The microbiome and personalized nutrition Eran Elinav, M.D., Ph.D. Immunology Department Weizmann Institute of Science, Israel AICR meeting November, 2016
The microbiome is a signaling hub Genetics Immunity
Studying host-microbiome interactions Circadian rhythmicity Thaiss et al, Cell 2014 Thaiss et al, Cell 2014, in press Nutrition Suez et al, Nature 2014 Zeevy et al, Cell 2015 Immunity Wladarska et al, Cell 2014 Levy et al, Cell 2015 Gury et al, Cell 2016 Eran Segal Relapsing obesity Thaiss et al, Nature 2016, in press Microbial Drivers of disease Korem et al, Science 2015
The obesity pandemic 1990 2000 2010 No Data <10% 10% 14% 15% 19% 20% 24% 25% 29% 30% 37.5% 78 million 17% 12.5 million Adults Children
Prevalence of Type II Diabetes worldwide
The obesity paradox Weight change (%) 2.0 0.0-2.0-4.0-6.0-8.0-10.0-12.0-14.0 Intervention -16.0-18.0 Males (average -7.0%) Weight Females change (average females -5.3%) Williams et al., Obesity Reviews, 2015
Our dieting failure Time following weight loss Months Exercise Meal replacement Very low energy diet Diet Exercise+diet Orlistat (pharma) Average weight loss Kilograms Source: McKinsey Global Institute November 2014
Current paradigm: defining diets by the food Low GI
Current paradigm: defining diets by the food High GI Low GI
Current paradigm: defining diets by the food
People have widely different glucose responses to the same food Percent of population 20 15 10 5 0 Response to Glucose 0 40 80 120 Glucose response (mg/dl) Glucose rise (mg/dl) 120 80 40 0 0 1 2 Time (hours)
People have widely different glucose responses to the same food Percent of population 25 20 15 10 5 0 Response to Bread 0 40 80 120 Glucose response (mg/dl) Glucose rise (mg/dl) 60 45 15 0 0 1 2 Time (hours)
one size fits all diet cannot be effective
Diet must be tailored to the individual
Factors driving personalized dietary responses can be measured Genetics Microbes Lifestyle
A novel sciencebased approach to nutrition: The Personalized Nutrition Project CRC Zeevi et al, Cell, 2015
The Personalized Nutrition Project: Experimental Setup Register online Log activities & food Analyze your bacteria Measure glucose response to food www.personalnutrition.or g Today Study Week Start During Study Week Study Week End View Bacteria Personal Diet Genetics Glucose Microbes Blood tests Questionnaires Connection meeting Study week analysis Plan personal diet
The Personalized Nutrition Project: Understanding personal glucose responses 1000 people profiled 50,000 meals measured 9,000,000 Kcal 12,000 Kg 5,000 standardized meals 2,000,000 glucose measures 150,000 hours 10 Billion metagenome reads 5,000 more people registered http://www.personalnutrition.org
The Personalized Nutrition Project: Cohort statistics Age BMI 25-70 years of age ~50% overweight ~20% obese ~25% pre-diabetic HbA1C%
Microbiome features overlap with HMP & MetaHIT
Postprandial (post-meal) response is correlated with multiple clinical markers Maximal post-meal response (mg/dl) 350 300 250 200 150 100 50 30 R=0.34 BMI 28 26 24 22 0 100 200 300 400 500 600 700 Participants (sorted)
Postprandial (post-meal) response is correlated with multiple clinical markers Maximal post-meal response (mg/dl) 350 300 250 200 150 100 50 6.2 6.0 5.8 R=0.56 HbA1c% 5.6 5.4 5.2 0 100 200 300 400 500 600 700 Participants (sorted) Maximal post-meal response (mg/dl) 350 300 250 200 150 100 50 110 R=0.65 Wakeup glucose (mg/dl) 100 90 80 0 100 200 300 400 500 600 700 Participants (sorted)
Postprandial (post-meal) response is correlated with multiple clinical markers Maximal post-meal response (mg/dl) 350 300 250 200 150 100 50 60 55 50 R=0.45 Age 45 40 35 30 0 100 200 300 400 500 600 700 Participants (sorted) Maximal post-meal response (mg/dl) 350 300 250 200 150 100 50 70 R=-0.22 HDL Cholesterol (mg/dl) 65 60 55 0 100 200 300 400 500 600 700 Participants (sorted)
Testing the cohort response to standardized meals 800 x
Testing the cohort response to standardized meals
Major inter-personal variability in the response to test foods
Major inter-personal variability in the response to real-life foods
Prediction scheme: current state of the art Measured iauc (mg/dl h) r=0.37 Meal carbohydrates (g)
New prediction scheme
Prediction scheme 0.7 0.64 0.67 Model Features R 0.6 0.5 0.55 0.4 0.37 0.41
Prediction scheme 0.7 0.64 0.67 Model Features Meal features Nutrient weights, water R 0.6 0.55 0.5 0.4 0.37 0.41
Prediction scheme 0.7 0.64 0.67 Model Features Meal features Nutrient weights, water R 0.6 0.5 0.55 Logged features Time of day, sleep, exercise 0.4 0.37 0.41
Prediction scheme 0.7 0.64 0.67 Model Features Meal features Nutrient weights, water R 0.6 0.5 0.55 Logged features Time of day, sleep, exercise 0.4 0.37 0.41 Microbiome features Gene levels, bacteria levels, 16S
Prediction scheme 0.7 0.64 0.68 Model Features Meal features Nutrient weights, water R 0.6 0.5 0.55 Logged features Time of day, sleep, exercise 0.4 0.37 0.41 Microbiome features Gene levels, bacteria levels, 16S Personal features Blood tests, questionnaires
Accurate predictions of personalized glucose responses State of the art Our prediction 800 participants Prediction validation 100 participants Measured PPGR (iauc; mg/dl*h) Meal carbohydrates (g) Predicted PPGR (iauc; mg/dl*h) Predicted PPGR (iauc; mg/dl*h)
Microbiome features affect the post-meal glucose response
Constructing personally tailored diets that achieve normal post-prandial glucose responses Expert (n=14) Good Diet: Low predicted responses Food diary Continuous glucose monitoring Daily microbiota profiling 6 x Predict 1 2 3 4 5 6 7 Days Bad Diet: High predicted responses Food diary One week profiling Algorithm (n=12) Continuous glucose monitoring Daily microbiota profiling 1 2 3 4 5 6 7 Days
Foods that appear in the good diet of one person may appear in the bad diet of another
Can you distinguish between the good and bad menus? Bad Good? Bad Good? Diet Diet Diet Diet Breakfast Lunch Snack Dinner Night snack Muesli Sushi Marzipan Corn and nuts Toblerone and coffee Egg with bread and coffee Hummus and pita Edamame Vegetable noodles with tofu Ice cream
Can you distinguish between the good and bad menus? Bad Diet Good Diet Breakfast Lunch Snack Dinner Night snack Muesli Sushi Marzipan Corn and nuts Toblerone and coffee Egg with bread and coffee Hummus and pita Edamame Vegetable noodles with tofu Ice cream
Personally tailored diets reduce the post-prandial glucose response Glucose levels (mg/dl) 180 140 100 60 1 2 3 4 5 6 7 Days Good Bad Glucose rise (mg/dl) 30 P<10-6 20 10 0 0 1 2 3 Time (hr)
Can you distinguish between the good and bad menus? Bad Good? Menu Menu Bad Good? Menu Menu Breakfast Lunch Snack Dinner Night snack Orange juice Schnitzel Peach Bread with butter Grapes Croissant Goulash with rice Halva Hummus Red wine
Can you distinguish between the good and bad menus? Bad Menu Good Menu Breakfast Lunch Snack Dinner Night snack Orange juice Schnitzel Peach Bread with butter Grapes Croissant Goulash with rice Halva Hummus Red wine
Personally tailored diets reduce the post-prandial glucose response Glucose levels (mg/dl) 180 140 100 60 1 2 3 4 5 6 7 Days Good bad Glucose rise (mg/dl) 30 P<10-6 20 10 0 0 1 2 3 Time (hr)
Personally tailored diets reduce the post-prandial glucose response Glucose levels (mg/dl) 180 140 100 60 Good bad 1 2 3 4 5 6 7 Days Glucose rise (mg/dl) 30 P<0.004 20 10 0 0 1 2 3 Time (hr)
Personally tailored diets reduce the post-prandial glucose response
Personalized dietary interventions induce consistent changes in microbiota
The diet- microbiome- metabolism axis can be diagnostically and therapeutically exploited
Thanks Elinav lab Meirav Pevsner-Fischer Hagit Shapiro Mally Bachash-Dori Nira Truzman Eran Blacher Karina Golberg Stavros Bashiardes Niv Zmora Sara Federici Chagai rott Maayan Levy Christoph Thaiss Jotham Suez Gili Zilberman-Shapira Timur Tuganbaev Ori Maza Shlomic Itav Shachar Rozin Evgeny Tatirovsky Leviel Fluhr Sofia Braverman Jemal Ali Mahdi Daniel Halstuch Maya Elazar Uria Mor Lenka Dohnalova Keren Uritzky Yotam Harnik Elise Taieb Rina Gudinski Career Development chair Segal Lab David Zeevi Tal Korem Adina Weinberger Daphna Rothschild Tali Avnit-Sagi Maya Pompan-Lotan Dar Lador Michal Rein Orly Ben Yaakov Noa Kossower Gal Malka Izhak Levy Hadas Elisar Anastasia Godneva Neta Levkovich Prof. Zamir Halpern Prof. Alon Harmelin