1. PulseNet USA: The National Molecular Subtyping Network for Foodborne Disease Surveillance Jennifer A. Kincaid Centers for Disease Control and Prevention May 30, 2008

2. What is PulseNet USA?  Established in 1996, The National Molecular Subtyping Network for Foodborne Disease Surveillance  National network of state and local public health/food regulatory agency laboratories (USDA, FDA) coordinated by CDC and APHL  Perform standardized molecular typing of foodborne disease-causing bacteria by Pulsed-field gel electrophoresis (PFGE)  Dynamic databases of DNA “fingerprints” at CDC—available on-demand to participants

3. The National Molecular Subtyping Network for Foodborne Disease Surveillance USDA-APHIS New York Ag. Lab NEW HAMPSHIRE Milwaukee MAINE FDA-ORA VERMONT WASHINGTON MI State MONTANANORTH DAKOTA MINNESOTA Vet Lab MASSACHUSETTS Philadelphia OREGONWISCONSIN Ohio Ag. Lab NEW YORK RHODE ISLAND CONNECTICUT IDAHOSOUTH DAKOTA FDA-ORA WYOMING IOWA New York City NEW JERSEY OHIO NEBRASKA FDA-CVM NEVADA DELAWARE ILLINOIS FDA-CFSAN UTAH FDA-ORA MARYLAND FDA-ORA COLORADO Washington D.C. Las KANSAS VIRGINIA Vegas KENTUCKY Santa Clara CALIFORNIA MISSOURI County USDA-AMS TENNESSEE OKLAHOMA ARKANSAS SOUTH FDA-ORA ARIZONA CAROLINA Los Angeles NEW MEXICO County USDA- FDA-ORA ARS/FSIS/FERN ALABAMA Orange County GEORGIA FDA-ORA Area Laboratories San Diego TEXAS County LOUISIANA Florida Ag. Lab PulseNet Headquarters ALASKA State/County/City/ Tarrant CountyDallas County Veterinary/ Agricultural Laboratories Tampa Houston PUERTO USDA Laboratories RICO HAWAII FDA Laboratories West Mountain South Central North Central Midwest Mid-Atlantic Southeast Northeast

4. PulseNet Objectives  To detect foodborne diseases case clusters that may be widespread outbreaks  Assist epidemiologists in investigating outbreaks  Separate outbreak-associated cases from other sporadic cases (case definition)  Assist in rapidly identifying the source of outbreaks (culture confirmation)  Act as a rapid and effective means of communication between public health laboratories (rapid alert system)  Attribution analyses

5. PulseNet: Communication and QA/QC  On-line databases  Standardized protocols  CDC Team postingsand molecular size  Cluster detection standards  Outbreak investigations  QA/QC Manual  Technical support  Standardized software  Quarterly/Annual Reports and nomenclature from CDC  Training workshops (lab  “PulseNet News” Newsletter & software)  Tri-annual publication  PulseNet Website  Certification and (www.cdc.gov/pulsenet) proficiency testing  Annual update meetings

6. The Three Basic Elements of PulseNet 2.Data analysis 1.Data acquisition 3. Data exchange

7. Pulsed Field Gel Electrophoresis  PFGE highly discriminatory  Universal, relatively simple technique that can be used in most laboratories  Definitive subtyping method, if highly standardized  Current “Gold Standard” for molecular subtyping

8. PFGE Process Bacterial Suspension Mix with Agarose Plug Mold Chemical Lysis and Washing DNA in Plugs Restriction Enzymes Electrophoresis (PFGE) Documentation (capture gel image) Data Analysis (BioNumerics)

9. PFGE Patterns of E. coli O157:H7 Fragment *** Sizes (in kilobas es) 1135 Kb 452.7 Kb 216.9 Kb 76.8 Kb 33.3 Kb DNA “fingerprints” *Global Reference Standard

10. PulseNet Laboratory Network Participating Labs PFGE Patterns & PulseNet National Demographic Data Databases (CDC) TAT from receipt to upload: ~4 working days Dbase Management Local Reports Databases

11. Local vs. National Databases LocalNational  Smaller, results from a few  Larger, results from many peopledifferent labs  Smaller representation of  Larger representation of patterns  Tends to be less diverse  More diverse  Sometimes different  Different “common” patterns “common” patterns than seenthan seen locally nationally or in other regions  See representation from all  Can only see localover, different regions representation, no national perspective

12. PulseNet Activity* *as of May 15, 2008 Over 325,000 PFGE patterns or DNA “fingerprints” submitted to PulseNet databases since 1996 DatabaseEntriesPatterns submitted Submitted 1st Enzyme2nd Enzyme Campylobacter5,4535,4001,834 E. coli31,63430,26216,674 Listeria9,2269,0078,002 Salmonella190,298188,24128,998 Shigella32,92232,2222,156 V. cholerae303282272 V. parahaemolyticus37 Y. pestis1,9761,97533

13. PulseNet Activity, 1996-2007 PFGE patterns submitted to PulseNet Databases 70000 6000 0 5000 0 4000 0 3000 0 2000 0 1000 0 0 1996 1997 1998 1999 2000 2001* 2002 2003 2004 2005 2006 2007

14. PulseNet is a cluster detection tool, not an outbreak detection system  A PulseNet CLUSTER is a group of patterns that are found indistinguishable by PFGE  CLUSTERS of cases identified by PulseNet are investigated by epidemiologists  If epidemiologic links are found between cases, the cluster is classified as an OUTBREAK

15. PulseNet Data Analysis: The Cluster Search Patterns submitted electronically 60- or 120-day cluster search performed Visually compare indistinguis hable Patterns and clusters are named by CDC Clusters communicated to the foodborne epidemiologists Cluster of indistinguishable patterns

16. Cluster Detection in PulseNet: PulseNet Workspace on CDC Team Subject:0802LACJPX-1c (JPXX01.0088)_LAC_Typhimurium LAC has a cluster of 3 Typhimurium isolates that are indistinguisable by XbaI and BlnI. The collection dates are from 12/10/2007 to 1/10/2008. Epi under investigation. The isolate numbers are: LAC_Z19766 2 year old boy, no travel LAC_Z19999, 43 year old female, no travel LAC_Z20072, 60 year old female, no travel We have seen this pattern before in LAC. Our epi contact is (name and phone) LAC08008PN.BDL Posted: 05 March 2008 09:52 AM

17. Cluster Detection  Local  National  Perform cluster search within local databasewithin national database or respond to local CDC  Compare to national Team posting database  Verify isolate information  Post message to CDC  Name patterns Team  Check band markings  Begin epi investigation  Assign cluster code  Look at pattern frequencies/tre nds

18. A large outbreak in one place may be obvious Detected and investigated locally

19. A dispersed outbreak in many places may be difficult to detect, unless…  Detect outbreaks centrally (or locally) through surveillance (widely dispersed, organism too common to notice small increase, identify related cases)  Investigation coordinated centrally  Distinguish from concurrent sporadic cases  Provide microbiological evidence of sources of outbreaks

20. Common CDC Epi Requests  Past trends or past association of cluster/outbreak pattern  Pattern Frequency graphs  PulseNet Cluster Log  Cluster/Outbreak codes  Updated line lists  Often daily for very active or high profile outbreaks  Additional information for rumor logs and conference calls

21. CDC Epi Requests: Additional Info  Second enzyme results  VetNet matches  USDA-ARS database  Connection to VetNet network established in 2007  All cluster patterns identified by PulseNet are routinely checked against VetNet  International matches  Connection to PulseNet Canada databases established in 2007  All cluster patterns identified by PulseNet are routinely checked against PulseNet Canada databases  PulseNet participants abroad (PulseNet Latin America, Europe, Asia Pacific, and Middle East)  Email alerts to PulseNet International

22. Interpreting Patterns within Outbreaks  For surveillance (cluster detection), allow no differences for possible matches  One-band differences are so common in large databases that virtually all submissions would be part of a cluster  For outbreaks (case definition), differences may be acceptable  May be necessary for shigellosis infections (person to person)  If the isolates with different patterns have epidemiologic links, assign different pattern names but include in reports (give same outbreak code)

23. Additional Restriction Enzymes  Adds to the discrimination of all organisms when using PulseNet protocols  makes cluster detection more precise  Can limit the need for epi follow-up of likely unrelated cases  Although it adds to the cost of PulseNet surveillance, running fingerprint profiles of both enzymes on the initial gel is often cost-effective and saves time

24. Additional Restriction Enzymes Example Salmonella Montevideo Indistinguishable 1 st enzyme patterns: need 2 nd enzyme results to distinguish clusters

25. Examples of PulseNet Involvement in Foodborne Outbreaks

26. E. Coli O157:H7 Outbreak in Colorado, July 2002  Outbreak pattern was rare  Closely related to a very common pattern  34 cases in 11 states  Outbreak strain found in ground beef from meat processing plant by USDA/FSIS  Outbreak stopped after recall of 18.4 million pounds of ground beef products from a plant

27. 1993 Western States E. coli O157 Outbreak 70 60 outbreak detected 1993726 ill, 4 deaths 504030504030 20 39 d 10 0 18152229364350576471 Day of Outbreak 2002 Colorado E. coli O157 Outbreak 70 60506050 40 outbreak detected 200244 ill, no deaths 30203020 10 18 d 0 18152229364350576471 Day of Outbreak

28. S. Tennessee outbreak numbers (April 11, 2007)  567 cases  47 states involved  No international cases  Median age 51 years (<1- 95 years)  73 % females  No deaths, 20% hospitalized  Among females: 1/3 UTI

29. Salmonella Tennessee Cases by State, (N=563) 1-9 cases10-19 cases20+ cases *Slide courtesy of EIS officer Anandi Sheth

30. Peanut Butter Samples  34 positive samples of peanut butter  Production dates from contaminated jars: 7/27/06 - 1/29/07  2 of the 3 outbreak patterns were found in the peanut butter

31. Notable Outbreaks  Salmonella Wandsworth and Typhimurium: Veggie Booty  Salmonella Heidelberg: Hummus  Salmonella I 4, 5, 12:i:- : Pot pies  Salmonella Enteritidis: Chicken Cordon Bleu  E. coli O157: ground beef  E. coli O157: frozen pepperoni pizza  Listeria monocytogenes: milk

32. Examples of Other Database Uses

33. Database Uses: Pattern Frequencies Long-term surveillance to aide in active surveillance—is this a true increase?

34. Database Uses: Pattern Trends Long-term surveillance to see if a pattern is still seen as commonly in the database

35. Database Uses: Attribution Analysis Long-term surveillance to see if certain patterns may be associated with specific foods Legend CDS 1 CDS 2 CDS 3 CDS 4 CDS 5

36. PulseNet S. Newport Attribution Analysis Figure 13: Proportional Distribution of Illnesses by food commodities using three attribution analysis methods 30.0% 25.0 % 20.0 % Method 1 15.0% Method 2 Method 3 10.0% 5.0 % 0.0 % Food Commodities Method 1: Considering three major clusters Method 2: Considering major clusters and sub-clusters Method 3: Considering sub-clusters with no non-human isolates of unknown

37. PulseNet International… A Family of Networks

38. Why are we going international with PulseNet?  We live in a global community  Foods produced on one continent may be consumed and cause disease in another - a global issue  Need an effective global early warning system  Global networking to utilize scarce public health resources effectively  Respond to other emerging infectious diseases or acts of bioterrorism

39. PulseNet Internationa l Collaboratio ns  Outbreak investigations  Laboratory and Analysis Troubleshooting  Protocol development and validation  Vibrio cholerae  Vibrio parahaemolyticus  Development, Evaluation and Validation of Next Generation Typing Methods  Canada, China, Hong Kong, Japan and U.K. MLVA subtyping protocol for E. coli O157

40. PulseNet Challenges and Future Improvements  Enhance support for outbreak investigations  New faster methods  Reduce the time it takes for isolates to go from clinical lab to the state/local public health lab  Reduce the time for PFGE testing of isolates  Expectation: ~4 days from receipt of isolate to upload to the national database(s)  Achieve real-time detection of clusters  Improve communication between laboratorians and epidemiologists at the state and national levels

41. PulseNet Challenges and Future Improvements  Strengthen collaborations with food industry  VetNet (USDA- ARS)  Currently collecting Salmonella and Campylobacter NARMS isolates  Attribution analysis  Protocols  Vibrio parahaemolyticus  Protocol has been developed and database scripts were included in the latest version of the MasterScripts (March 2007)  The number of submissions to PulseNet continues to rise…..

42. New Subtyping Methods  PFGE data is complex  Need for simple, non-image based as discriminatory supplementary methods  Sequencing based methods  Multi Locus Variable number of tandem repeats Analysis (MLVA)  SNP analysis

43. MLVA Analysis  Sequence-based subtyping  Can further discriminate common PFGE patterns through highly variable target sequences  Data may be epidemiologically more relevant than PFGE data  Results more straightforward  Don’t have the same interpretation issues as with PFGE band marking

44. Thank you for your attention The findings and conclusions in this presentation are those of the author and do not necessarily represent the views of the Centers for Disease Control and Prevention.