene expression based classifier correctly predicted disease and axonal loss; they may cause new acute inflammatory progression for two years. Another relevant work is the lesions or can activate old lesions within the CNS [5-7]. work of Baranzini et al. [27], they showed that PBMC gene Accordingly, relapses are the visible clinical expression of expression can be used to predict the response of MS the complicated immunopathological mechanisms oper- patients to recombinant human interferon beta (rINFβ). ating in the CNS and peripheral blood. The ability to pre- dict the occurrence of a subsequent relapse (yes/no) and The aim of current study was to evaluate whether it is pos- to estimate the time when that process will occur has sible to use peripheral blood gene expression to predict important clinical and practical implications. This knowl- the time to next acute relapse in CIS and relapsing-remit- edge can help in decisions related to treatment – e.g. either ting MS patients. treat patients with more aggressive disease or avoid over- treatment of patients with a more favorable disease Most of the new radiological MRI lesions are clinically course. Prediction of the time to next relapse can also be silent. The frequency of new radiological lesions is ten useful in the design of clinical trials as an additional crite- times higher than the frequency of clinical relapses; i.e., rion for selecting active patients. For patients with clini- on average, a cumulative effect of about 10 new radiolog- cally isolated syndrome (CIS), who have just experienced ical MRI lesions is equivalent to one clinical relapse [28]. the first relapse, such a tool can be used for estimating the Since most clinical relapses are associated with new MRI probability to convert to definite MS by predicting the lesions, and since MRI measurements was available only time until the second relapse. for small fraction of the patients, in the current study we focused only on clinically definite MS relapses. Biologically, analysis of genes and pathways that are related to predicting relapses may help to better under- We designed a comprehensive feature selection procedure stand the mechanisms underlying the progression of the that was implemented on different sets of feature includ- disease, and more specifically the processes that trigger ing: 1) all genes represented in the microarray; 2) set of and operate in acute MS relapses. genes significantly discriminated between groups of dif- ferent classes of time to next relapse; 3) genes significantly Various demographic and disease-related variables have correlated with time to next relapse, and 4) clinical and been utilized for predicting clinical outcome. Late age at demographical confounders. disease onset, poly-symptomatic symptomatology at onset, higher annual relapse rate and short time-interval This approach enabled us to identify a PBMC gene expres- between attacks are correlated with poor outcome, while sion based predictors that envisioned the time until the onset with the presentation of optic neuritis or sensory next relapse with high accuracy. symptomatology have been associated with a good out- come [8-15]. Disease-related variables, measurements of Methods autoantibodies, and gene expression were found to be useful for diagnosis and prognosis in MS and in other Subjects autoimmune diseases [16-18]. For example, in the case of The study was approved by the Sheba Medical Center MS, Martinez-Yelamos et al [19,20] showed that CSF-TAU Institutional Review Board, and all patients gave written and 14-3 3 proteins are independent predictive factors for informed consent for participation. short time conversion to clinical definite MS. On the other hand, the correlation between anti-myelin antibodies and It was a prospective collection of data. The data set time to next relapse in CIS patients has produced contro- includes 94 patients, 62 patients diagnosed with definite versial results [21-23]. MS according to McDonald criteria [29] and 32 patients Page 2 of 14 (page number not for citation purposes) 23

BMC Medical Genomics 2009, 2:46 http://www.biomedcentral.com/1755-8794/2/46 with CIS. All patients were free of steroids and immu- scores were recorded consecutively. Time from baseline nomodulatory treatments for at least 30 days before gene expression analysis to next acute relapse was blood withdrawal, and were at least one year after treat- recorded and used as a variable for clinical outcome pre- ment with cyclophosphamide (see Additional file 1). As diction. can be seen, 62 patients had not experienced previous treatment; the rest of the patients had average distance of RNA isolation and microarray expression profiling 300.4 ± 68.3 days from previous immunomodulatory The blood samples were collected for this analysis. After treatment. The time to next relapse was not a selection cri- blood withdrawal, PBMC was immediately purified and terion for inclusion patients in study; we randomly sam- frozen in liquid nitrogen for the future microarray analy- pled the blood of 100 patients (94 microarrays passed sis. Microarray analysis was performed each time that a quality control criteria). large enough set of samples was collected (between 10–12 samples for a working set). We excluded from the study patients with Neuromylitis Optica (NMO) according to the criteria of Wingerchuk et PBMC were separated on Ficol-Hypaque gradient, total al. [30]. RNA was purified, labelled, hybridized to a Genechip array The demographical characteristics of the patients are pre- sented in Table 1. Patients were followed-up prospectively (either HU-133A or HU133A-2) and scanned (Hewlett for a maximal period of 3.5 years (1264 days) or up to the Packard, GeneArray-TM scanner G2500A) according to first next acute relapse during the follow up period. Neu- the manufacturer's protocol (Affymetrix Inc, Santa Clara, rological examination was performed every 3 months and CA, USA). All microarrays used in analysis passed all the at the time of a suspected relapse, Expanded Disability stringent quality control criteria. The gene expression Status Scale (EDSS) assessment was completed accord- measurements used in this study are available and can be ingly. downloaded from Gene Expression Omnibus (http:// www.ncbi.nlm.nih.gov/geo/; accession number During the follow-up period, 33 definite MS patients ini- GSE15245). tiated various immunomodulatory treatments (Table 2) while 61 patients remained untreated. Data Analysis Data analysis was performed by Partek Genomics Solu- As the aim of this study is the prediction of the time till the tion software http://www.partek.com. Expression values next relapse we gathered patients with large range of this were computed from raw CEL files by applying the Robust parameter; our dataset included patients with long period Multi-Chip Average (RMA) background correction algo- between relapses. Some of these patients have benign MS rithm. The RMA correction included: 1) values back- (if they have EDSS < 3.0 after 15 years with the disease). ground correction; 2) quantile normalization; 3) log2 Patients with benign MS are not necessarily treated in our transformation; 4) median polish summarization. In country. Additionally, CIS patients (32 of the patients) are order to avoid the noise caused by variable set effects we not treated according medical regulations in our country. normalized each set to pre-saved distribution pattern of a Thus, in summery, our study included relatively high well balanced set used as a reference distribution. number untreated patients To reduce batch effect, Analysis Of Variance (ANOVA) Definition of Relapses multiple model analysis was applied. Source of variation MS relapse was defined as the onset of new objective neu- was analyzed; non-relevant batches effects such as array rological symptoms/signs or worsening of existing neuro- type, working batch, patient age and gender were elimi- logical disability, not accompanied by metabolic changes, nated. Most Informative Genes (MIGs) were defined as fever or other signs of infection, and lasting for a period of genes that distinguished between the different classes of at least 48 hours accompanied by objective change of at time to next acute relapse with p < 0.01 by ANOVA test. least 0.5 in the EDSS score. Confirmed relapses and EDSS Most Correlated Genes (MCGs) were defined as genes that Table 1: Clinical and demographical characteristic of the analyzed patients. F/M Yes/No Future IMD Treatment Age (Years) Disease Duration (Years) Annual Relapse Rate EDSS CIS 32.1 ± 1.5 0.20 ± 0.02 ------ 0.9 ± 0.2 19/13 0/32 Definite MS 38.5 ± 1.4 8.50 ± 1.09 0.92 ± 0.1 2.4 ± 0.2 41/21 33/29 Page 3 of 14 (page number not for citation purposes) 24

BMC Medical Genomics 2009, 2:46 http://www.biomedcentral.com/1755-8794/2/46 Table 2: The immunomodulatory treatments the analyzed patients underwent after blood sampling. CIS Definite MS Non treated Interferon β- Interferon β- Interferon β- Glatiramer Intravenous patients 1a Avonex 1b Betaferon 1a Rebif acetate Immunoglobulins Iv-Ig Copaxone 6 No of 32 62 61 5 2 10 10 Patients were correlated with time to next acute relapse by Spear- ate linear predictor gave better results than SVM regressor man, Pearson or Kendal method with p < 0.05. (the Partek implementation of SVM regressor gave error rate that was much higher, around 0.8). Dividing the patients to classes We divided the patients to three classes comprising at least In the case of the FTP, the gene expression of each predic- 20 patients. Each class corresponded to relatively similar tive gene is multiplied by a weight (positive or negative time ranges until their next relapse: a) 31 patients that had real number) and the results are summed. This sum is not experienced relapse in the 1264 days of the follow up used as a prediction for the number of days till the next period; b) 40 patients that experienced relapse in less than relapse. Thus, the expression level of each gene may have 500 days and c) 23 patients that experienced relapse in a positive or a negative affect on the prediction. In the case 500 to 1264 days. The boundaries of the classes were cho- of the FLP, which is based on SVM classifier [31], the gen- sen according to various constrains: 1) the upper bound eral idea is similar but more complicated – each gene may (1264) reflects that the patients' follow up period after have a positive effect on the prediction in some cases and blood withdrawal. 2) We wanted to divide the time rang negative effect in other cases. 0–1264 to relatively similar ranges such that the number of patients in each group will be similar (at least 20 The feature selection procedure for finding the most patients). Thus, we decided on the second boundary predictive genes (500), a larger boundary would decrease the number of We used different sets of genes as input to our feature patients in the second group while a smaller one would selection procedure: a) genes that were differentially result in a too small time range of the first group. expressed between the three different classes of time till the next relapse, b) genes that were expressed with corre- The distribution of subjects with CIS and Definite MS lation (p < 0.05) with the time till the next relapse, c) all across these three categories is [30:16:15] for definite MS unique genes presented on the microarray. and [9: 5:14] for CIS for the three classes (less than 500 days: between 500 and 1264 days: more than 1264 days) Using a very large set of genes for constructing the predic- respectively. As can be seen, in both groups there are sub- tors elevates the risk of overfitting (e.g. see [32]). The solu- jects in each of the three categories. tion for overfitting problem is the Leave 20% Out Cross validation (L20OCV) procedure (a version of leave one In the case of the FTP, we used resolution of 50 days since out cross validation [33,34]) that is described in this sub- we wanted to divide the range related to the first group to section. equally spaced sub-ranges (thus, for example, 30 or 90 days were inappropriate). It is important to note that the A general flow diagram of the forward selection algorithm concept used here could be used with different resolutions for choosing the set of genes consisting each of the predic- and still give qualitatively similar results (for example tors is described in Figure 1. The same feature selection lower resolution of 100 days gave error-rate of less than algorithm was used both for the FLP and for the FTP. 0.2, p < 0.0001). We started with a set of 22,215 gene-transcripts present on Implementations of the predictors each microarray. The expression levels of multiple probes The predictor has two major parts: 1) a Support Vector related to the same gene were averaged, resulting in a set Machine (SVM) classifier, that we named First Level Pre- of 10,594 potential features (genes and annotated dictor (FLP), and 2) a multivariate linear regressor that we sequences). At the initial stage, we computed a vector of named Fine Tuning Predictor (FTP). We used the Spider errors in 100 (80% training set, 20% test set) leave 20% http://www.kyb.tuebingen.mpg.de/bs/people/spider/ out iterations for each gene. We chose the gene with the implementation for the SVM multi-class classifier. We lowest mean error rate and all the genes whose mean error used a multivariate linear regressor for the FTP (for rate that is not significantly higher than the error rate of patients with time till next relapse < 500). This multivari- the first gene (p-value < 0.05 by t-test). Page 4 of 14 (page number not for citation purposes) 25

BMC Medical Genomics 2009, 2:46 http://www.biomedcentral.com/1755-8794/2/46 The initial predictor includes 0 genes. Evaluate each predictor in the Have adding the last gene yielded NO set of the most predictive a significant improvement of the predictors by averaging the performances? error of K'>>K L20OCV of the Yes data to 20% training and 80% Expand the Mi best predictors of i genes with N/Mi most test set. predictive genes i stages Plot the set of predictors that Evaluate each predictor by are not significantly worse than averaging the error of K the best predictor L20OCV of the data to 20% training and 80% test set. Legend: N - the total no of genes Choose the Mi+1 predictors with Mi- the no of best predictors in stage i i+1 genes that are not K - no of L20OCV at the initial stages K' - no of L20OCV at the final stage significantly worse than the best predictor with i+1 genes FGiegnuerreal1flow diagram of the procedure for finding the predictors (FLPs or the FTPs, we used k = 100 and k' = 10,000) General flow diagram of the procedure for finding the predictors (FLPs or the FTPs, we used k = 100 and k' = 10,000). In the following stages, we expanded the initial sets while rest of the patients). In this case, the mean error rate of the getting significantly better predictors which were based on FLPs was less than 0.11 on the training set and 0.13 on the 2, 3 or more genes. In each step, we tried to expand each test set (all p-values < 0.001). In the case of the FTPs, the predictor in the current set of best predictors by adding mean error rates on the training set and test set were 0.44 more genes to the predictors, and while keeping all the and 0.66 respectively (all p-values < 0.001). predictors whose error rates were not significantly worse than the error rate of the best predictor. This was done by Evaluating the Performance of the Predictors on an iterative cross validation procedure in which 20% of Subgroups of Patients the initial data set was left apart and the remaining 80% For evaluating the performances of the predictors on sub- were used as temporary training set in each of the itera- groups of the patients (e.g. MS vs. CIS, or non-treated vs. tions. We performed 100 iterations for the cross valida- patients under the various treatments), we performed tion procedure at each stage for selecting sets of potential 1000 Leave 20% Out Cross Validation (L20OCV) proce- predictors. The output of the initial stage was a set of pre- dures where in each L20OCV step we randomly chose dictors with similar performances [i.e. according to the subsets of 80% of each of the patient subgroups for train- 100 leave 20% outs, none of them was significantly better ing the predictors (all these subsets were unified to one than the best predictor; namely, all the p-values were > subset), and tested the predictors on the rest of the sam- 0.05]. ples (in the cases where the dataset was very small, at least one patient was chosen for the training set and for the test At the final stage, to discriminate between the predictors set). The predictors were based on the sets of genes that that have similar performances as the best predictor, we were found by the procedure that was described in the performed 10,000 cross validation iterations and selected previous subsection, but in each iteration a different train- the predictor(s) that was/were significantly better than the ing set for inferring the weights of the different genes in others. the predictor was used; and in each iteration the predic- tors were implemented on different test set. The final error Note that qualitatively similar results were gained when rate of each subgroup is the average error rate across all the we learned the predictors based on 85 of the patients and 1000 L20OCV procedures for patients from the subgroup. afterward test them on a different group of 9 patients (the Page 5 of 14 (page number not for citation purposes) 26

BMC Medical Genomics 2009, 2:46 http://www.biomedcentral.com/1755-8794/2/46 Another Set of Patients for Additional Validation Resources http://david.abcc.ncifcrf.gov/home.jsp and For further evaluation of the FLP performances, we col- Ingenuity Pathways Analysis web-software http:// lected an additional dataset of 10 patients (3 from the first www.ingenuity.com. Information about the most predic- group, 2 from the second group, and 5 from the third tive genes was extracted from NCBI http:// group). The gene expression of each of the patients was www.ncbi.nlm.nih.gov/sites/entrez?db=gene. normalized separately with the original dataset of 94 patients. Then, the FLPs were implemented on the nor- Results malized expression levels. The demographical and clinical characteristics of these patients are presented in Table 3. To learn about the three groups of patients, we performed Principal Component Analysis (PCA) and clustering anal- The Role of Clinical and Demographical Variables in the ysis of the patients based on 1359 MIGs (Additional file Predictors' Performance 2). The patients with time until next relapse < 500 days We examined if clinical parameters are helpful for predict- exhibit a relatively coherent clustering where 29/40 of the ing the time to next relapse by 1) evaluating the perform- patients appear in the same cluster. The clustering results ance of a FLP and FTP predictors that are based only on of the other groups were much worse as they were parti- these parameters, and 2) examining if these variables can tioned among many clusters with up to 5 patients in the improve the performances of our predictors. To this end, same cluster. This result demonstrates that it is much we checked the following clinical parameters: age, MS more complicated to cluster sub-groups of patients that stage (CIS or Definite), gender, annual relapse rate, EDSS will experience their next relapse in more than 500 days. at time of blood sampling, disease duration, age at onset, Thus, an additional, finer predictor only for the first group EDSS change in the last relapse. (time until next relapse < 500) was justified. Empirical p-values for the predictors We named the three groups classifier First Level Predictor We computed empirically p-values for the best FLP and (FLP). The FLP classified the time till next relapse of a FTP by performing random permutation of the labels, patient to one of the following groups: <500 days, learning best predictors for each such permutation, and between 500 and 1264 days, and >1264 days. The finer computing the fraction of cases (out of 1000 permuta- predictor, for the patients with time till next relapse < 500 tions) that a predictor for permutated labels gave better days was named Fine Tuning Predictor (FTP). error rate than the original predictor. All the 1000 random predictors were much worse than the original one (p- Based on expression of all 10,594 genes, we performed a value < 0.001). For example, the average error rates of the Leave 20% Out Cross Validation (L20OCV) procedure to random FLP were 0.67 and the average error rates of the evaluate FLP that simultaneously discriminated between random FTP were 0.8. patients that experienced acute relapse in one of the three periods mentioned above during clinical follow up (as The Role of Relapse Severity in the Predictors' was described in the Methods section). The cases where Performance the FLP does not classify patients to their correct groups We examined if the performances of the predictors are defined as errors. The output L20OCV was a group of depend on the severity of the relapse (measured as the predictors that are based on the gene expression of sets of change in EDSS in the last relapse). To this end, we 1 to 10 genes with error rate range between 0.37 (for sin- divided the patients to 8 groups according to their change gle gene) to 0.079 (for 10 genes), see Figure 2A. There are in EDSS in the last relapse (the range was 0.5 – 6.5). We a few dozen predictors that gave similar results. For exam- computed the error rate of the predictors for each of these ple, Additional file 3 includes 36 FLP with error rate < 0.1. groups (as was performed for the treatment groups). To differentiate a smaller set of statistically significant pre- Biological functional analysis dictors from those with similar predictive ability we per- Gene functional annotation was performed using func- formed 10000 additional L20OCV iterations which tional classification tools such as David Bioinformatics enabled us to choose three best FLP (each is based on 10 genes; see Table 4). Table 3: Clinical and demographical characteristic of the additional validation set of patients (4 CIS and 6 Definite MS). Age (Years) Disease Duration (Years) Annua Relapse Rate EDSS F/M Yes/No Future IMD Treatment CIS 24 ± 5.01 0.34 ± 0.09 6.1 ± 2.05 2.58 ± 0.15 0/6 6/0 Definite MS 36 ± 7.61 5.3 ± 2.39 1 ± 0.51 5.3 ± 2.39 2/2 3/1 Page 6 of 14 (page number not for citation purposes) 27

BMC Medical Genomics 2009, 2:46 http://www.biomedcentral.com/1755-8794/2/46 A. 0.4 Error Rate First Level Predictor B. 0.7 First Level 1 2 N3 um4 b5er6of7Ge8 ne9 s 10 Predictor 0.35 0.6 0.3 0.5 Probability 0.25 0.4 0.2 0.3 0.15 0.2 0.1 0.05 0.1 0 Error Rate0 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 C. 0.7 D. 0.35 Fine Tuning Predictor 0.6 0.3 0.5 Error Rate Fine Tuning Predictor Probability 0.25 0.4 0.2 0.3 0.15 0.2 0.1 0.1 0.05 0 00 0.1 0.2 0E.3rro0.4r R0.a5 te0.6 0.7 0.8 1 2 N3 u4mb5er6 of7 G8en9es10 FErigrourrepr2obabilities of the FLP and the FTP Error probabilities of the FLP and the FTP. A. The improvement in the error probability of the best FLP as function of the number of predicting genes. Each improvement was significantly better (p-value < 0.05) than the previous one (see Meth- ods). B. The error rate distribution of the best FLP on the test sets. C. The improvement in the error probability of the best FTP. Each improvement was significantly better (p-value < 0.05) than the previous one (see Methods). D. The error rate distri- bution of the best FTP on the test sets. Table 4: The performances (error rate) of the predictors in different subgroups of patients. Predictor CIS Definite Non Interferon Interferon Interferon Glatiramer Intravenous Average MS treated beta 1a beta 1b beta 1a acetate Immunoglobulines error r ate Avonex Rebif Copaxone Iv-Ig Betaferon FLP1 0.10 0.08 0.087 0.01 0.01 0.09 0.18 0.175 0.079 FLP2 0.10 0.08 0.085 0.02 0.03 0.05 0.215 0.18 0.0791 FLP3 0.088 0.071 0.082 0.01 0.02 0.105 0.165 0.17 0.0792 FTP1 0.21 0.355 0.332 -- -- 0.42 0.54 0.34 0.345 FTP2 0.29 0.38 0.366 -- -- 0.345 0.51 0.3 0.349 3FTP 0.27 0.37 0.366 -- -- 0.245 0.535 0.41 0.349 FTP4 0.31 0.38 0.352 -- -- 0.235 0.53 0.37 0.349 FLP1 – FLP3 denote FLPs and FTP1 – FTP4 denote FTPs; see Table 5 for the genes in each predictor) on each sub-group (when applicable), and the average error-rate (last column). The average error rate is the error rate when considering the entire dataset. Page 7 of 14 (page number not for citation purposes) 28

BMC Medical Genomics 2009, 2:46 http://www.biomedcentral.com/1755-8794/2/46 The improvement in the error rate of the FLP as a function were not improved (see Supplementary Note 1 in Addi- of the number of genes that were used by the predictor is tional file 5). demonstrated in Figure 2A. Genes were added iteratively to the FLP until there was no significant improvement in To visualize boundaries of the FLP decision regions we the performances of the predictor (a flow diagram appears performed a plot of the expression levels of the 2 most in Figure 1). predictive genes (FLJ10201 and PDCD2) of the best FLP and the boundaries of the decision regions of the predic- The error rate distribution of the best FLP is depicted in tor (Figure 4). As can be seen, the boundaries of these Figure 2B. As can be seen, for 50% of patients the predic- regions are non-linear (see Additional file 6 for graphs of tor gave absolutely correct classifications (i.e. it correctly other pairs of predictive genes), and resemble the results classified all the patients to one of three groups), in 20% that were reported in [26]. Finally, the error rate of the of the cases the error rate was 0.05, and only 30% of best FLPs on an additional independent dataset of 12 patients were predicted with error rate more than 0.05. patients was 0.3 (p-value < 0.001; see Methods); further The 10 genes involved in each of the 3 best FLPs are pre- supporting the viability of the FLPs. sented in Table 5. The functional annotations of those genes appear in Additional file 4. Next, we designed a more accurate predictor that was named Fine Tuning Predictor (FTP). It predicts the time The probabilities of the different types of classification until the next relapse only for patients that experience errors of the best FLP are depicted in Figure 3. For exam- acute relapse during a period of 500 days. As a FTP we ple, as can be seen in Figure 3, a patient that belongs to used a multivariate regressor (see Methods) that can pre- group 1 (relapse in less than 500 days) has a probability dict the time until the next relapse with a resolution of a of 0.03 of being misclassified by the FLP and to be few days. In the case of the FTP, we defined a prediction included in group 2 (relapse in 500 – 1264 days), and a error as a prediction that is more/less than 50 days (± 50) probability of 0.023 to be misclassified and to be included from the real date of relapse onset. We found 240 gene in group 3 (relapse in more than 1264 days). In compari- sets that gave error rate < 0.36. Our feature selection pro- son, a random assignment to one of these groups gave an cedure combined with 10000 permutations of Leave One error rate of 0.67. 20% Cross Validation (L20OCV) procedure found four FTP s; each FTP was based on 9 genes. The minimal error- The performances of the predictor were significant (p < rate of each FTP was 0.35 (p-value < 0.001); and was sig- 0.001; details about the p-value calculation appear in the nificantly better than the other gene sets. The error rate of Methods section). the FTP after random permutations of the labels was 0.8; this is 2.3 folds higher than the error rate of the inferred When we implemented the feature selection procedure FTP (see Methods for description about the p-value). The using only MIG genes the performances of the result FLP error rates of best 9-genes-FTPs are demonstrated in Table 4. Table 5: The genes that were selected for the best FLPs (FLP1 – FLP3) and best FTPs (FTP1 – FTP4). Predictor Predictor Error Gene 1 Gene 2 Gene 3 Gene 4 Gene 5 Gene 6 Gene 7 Gene 8 Gene 9 Gene 10 name Rate FLP FLP1 0.079 FLJ10201 PDCD2 IL24 MEFV CA2 SLM1 CLCN4 SMARCA1 TRIM22 TGFB2 FLP2 0.0791 FLJ10201 PDCD2 IL24 MEFV CA2 SLM1 CLCN4 SMARCA1 TRIM22 SPN FLP3 0.0792 FLJ10201 PDCD2 IL24 MEFV CA2 SLM1 CLCN4 SMARCA1 TRIM22 TP73L FTP FTP1 0.345 KIAA1043 LOC51145 PPFIA1 MGC8685 DNCH2 PCOLCE2 FPRL1 G3BP RHBG --- FTP2 0.349 KIAA1043 LOC51145 PPFIA1 MGC8685 DNCH2 TAF4B FPRL1 PCOLCE2 FLJ21802 FTP3 0.349 KIAA1043 LOC51145 PPFIA1 MGC8685 DNCH2 PCOLCE2 FPRL1 FLJ21802 TAF4B FTP4 0.349 KIAA1043 LOC51145 PPFIA1 MGC8685 DNCH2 PCOLCE2 FPRL1 TAF4B FLJ21802 Full description of the genes and their gene bank ID appears in Additional file 4. The expression levels of the genes appear in Additional file 1. MEFV is an abbreviation of Homo sapiens familial Mediterranean fever locus region, mRNA sequence. Page 8 of 14 (page number not for citation purposes) 29

BMC Medical Genomics 2009, 2:46 http://www.biomedcentral.com/1755-8794/2/46 0.023 Less 0.03 Over 1264 than 500 Days Predicted time Days 0.05 0.003 0.018 0.15 500 - 1264 Days PDCD2FThigeudrieffe3rent types of errors of the best FLP (FLP1) Observed time The different types of errors of the best FLP (FLP1). The number on an arrow from state x to state y is the prob- TFthihegeunprerexetd5ricetleapdsteim(ien tdoaynse)xftoretlhaepsbeevset rFsTuPs observed time of ability that the predictor miss-classify a patient whose true The predicted time to next relapse versus observed state is x and will put it in state y. This is an extension of the time of the next relapse (in days) for the best FTP. widely used two states sensitivity and specificity measures. The graph demonstrates a very high correlation between these two values (Spearman correlation 0.82, p-value = 10- As in the case of the FLP, we performed a similar analysis 10). of the improvement in the error rate of the best FTP (see Table 4) as function of the number of predictive genes (Spearman correlation 0.82, p-value = 10-10). The analysis (from 1 to 9; Figure 2C). Every time a gene was added to of error rate distribution of the best FTP appears in Figure the FTP, the performances of the FTP were significantly 2D. In this case, the error rate has normal distribution improved (see Methods). with mean error rate of 0.35; for example, 20% of the The plot of best FTP performances vs. observed time to patients have error rate < 0.2 (Figure 2D). next relapse during 500 days of follow up appears in Fig- ure 5. As can be seen, the two values are very correlative When we implemented the feature selection procedure only on genes whose expression is correlative with time FLJ10201 until next relapse the FTP results were not improved (see Supplementary Note 2 in Additional file 5). TrdmFeahiolgyaesupttrs–hpeerreie4ned5)ic0lata0issv–sef1uifg2nice6cant4tieioodsnnaoyorfsfet–gthhieogernFegsLeeP(nnr,,eFrlLaeJpxl1asp0per2seei0ns1sin5ioa0mn0doodrPfaeDtyhtsCheD–atn2bwl1uo2e6, 4 The three classification regions (relapse in 500 days – To further evaluate the performances of the predictors, we blue, relapse in 500–1264 days – green, relapse in divided the patients in two ways. First, the dataset was more than 1264 days – red) as function of the gene divided to CIS and definite MS patients (32 and 62 expression of the two most predictive genes of the patients respectively); second, the dataset was divided FLP, FLJ10201 and PDCD2. Each point denotes a patient. into 6 groups according to their future treatment after their blood was withdrawn (Non-treated, Interferon β-1a (Avonex), Interferon β-1b (Betaferon), Interferon β-1a (Rebif), Glatiramer Acetate (Copaxone) and Intravenous Immunoglobulins (Iv-Ig) with 61, 5, 2, 10, 10, and 6 patients respectively (see Table 2). The performances of the best predictors in different subgroups of patients were evaluated as was described in the Methods section. Table 4 depicts the performances of each of the best pre- dictors (FLPs and FTPs) on each of these groups. The error rates remain significant for each of these groups. The FLP error rates were 0.1 and 0.08 (p-value < 0.001) for the CIS and MS groups respectively; the FTP error rates were 0.21 and 0.355 (p value < 0.001), for the CIS and MS groups respectively. Page 9 of 14 (page number not for citation purposes) 30

BMC Medical Genomics 2009, 2:46 http://www.biomedcentral.com/1755-8794/2/46 In the analysis of the different treatments, the FLP error ingly good. For comparison, our simulation showed that rate ranged between 0.087 (non-treated) to 0.18 (Copax- a random FTP (i.e. the best FTP after a random shuffling one); the FTP error rate ranged between 0.235 (Rebif) and the input labels) gave an error rate that was close to 0.8 – 0.53 (Copaxone), see Table 4. The probabilities of the dif- 2.3 fold higher than our error rate. ferent types of classification errors for each treatment group appear in Additional file 1. In all sub-groups the This study includes 94 patients for evaluating the predic- best predictor exhibited significant performances (p-value tors. It is clear that a larger dataset will give better perform- < 0.001); this fact suggests that the signal of the next ances. In order to estimate the potential improvement in relapse is usually strong enough to be detected by our pre- the error rate when using larger datasets we performed the dictor even when the patients undergo various Immu- following analysis: we computed the error rate of the FLP nomodulatory Drugs (IMD) treatments after blood and the FTP as function of the dataset size (% of the orig- withdrawal. inal dataset; see Additional file 8). The figure shows that the error rates decrease for larger datasets. This fact sug- For additional validation of the FLP, we collected an addi- gests that with the accumulation of more gene expression tional independent dataset of 10 untreated RRMS patients measurements we can design better predictors. Specifi- (see more details about this dataset in the Methods sec- cally, enlargement of the dataset to 200 patients (instead tion and Table 3). The total error rate of the FLP on these of 94) will give a classification error of about 0.05 and a patients was 0.25. This result further supports the viability regression error of about 0.2 (Additional file 8). of our approach. Another interesting conclusion from this work is that Clinical and demographical confounders did not improve there are multiple predictors (FLPs and FTPs) that have the performance of the best gene expression based FLP similar performances. The predictors that were described and FTP (see Methods for the exact list of Clinical and in this work were significantly better than the other pre- demographical confounders we checked). FLP and FTP dictors; however, there were a few dozen predictors that predictors that were based on combination of these con- gave similar results. For example, Additional file 3 founders and that were found by our approach had much includes 36 FLP with error rate < 0.1, and 240 FTP with higher error rates than the predictors that were based only error rate < 0.36. This means that the best predictors on gene expression (FLP error rate > 0.63, FTP error rate > appear in this work can be replaced by other predictors 0.74; for all the predictors that were based on the above with a relatively small influence on the error rate. mentioned confounders). We also did not get a significant correlation between the error rate of a predictor and the Finding a good predictor for the time to next relapse and severity of relapse (as was measured by the change in finding a molecular explanation for relapses are different EDSS), Spearman correlation -0.29, p-value = 0.5, for FTP; tasks with a possible overlap. There are a few explanations Spearman correlation -0.6, p-value = 0.12, for FLP, see why the connection between the predictive genes and Methods for more details). relapse associated mechanism is not necessarily immedi- ate: First, the predictors were designed to include rela- Discussion tively small number of genes while the actual mechanisms may include dozens of signaling pathways. Second, our In this work, we demonstrated that gene expression in study was based on changes in gene transcription levels; it PBMC can be used for predicting the time of the upcom- is possible that major parts of the relapse associated regu- ing relapse in MS. Using different prediction strategies to latory mechanisms are post transcriptional. In such cases, determine an appropriated gene set for accurate relapse the most relevant genes are useless in terms of improving prediction we found that the classifier that was based on the predictions and the feature selection procedure finds all microarray genes had the best prediction. We describe genes that are less relevant but that exhibit significant a FLP, which is based on the expression levels of ten genes, change in their mRNA levels (e.g. genes that are regulated/ which can predict the time till next relapse in a resolution regulate the genes that are directly related to relapse). of 500 days during 3.5 years of disease progression. An additional FTP, which is based on different set of nine However, many of the FLP genes are linked to MS. This is genes, can be used for a prediction of a higher resolution an additional support of their predictive ability. For exam- (e.g. a resolution of 50 days). ple, the gene TGFB2 is a one of the master genes in MS; it is closely related to a rapid recovery from relapses that is At first glance, the error rate (about 0.35) of the FTP seems mediated by Th2/Th3 lymphocytes. Th2/Th3 lymphocytes relatively high. However it is important to remember that produce anti-inflammatory cytokines (like IL10) [35]. the definition of error in this case was very tight (more TGFB inhibits IL12 mediated inflammatory response, and than 50 days from the real value). As mentioned, this error it virtually decreases T cells proliferation and IFNgamma rate was very significant (p-value < 0.001) and surpris- Page 10 of 14 (page number not for citation purposes) 31

BMC Medical Genomics 2009, 2:46 http://www.biomedcentral.com/1755-8794/2/46 production [36]. TGFB prevents induction of pro-inflam- FTP FLP matory gene-program by inhibiting the expression of 25% of the TNFalpha/IFNgamma induced genes [35]. The target PCOLCE2 IL24 genes that TGFB inhibits are various genes that are involved in MS pathogenesis processes (e.g. chemotaxis, FPRL1 TGFB2 IFNG SLM1 adhesion and cell migration). G3BP CA2 MEFV PDCD2 TRIM22 The gene for Familial Mediterranean Fever (MEFV) is expressed in early leukocyte development and is regulated Stimulation in response to inflammatory mediators. Stimulation of Suppression cells with the proinflammatory agents interferon (IFN) gamma, tumor necrosis factor, and lipopolysaccharide FRFDLeiigPsgcuaurlnasetdsoio6FrnTy Psne;ectthtwieoonr)ektowfotrhkeisgebnaessedthoant atrhee pliatertraotfutrhee(sbeeestthe induced MEFV expression, whereas the anti inflammatory Regulatory network of the genes that are part of the cytokines (IL4, IL-10) and especially TFGB inhibited such best FLP and FTP; the network is based on the litera- expression [37]. ture (see the Discussion section). The CA2 gene (carbonic anhydrase II) supports the trans- strated in activated brain prion plaques and brain lesions port of bicarbonate ions, sodium ions, and water from in Alzheimer disease [46]. PPFIA1 receptor gene is blood to the CSF; and in the myelin sheath CA2 supports involved in cell motility, cells spreading, migration and compaction of myelin by stimulating co-transport of ions adhesion. Up regulation of human PPFIA1 (LIPRIN) gene between the myelin membranes. The double mutant mice in peripheral blood is associated with psoriatic arthritis deficient by CA2 and myelin displayed tremors and sei- [47]. Another interesting gene is G3BP that encodes a zures [38]. Interestingly, the onset of seizures was delayed downstream effector protein of the Ras signaling pathway significantly in the double mutants, and the lifespan [48]. Interestingly, as in the case of best FLP, the genes increased by several months, this fact corroborates with FPRL1 and G3BP are regulated by TGFB [49,50]. This CA2's activity as predictor of acute onset in MS. result suggests that a single pathway can explain both the FLP and the FTP genes. Another important group of genes that are part of the best FLP is related to the interferon regulation mechanism. To summarize, most of the predictive genes seem very rel- This group includes RNA binding and signal transduction evant to the pathophysiology of MS. We have constructed SLM1 gene, that is associated with Interferon Receptor 1 a schematic regulatory network which unifies many of Binding Protein 4 (IR1B4) [39] and TRIM22, an important these genes to a single regulatory network (Figure 6). One member of interferon related genes, that is involved in important goal for further research is to better understand transduction of IFN activity [40]. how these genes are involved in the biological mecha- nisms that lead to a clinical relapse. The IL24 gene is a member of anti-inflammatory IL10 family cytokines involved in immune response. The over- We also found a few predictive genes whose their biolog- expression of IL24 stimulates pro-apoptotic CADD family ical roles are unknown (see Additional file 4). We thus genes and activation of apoptosis. On the other hand, think that the potential connection of these genes to MS is IL24 can increase secretion of IFNG in human PBMC. The a natural target for a further study. IFNG by himself is able to repress TGFB mRNA expression as demonstrated in CD18 positive cells [41] and in As relapses have different level of severity [e.g. it can be human lymphocytes it increases mRNA expression of measured by the increase in EDSS] one may think that in MEFV [42]. Additionally IFNG involved in regulation of the cases of a more severe next relapse the performances the protein NFKBIB (corresponding to the gene PDCD2) of the predictors will be better. Analysis of the predictor that is associated with programmed death of lymphocytes error probability for the FLP and FTP as a function of the [43-45]. Based on the above relations we reconstructed a change in EDSS levels showed a negative relation (as unified regulatory network for most of the predictive expected – i.e. larger changes in EDSS are easier to detect). genes that appear in the FLP (see Figure 6). However, this correlation was not significant. Thus, a final answer to this question should be deferred till a larger The best FTP includes 2 inflammatory related genes gene expression dataset will be accumulated. (FPRL1 and PPFIA1). FPRL1 functions as a receptor com- ponent of inflammatory response [45], activation of FPRL1 results in leukocytes activation. FPRL1 involved in direct monocytes/microglia migration as was demon- Page 11 of 14 (page number not for citation purposes) 32

BMC Medical Genomics 2009, 2:46 http://www.biomedcentral.com/1755-8794/2/46 We used a heterogeneous dataset for inferring the predic- on clinically definite MS relapses. In the future, when such tors. The dataset included both CIS and Definite MS information will be available, it can be used for improving patients, and patients that underwent different immu- the performances of our predictor. In addition, based on nomodulatory treatments after their blood was sampled. such data it will be feasible to study the possibility to pre- A survival analysis showed that the disease stage (CIS or dict radiological MRI lesions (that are possibly clinically Definite MS) had a statistical significant influence on the silent) from gene expression in PBMC. probability to experience next acute relapse (Additional file 9, Supplementary Note 3 in Additional file 5). Further, Finally, it is possible that the techniques described here statistical analysis of the gene expression of these two will be valuable not only in future MS research but also in groups of patients showed that there are dozens of genes other autoimmune disease with relapsing-remitting that are differentially expressed in these groups (data not nature. shown). Abbreviations However, our predictor was insensitive to the disease stage and successfully dealt with this issue. The error rates for CIS: Clinically Isolated Syndrome; PBMC: Peripheral the two groups were significantly low (less than 0.1 for the Blood Mononuclear Cells; MS: Multiple Sclerosis; MIG: FLP and less than 0.37 for the FTP). This fact may suggest Most Informative Genes; MCG: Most Correlated Genes; that, in the case of the small sets of predictive genes, the SVM: Support Vector Machine; L20OCV: Leave 20% Out changes in PBMC gene expression before the second Cross Validation; FLP: First Level Predictor; FTP: Fine Tun- relapse (CIS patients) or before any other relapse (Defi- ing Predictor; RRMS: Relapsing Remitting MS; EDSS: nite MS patients) are similar. Expanded Disability Status Scale; RMA: Robust Multi-chip Analysis; ANOVA: Analysis of Variance; IMD: immu- Our dataset includes patients that underwent various nomodulatory drugs. treatments after we sampled their gene expression. We believe that these patients are the major source of error for Competing interests our predictors. On the other hand, we decided to include them in the analysis since they improved its statistical sig- The authors declare that they have no competing interests. nificance. We demonstrated that our predictor gave signif- icantly good results, also when considering each of these Authors' contributions datasets separately. This was unexpected since it is known that in general drugs change the relapse frequency. The MG, TT, and AA participated in the design of the study. TT explanation of this result is simple: Most of the treatments and UR designed the predictors. MG, TT, and RO analyzed delay the next relapse by about 30% [51-55], and this fact the data. MG, TT, and AA wrote the paper. All the authors increases the prediction error primarily for patients whose approved the final manuscript. real time to next relapse is close to the boundaries of their classification group (e.g. close to 500 or close to 1264). Additional material Since the number of such patients is relatively low the error rate remains significantly low (see details in Supple- Additional file 1 mentary Note 4 in Additional file 5). Supplementary Table 1. The previous treatment of each patient and the Conclusion number of days between the previous treatment and blood sampling; the expression levels of the predictive genes across all analyzed patients and We conclude that gene expression in PBMC can be used to the number of days till the next relapse for each of the patients. accurately predict the time until the next acute relapse. In Click here for file this work, we described a few sets of predictive genes that [http://www.biomedcentral.com/content/supplementary/1755- can be used for this purpose and demonstrated that other 8794-2-46-S1.xls] combinations may also yield significant results. It is pos- sible that different technology for measuring the gene Additional file 2 expression will yield different sets of the most predictive genes. Thus, our next goal is to find sets of predictive genes Supplementary Figure 1. Clustering and Principal Component Analysis that give significant results when their gene expression is (PCA) analysis of the patients based on gene expression of 1359 MIGs. measured by cheaper, small-scale, technologies such as Click here for file kinetic RT-PCR. [http://www.biomedcentral.com/content/supplementary/1755- 8794-2-46-S2.doc] In this work, as information about clinically silent lesions was not available for most of the patients, we focused only Additional file 3 Supplementary Table 2. FLPs/FTPs whose performances are similar to the performances of the best FLP/FTP described in the main text. Click here for file [http://www.biomedcentral.com/content/supplementary/1755- 8794-2-46-S3.xls] Page 12 of 14 (page number not for citation purposes) 33

BMC Medical Genomics 2009, 2:46 http://www.biomedcentral.com/1755-8794/2/46 Additional file 4 7. McDonald WI: Relapse, remission, and progression in multiple sclerosis. N Engl J Med 2000, 343(20):1486-1487. Supplementary Table 3. Description of the predictive genes. Click here for file 8. Scott TF, Schramke CJ, Novero J, Chieffe C: Short-term prognosis [http://www.biomedcentral.com/content/supplementary/1755- in early relapsing-remitting multiple sclerosis. Neurology 2000, 8794-2-46-S4.xls] 55(5):689-693. Additional file 5 9. Held U, Heigenhauser L, Shang C, Kappos L, Polman C: Predictors of relapse rate in MS clinical trials. Neurology 2005, Supplementary notes 1–4. All the Supplementary notes. 65(11):1769-1773. Click here for file [http://www.biomedcentral.com/content/supplementary/1755- 10. Bergamaschi R, Berzuini C, Romani A, Cosi V: Predicting second- 8794-2-46-S5.doc] ary progression in relapsing-remitting multiple sclerosis: a Bayesian analysis. J Neurol Sci 2001, 189(1–2):13-21. Additional file 6 11. Bergamaschi R: Prognosis of multiple sclerosis: clinical factors Supplementary Figure 2. The region of each of the classes as function of predicting the late evolution for an early treatment decision. the gene expression of the predictive genes. Expert Rev Neurother 2006, 6(3):357-364. Click here for file [http://www.biomedcentral.com/content/supplementary/1755- 12. Bergamaschi R, Quaglini S, Trojano M, Amato MP, Tavazzi E, Paolicelli 8794-2-46-S6.doc] D, Zipoli V, Romani A, Fuiani A, Portaccio E, et al.: Early prediction of the long term evolution of multiple sclerosis: the Bayesian Additional file 7 Risk Estimate for Multiple Sclerosis (BREMS) score. J Neurol Neurosurg Psychiatry 2007, 78(7):757-759. Supplementary Figure 3. The different types of errors of the best FLP (FLP1) for different stages of the disease and for different future IMD 13. Weinshenker BG, Ebers GC: The natural history of multiple treatment. sclerosis. Can J Neurol Sci 1987, 14(3):255-261. Click here for file [http://www.biomedcentral.com/content/supplementary/1755- 14. Phadke JG: Clinical aspects of multiple sclerosis in north-east 8794-2-46-S7.doc] Scotland with particular reference to its course and progno- sis. Brain 1990, 113(Pt 6):1597-1628. Additional file 8 15. Runmarker B, Andersson C, Oden A, Andersen O: Prediction of Supplementary Figure 4. The prediction errors as function of the dataset outcome in multiple sclerosis based on multivariate models. size (% of the used dataset) for the FLP and the FTP. J Neurol 1994, 241(10):597-604. Click here for file [http://www.biomedcentral.com/content/supplementary/1755- 16. Allantaz F, Chaussabel D, Stichweh D, Bennett L, Allman W, Mejias A, 8794-2-46-S8.doc] Ardura M, Chung W, Wise C, Palucka K, et al.: Blood leukocyte microarrays to diagnose systemic onset juvenile idiopathic Additional file 9 arthritis and follow the response to IL-1 blockade. J Exp Med 2007, 204(9):2131-2144. Supplementary Figure 5. Survival analysis of MS patients with definite MS and patients with Clinically Isolated Syndrome. 17. Oken O, Batur G, Gunduz R, Yorgancioglu RZ: Factors associated Click here for file with functional disability in patients with rheumatoid arthri- [http://www.biomedcentral.com/content/supplementary/1755- tis. Rheumatol Int 2008, 29(2):163-166. 8794-2-46-S9.doc] 18. Vojdani A: Antibodies as predictors of complex autoimmune Acknowledgements diseases. Int J Immunopathol Pharmacol 2008, 21(2):267-278. T.T. was supported by the Edmond J. Safra Bioinformatics program at Tel 19. Martinez-Yelamos A, Rovira A, Sanchez-Valle R, Martinez-Yelamos S, Aviv University and the Yeshaya Horowitz Association through the Center Tintore M, Blanco Y, Graus F, Montalban X, Arbizu T, Saiz A: CSF for Complexity Science. 14-3-3 protein assay and MRI as prognostic markers in patients with a clinically isolated syndrome suggestive of MS. References J Neurol 2004, 251(10):1278-1279. 1. Conti-Fine BM, Milani M, Kaminski HJ: Myasthenia gravis: past, 20. Martinez-Yelamos A, Saiz A, Bas J, Hernandez JJ, Graus F, Arbizu T: present, and future. J Clin Invest 2006, 116(11):2843-2854. Tau protein in cerebrospinal fluid: a possible marker of poor outcome in patients with early relapsing-remitting multiple 2. Hutchinson M, Bresnihan B: Neurological lupus erythematosus sclerosis. Neurosci Lett 2004, 363(1):14-17. with tonic seizures simulating multiple sclerosis. J Neurol Neu- rosurg Psychiatry 1983, 46(6):583-585. 21. Berger T, Rubner P, Schautzer F, Egg R, Ulmer H, Mayringer I, Dilitz E, Deisenhammer F, Reindl M: Antimyelin antibodies as a predic- 3. Majithia V, Geraci SA: Rheumatoid arthritis: diagnosis and man- tor of clinically definite multiple sclerosis after a first demy- agement. Am J Med 2007, 120(11):936-939. elinating event. N Engl J Med 2003, 349(2):139-145. 4. Hanauer SB: Inflammatory bowel disease. N Engl J Med 1996, 22. Berger T, Reindl M: Lack of association between antimyelin 334(13):841-848. antibodies and progression to multiple sclerosis. N Engl J Med 2007, 356(18):1888-1889. 5. McDonald E: Multiple sclerosis: common management issues. Aust Fam Physician 1992, 21(10):1421-1424. 23. Kuhle J, Lindberg RL, Regeniter A, Mehling M, Hoffmann F, Reindl M, Berger T, Radue EW, Leppert D, Kappos L: Antimyelin antibodies 6. McDonald WI, Ron MA: Multiple sclerosis: the disease and its in clinically isolated syndromes correlate with inflammation manifestations. Philos Trans R Soc Lond B Biol Sci 1999, in MRI and CSF. J Neurol 2007, 254(2):160-168. 354(1390):1615-1622. 24. Bergamaschi R: Prognostic factors in multiple sclerosis. Int Rev Neurobiol 2007, 79:423-447. 25. Sailer M, O'Riordan JI, Thompson AJ, Kingsley DP, MacManus DG, McDonald WI, Miller DH: Quantitative MRI in patients with clinically isolated syndromes suggestive of demyelination. Neurology 1999, 52(3):599-606. 26. Achiron A, Gurevich M, Snir Y, Segal E, Mandel M: Zinc-ion binding and cytokine activity regulation pathways predicts outcome in relapsing-remitting multiple sclerosis. Clin Exp Immunol 2007, 149(2):235-242. 27. Baranzini SE, Mousavi P, Rio J, Caillier SJ, Stillman A, Villoslada P, Wyatt MM, Comabella M, Greller LD, Somogyi R, et al.: Transcrip- tion-based prediction of response to IFNbeta using super- vised computational methods. PLoS Biol 2005, 3(1):e2. 28. Filippi M, De Stefano N, Dousset V, McGowan JCE: MR imaging in white matter diseases of the brain and spinal cord. Berlin: Springer; 2005:219. 29. McDonald WI, Compston A, Edan G, Goodkin D, Hartung HP, Lublin FD, McFarland HF, Paty DW, Polman CH, Reingold SC, et al.: Rec- Page 13 of 14 (page number not for citation purposes) 34

BMC Medical Genomics 2009, 2:46 http://www.biomedcentral.com/1755-8794/2/46 ommended diagnostic criteria for multiple sclerosis: guide- 48. Barnes CJ, Li F, Mandal M, Yang Z, Sahin AA, Kumar R: Heregulin lines from the International Panel on the diagnosis of induces expression, ATPase activity, and nuclear localization multiple sclerosis. Ann Neurol 2001, 50(1):121-127. of G3BP, a Ras signaling component, in human breast 30. Wingerchuk DM, Lennon VA, Pittock SJ, Lucchinetti CF, Weinsh- tumors. Cancer research 2002, 62(5):1251-1255. enker BG: Revised diagnostic criteria for neuromyelitis optica. Neurology 2006, 66(10):1485-1489. 49. Le Y, Iribarren P, Gong W, Cui Y, Zhang X, Wang JM: TGF-beta1 31. Burges CJC: A Tutorial on Support Vector Machines for Pat- disrupts endotoxin signaling in microglial cells through tern Recognition. Data Mining and Knowledge Discovery 1998, Smad3 and MAPK pathways. J Immunol 2004, 173(2):962-968. 2:121-167. 32. Dietterich T: Overfitting and undercomputing in machine 50. Stasyk T, Dubrovska A, Lomnytska M, Yakymovych I, Wernstedt C, learning. ACM Computing Surveys 1995, 27(3):326-327. Heldin CH, Hellman U, Souchelnytskyi S: Phosphoproteome pro- 33. Efron B: Estimating the error rate of a prediction rule filing of transforming growth factor (TGF)-beta signaling: improvement on cross-validation. Journal of the American Statis- abrogation of TGFbeta1-dependent phosphorylation of tran- tical Association 1983, 78(382):316-330. scription factor-II-I (TFII-I) enhances cooperation of TFII-I 34. Geisser S: Predictive Inference: An Introduction. New York: and Smad3 in transcription. Mol Biol Cell 2005, CRC Press; 1993. 16(10):4765-4780. 35. Paglinawan R, Malipiero U, Schlapbach R, Frei K, Reith W, Fontana A: TGFbeta directs gene expression of activated microglia to 51. IFNB Multiple Sclerosis Study Group E-U: Interferon beta-1b is an anti-inflammatory phenotype strongly focusing on chem- effective in relapsing-remitting multiple sclerosis. I. Clinical okine genes and cell migratory genes. Glia 2003, results of a multicenter, randomized, double-blind, placebo- 44(3):219-231. controlled trial. The IFNB Multiple Sclerosis Study Group. 36. Bright JJ, Sriram S: TGF-beta inhibits IL-12-induced activation Neurology 1993, 43(4):655-661. of Jak-STAT pathway in T lymphocytes. J Immunol 1998, 161(4):1772-1777. 52. (Commanditaire) IMSG, University of British Columbia MS/MRI Anal- 37. Centola M, Wood G, Frucht DM, Galon J, Aringer M, Farrell C, ysis Group E-UC: Interferon beta-1b in the treatment of mul- Kingma DW, Horwitz ME, Mansfield E, Holland SM, et al.: The gene tiple sclerosis: final outcome of the randomized controlled for familial Mediterranean fever, MEFV, is expressed in early trial. The IFNB Multiple Sclerosis Study Group and The Uni- leukocyte development and is regulated in response to versity of British Columbia MS/MRI Analysis Group. Neurol- inflammatory mediators. Blood 2000, 95(10):3223-3231. ogy 1995, 45(7):1277-1285. 38. Cammer W, Zhang H, Tansey FA: Effects of carbonic anhydrase II (CAII) deficiency on CNS structure and function in the 53. Group PPoRaDbIb-aSiMSS: Randomised double-blind placebo- myelin-deficient CAII-deficient double mutant mouse. J Neu- controlled study of interferon beta-1a in relapsing/remitting rosci Res 1995, 40(4):451-457. multiple sclerosis. Lancet 1998, 352(9139):1498-1504. 39. Cote J, Boisvert FM, Boulanger MC, Bedford MT, Richard S: Sam68 RNA binding protein is an in vivo substrate for protein 54. Group UMMS, IFNB Multiple Sclerosis Study Group E-U: PRISMS- arginine N-methyltransferase 1. Mol Biol Cell 2003, 4: Long-term efficacy of interferon-beta-1a in relapsing MS. 14(1):274-287. Neurology 2001, 56(12):1628-1636. 40. Tissot C, Mechti N: Molecular cloning of a new interferon- induced factor that represses human immunodeficiency 55. Jacobs LD, Cookfair DL, Rudick RA, Herndon RM, Richert JR, Salazar virus type 1 long terminal repeat expression. J Biol Chem 1995, AM, Fischer JS, Goodkin DE, Granger CV, Simon JH, et al.: Intramus- 270(25):14891-14898. cular interferon beta-1a for disease progression in relapsing 41. Andrianifahanana M, Agrawal A, Singh AP, Moniaux N, van Seuningen multiple sclerosis. The Multiple Sclerosis Collaborative I, Aubert JP, Meza J, Batra SK: Synergistic induction of the MUC4 Research Group (MSCRG). Ann Neurol 1996, 39(3):285-294. mucin gene by interferon-gamma and retinoic acid in human pancreatic tumour cells involves a reprogramming of signal- Pre-publication history The pre-publication history for this paper can be accessed here: http://www.biomedcentral.com/1755-8794/2/46/prepub ling pathways. Oncogene 2005, 24(40):6143-6154. 42. Baron BW, Anastasi J, Thirman MJ, Furukawa Y, Fears S, Kim DC, Simone F, Birkenbach M, Montag A, Sadhu A, et al.: The human pro- grammed cell death-2 (PDCD2) gene is a target of BCL6 repression: implications for a role of BCL6 in the down-reg- ulation of apoptosis. Proc Natl Acad Sci USA 2002, 99(5):2860-2865. 43. Bouwmeester T, Bauch A, Ruffner H, Angrand PO, Bergamini G, Croughton K, Cruciat C, Eberhard D, Gagneur J, Ghidelli S, et al.: A physical and functional map of the human TNF-alpha/NF- kappa B signal transduction pathway. Nat Cell Biol 2004, 6(2):97-105. 44. Deb A, Haque SJ, Mogensen T, Silverman RH, Williams BR: RNA- dependent protein kinase PKR is required for activation of NF-kappa B by IFN-gamma in a STAT1-independent path- way. J Immunol 2001, 166(10):6170-6180. Publish with Bio Med Central and every 45. Klein C, Paul JI, Sauve K, Schmidt MM, Arcangeli L, Ransom J, True- scientist can read your work free of charge heart J, Manfredi JP, Broach JR, Murphy AJ: Identification of surro- gate agonists for the human FPRL-1 receptor by autocrine selection in yeast. Nature biotechnology 1998, 16(13):1334-1337. \"BioMed Central will be the most significant development for 46. Le Y, Yazawa H, Gong W, Yu Z, Ferrans VJ, Murphy PM, Wang JM: disseminating the results of biomedical researc h in our lifetime.\" The neurotoxic prion peptide fragment PrP(106–126) is a Sir Paul Nurse, Cancer Research UK chemotactic agonist for the G protein-coupled receptor Your research papers will be: formyl peptide receptor-like 1. J Immunol 2001, 166(3):1448-1451. available free of charge to the entire biomedical community 47. Batliwalla FM, Li W, Ritchlin CT, Xiao X, Brenner M, Laragione T, peer reviewed and published immediately upon acceptance Shao T, Durham R, Kemshetti S, Schwarz E, et al.: Microarray anal- yses of peripheral blood cells identifies unique gene expres- cited in PubMed and archived on PubMed Central sion signature in psoriatic arthritis. Molecular medicine yours — you keep the copyright (Cambridge, Mass) 2005, 11(1–12):21-29. BioMedcentral Submit your manuscript here: http://www.biomedcentral.com/info/publishing_adv.asp Page 14 of 14 (page number not for citation purposes) 35



Laquinimod suppress antigen presentation in relapsing-remitting multiple sclerosis: In-vitro high-throughput gene expression study Journal of neuroimmunology | 2010 ‫ דר' מיכאל גורביץ‬:‫מנחה‬ ‫ פרופ' ענת אחירון‬:‫מנחה‬ ,‫מנהל המעבדה הנוירואימונולוגית‬ ‫ מנהלת המרכז‬,‫מייסדת פרויקט ח\"ץ‬ ‫המרכז לטרשת נפוצה‬ ‫לטרשת נפוצה ואחראית הקתדרה‬ ‫למחלות אוטואימוניות אוניברסיטת ת\"א‬ [email protected] [email protected] ‫רותם אורבך‬ ‫אונ' תל אביב‬ ‫השתתפה כסטודנטית בפרויקט ח״ץ‬ 2007-2009 ‫בין השנים‬ [email protected] 37

Journal of Neuroimmunology 221 (2010) 87–94 Contents lists available at ScienceDirect Journal of Neuroimmunology j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j n e u r o i m Laquinimod suppress antigen presentation in relapsing–remitting multiple sclerosis: In-vitro high-throughput gene expression study M. Gurevich ⁎, T. Gritzman, R. Orbach, T. Tuller, A. Feldman, A. Achiron Multiple Sclerosis Center, Sheba Medical Center, Tel-Hashomer, Israel Sackler School of Medicine, Tel-Aviv University, Israel article info abstract Article history: Laquinimod (LAQ) is a new immunomodulatory drug shown to be effective in the treatment of relapsing– Received 16 December 2009 remitting multiple sclerosis (RRMS); however, its molecular target pathways are not well recognized. In this Received in revised form 9 February 2010 study we characterized in-vitro the molecular effects of LAQ in peripheral blood mononuclear cells (PBMC) Accepted 11 February 2010 of healthy subjects and RRMS patients by gene expression microarrays. We demonstrated that LAQ induced suppression of genes related to antigen presentation and corresponding inflammatory pathways. These Keywords: findings were demonstrated mainly via the NFkB pathway. Analysis of PBMC subpopulations identified Multiple sclerosis activation of Th2 response in CD14+ and CD4+ cells and suppression of proliferation in CD8+ cells. Laquinimod Microarray gene expression © 2010 Elsevier B.V. All rights reserved. PBMC Cell subpopulation 1. Introduction Underlying the molecular mechanisms induced by LAQ is of great importance in order to better understand how the drug affects the Laquinimod (quinolin-3-carboxamide, LAQ) is a novel oral immune system and how it induces significant changes over-time that immunomodulatory drug, developed as a disease-modifying treat- could lead to modulation of the active disease process in MS. ment for relapsing–remitting multiple sclerosis (RRMS). In MS- related animal models, LAQ has been demonstrated to inhibit the High-throughput gene expression microarray technology enables development of acute experimental autoimmune encephalomyelitis a reproducible detection of gene expression profiles for thousands (EAE) (Brunmark et al., 2002), and to suppress experimental of genes simultaneously. In recent years this technology was applied autoimmune neuritis (Zou et al., 2002). to comprehensive signature evaluations in various diseases, to identify the effects of in-vivo and in-vitro therapeutic treatments, LAQ suppressed the development of EAE through modulation of and to discover single genes that are critical to a specific disease pat- Th1/Th2 response by the inhibition of IL-12 and TNF-α and the tern or operate in a specific biologic process (Achiron and Gurevich, enhancement of IL-4, IL-10, and TGFβ (Yang et al., 2004). 2006). In a multicenter, double-blind, randomized phase II clinical trial In the current study, we aimed to characterize the molecular over 24 weeks that included 209 RRMS patients, LAQ (0.3 mg) induced mechanisms induced by LAQ. We conducted a large-scale gene 44% reduction in the mean cumulative number of active brain lesions expression microarray analysis in RRMS patients and healthy subjects. as compared to placebo (Polman et al., 2005). In an additional large We performed in-vitro analysis of peripheral blood mononuclear cells phase IIb study over 36 weeks, that included 306 RRMS patients, LAQ (PBMC) incubated with LAQ, and also analyzed the specific gene (0.6 mg) demonstrated a statistically significant 40% and 50% decrease expression signatures in peripheral blood subpopulations: CD4+, in the cumulative number of Gd-enhancing and new T1 brain lesions, CD8+, CD14+, CD19+ and NK cells. We identified transcriptionally respectively (Comi et al., 2008). Currently, LAQ is evaluated in two altered key genes involved in LAQ-induced pathways in RRMS and large phase III clinical trials aimed to assess its effects in comparison to studied the biological networks that significantly present transcrip- placebo or interferon beta 1-a in RRMS patients. tional alteration (up and/or down-regulated) genes or pathways which contribute to LAQ effects. In addition, the protein verification ⁎ Corresponding author. Multiple Sclerosis Center, Sheba Medical Center, Tel- experiments further support the gene expression molecular findings Hashomer, 52621, Israel. Tel.: +972 3 5303811; fax: +972 3 5348186. of LAQ operating networks. The results provide a meaningful insight into the mechanisms of action orchestrated by LAQ in RRMS and E-mail address: [email protected] (M. Gurevich). their impact on immune cell functions. 0165-5728/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.jneuroim.2010.02.010 Downloaded for Anonymous User (n/a) at Sheba Medical Center from ClinicalKey.com by Elsevier on October 19, 382021. For personal use only. No other uses without permission. Copyright ©2021. Elsevier Inc. All rights reserved.

88 M. Gurevich et al. / Journal of Neuroimmunology 221 (2010) 87–94 2. Materials and methods and scanned according to the manufacturer's protocol (Affymetrix, Inc., USA). 2.1. Study design 2.3. Data analysis All participants gave written informed consent and the study was approved by the Sheba IRB committee. Peripheral blood samples were Data analysis was performed on Partek Genomics Solution software obtained from five healthy female subjects (mean ± SE age 27.8 ± (www.partek.com). Expression values were computed from raw CEL 4.1 years) and 10 RRMS female patients (mean ± SE age 38.9 ± files by applying the Robust Multi-Chip Average (RMA) background 4.5 years, disease duration 7.5 ± 3.5 years, neurological disability by correction algorithm. RMA correction included: 1) values background EDSS score 2.0 ± 0.5) for gene microarray analysis. All patients were correction; 2) quintile normalization; 3) log2 transformation; and 4) free of immunomodulatory or corticosteroid treatments at least median polish summarization. High score genes (HSGs) were defined as 30 days before blood withdrawal. Microarray analysis of in-vitro LAQ those that differentiated between samples (incubated with or without effects was performed after incubation for 24 h of PBMC in the LAQ) with a p b 0.05 by paired t-test analysis. presence or absence of two concentrations of LAQ: 0.1 µM and 1.0 µM, that correspond to the oral 0.3 mg and 0.6 mg used in LAQ clinical 2.3.1. Significance of gene expression signature trials. Gene expression experiments on PBMC cell subtypes were Due to the relatively small datasets size and taking into account performed after incubation for 24 h with LAQ 0.1 µM. The study design is demonstrated in Fig. 1. multiple comparison correction, instead of analyzing transcriptionally altered individual genes we used the module map analysis method. 2.2. Cell cultures, RNA isolation and hybridization This was done by organizing genes into a network based on gene expression and pathway analysis software. The advantage of such PBMC were extracted from 15 ml peripheral blood, separated by approach is the combination of classical scoring analysis with Ficoll–Hypaque gradient and incubated at 37 °C in a humidified CO2 organization of genes in co-regulated modules related to distinct incubator with or without 0.1 µM or 1.0 µM LAQ for 24 h. biological functions and taking into consideration known relationship between genes (Segal et al., 2005). A module was defined as a set of Cell viability was measured on total PBMC by propidium iodide (PI) co-regulated genes with similar expression patterns. To infer staining using an automated cell counter (Nucleocounter, Chemometeck). regulatory modules, we scored the functional/biological coherence of each set according to the percentage of its genes covered by After incubation, total RNA was extracted using both Trizol (Invitro- annotations significantly enriched in the division. A global view of the gen, USA) and Phase-Lock-Gel columns (Eppendorf, Germany) including various functions was created by compiling all gene annotations in a DNase digestion step. RNA integrity was assessed by RNA Experion each module into a single matrix using probabilistic graphical models automated electrophoresis system. implemented in Genomica software (http://genomica.weizmann.ac. il). Module analysis was performed on HSGs across data sets including Probe synthesis using 3 μg total RNA, hybridization, detection, and Human BioCarta, KEGGS, and Human GO. Biological functional scanning was performed according to the standard Affymetrix, Inc. analysis was completed on modules that passed a false discovery USA protocols; cDNA was synthesized using the Two-Cycle cDNA rate correction test at p b 0.05. The reconstruction of LAQ related Synthesis Kit (Affymetrix, Inc., USA), and in-vitro transcription biological mechanisms was performed by Ingenuity software, (www. performed with the GeneChip IVT Labeling Kit (Affymetrix, Inc., ingenuity.com). The detailed descriptions of the module analysis and USA). The biotin-labeled IVT–RNA was hybridized to HG-U133A-2 pathway reconstruction are represented in the methods Supplemen- arrays containing 18,400 gene transcripts, each corresponding to tary material. 14,500 well-annotated human genes, washed in a GeneChip Fluidics Station 450 (Hewlett Packard, USA, GeneArray-TM scanner G2500A) Fig. 1. Schematic illustration of the study design. LAQ-0.1(1.0) — cells incubated with 0.1 μM or 1.0 μM of LAQ; LAQ-0 — incubation without LAQ. Downloaded for Anonymous User (n/a) at Sheba Medical Center from ClinicalKey.com by Elsevier on October 19, 392021. For personal use only. No other uses without permission. Copyright ©2021. Elsevier Inc. All rights reserved.

M. Gurevich et al. / Journal of Neuroimmunology 221 (2010) 87–94 89 2.4. LAQ effects by gene expression analysis of immune cell subtypes: 3.3. Module analysis of LAQ (0.1 μM) effects on PBMC from RRMS CD4+, CD8+, CD14+, CD19+, and NK cells patients PBMC were collected from five female RRMS patients and In MS patients, LAQ induced differential expression of 698 HSGs. incubated in the presence or absence of 0.1 µM LAQ for 24 h. We Similar to the LAQ (0.1 µM) effects on PBMC from healthy subjects, the intentionally incubated PBMC with LAQ before separation of the cells most significant module (3.97 × 10−3, 174 genes, Supplementary to subtypes in order to preserve natural cell-to-cell interactions Fig. 2A, B) was associated with genes related to antigen presentation during LAQ incubation. After incubation CD4+, CD8+, CD14+, pathways. We demonstrated under-expression of MHC class II family CD19+ and NK cells were isolated by magnetic activated cell genes (HLA-DQA1, HLA-DQB1 and HLA-DQB2) together with sup- sorting method (Dynal Biotech ASA, Norway). Total RNA was pression of TLR6, a key player in the detection of pathogens via isolated from each cell population and hybridized to a U133A-2 presentation of antigen fragments to T-helper cells (Akira et al., 2001). array (Affymetrix) as described above. Following pathogen recognition, TLR signaling leads to NFκB and MAPK transcription factor activation (Kawai and Akira, 2006). 2.5. Protein verification Stimulation of cytoplasmic NFκB is initiated by the signal-induced degradation of IκBs (inhibitors of κB) (Xiao and Ghosh, 2005). 2.5.1. Western blot analysis Accordingly, in our study we demonstrated over-expression of the IκB Protein expression level was assessed in PBMC samples by NFκBIE and suppression of BTRC encoding gene that increases ubiquitination of NFκB inhibitor protein (Strack et al., 2000). In Western blot using anti-HLA-DQA/DQB monoclonal antibodies concert with the increased expression of the NFκB inhibitor, we also (Santa Cruz Lab., USA). Analysis was performed in samples obtained found suppressed expression of NFκB downstream signaling genes from five RRMS patients and five healthy subjects after 24 and 48 h like CXCL9, a well known Th1 chemokine involved in T-cell trafficking incubation with or without 0.1 µM LAQ. (Teixeira et al., 2004), and under-expression of LY9, a cell surface protein involved in lymphocyte activation (Del Valle et al., 2003). 2.5.2. Protein microarray analysis Moreover, suppression of apoptosis mediators such as TNFSF10 and Protein microarray analysis was performed in samples obtained EIF2AK1 was observed. Our analysis also identified a significant activation of MAPK14 (P38), a second key player in TLRs downstream from five RRMS patients. Aliquots of supernatants were obtained from signaling that mediates the initiation phase of innate immunity and PBMC incubated in the presence or absence of 0.1 μM LAQ for 48 h. cell death (Dong et al., 2002). MAPK14 regulates apoptosis related The aliquots were assayed for 507 human proteins using the RayBio genes including HSPB3, BID and MDM2, which were found to be over- Biotin Label-based Human Antibody Array. The samples were expressed. Additionally, CASP9 that is required to induce BID biotinylated, dialyzed, incubated with array membrane and visualized processing, and TNFSF11 which induces apoptosis associated with with HRP-stretavidin. Cytokines were quantified by densitometry activation of CASP9 (Izawa et al., 2007), were activated. Signaling using ImageJ software, background was subtracted and each sample through the MAPK14 pathway can operate with MAP2K1 and FCER1B was then normalized to an internal positive control. (MS4A2) that were both over-expressed, in order to activate the Fc epsilon RI signaling pathway. 3. Results 3.4. Module analysis of LAQ (1.0 μM) effects on PBMC from RRMS 3.1. Cell viability patients Cell viability was not affected by LAQ. The number of PI stained LAQ induced differential expression of 821 HSG. The most PBMC from RRMS patients was not altered by incubation with either significant (p = 0.02) module included 88 genes (Supplementary 0.1 μM or1.0 μM LAQ. The percent of viable cells was 97.5 ± 0.4% and Fig. 3A,B), and was characterized by suppression of genes involved in 96.1 ± 1.1%, respectively. macrophage phagocytosis such as CSF1 and SFTPD (Stanley et al., 1997; Sano and Kuroki, 2005), suppression of IFNA16, an interferon 3.2. Module analysis of LAQ (0.1 µM) effects on PBMC from healthy inducible gene that is produced by antigen presenting cells (APC) and subjects PTN encoding gene that induces expression of inflammatory cytokines (Achour et al., 2008). Additionally, BMP2 gene that initiates signal In healthy subjects LAQ induced differential expression of 1140 transduction pathways via NFκB to induce inflammation (Csiszar HSGs. The most significant module (p = 3.79⁎10− 3) included 284 et al., 2006), was suppressed. Accordingly, SPN (CD43) and KIR2DS3 genes. A global view of the modules is demonstrated in Supplemen- genes that are known to be involved in T-cell activation were under- tary Fig. 1A,B.This module was characterized by under-expression of expressed (Parham, 2005; Fierro et al., 2006). genes related to antigen presentation pathway including degradation of antigenic proteins (MSR1 and CTSS), MHC class II molecules (HLA- The significant role of LAQ on APC was indicated also by under- DRB, HLA-DQB, and HLA-DPB), component of complement (C6) and expression of chemotaxis associated genes including CCL17 (TARC) and genes associated with presentation of antigens to T-cells (CD1A, DEFB1, both expressed by APC and display chemotactic activity for CD1E). In this module the suppression of antigen presentation was T lymphocytes (Chertov et al., 1996; Imai et al., 1997; Duits et al., 2002). associated with under-expression of inflammatory related molecules Moreover, the MDK encoding gene, a product that promotes neutrophil like interferon gamma-inducible protein (IFI30) and leukocyte chemotaxis (Takada et al., 1997), was also under-expressed. specific transcript (LST1). Moreover, under-expression of chemokine signaling molecules (CCR1, CCR2, and CCL13), integrins, adhesion In support of a previous study that demonstrated Th2 shift and related molecules (ITGA2, ITGB6, MMP13, and ADAMTS2) and CXC suppression of inflammatory encoding genes (Yang et al., 2004), in sub-family of cytokines (CXCL9 and CXCL14) were observed. In our study we observed an over-expression of Jak3, which plays an addition, CD24, which has been proven to be essential for the important role in Th2 cell development (Yamashita et al., 2000). development of EAE in mice (Bai et al., 2000) was found to be under- expressed. The combined molecular effects induced by LAQ in-vitro in both healthy subjects and RRMS patients were found to be related to down regulation of antigen presentation MHC class II genes and suppression of inflammatory pathways and chemokine signaling related mole- cules and are presented in Fig. 2A, B. Downloaded for Anonymous User (n/a) at Sheba Medical Center from ClinicalKey.com by Elsevier on October 19, 402021. For personal use only. No other uses without permission. Copyright ©2021. Elsevier Inc. All rights reserved.

90 M. Gurevich et al. / Journal of Neuroimmunology 221 (2010) 87–94 Fig. 2. Most significant regulatory networks induced by LAQ 0.1 μM, 24 h incubation in PBMC. (A) Healthy subjects and (B) RRMS patients. The relations between the genes were inferred from the relationships known in the scientific literature using data-mining Ingenuity software. Each node represents a gene; red color denotes over-expressed genes; green color denotes down-expressed genes. Connections indicate regulatory interactions as follows: arrows = direct activation; dashed arrows = indirect activation. The color areas present the following biological pathways: Pink — antigen presentation, light blue — inflammation (chemokine, adhesion), yellow — apoptosis. Key symbols for specific molecules are shown in the blue box. 3.5. Module analysis of LAQ (0.1 µM) effects on PBMC subpopulations 3.5.1. CD14+ cells Module analysis of LAQ effects on CD14+ cells identified a LAQ induced differential expression of 487, 942, 640, 900 and 703 HSG that correspond to CD14+, CD8+, CD4+, NK and CD19+ cells. significant (p = 0.02) module that included 45 genes. This module The comprehensive LAQ operating regulatory molecular mechanisms was characterized by differential expression of genes related to the in PBMC subpopulations are demonstrated in Fig. 3. inflammatory response of macrophages (p = 2.2 × 105) leading to Th2 shift. We found activation of SFTPD, a macrophage related gene that is Downloaded for Anonymous User (n/a) at Sheba Medical Center from ClinicalKey.com by Elsevier on October 19, 412021. For personal use only. No other uses without permission. Copyright ©2021. Elsevier Inc. All rights reserved.

M. Gurevich et al. / Journal of Neuroimmunology 221 (2010) 87–94 91 Fig. 3. LAQ induced molecular immuno-modulation in PBMC subpopulations. Over-expressed genes are depicted in red, down-expressed in green. characterized as a stimulator of phagocytosis and suppressor of IL2 et al., 2008). As expected, CD4 and TNFRSF4 receptor genes that play a production and T-cell proliferation (Borron et al., 1998; Kishore et al., key role in CD4+ T-cell response and IL4 secretion (Noben-Trauth 2006). Accordingly, we demonstrated suppression of two positive et al., 1997; Roos et al., 1998) were over-expressed. The module was regulators of cellular proliferation, MAP3K7 — a strong positive also characterized by suppression of the important chemoattractant regulator of IL2 and TGF-beta activation (Wan et al., 2006), and KRAS a CCL24 (Menzies-Gow et al., 2002). positive regulator of proliferation (Chang et al., 2003). 3.5.4. NK cells Moreover, we identified a decreased expression of the PTGER4 In NK cells, incubation with LAQ induced significant (p = 4.0 × 10−3) gene known to be involved in T-cell signaling and to mediate inflammation (Kabashima et al., 2003; Caristi et al., 2005), over- module containing 79 genes. This module was characterized by an expression of B7-H2 gene that acts as a T-cell co-stimulatory ligand enrichment of genes related to the cellular immune response and elicits secretion of Th2 cytokines such as IL-4 and IL-10 (Wang (p = 9.9 × 10−6). Specifically, NK activating receptors like KLRC3 and and Chen, 2004), and over-expression of NTF3 gene that directly NCR1 (Moretta et al., 2000; Lanier, 2005) and GNLY encoding for the regulates the neurotrophic factors NGF and BDNF (Paul et al., 2001; cytotoxic granularly protein and known to be a significant effector of Randolph et al., 2007). cytotoxic T-cell activity upon antigen stimulation (Stenger et al., 1998), were suppressed. This transcription modulation indicates suppression 3.5.2. CD8+ cells of specific NK cytotoxic related mechanism. In this context, the negative The most significant module (p = 3.97 × 10−3) included 62 genes, regulator of Th1 cytokine signaling, the anti-inflammatory gene SOCS2 (Hanada et al., 2003), was up-regulated. and was enriched by genes that lead to suppressed cellular growth and proliferation (p = 2.0 × 10− 6). This module pathway included 3.5.5. CD19+ cells down-regulation of the E2F3-transcription factor gene that plays a In CD19+ cells LAQ induced a significant (p = 0.03) module that critical role in cell cycle and cell differentiation (Leone et al., 1998) and the CDK3 gene — a downstream member of the same pathway. included 76 genes. We observed over-expression of RALGDS, that has Additionally, SYK and PTPRC (CD45) genes typically involved in T-cell been shown to bind and suppress the activity of cell adhesion proliferation and regulation of cytokine signaling (Minami et al., molecules like LFA1 and VL4, both associated with T-cell invasion 1995; Johnson et al., 2000) were decreased. through the BBB (de Bruyn et al., 2002). Additionally, over-expression of CSF2, a granulocyte/macrophage colony-stimulating factor that 3.5.3. CD4+ cells interferes with the effect of TGF-beta1 on APCs, and suppression of Module analysis of LAQ effects on CD4+ cells identified a PRKCQ, a gene implicated in the activation of NFκB (Gruber et al., 2008) and T cells (Letschka et al., 2008) was identified. significant (p = 3.9 × 10− 3) module that integrated 72 genes. This module was enriched by Fc epsilon RI component of the TCR signaling 3.6. Verification of the gene expression findings on the protein level pathway (p = 9.86 × 10− 5) (Krishnan et al., 2003; Okoye et al., 2007), including over-expression of FCER1A and FCGR2B genes that lead to The effect of suppression of antigen recognition by LAQ was over-expression of Th2 anti-inflammatory response (Kaneko et al., verified by Western blot analysis, Fig. 4A. The results demonstrated 2006). In this module we also observed activation of IL-4 cytokine that the protein expression of MHC class II related molecules (HLA- gene that is important for the generation of anti-inflammatory Th2 DQA/B) in PBMC of healthy subjects and RRMS patients did not response (Opal and DePalo, 2000), and identified the suppression of change after 24 h of incubation with LAQ 0.1 μM. However, after 48 h ALOX5 gene, that has an important role in the development of T cells of incubation the expression level of HLA-DQA/B decreased in healthy and when suppressed also increases IL10 and Th2 skewing (DiMeo Downloaded for Anonymous User (n/a) at Sheba Medical Center from ClinicalKey.com by Elsevier on October 19, 422021. For personal use only. No other uses without permission. Copyright ©2021. Elsevier Inc. All rights reserved.

92 M. Gurevich et al. / Journal of Neuroimmunology 221 (2010) 87–94 Fig. 4. Verification of LAQ biological effects on the protein level. (A) The HLA-DQA/DOB expression level by Western blot in PBMC from healthy subjects and MS patients after incubation with LAQ (0.1 μM) for 24 or 48 h. LAQ-0 — incubation without LAQ. PC-positive control at 29 kDa. (B) Relative protein array expression levels of inflammatory cytokines and adhesion molecules of MS patients after 48 h incubation with 0.1 μM LAQ. Expression level without LAQ was considered as 100%. subjects and was completely suppressed in RRMS patients. Interest- and Miller, 1996). Furthermore, this operating network was more ingly, in the microarray gene expression data, suppression of MHC evident at higher dose of LAQ (1.0 µM) as demonstrated by class II molecules was already detected within 24 h, while the suppression of pathogen recognition receptor (SFTPD), differentiation translated protein change was observed only after 48 h. and function of macrophages (CSF1), macrophage products (IFNA16) and suppression of chemotaxis associated genes (CCL17, DEFB1). Protein microarray analysis performed after 48 h of incubation These findings indicate suppression of macrophage activity through with LAQ, demonstrated significant suppression of six pro-inflamma- decreased Th1 immune response and in addition the increased tory molecules including the cytokines IL6 (p = 0.05), IL1F9 (p = 0.03) expression of JAK3 encoding gene, an inducer of Th2 response and IL1SRI (p = 0.02); the chemokines IL8RA (p = 0.05) and RANTES (Verbsky et al., 2002) further promotes LAQ induced immune reaction (p = 0.008); and the adhesion molecule ICAM1 (p = 0.04), Fig. 4B. The towards anti-inflammatory response. findings confirmed the anti-inflammatory effects of LAQ demonstrat- ed by the gene expression experiments. The second network demonstrates novel interactions between up- regulation of NFκBIE facilitating anti-inflammatory response associ- 4. Discussion ated with the expression of early transcription genes via MAPK pathway. Over-expression of the NFkB inhibitor NFκBIE and suppres- Laquinimod is an immunomodulatory drug with potential thera- sion of BTRC that increases ubiquitination of NFκBIE, inactivate the peutic effects in RRMS. We have used an integrated functional module NFκB transcription factor by preventing its translocation to the concept by merging co-expressed genes and reconstruction of nucleus (Hayden and Ghosh, 2008). This distinct effect leads to interacted regulatory networks to detect transcriptional altered suppression of inflammatory associated genes such as CXCL9 pathways and interactive regulations operated under LAQ treatment. (Teixeira et al., 2004). The suppression of NFkB pathway by LAQ This approach provides the ability to construct relationship of was mainly evident in RRMS patients as compared with healthy molecules across biological functions by integrating functional subjects indicating that the underlying pathogenic inflammatory interactions with co-expressed information. Our analysis of the process in MS is responding to the anti-inflammatory NFkB dependent transcriptional gene expression profiles induced by LAQ in-vitro in LAQ effects. Inhibition of NFκB has been linked directly to apoptosis, PBMC derived from healthy subjects and RRMS patients demonstrated inappropriate immune cell development and delayed cell growth that, LAQ induced three novel significantly altered networks related to (Baldwin, 1996; Chen et al., 1999). Furthermore, downstream genes down regulation of MHC class-II antigen presenting genes and as a encoding chemotaxis and adhesion molecules like CCL13, CCR2 and result provoked suppression of downstream inflammatory pathways CXCL9 were found to be down-regulated. These chemokines play and enhanced apoptosis. critical role in directing myelin-reactive T cells into MS lesion sites (Zhang et al., 2000), thus their suppressed expression by LAQ The first network illustrates the suppression of MHC class II contributes to decreased inflammatory pathways in MS. antigen presentation impeding immune response and down-regulat- ing epitopes spreading. The suppression of expression of early The third LAQ-induced identified regulatory network demonstrat- transcription genes associated with antigen recognition via TLR ed over-expression of apoptotic genes like CASP9, BID and MDM2. We pathway was further associated with degradation of antigenic inferred unique interactions between the down-regulated NFkB- proteins to peptides (CTSS) and MHC class II molecules like CD1D, MAPK pathway and apoptotic activation, both contributing to the HLA-DP/DR/DQ. The alteration in MHC class II signaling pathways diminution of the inflammatory response. In a previous study (Achiron suggests that LAQ could significantly down-regulates the autoim- et al., 2007) we demonstrated a mirror image of the process, with the mune process in RRMS, preventing reactivation of autoimmune T cells occurrence of over-expression of the negative regulator of FAS- in the periphery by APC (Stoeckle and Tolosa, 2009). The possibility to induced apoptosis (TOSO) and the BCL2 anti-apoptotic family suppress this mechanism may be beneficial also to prevent diversi- members (BCL2 and BCL2), as well as down expression of pro- fication of MS specific antigens and reduce tissue damage (Vanderlugt apoptotic genes like BAX, APAF1 and caspases 1, 2, 8, 9 and 10, in RRMS Downloaded for Anonymous User (n/a) at Sheba Medical Center from ClinicalKey.com by Elsevier on October 19, 432021. For personal use only. No other uses without permission. Copyright ©2021. Elsevier Inc. All rights reserved.

M. Gurevich et al. / Journal of Neuroimmunology 221 (2010) 87–94 93 patients during acute relapse. The current findings suggest that the Chertov, O., Michiel, D.F., Xu, L., Wang, J.M., Tani, K., Murphy, W.J., Longo, D.L., Taub, D.D., inflammatory process in the active disease is targeted by inhibition of Oppenheim, J.J., 1996. Identification of defensin-1, defensin-2, and CAP37/ pro-apoptotic and repression of anti-apoptotic genes indicating a new azurocidin as T-cell chemoattractant proteins released from interleukin-8- link between LAQ induced NFkB suppression and activation of stimulated neutrophils. J. Biol. Chem. 271, 2935–2940. apoptosis. This was further portrayed by the findings of LAQ-induced effects in PBMC subpopulations. In CD14+ macrophage we observed Comi, G., Pulizzi, A., Rovaris, M., Abramsky, O., Arbizu, T., Boiko, A., Gold, R., Havrdova, E., an activation of genes related to the suppression of proliferation of Komoly, S., Selmaj, K., Sharrack, B., Filippi, M., 2008. Effect of laquinimod on MRI- T cells and activation of inducers of Th2 cytokine response. monitored disease activity in patients with relapsing-remitting multiple sclerosis: Accordingly, the gene expression pattern expressed in CD8+ a multicentre, randomised, double-blind, placebo-controlled phase IIb study. lymphocytes was related to suppression of T-cell proliferation and Lancet 371, 2085–2092. cytokine signaling. The failure of CD8+ cells to proliferate following LAQ exposure is explained by suppression of the cell cycle transcrip- Csiszar, A., Ahmad, M., Smith, K.E., Labinskyy, N., Gao, Q., Kaley, G., Edwards, J.G., Wolin, tion factor genes E2F3 and CDK3 (Leone et al., 1998; Sage, 2004), and M.S., Ungvari, Z., 2006. Bone morphogenetic protein-2 induces proinflammatory may invert the known autoimmune pathogenic effects of CD8+ T cells endothelial phenotype. Am. J. Pathol. 168, 629–638. in MS (Johnson et al., 2007). In CD4+ T-cells, LAQ activated the expression of Fc epsilon related genes. Fc epsilon receptor was recently de Bruyn, K.M., Rangarajan, S., Reedquist, K.A., Figdor, C.G., Bos, J.L., 2002. The small reported to constitute a part of the T-cell receptor complex and to GTPase Rap1 is required for Mn(2+)- and antibody-induced LFA-1- and VLA-4- contribute to T-cell activation (Krishnan et al., 2003; Okoye et al., mediated cell adhesion. J. Biol. Chem. 277, 29468–29476. 2007). LAQ induced CD4+ T-cell activation resulted in over- expression of genes like TNFRSF4 and IL4 related to Th2 anti- Del Valle, J.M., Engel, P., Martin, M., 2003. The cell surface expression of SAP-binding inflammatory response. The effect of LAQ on NK cells further supports receptor CD229 is regulated via its interaction with clathrin-associated adaptor the process of Th1 cytokine signaling suppression, while in CD19+ complex 2 (AP-2). J. Biol. Chem. 278, 17430–17437. cells, repressions of adhesion molecules and of NFkB pathway were dominant. DiMeo, D., Tian, J., Zhang, J., Narushima, S., Berg, D.J., 2008. Increased interleukin-10 production and Th2 skewing in the absence of 5-lipoxygenase. Immunology 123, Our findings demonstrate that LAQ effects on immune modulation 250–262. are related to the suppression of antigen presenting mechanism followed by a decrease of chemotaxis and adhesion. The anti- Dong, C., Davis, R.J., Flavell, R.A., 2002. MAP kinases in the immune response. Annu. Rev. inflammatory potency of LAQ was realized through the suppression Immunol. 20, 55–72. of the NFkB pathway that concordantly led to the activation of apoptosis of immuno-competent cells. The intensive molecular Duits, L.A., Ravensbergen, B., Rademaker, M., Hiemstra, P.S., Nibbering, P.H., 2002. analyses of these interactions bring innovative insights into the Expression of beta-defensin 1 and 2 mRNA by human monocytes, macrophages and understanding of LAQ effects in RRMS. dendritic cells. Immunology 106, 517–525. Appendix A. Supplementary data Fierro, N.A., Pedraza-Alva, G., Rosenstein, Y., 2006. TCR-dependent cell response is modulated by the timing of CD43 engagement. J. Immunol. 176, 7346–7353. Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.jneuroim.2010.02.010. Gruber, T., Fresser, F., Jenny, M., Uberall, F., Leitges, M., Baier, G., 2008. PKCtheta cooperates with atypical PKCzeta and PKCiota in NF-kappaB transactivation of T References lymphocytes. Mol. Immunol. 45, 117–126. Achiron, A., Gurevich, M., 2006. Peripheral blood gene expression signature mirrors Hanada, T., Kinjyo, I., Inagaki-Ohara, K., Yoshimura, A., 2003. Negative regulation of central nervous system disease: the model of multiple sclerosis. Autoimmun. Rev. cytokine signaling by CIS/SOCS family proteins and their roles in inflammatory 5, 517–522. diseases. Rev. Physiol. Biochem. Pharmacol. 149, 72–86. Achiron, A., Feldman, A., Mandel, M., Gurevich, M., 2007. Impaired expression of Hayden, M.S., Ghosh, S., 2008. Shared principles in NF-kappaB signaling. Cell 132, peripheral blood apoptotic-related gene transcripts in acute multiple sclerosis 344–362. relapse. Ann. N.Y. Acad. Sci. 1107, 155–167. Imai, T., Baba, M., Nishimura, M., Kakizaki, M., Takagi, S., Yoshie, O., 1997. The T cell- Achour, A., M'Bika, J.P., Baudouin, F., Caruelle, D., Courty, J., 2008. Pleiotrophin induces directed CC chemokine TARC is a highly specific biological ligand for CC chemokine expression of inflammatory cytokines in peripheral blood mononuclear cells. receptor 4. J. Biol. Chem. 272, 15036–15042. Biochimie 90, 1791–1795. Izawa, T., Ishimaru, N., Moriyama, K., Kohashi, M., Arakaki, R., Hayashi, Y., 2007. Akira, S., Takeda, K., Kaisho, T., 2001. Toll-like receptors: critical proteins linking innate Crosstalk between RANKL and Fas signaling in dendritic cells controls immune and acquired immunity. Nat. Immunol. 2, 675–680. tolerance. Blood 110, 242–250. Bai, X.F., Liu, J.Q., Liu, X., Guo, Y., Cox, K., Wen, J., Zheng, P., Liu, Y., 2000. The heat-stable Johnson, K.G., Bromley, S.K., Dustin, M.L., Thomas, M.L., 2000. A supramolecular basis for antigen determines pathogenicity of self-reactive T cells in experimental CD45 tyrosine phosphatase regulation in sustained T cell activation. Proc. Natl. autoimmune encephalomyelitis. J. Clin. Invest. 105, 1227–1232. Acad. Sci. U. S. A. 97, 10138–10143. Baldwin Jr., A.S., 1996. The NF-kappa B and I kappa B proteins: new discoveries and Johnson, A.J., Suidan, G.L., McDole, J., Pirko, I., 2007. The CD8 T cell in multiple insights. Annu. Rev. Immunol. 14, 649–683. sclerosis: suppressor cell or mediator of neuropathology? Int. Rev. Neurobiol. 79, 73–97. Borron, P.J., Crouch, E.C., Lewis, J.F., Wright, J.R., Possmayer, F., Fraher, L.J., 1998. Recombinant rat surfactant-associated protein D inhibits human T lymphocyte Kabashima, K., Sakata, D., Nagamachi, M., Miyachi, Y., Inaba, K., Narumiya, S., 2003. proliferation and IL-2 production. J. Immunol. 161, 4599–4603. Prostaglandin E2-EP4 signaling initiates skin immune responses by promoting migration and maturation of Langerhans cells. Nat. Med. 9, 744–749. Brunmark, C., Runstrom, A., Ohlsson, L., Sparre, B., Brodin, T., Astrom, M., Hedlund, G., 2002. The new orally active immunoregulator laquinimod (ABR-215062) effec- Kaneko, Y., Nimmerjahn, F., Ravetch, J.V., 2006. Anti-inflammatory activity of tively inhibits development and relapses of experimental autoimmune encepha- immunoglobulin G resulting from Fc sialylation. Science 313, 670–673. lomyelitis. J. Neuroimmunol. 130, 163–172. Kawai, T., Akira, S., 2006. TLR signaling. Cell Death Differ. 13, 816–825. Caristi, S., Piraino, G., Cucinotta, M., Valenti, A., Loddo, S., Teti, D., 2005. Prostaglandin E2 Kishore, U., Greenhough, T.J., Waters, P., Shrive, A.K., Ghai, R., Kamran, M.F., Bernal, A.L., induces interleukin-8 gene transcription by activating C/EBP homologous protein in human T lymphocytes. J. Biol. Chem. 280, 14433–14442. Reid, K.B., Madan, T., Chakraborty, T., 2006. Surfactant proteins SP-A and SP-D: structure, function and receptors. Mol. Immunol. 43, 1293–1315. Chang, F., Steelman, L.S., Shelton, J.G., Lee, J.T., Navolanic, P.M., Blalock, W.L., Franklin, R., Krishnan, S., Warke, V.G., Nambiar, M.P., Tsokos, G.C., Farber, D.L., 2003. The FcR gamma McCubrey, J.A., 2003. Regulation of cell cycle progression and apoptosis by the Ras/ subunit and Syk kinase replace the CD3 zeta-chain and ZAP-70 kinase in the TCR Raf/MEK/ERK pathway (Review). Int. J. Oncol. 22, 469–480. signaling complex of human effector CD4 T cells. J. Immunol. 170, 4189–4195. Lanier, L.L., 2005. NK cell recognition. Annu. Rev. Immunol. 23, 225–274. Chen, F., Castranova, V., Shi, X., Demers, L.M., 1999. New insights into the role of nuclear Leone, G., DeGregori, J., Yan, Z., Jakoi, L., Ishida, S., Williams, R.S., Nevins, J.R., 1998. E2F3 factor-kappaB, a ubiquitous transcription factor in the initiation of diseases. Clin. activity is regulated during the cell cycle and is required for the induction of S Chem. 45, 7–17. phase. Genes Dev. 12, 2120–2130. Letschka, T., Kollmann, V., Pfeifhofer-Obermair, C., Lutz-Nicoladoni, C., Obermair, G.J., Fresser, F., Leitges, M., Hermann-Kleiter, N., Kaminski, S., Baier, G., 2008. PKC-theta selectively controls the adhesion-stimulating molecule Rap1. Blood 112, 4617–4627. Menzies-Gow, A., Ying, S., Sabroe, I., Stubbs, V.L., Soler, D., Williams, T.J., Kay, A.B., 2002. Eotaxin (CCL11) and eotaxin-2 (CCL24) induce recruitment of eosinophils, basophils, neutrophils, and macrophages as well as features of early- and late- phase allergic reactions following cutaneous injection in human atopic and nonatopic volunteers. J. Immunol. 169, 2712–2718. Minami, Y., Nakagawa, Y., Kawahara, A., Miyazaki, T., Sada, K., Yamamura, H., Taniguchi, T., 1995. Protein tyrosine kinase Syk is associated with and activated by the IL-2 receptor: possible link with the c-myc induction pathway. Immunity 2, 89–100. Moretta, A., Biassoni, R., Bottino, C., Mingari, M.C., Moretta, L., 2000. Natural cytotoxicity receptors that trigger human NK-cell-mediated cytolysis. Immunol. Today 21, 228–234. Noben-Trauth, N., Shultz, L.D., Brombacher, F., Urban Jr., J.F., Gu, H., Paul, W.E., 1997. An interleukin 4 (IL-4)-independent pathway for CD4+ T cell IL-4 production is revealed in IL-4 receptor-deficient mice. Proc. Natl. Acad. Sci. U.S.A. 94, 10838–10843. Okoye, F.I., Krishnan, S., Chandok, M.R., Tsokos, G.C., Farber, D.L., 2007. Proximal signaling control of human effector CD4 T cell function. Clin. Immunol. 125, 5–15. Opal, S.M., DePalo, V.A., 2000. Anti-inflammatory cytokines. Chest 117, 1162–1172. Downloaded for Anonymous User (n/a) at Sheba Medical Center from ClinicalKey.com by Elsevier on October 19, 442021. For personal use only. No other uses without permission. Copyright ©2021. Elsevier Inc. All rights reserved.

94 M. Gurevich et al. / Journal of Neuroimmunology 221 (2010) 87–94 Parham, P., 2005. MHC class I molecules and KIRs in human history, health and survival. responses: chemotactic activity to neutrophils and association with inflammatory Nat. Rev. Immunol. 5, 201–214. synovitis. J. Biochem. 122, 453–458. Paul, J., Gottmann, K., Lessmann, V., 2001. NT-3 regulates BDNF-induced modulation of Teixeira Jr., A.L., Cardoso, F., Souza, A.L., Teixeira, M.M., 2004. Increased serum synaptic transmission in cultured hippocampal neurons. Neuroreport 12, 2635–2639. concentrations of monokine induced by interferon-gamma/CXCL9 and interferon- gamma-inducible protein 10/CXCL-10 in Sydenham's chorea patients. J. Neuroim- Polman, C., Barkhof, F., Sandberg-Wollheim, M., Linde, A., Nordle, O., Nederman, T., 2005. Treatment with laquinimod reduces development of active MRI lesions in munol. 150, 157–162. relapsing MS. Neurology 64, 987–991. Vanderlugt, C.J., Miller, S.D., 1996. Epitope spreading. Curr. Opin. Immunol. 8, 831–836. Randolph, C.L., Bierl, M.A., Isaacson, L.G., 2007. Regulation of NGF and NT-3 protein Verbsky, J.W., Randolph, D.A., Shornick, L.P., Chaplin, D.D., 2002. Nonhematopoietic expression in peripheral targets by sympathetic input. Brain Res. 1144, 59–69. expression of Janus kinase 3 is required for efficient recruitment of Th2 Roos, A., Schilder-Tol, E.J., Weening, J.J., Aten, J., 1998. Strong expression of CD134 lymphocytes and eosinophils in OVA-induced airway inflammation. J. Immunol. (OX40), a member of the TNF receptor family, in a T helper 2-type cytokine environment. J. Leukoc. Biol. 64, 503–510. 168, 2475–2482. Wan, Y.Y., Chi, H., Xie, M., Schneider, M.D., Flavell, R.A., 2006. The kinase TAK1 Sage, J., 2004. Cyclin C makes an entry into the cell cycle. Dev. Cell 6, 607–608. Sano, H., Kuroki, Y., 2005. The lung collectins, SP-A and SP-D, modulate pulmonary integrates antigen and cytokine receptor signaling for T cell development, survival innate immunity. Mol. Immunol. 42, 279–287. and function. Nat. Immunol. 7, 851–858. Segal, E., Friedman, N., Kaminski, N., Regev, A., Koller, D., 2005. From signatures to Wang, S., Chen, L., 2004. Co-signaling molecules of the B7-CD28 family in positive and models: understanding cancer using microarrays. Nat. Genet. 37, S38–S45 negative regulation of T lymphocyte responses. Microbes Infect. 6, 759–766. Suppl. Xiao, C., Ghosh, S., 2005. NF-kappaB, an evolutionarily conserved mediator of immune Stanley, E.R., Berg, K.L., Einstein, D.B., Lee, P.S., Pixley, F.J., Wang, Y., Yeung, Y.G., 1997. Biology and action of colony-stimulating factor-1. Mol. Reprod. Dev. 46, 4–10. and inflammatory responses. Adv. Exp. Med. Biol. 560, 41–45. Stenger, S., Hanson, D.A., Teitelbaum, R., Dewan, P., Niazi, K.R., Froelich, C.J., Ganz, T., Thoma-Uszynski, S., Melian, A., Bogdan, C., Porcelli, S.A., Bloom, B.R., Krensky, A.M., Yamashita, M., Katsumata, M., Iwashima, M., Kimura, M., Shimizu, C., Kamata, T., Shin, Modlin, R.L., 1998. An antimicrobial activity of cytolytic T cells mediated by T., Seki, N., Suzuki, S., Taniguchi, M., Nakayama, T., 2000. T cell receptor-induced granulysin. Science 282, 121–125. Stoeckle, C., Tolosa, E., 2009. Antigen processing and presentation in multiple sclerosis. calcineurin activation regulates T helper type 2 cell development by modifying the Results Probl. Cell Differ. interleukin 4 receptor signaling complex. J. Exp. Med. 191, 1869–1879. Strack, P., Caligiuri, M., Pelletier, M., Boisclair, M., Theodoras, A., Beer-Romero, P., Glass, Yang, J.S., Xu, L.Y., Xiao, B.G., Hedlund, G., Link, H., 2004. Laquinimod (ABR-215062) S., Parsons, T., Copeland, R.A., Auger, K.R., Benfield, P., Brizuela, L., Rolfe, M., 2000. suppresses the development of experimental autoimmune encephalomyelitis, SCF(beta-TRCP) and phosphorylation dependent ubiquitinationof I kappa B alpha catalyzed by Ubc3 and Ubc4. Oncogene 19, 3529–3536. modulates the Th1/Th2 balance and induces the Th3 cytokine TGF-beta in Lewis Takada, T., Toriyama, K., Muramatsu, H., Song, X.J., Torii, S., Muramatsu, T., 1997. Midkine, a retinoic acid-inducible heparin-binding cytokine in inflammatory rats. J. Neuroimmunol. 156, 3–9. Zhang, G.X., Baker, C.M., Kolson, D.L., Rostami, A.M., 2000. Chemokines and chemokine receptors in the pathogenesis of multiple sclerosis. Mult. Scler. 6, 3–13. Zou, L.P., Abbas, N., Volkmann, I., Nennesmo, I., Levi, M., Wahren, B., Winblad, B., Hedlund, G., Zhu, J., 2002. Suppression of experimental autoimmune neuritis by ABR-215062 is associated with altered Th1/Th2 balance and inhibited migration of inflammatory cells into the peripheral nerve tissue. Neuropharmacology 42, 731–739. Downloaded for Anonymous User (n/a) at Sheba Medical Center from ClinicalKey.com by Elsevier on October 19, 452021. For personal use only. No other uses without permission. Copyright ©2021. Elsevier Inc. All rights reserved.



‫טרשת נפוצה טבה‬ ‫הרפואה | ‪2011‬‬ ‫פרופ' ענת אחירון‬ ‫מייסדת פרויקט ח\"ץ‪ ,‬מנהלת המרכז‬ ‫לטרשת נפוצה ואחראית הקתדרה‬ ‫למחלות אוטואימוניות אוניברסיטת ת\"א‬ ‫‪[email protected]‬‬ ‫גלעד וינדר‬ ‫אונ' תל אביב‬ ‫השתתף כסטודנט בפרויקט ח״ץ‬ ‫בין השנים ‪2007-2009‬‬ ‫‪47‬‬

‫הרפואה • כרך ‪ • 150‬חוב' ‪ • 5‬מאי ‪2011‬‬ ‫סקירות‬ ‫טרשת נפוצה טבה‬ ‫ענת אחירון‪2,1‬‬ ‫גלעד וינדר*‪2,‬‬ ‫בסקירה זו נדון במאפיינים השונים של טרשת נפוצה טבה (‪Benign multiple‬‬ ‫תקציר‪:‬‬ ‫‪ ,)sclerosis‬בהגדרת הנכות על פי קריטריונים קוגניטיביים ומוטוריים‪ ,‬במשך‬ ‫‪1‬המרכז לטרשת נפוצה‪ ,‬מרכז רפואי שיבא‪ ,‬תל השומר‪ ,‬רמת גן‬ ‫המחלה‪ ,‬במאפיינים בתהודה מגנטית (‪ ,)MRI‬בהתקדמות המחלה הטבה‬ ‫‪2‬הפקולטה לרפואה סאקלר‪ ,‬אוניברסיטת תל אביב‪ ,‬רמת אביב‬ ‫למחלה מתקדמת ובקריטריונים לטיפול‪.‬‬ ‫*סטודנט לרפואה שנה ה'‪.‬‬ ‫טרשת נפוצה; מהלך טב; תהודה מגנטית; תפקוד קוגניטיבי; מדד מצב מוגבלות ‪.EDSS -‬‬ ‫מילות מפתח‪:‬‬ ‫‪Multiple sclerosis; Benign; Magnetic resonance imaging; Cognition; Expanded Disability Status Scale.‬‬ ‫‪:KEY WORDS‬‬ ‫קוגניטיביים)‪ .‬טווח תוצאות ה–‪ EDSS‬נע בין ‪ ,10-0‬כאשר ציון ‪0‬‬ ‫הקדמה‬ ‫‪463‬‬ ‫מסמן בדיקה נוירולוגית תקינה‪ ,‬וככל שהתוצאה גבוהה יותר היא‬ ‫טרשת נפוצה היא מחלה דלקתית‪ ,‬אוטואימונית דה–מיאלינטיבית‬ ‫מעידה על פגיעה חמורה יותר [‪.]11‬‬ ‫כרונית‪ ,‬הפוגעת במערכת העצבים המרכזית‪ .‬טרשת נפוצה היא‬ ‫טרשת נפוצה טבה מוגדרת כמחלה עם מוגבלות נמוכה‪ ,‬כלומר‬ ‫השכיחה מבין המחלות הדה–מיאלנטיביות‪ .‬במחלות אלו מתרחשת‬ ‫ערך נמוך מ–‪ ]10[ 3.5‬או קטן‪/‬שווה ל–‪ 3.0‬בסולם ה–‪.]7[ EDSS‬‬ ‫פגיעה במעטפת המיאלין של אקסונים במערכת העצבים המרכזית‪,‬‬ ‫המשמעות היא פגיעה נוירולוגית קלה ללא הגבלה בניידות‪ ,‬דהיינו‬ ‫המביאה להפרעה בהעברת גירוי עצבי דרכם‪ .‬מהלך המחלה מתאפיין‬ ‫המטופל מסוגל ללכת מרחק של לפחות ‪ 500‬מטר ללא אמצעי עזר‪.‬‬ ‫באירועים מוגדרים של חסר נוירולוגי (התקפים חדים)‪ ,‬ומלווה לעיתים‬ ‫במהלך של פגיעה נוירולוגית מתקדמת לאורך זמן‪ .‬לשם הגדרת‬ ‫ג‪ .‬תפקוד קוגניטיבי‬ ‫המחלה דרושים לפחות שני אירועים קליניים של התקפים חדים‬ ‫‪ EDSS‬מהווה מדד לחומרת התסמינים ולהתקדמות המחלה‪ ,‬וניתנת בו‬ ‫[‪ .]1‬המחלה מופיעה בגיל צעיר‪ ,‬שיאה בסביבות גיל ‪ 30‬שנה‪ ,‬וטווח‬ ‫חשיבות גדולה למדדים הקשורים לתנועה‪ .‬פגיעה קוגניטיבית משפיעה‬ ‫ההופעה בין הגילים ‪ 40-20‬שנה‪ .‬כברוב המחלות האוטואימוניות‪,‬‬ ‫פחות על מדד ה–‪ ,EDSS‬למרות שירידה קוגניטיבית מתרחשת בחלק‬ ‫ניכר מחולי טרשת נפוצה‪ Amato .‬וחב' [‪ ]12‬דיווחו על ליקוי קוגניטיבי‬ ‫נשים לוקות במחלה בשיעור הגבוה פי שניים מגברים‪.‬‬ ‫ב–‪ 45%‬מהלוקים בטרשת נפוצה טבה‪ .‬הליקוי הקוגניטיבי הוא אחד‬ ‫טרשת נפוצה היא מחלה הטרוגנית מבחינת ביטוייה הקליניים‪,‬‬ ‫המאפיינים העיקריים המשפיעים על איכות חייהם של חולי טרשת‬ ‫הגנטיים‪ ,‬האימונולוגיים ומידת הנזק המשתקף בתהודה מגנטית‬ ‫נפוצה טבה [‪ .]13‬כמו כן‪ ,‬תפקוד קוגניטיבי פגוע הוא אחד המדדים‬ ‫(‪ .]2[ )MRI‬מקובל בספרות‪ ,‬כי ב–‪ 20%-10%‬מכלל חולי טרשת נפוצה‬ ‫המגדילים את הסיכוי להתקדמות המחלה ממהלך טב למתקדם יותר‬ ‫מהלך המחלה טב ‪ -‬מצב הקרוי 'טרשת נפוצה טבה' [‪ .]4,3‬אולם‬ ‫[‪ Rovaris .]9‬וחב' [‪ ]14‬מצאו‪ ,‬כי בחולי טרשת נפוצה טבה עם ליקוי‬ ‫עקב היעדר הסכמה גורפת על הקריטריונים להגדרת המחלה‪ ,‬הרי‬ ‫קוגניטיבי נמצא נזק מבני חמור בחומר הלבן‪ ,‬הדומה בצורה משמעותית‬ ‫ששיעור הלוקים בטרשת נפוצה טבה מכלל החולים בטרשת נפוצה‪,‬‬ ‫לפגיעה בחולי טרשת נפוצה עם מהלך מחלה מתקדם שניוני (‪.)SPMS‬‬ ‫‪ Mesaros‬וחב' [‪ ]16,15‬הראו‪ ,‬כי ליקוי קוגניטיבי בחולי טרשת נפוצה‬ ‫נע על פי פרסומים בספרות בין ‪.]6,5[ 40%-5%‬‬ ‫טבה נמצא במיתאם עם נזק נרחב יותר בקורה (‪- Corpus callosum‬‬ ‫אגד הסיבים הלבנים הגדול ביותר המחבר בין ההמיספרות ומכאן‬ ‫קשיים בהגדרת טרשת נפוצה טבה‬ ‫בעל חשיבות בתפקוד קוגניטיבי)‪ ,‬הן בנזק מוקדי כמו נגעים‪ ,‬והן בנזק‬ ‫א‪ .‬ממד הזמן ‪ -‬משך מחלה‬ ‫מפושט כמו ‪.]16,15[ )Normal-Appearing White Matter( NAWN‬‬ ‫הגדרת טרשת נפוצה טבה על פי מדד משך מחלה היא‬ ‫אחת ההמלצות היא להכליל את ההערכה הקוגניטיבית כמדד‬ ‫רטרוספקטיבית‪ ,‬לאחר משך מחלה ארוך בליווי מוגבלות בדרגה‬ ‫לטרשת נפוצה‪ ,‬ואם קיים ליקוי קוגניטיבי לא יוגדר מצב של טרשת‬ ‫נמוכה‪ ,‬כפי שבאה לידי ביטוי ב–‪Expended Disability Status( EDSS‬‬ ‫נפוצה טבה‪ .‬בהתאמה‪ ,‬הגדרה זו גורסת כי תפקוד קוגניטיבי שמור‬ ‫‪ .)Scale‬משך מחלה ארוך הוא עשור ומעלה [‪ ]8,7‬או ‪ 15‬שנים ומעלה‬ ‫[‪ .]10,9‬עם זאת‪ ,‬כיוון שדווח בספרות על חולים שמצבם החמיר‬ ‫מהווה קריטריון באבחון טרשת נפוצה טבה [‪.]3‬‬ ‫גם לאחר תקופה ארוכה‪ ,‬תיתכן בעייתיות ביישום המדד של משך‬ ‫ד‪ .‬פעילות המחלה על פי בדיקת תהודה מגנטית של המוח‬ ‫מחלה בלבד לצורך הגדרת טרשת נפוצה טבה‪.‬‬ ‫(‪)Brain MRI‬‬ ‫ב‪ .‬נכות נוירולוגית‬ ‫תהודה מגנטית של המוח תורמת תרומה משמעותית לאבחון‬ ‫מצב הנכות הנוירולוגית הוא המרכיב השני בהגדרת מהלך טב‬ ‫טרשת נפוצה כחלק מהמדדים של ‪ ,McDonald‬שהם מדדי האבחון‬ ‫של טרשת נפוצה‪ .‬הערכת הנכות נקבעת על פי תוצאות בדיקה‬ ‫המומלצים לטרשת נפוצה [‪ .]17‬את הנגעים האופייניים בחומר הלבן‬ ‫נוירולוגית מפורטת וסולם ה–‪ .EDSS‬סולם זה הוא מדד שנועד‬ ‫מקובל למדוד באופן כמותי מבחינת מספר הנגעים ונפחם‪ .‬למיקום‬ ‫לכמת את הנכות הנוירולוגית על פי מערכות תפקודיות (המערכת‬ ‫המוטורית‪ ,‬המוחון ותפקודי קואורדינציה‪ ,‬גזע המוח ‪ -‬דיבור‬ ‫הנגעים ומספרם יש משמעות פרוגנוסטית למהלך המחלה [‪.]18‬‬ ‫ובליעה‪ ,‬תחושה‪ ,‬ראייה‪ ,‬תפקוד המעי ושלפוחית השתן‪ ,‬ותפקודים‬ ‫בעשור האחרון‪ ,‬במסגרת יישום שיטת ‪Diffusion Tensor Imaging‬‬ ‫(‪ )DTI‬בתהודה מגנטית‪ ,‬התגלה כי גם באזורי חומר לבן שנראים‬ ‫‪48‬‬