Unraveling the Contact Network Patterns between Commercial Turkey Operation in North Carolina and the Distribution of Salmonella Species

Salmonellosis originating from poultry poses a significant threat to human health. Surveillance within production is thus needed to minimize risk. The objectives of this work were to investigate the distribution of Salmonella spp. from a commercial turkey operation and describe the animal movement patterns to investigate the association between contact network structure and Salmonella infection status. Four years of routine growout farm samples along with data on facility location, time since barns were built, production style, and bird movement data were utilized. From all of the surveillance samples collected, Salmonella serotyping was performed on positive samples and results showed that the most represented groups were C1 (28.67%), B (24.37%) and C2 (17.13%). The serovar Infantis (26.44%) was the most highly represented, followed by Senftenberg (12.76%) and Albany (10.93%). Results illustrated the seasonality of Salmonella presence with a higher number of positive samples being collected in the second half of each calendar year. We also demonstrated that Salmonella was more likely to occur in samples from older farms compared to farms built more recently. The contact network connectivity was low, although a few highly connected farms were identified. Results of the contact network showed that the farms which tested positive for Salmonella were not clustered within the network, suggesting that even though Salmonella dissemination occurs via transferring infected birds, for this study case it is unlikely the most important route of transmission. In conclusion, this study identified seasonality of Salmonella with significantly more cases in the second half of each year and also uncovered the role of between-farm movement of birds as not a major mode of Salmonella transmission.


Introduction
Salmonella infections in poultry are a health risk to the birds and later on to humans as a foodborne pathogen. The consumption of turkey in the U.S. has increased 17-fold since 1909 [1]. Scharff estimates Salmonella species in poultry to cost approximately $2.8 billion annually in the U. S. [2]. Consumption of undercooked contaminated products can cause acute, severe gastroenteritis in humans [3]. This straight, non-spore-forming Gram-negative rod can be quite challenging to control as the introduction to and maintenance within poultry operations is multifaceted. More than 2500 serovars exist within the subspecies enterica with about 10% of those being found in poultry production [3]. Groups are designated based on the somatic (O) and flagellar (H) antigens [4].
Adult turkeys are often asymptomatic with Salmonella infections but can transmit the bacteria through eggs to poults [5]. Vaccination is available and often implemented in breeder turkey flocks as a means to control certain serovars that pose a higher risk to human health. A 2014 study utilizing data from another commercial production company in the US showed that 90% of turkey farms that sampled positive for Salmonella via drag/bootie swabs tested positive for Salmonella at the processing plant [6], emphasizing the need to identify and mediate this pathogen prior to the production of a product that is destined for human consumption.
Incidences of Salmonella in turkey flocks has been estimated at between 16% and 54% [5]. The introduction of Salmonella to the flock can occur from many sources since it is a bacterium transmitted both horizontally and vertically (internal and/or external contamination of eggs) [5]. From the hatchery to the processing plant, Salmonella is a constant threat to poultry production systems. The main modes of transmission include contaminated feed, biologic vectors (birds, rats, mice, mites, insects), contaminated water and mechanical vectors/fomites (trucks, personnel, equipment) [5]. Transmission may occur between farms and between houses on a farm when on-farm biosecurity (use of personal protective equipment including bouffant, clean coveralls, clean boots or boot cover and gloves, disinfection of vehicle prior crossing established clean-dirty line of property, cleaning and disinfection of equipment used within multiple houses during visit, use of boot disinfection stations between entering houses, and following of all company protocols for personal visits to farms) is not being followed, as shown by several other studies that analyze the movement of other diseases such as avian influenza through the use of a network model [7,8]. Current literature specifically on the analysis of the movement networks and the dissemination of Salmonella within turkey production is limited. In the present study, a contact network was created by following the movement of turkey flocks from their brooding location to their growout location. Following these pairs would enable the identification of brooder farm locations that were more or less likely to repeatedly test positive for Salmonella at the surveillance sampling time point.
The transportation of live turkeys to the processing plant is the final opportunity to expose the birds to Salmonella once they leave a growout facility [9]. Proper cleaning and disinfection of transport coops is essential to ensure the minimal transmission of bacteria. The ability of Salmonella to persist in products deemed for human consumption has forced processing plants to consider decontamination treatments as the easiest way to control this pathogen during processing [3].
Multiple studies have shown the value of using contact networks paired with betweenflock movement of personnel to implement improved surveillance and intervention to combat disease [10,11]. However, there is limited information on network analysis [7,[10][11][12][13][14][15] and disease transmission models in poultry systems [7,10]. Furthermore, no information is available in the literature on the transmission of Salmonella among turkey flocks. In this study, we described the distribution of Salmonella among the sites of one turkey producing company located in North Carolina and the spread along the movement of birds from the brooder to the growout farms' network. This work is based on Salmonella cases collected approximately two weeks prior to harvest from growout farms from 2017 to 2020. Additionally, we assessed the distribution by serotype and group.

General Description of the Collected Database and the Distribution of Salmonella spp. Groups and Serotypes
In this study, we considered an operation composed of 50 brooder and 76 growout turkey farms that range in age from 1-30 years and are geographically spread throughout 8 counties in the state of North Carolina. The following tables and figures show results from routine surveillance samples collected from growout farms approximately two weeks prior to harvest. North Carolina as a state is second, behind Minnesota, when it comes to the number of turkeys produced annually. The production system considered in this study has protocols that parallel many others, making it an ideal system to analyze and provide conclusions which apply to other systems in similar geographic areas. Positive samples from bootie swabs (i.e., "shoe cover swabs" as defined by the National Poultry Improvement Plan Program Standards Section 3(a)(1)(iii)) consisted of 30 separate Salmonella serovars encompassing 11 different groups. Figures 1 and 2 show the quantification of collected samples and their proportion within the total number of collected positive samples. The    Supplementary Table S1.    Supplementary Table S1.

Facility Age Influence on Salmonella
Company data on the year each facility was built was used to determine the farm or facility's age. Age ranges included greater than 30 years, 20-30 years and less than 20 years. The likelihood of a farm having a positive Salmonella surveillance sample increased with the age of the farm, with positive and significant correlation rho = 0.29 (p value < 0.005). Farms built 30 years ago or more had a 57% chance of collecting a positive sample; farms built between 20-30 years ago have 54% chance of collecting a positive sample; farms less than 20 years of age have a 29% chance of collecting a positive sample.

Between-Farm Contact Networks
The total number of unique between-farm movements were 358 transported groups, with a maximum out-degree of 28 and maximum in-degree of 14. As expected, the betweenness of this poultry network was zero, which meant there were no intermediate farms from brooder to grower, thus only direct pairs of movements were recorded. Data on the number of nodes, edges and network parameters of connectivity and cohesiveness (GSCC and GWCC) for the entire period (2017-2020) are shown in Table 3. The proportion of the size of GWCC included all farms in the network, and because only direct movements from brooder to growers were recorded, the GSCC was one. In addition, we evaluated the degree distribution of the entire network while also considering the number of contacts of each farm type ( Figure 3). Please see Supplementary Material ( Figure S3) for the description of in and out-degree of each farm type.

The Contact Network for the Movements Involving Infected Movements between Brooder and Growout Farms
The Salmonella positive degree, in-degree and out-degree was not significantly different (p > 0.05) when compared with the entire contact network, in which mean total degree was 9.08, in-degree and out-degree of 4.54, while the whole contact network exhibits a total degree of 12.63, in-degree and out-degree of 6.31. This suggests that the distribution of Salmonella among farms is not compartmentalized between infected and non-infected movement flows, though examining the direct links between brooder and growers reveals that there is substantial mixing of birds at growers coming from brooders that were identified as infected and uninfected ( Figure 4). Additionally, we have also run the K-test over the infected (at least one positive Salmonella result). The direct association between bird movement and Salmonella was not significant based on the k-test (p-value > 0.05), therefore, farms were not significantly more connected than expected by chance (randomly). This suggests that the contact network does not provide evidence that Salmonella is spreading by the movement of infected birds. on the number of nodes, edges and network parameters of connectivity and cohesiveness (GSCC and GWCC) for the entire period (2017-2020) are shown in Table 3. The proportion of the size of GWCC included all farms in the network, and because only direct movements from brooder to growers were recorded, the GSCC was one.
In addition, we evaluated the degree distribution of the entire network while also considering the number of contacts of each farm type (Figure 3). Please see supplementary material ( Figure S3) for the description of in and out-degree of each farm type.

Discussion
The purpose of this study was to identify contributing factors to the spread of Salmonella within an eastern U.S. turkey production system by analyzing the prevalence of positive samples collected during routine surveillance of growout flocks and between-farm bird movements in mediation of Salmonella dissemination. While the human-health impact of Salmonella is important, it was not evaluated in this study, as information on the impact of working in poultry production on each employee was not available. Rather, the focus Pathogens 2021, 10, 1539 7 of 12 was placed on evaluating the incidence and potential movements of this pathogen within this system. In this study, factors that were considered included rearing style (antibiotic-free vs. conventional) over time, impact of brooder location on growout location and age of the farm according to company documentation. Rearing style and seasonality had varying trends and the older a facility was the more likely it was to be positive for Salmonella on surveillance sampling in this system. Group C1 was the serogroup with the highest prevalence of Salmonella, with almost one third of all samples collected falling into this group, including the most frequently isolated serovar: Infantis. We have also demonstrated that the turkey farms in the evaluated production company are weakly connected through the movement of bird's brooders directly into growouts. The Salmonella infected network showed a lower degree when compared with the entire network, additionally the cases of Salmonella did now show a clear association between movement of birds and positive growout sites. Despite the fact that, for Salmonella, the infected networks were similar to the non-infected network, we remark that the use of contact networks could become a useful tool to assess the movement-associated risk of other poultry diseases such as avian influenza and Mycoplasmosis. Infected networks were not more infected than farms in the whole contact network.
A 2010 study analyzing Salmonella serotypes isolated from turkey cuts at different steps of processing at a processing plant identified 83% (101/122) of their isolates were Salmonella Derby, rendering it the overwhelming majority [16] from samples collected. In the present study, we demonstrated that the same serotypes were found in the live production stage, however at much lower rates, i.e., Salmonella Derby accounted for less than 1% (0.29%) of the positive samples collected. This suggests that certain serovars may be more likely to be harbored and introduced to incoming carcasses at the processing level rather than in live production. In comparison to broiler live production and processing, the most commonly isolated group is C3, especially Salmonella Kentucky [17]. The present study considering live production of turkeys did not have any positive samples from group C3. While there are many serovars of Salmonella that affect both species, trends from these surveillance samples indicate that some may be more species-specific than others, including those that fall into group C3 being better adapted to chickens.
We found clear patterns in the occurrence of Salmonella in ABF production with peaks in the second half of each calendar year. In contrast, a previous study, also from North Carolina, determined no association between Salmonella and seasonality [18] while we found trends in the occurrence of Salmonella with peaks from September to November in ABF production as seen in Figures 1 and 2. Analysis of Supplemental Figure S2 shows peaks of positive samples in the months of October and November in years 2018, 2019 and 2020. The entrance of biological vectors into the poultry house is more likely to occur during this time of the year as these animals attempt to evade the cooler temperatures by seeking refuge in the climate-controlled poultry house that also has a readily available supply of food and water. This system determines rearing style (conventional vs. ABF) of each flock at each location based on market demands so this study is unable to assess Salmonella prevalence solely based on the rearing style of each location, as each location grows turkeys under both styles. Figure 2 shows an overall increase in the collected positive samples over time from ABF farms over the course of the four years. This trend was consistent with the goals of this production system, as the national demand for ABF turkey at the level of processing has increased over the four years analyzed. In the years 2018-2020 during the months of September through December, many more positive samples were collected from ABF flocks compared to conventional flocks. Conventionally reared flocks appear to show a decline in negative samples collected over the course of the four years, with peaks of positive samples collected inconsistently dispersed throughout the year with occasional peaks in summer months, especially during the years 2018 and 2019. During the summer, naturally ventilated, curtain-sided facilities will often utilize a fogger system to aid in cooling of the birds. These cooling systems can increase relative humidity in the environment 4-5% [19], facilitating the creation of the ideal environment for bacteria growth [20]. In summating all four years of production, this system has a higher prevalence of Salmonella in ABF production flock compared to conventional production flock of over 10%, indicating that the use of some antimicrobials may aid in the reduction in Salmonella even though therapy may not be targeted specifically at this pathogen.
Since Salmonella can be horizontally transmitted within and between flocks, a flock transmitting the bacteria at the time of transfer would continue to pass it from bird to bird after arrival at the growout facility [5]. The ability of a location to harbor a pathogen, specifically Salmonella, was considered in this study by identifying flocks sampled at multiple growout locations that all stemmed from the same brooder location. Through our analysis we determined that the presence of Salmonella at each growout location is independent of the brooder from which the flock originated within this operation. The brooder farms do not appear to be a source of spread for Salmonella. Similarly, the proximity of one location of a farm to another did not appear to be a risk factor for testing positive for Salmonella. Salmonella was widely distributed across eight counties in which this system grows birds with clusters of farms in the same geographical area not necessarily consistently testing positive for the same serovars confirming that other methods of transmission (such as biologic vectors, fomites, contaminated feed) are a more important source of transmission than geographic proximity with airborne transmission via contaminated dust.
A study in a laying hen system determined that the older the infrastructure of the housing system, the more likely they were to test positive for Salmonella [21]. Similar results were found in this study consisting of all curtain-sided, naturally ventilated houses. This result is likely due to better overall infrastructure of newer houses with cleaner and modern construction materials that are designed with good structural biosecurity in mind. It would be expected that all curtain-sided, naturally ventilated houses have more physical openings compared to solid-sided, tunnel-ventilated houses, and consequently a higher chance for the introduction of pathogens via biologic vectors. Additionally, older houses are more likely to have structural deterioration that allows the entrance of biological vectors of Salmonella including birds and rodents. It can be concluded that farm age is a risk factor for Salmonella presence.
For the North America poultry industry, our results from the between-farm movement suggested that even though there were insignificant associations between movements and Salmonella positive cases, key network centrality metrics such as degree could be considered in order to implement risk-based disease surveillance. The utilization of the ranked farms by total degree could be used to disrupt the contact network between infected and uninfected movement flows in the case of Salmonella and also in the case of a foreign animal disease (FAD) epidemic [7,8,10]. On the other hand, the extremely low level of clustering of the infected network compared to a random network is suggestive that Salmonella might not maintain itself and spread within the poultry network. Therefore, other unmeasured or unknown transmission routes such as proximity of mortality disposal (incineration, on-farm composting) to houses, feed truck time at each facility and movement of company personnel should be further investigated while targeting between-farm Salmonella control.
In an ideal study of a poultry production system in which Salmonella prevalence and risk factors were being assessed, the management method of rearing turkeys (ABF vs. conventional) would be set by location in order to allow for an analysis of each type of style's influence on the likelihood of Salmonella presence. Similarly, our study focused on surveillance data from growout locations only. One study found the bacteria persist in the dust of a depopulated poultry facility up to one year after the birds were removed [22]. Further work could build on the present study by collecting samples at the brooder level in order to determine if there is true maintenance and/or movement of the bacteria from one location to the next when the flock transfers at five weeks of age within this particular system. The lack of data collection from the brooder location is a limitation of this study as we cannot determine the origin of the bacteria in each flock. Moreover, a consideration of sampling the growout houses multiple times, to investigate whether there is "replacement" of Salmonella serotypes over time would be ideal. It would also be beneficial to consider data from breeder farms in an effort to identify any vertical transmission of Salmonella within the system. Future work should also include feed truck movement data along with information on movement of service and maintenance personnel. Litter age and management should be included as well. Salmonella is a complex and ever-present pathogen within poultry production. While many factors exist that may be contributing to the movement of Salmonella within poultry flocks, an analysis of these factors to identify the major risk factors can be accomplished.

General Description of Data and Mathematical Calculations Used to Aid in Data Interpretation
Description of Production System: Poults are placed at day of age on brooder farms typically comprising two houses with a total capacity of 10,500-16,000 poults total. At approximately five weeks of age, birds are transferred from their brooder farm to a growout farm in order to facilitate the expected change in size of each bird through the growing period. Growout farms are typically composed of between 3 and 6 houses with a total capacity of 16,000-21,000 turkeys total, depending on the sex being reared (heavy hens vs. light hens, vs. heavy toms). The growout facility will house the birds until harvest age. At transfer, either an entire flock is moved from a brooder location to a growout location, or two brooder locations are moved to the same growout location to fulfill capacity as explained above. All hen farms are equipped with an incinerator to dispose of dead birds while the tom farms have composting capability. All tom turkeys are raised within the conventional style (reared without the constraints of having to maintain antibiotic-free status) while heavy and light hens are reared either conventional-style or antibiotic-free (ABF) depending on market demands. During the life of the flock, on-farm movements of personnel frequently occur. Feed trucks deliver rations to each farm approximately once each week and perform deliveries to approximately four farms each day. Service personnel assess farms once or twice weekly for growout and brooder age turkeys, respectively. Downtime between flocks on farms varies from 7 to 21 days. Breeder flocks from which hatched poults originate are vaccinated six (toms) to eleven (hens) times prior to going into production with a Salmonella vaccine (Poulvac ® ST, Zoetis, Parsippany-Troy Hills, NJ, USA).
Sample Collection: Routine samples collected via "bootie swabs" are acquired and analyzed approximately two weeks prior to harvest from growout farms. An individual farm consists between three and six houses. For routine collection, all houses at each location are sampled, then samples are pooled together for submission. To perform a "bootie swab" (i.e., "shoe cover swabs" as defined by the National Poultry Improvement Plan Program Standards Section 3(a)(1)(iii)) a plastic shoe bootie is applied to one of both shoes. After walking through the house, the plastic bootie is removed and submitted for analysis in a whirl pak bag.
Sample analysis: Collected samples are submitted to a third-party lab for analysis and serotyping. Traditional benchwork testing is performed using a series of cultures and overnight incubation periods to initially identify positive and negative samples, then differentiate positive samples into serogroups. Lastly, serotyping is performed on positive samples to define the serovar.
Facility age calculation: The total number of times each farm tested positive on surveillance sampling was divided by the total number of samples collected from each farm then multiplied by 100 to calculate the percent chance a farm would test positive. Age was determined using company records and divided into the following categories: old = greater than 30 years, middle = between 20 and 30 years, and new = less than 20 years. In addition, we have evaluated the correlation (Spearman's rank correlation) between tested negative farms and each facility age.
Descriptive analysis: Tables 1 and 2 were created by determining the total number of times a particular group/serovar was represented (total number of positive) in data containing all collected samples. The total number of positives for each group/serovar was then divided by the total number of positives within the entire collection to calculate a proportion for each serovar/group. Figures 1 and 2 show the prevalence, or number of positive samples collected according to historical data provided by the company, of Salmonella over the course of the four years. Figure 1 shows each group represented by a different color and displayed according to the frequency of collection. Figure 2 shows the prevalence of Salmonella over four years according to the production style, antibiotic-free (ABF) and conventional. Quantification of the number of samples collected by each rearing style is shown by the two different colors of the bars within the figure.
Description of nomenclature: Salmonella groups are defined using the Kauffmann-White scheme, which uses serologic identification of the O (somatic) and H (flagellar) antigens to assign groups then further separates the organism into serovars [4].

Between Farm Contact Networks
To analyze the between-farm contacts, we reconstructed a directed static network. To apply a static network analysis, repeated between-farm contacts between two farms throughout the observation period, from August 2016 to December 2020 were aggregated into a single one. Here, directed 'edges' are represented by bird shipments between two 'nodes', where nodes correspond to turkey farms. For the static and the time-series networks, we calculated several centrality indicators for farms within the network: outdegree, in-degree, closeness centrality, betweenness, giant weakly connected component (GWCC) and giant strongly connected component (GSCC) (Table 4) [23]. Table 4. Description of network analysis terminology and metrics.

Parameter
Definition Reference

Nodes
The unit of interest in network analysis. For example, premises or slaughterhouses. [24] Edge The link between two nodes in a network. [24] Degree (k) A number of unique contacts to and from a specific premise. When the directionality is considered, the ingoing and outgoing contacts are defined: out-degree is the number of contacts originating from a specific premise, and in-degree is the number of contacts coming into a specific premise. [24]

Movements
The number of movements as batches are recorded over a certain period of time. [24] Betweenness The extent to which a node lies on a path connecting other pairs of nodes, defined by the number of geodesics (shortest paths) going through a node. [24] Giant weakly connected component (GWCC) The proportion of nodes that are connected in the largest component when directionality of movement is ignored. [24] Giant strongly connected component (GSCC) The proportion of the nodes that are connected in the largest component when directionality of movement is considered. [24] 4.

The Contact Network for the Movements Involving Infected Movements
A transmission network is typically a subset of the complete contact network, as not all contacts lead to disease transmission [25]. Therefore, we considered the infected movements as a subdivision of the full network because we were interested in the disease dynamics and how this information can be directly translated into action in the field. The infected network included 27 farms. The network is represented in Figure 4, where infected and uninfected farms are color-coded, while also presenting each farm type. To evaluate whether there was a significant relation between the observed network and the distribution of Salmonella-positive cases across the network we performed a path-based k-test, described elsewhere [26]. The Salmonella-positive farms were labeled as infected if at least one positive diagnostic was reported. We ran the path-based k-tests [26] with 10,000 iterations. These analyses assessed whether the mean network path length between infected farms was smaller than expected by chance, and therefore the extent to which the social network predicted the occurrence of Salmonella cases. All analysis and visualization were performed using R statistical software [27] version (3.4.2).

Conclusions
In conclusion, this study shows some seasonality of Salmonella surveillance samples from growout farms. There appears to be an increase in positive samples collected during the second half of each year. Rearing style does appear to influence this as the results of collected samples as more positive samples were collected from farms utilizing an ABF production style compared to a conventional production style. The movement of birds from brooder farm to growout farm did not have any effect on the likelihood of testing positive for Salmonella on surveillance bootie swabs prior to harvest. The array of samples collected is widely distributed across the network with no indication that a single brooder location is contributing to the spread of Salmonella to multiple growout locations. Lastly, the factor that seems to have the largest effect on the likelihood of testing positive for Salmonella in this system is farm age, with farms less than 20 years of age more likely to test negative compared to farms 20-30 years of age and farms 30+ years of age. Turkey movement data analysis results presented here will be helpful in informing the identification of farms to be targeted as part of Salmonella control and surveillance activities. Given the characteristics of the network, directing surveillance, awareness, and biosecurity enforcement targeted at sites with top degree may increase the number of positive samples but is less likely to reduce between farm transmissions.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/pathogens10121539/s1, Figure S1: Distribution of Salmonella serotypes overtime, Figure S2: Distribution of Salmonella serotypes over times in antibiotic-free and conventional rearing styles, Figure S3: In and out-degree distribution of brooder and growout farms, Data S1: Supplementary data for figures.

Data Availability Statement:
The data that support the findings of this study are not publicly available and are protected by confidential agreements, therefore, are not available.