Food Protection Trends - November/December 2019 - 483

three other states, and then rapidly growing as more states
joined the network. We started with E. coli O157, then expanded
to include Listeria monocytogenes, Salmonella and other
enteric bacteria, and the initial effort in four pilot states expanded
to include all 50 states, and several large cities. The central
premise was that if we found a cluster of infections caused by
strains of E. coli O157 that were indistinguishable by the PFGE
methods, they are likely to share a source; if we interviewed
those patients, we might very well be able to identify what it
was, something we wouldn't have known otherwise. The PFGE
subtyping enhanced signal detection against the background
noise of sporadic cases. PulseNet is now over 20 years old,
with participating laboratories testing over 50,000 bacteria
a year, using a standard subtyping method and storing the
information into a common database that all participants can
review and use. FDA and USDA/FSIS participated from the
beginning, testing isolates from foods and animals and putting
the results into the same database so they can be compared
immediately with the clinical isolates.
As PulseNet grew, the number of multistate foodborne
outbreaks reported in our outbreak surveillance surged, from
2-3 per year in the decades preceding the launch, to 23 such
outbreaks in 2010. By linking cases together across the
sub-typing network, we could now detect and investigate a
dispersed cluster spread across several states, which meant
we were able to stop an ongoing outbreak, identify food safety
gaps early in the production of a food that was then distributed
across the country, and drive improvements in prevention
across the system. This depended on finding that signal by
subtyping strains, investigating those clusters by interviewing
patients that have bacteria with the same subtype, and linking
that information to food source traceback information and to
any non-human isolates that may be relevant.
PulseNet and other major improvements in surveillance
and in prevention followed in the wake of the large and devastating outbreak of E. coli O157 infections in that they were
linked to ground beef in 1993. This was a time of great improvement in our food safety systems, including modernization
of inspection of meat and poultry with a new focus on reducing
pathogen contamination and great attention to the safety of
processed meats. Also in 1996, we at CDC with 10 state
partners, FDA, and USDA/FSIS began a comprehensive
surveillance program called FoodNet, to track the frequency
of several infections often spread by food. Between 1996
and the mid 2000s, FoodNet documented major reductions
of between 43% to 49% in infections caused by E. coli O157,
Listeria monocytogenes, and Campylobacter though Salmonella infections that remained essentially unchanged. Since the
mid-2000s, FoodNet has not seen further substantial reductions in these infections, which makes it unlikely that we will
reach the desired 25% reductions from 2007 that are part of
the Healthy People 2020 goals, unless something changes.
What we decided we could do at CDC was to make surveillance and investigations even more powerful by applying
whole genome sequencing for routine subtyping in PulseNet.
The cost of sequencing had dropped dramatically, and the
technology could now be applied in public health laboratories.
This looked like it would have several benefits. Perhaps this
method would let us detect, investigate and control outbreaks
we may have been missing and identify other emerging problems that need to be addressed. We could also use our surveillance to attribute illnesses to specific food categories and
target interventions. New tools applied to the sequence data
could give us the ability to predict important strain characteristics directly from sequence, including serotype, antimicrobial
resistance and Shiga toxin type, making that information available early in an investigation. Finally, building a large database
of sequences is a critical step to developing the metagenomic
tools for public health that we will need to have in the future.

How do we determine that a food is the source of an
outbreak of illness? We assemble three main types of information to make that judgement. First, we use epidemiologic
evidence, showing that the persons who are infected with the
same strain report consuming a particular food at a far higher
frequency than would be expected. Second, traceback information can help to show that the suspect food item eaten in
several locations came from the same source, suggesting
where contamination might have been introduced. Third, microbiological assessment of that food, facility or other source may
show that the same pathogens are found in that food, facility or
farm. Having information of all three types is ideal; two types
of information can together be enough to take action
These types of evidence are evolving and in transition. For
example, the new culture independent clinical diagnostic tools
mean that some infections that were formerly rarely reported
are now being identified more frequently, such as those caused
by Cyclospora, Vibrio, Yersinia or enterotoxigenic E. coli. Traceback is improving, using digital supply chain systems that industry
is developing to identify sources more quickly and definitively.
Traceability is accountability, a critical part of the whole food
safety system. And then there is the transition from the over
20 year old PFGE method for subtyping to whole genome
sequencing for routine subtyping in PulseNet.
Starting in 2013, we started a collaborative pilot effort to
apply WGS to infections with Listeria monocytogenes, the
cause of rare but severe infections in the elderly, immunocompromised or pregnant individuals. We at CDC tested all the
isolates from clinical sources, approximately 800 per year,
while colleagues at FDA and USDA/FSIS tested isolates from
foods, food animals and other sources, with a new partner, the
National Center for Bioinformatic Information, serving as a
publicly available repository for DNA sequence information.
After three first years, the results were striking. The clusters
under investigation increased, the average size of those
clusters decreased, and the number of clusters that were
solved, i.e., for which the source was determined, jumped
three-fold. When PFGE itself was introduced, the number
of listeriosis outbreaks detected and solved had increased
from two per decade to two per year, and now with WGS, the
number increased again to six per year. These outbreaks were
traced to familiar sources like soft cheeses and sprouts, and
to new and previously unsuspected risks, like caramel apples,
stone fruits and ice cream, leading to new prevention efforts in
those sectors. During this pilot effort, we saw that events were
not always a group of highly-related cases clustered tightly
in space and time. Our appreciation for what an outbreak is
began to broaden.
For example, one large outbreak detected in 2015 was
caused by Listeria monocytogenes strains with five different
PFGE patterns but shown by WGS to be highly related. Had
we been relying on PFGE to detect the outbreak and guide its
investigation, we likely would not have realized it was a cluster
at all, and not launched an investigation. The final count was
30 human infections in residents of 10 states, of which 28
were hospitalized and three died, along with one fetal loss. Of
those interviewed, 75% reported eating a soft cheese of Middle
Eastern, Mediterranean or Hispanic style, that was traced back
to one cheese plant in California, where the same strain was
found. This led to a voluntary recall of the cheeses from that
plant. Interestingly, review of WGS found that human infections
from closely-related strains were identified as far back as 2010,
growing slowly in number to 2015, and matching a strain isolated at the same plant in 2010. This outbreak may thus have
been the result of in-plant harborage lasting for years, that had
now been detected and stopped.
(Continued on next page)

November/December Food Protection Trends

483

Food Protection Trends - November/December 2019

Table of Contents for the Digital Edition of Food Protection Trends - November/December 2019