To meet the constantly growing demand for poultry meat, high production standards must be met to ensure optimum feed efficiency, weight gain, and subsequent meat yield (Daniel et al., 2011). A healthy and functional digestive tract in commercially raised broiler chickens is important for promoting such conditions (Van Immerseel et al., 2004). Antibiotic growth promoters (AGPs) and anticoccidial drugs have long been used to help balance the precarious intestinal health of broiler chickens and to prevent a disease state under intensive commercial production conditions (Porter, 1998; Daniel et al., 2011). However, the large-scale use of antimicrobials in animal production is no longer considered wise due to concerns about food-producing animals acting as a reservoir for antibiotic resistant pathogenic bacteria transmissible to humans (Perron et al., 2007; Seal et al., 2013). Unfortunately, antimicrobial withdrawal has a major impact on gut health in intensively reared broiler chickens (Timbermont et al., 2011). Indeed, following the European ban on AGPs in 1999, production performances were negatively impacted, which correlated with an increased incidence of digestive problems in poultry flocks, such as necrotic enteritis (NE) (Van Immerseel et al., 2009). These regulatory restrictions forced the poultry industry to look for sustainable alternatives to preserve gut health and to minimize animal loss (Yegani and Korver, 2008). NE is a disease caused by Clostridium perfringens and impacts the economic, welfare, as well as the food safety aspects of the broiler chicken production (Van Immerseel et al., 2004, 2016). Until now, no alternative strategy has been completely cost-effective in replacing AGPs. This situation makes any research efforts aiming at better understanding the role of C. perfringens in necrotic enteritis highly relevant (Huyghebaert et al., 2011; Smith, 2011).

A previous field study was conducted by our group and compared two different rearing protocols for commercial broiler chickens (Gaucher et al., 2015). We compared a conventional protocol where birds were fed a diet containing antimicrobials and a drug-free program where antimicrobials were replaced by a combination of alternative strategies, including essential oil-based products added to poultry feed, acidification of the drinking water, anticoccidial vaccination at the hatchery, and improved brooding conditions. For all farms participating in the drug-free program, nutritional guidelines pertaining to feed composition were provided (Gaucher et al., 2015). We found that the drug-free program was associated with an increased incidence of clinical NE outbreaks (Gaucher et al., 2015). Indeed, of the eight participating farms, recurring NE was observed in the same two farms over the course of our study, even if the drug-free program was uniformly applied to all. This trend has also been previously reported by Smith (2011). Thus, it is important to better characterize the C. perfringens community found in these healthy and NE-affected farms and to test how C. perfringens population richness correlates with their health status (Dumas et al., 2011).

Here we compare the C. perfringens richness and virulence genes carriage initially found in these healthy and NE-affected farms, and track how the bacterial population changed over time according to flock health status.

This study provides new insights into the dynamics of C. perfringens populations in drug-free farms, according to their digestive health clinical profile: healthy vs. diseased (clinical NE-positive). Although many studies have been conducted to better describe the role of C. perfringens in necrotic enteritis pathogenesis, little work has been done to better characterize this microorganism population structure in commercial broiler chicken farms, where the use of antimicrobials has been discontinued (Prescott et al., 2016; Rood et al., 2016). More importantly, much remains to be explained to better understand why raising drug-free birds would be more challenging on some farms in regard to digestive health management and NE issues (Brady et al., 2010; Smith, 2011). Here, we show that farms presenting with a higher genotypic richness of C. perfringens were more likely to develop NE over a 14-month period. Furthermore, we found that following recurring NE outbreaks, diseased farms harbored a decreased genotypic richness, suggesting that one or more genotypes were selectively increased in frequency during the outbreaks.

In healthy and well performing farms, C. perfringens species richness was lower and relatively well assessed during the sampling done at the initial time point of the study, an affirmation supported by the more rapidly plateauing appearance of the rarefaction and completedness curves (Gotelli and Colwell, 2001; Hughes and Hellmann, 2005). Conversely, one unexpected result was the presence of a higher and not completely revealed genotypic richness found at the beginning of the field study in farms that would experience recurring NE outbreaks during the study (Figure 1B). This higher richness could have increased the likelihood of one or many virulent clones being present in the C. perfringens population at the start of the study. Indeed, it has been reported that a mild sub-clinical form of NE can exist even in the presence of antibiotics and more importantly, can be correlated with a larger and a more diverse C. perfringens population (Johansson et al., 2010; Lyhs et al., 2013). Given the absence of antibiotics, a less well-controlled C. perfringens population has likely played a role in the occurrence of NE outbreaks on the diseased farms of our study, as has been reported previously by others (Watkins et al., 1997; Johansson et al., 2004; Martel et al., 2004; Silva et al., 2009; Gharaibeh et al., 2010; Charlebois et al., 2012). Both antimicrobial resistance and virulence genes can contribute to the adaptation process of C. perfringens, and although the presence of antibiotic resistance genes in NE-causing strains was most probably not the cause of the NE outbreaks in the diseased farms of our field study, it may have contributed to it by fostering the colonization of these farms by C. perfringens strains harboring these plasmids (Beceiro et al., 2013). Assessing the antibiotic resistance profile of the recovered C. perfringens strains from these farms at the end of the study could then be interesting, as the genes encoding antibiotic resistance can be found on the same plasmids carrying the toxin genes that showed a significant increase in frequency over time in the diseased farms of our study (Harrison and Brockhurst, 2012; Palliyeguru and Rose, 2014). If that is the case, the recurring nature of NE outbreaks observed on these farms could have contributed to an increased prevalence of strains showing antibiotic resistance at the end of the experiment (Beceiro et al., 2013).

Based on rarefaction curves, the changes in C. perfringens richness over the study period also showed great differences between the farm clinical profiles. Interestingly, in healthy and well-performing farms, rarefaction and completedness curves showed that C. perfringens richness was relatively well assessed by our sampling procedure at the initial and final time points. However, only the final sampling revealed the complete species richness for the NE-affected farms; the initial sampling in these farms showing only a proportion of the genotypic richness (Figure 2D). Rarefaction curves related to the health status of the farms suggested that the implementation of the drug-free program was associated with the fostering and persistence of a beneficial C. perfringens population for the birds, an affirmation supported by the performance results (Table 1; Heuer et al., 2008). Moreover, the conservation of a limited genetic richness among the C. perfringens population living in these healthy farms over time might have heightened the capacity of this bacterial community to respond to further selective pressure. The richness observed could also represent the successful establishment of a favorable and well-adapted C. perfringens population in this novel and challenging environment, generated by the implementation of the drug-free program (Rius and Darling, 2014). The presence of multiple C. perfringens clones in healthy birds and healthy flocks has also been reported by other authors (Engstrom et al., 2003; Nauerby et al., 2003; Barbara et al., 2008; Chalmers et al., 2008). The marked reduction in C. perfringens species richness observed at the end of the study in farms that experienced recurring clinical outbreaks of NE is informative (Figure 2B). Indeed, as previously proposed, birds living in NE-affected farms may have acquired the virulent clone from their environment, potentially contaminated following earlier episodes of subclinical or clinical NE (Palliyeguru and Rose, 2014). Accordingly, given the proper conditions engendered by antimicrobial withdrawal, clinical NE outbreaks occurred in these farms, further increasing the available population of disease-causing strains (Cooper and Songer, 2009). This is also in line with the assumption stating that in a finite population, under specific conditions, a clone that is better adapted and able to use the niche’s resources more appropriately can evolve toward a population in which all organisms are direct descendants of this fitter and favored ancestor (Doolittle, 2012; Maharjan et al., 2012). As a clinical NE outbreak is usually related to the selective multiplication of one or few virulent clones (Sawires and Songer, 2006; Barbara et al., 2008; Timbermont et al., 2009), we propose that, in diseased farms, the C. perfringens population was gradually replaced by the clonal expansion of some virulent strains. A similar expansion of a few virulent clones has bene reported by others (Sawires and Songer, 2006; Barbara et al., 2008; Timbermont et al., 2009). Again, it is relevant to mention that antibiotic therapy with penicillin was used to control severe NE outbreaks and avoid further NE-associated complications in two of the eleven (one treatment on each affected farm) NE-affected flocks considered in this study. Though no C. perfringens strains from these treated flocks were included in the current study, this treatment might also have, to a limited extent, contributed to the observed decrease in C. perfringens genotypic richness by selectively eliminating specific sensitive clones. Our sampling procedure most likely included these virulent clones, as indicated by the loss of species richness over time, as well as the trend toward a more rapidly plateauing curve observed for these diseased farms at the end of the experiment (Figure 2C). However, even though rarefaction curves show a loss of species richness over time for NE-affected farms, patterns obtained using a PFGE typing approach indicate that these virulent clones continue to undergo recombination, confirming the dynamic nature of this bacterial species, as previously described (Lacey et al., 2016; Parent et al., 2016). The number of genotypes identified for C. perfringens strains obtained from the final sampling in our study is also in line with the literature reporting an increasing C. perfringens population richness following a NE outbreak (Nauerby et al., 2003; Barbara et al., 2008; Chalmers et al., 2008).

Some factors might have influenced the ability of these virulent clones to colonize and persist in the NE-affected farms of our study. One explanation could come from the toxinotyping results. It has been reported that C. perfringens strains associated with NE outbreaks would not only possess a chromosomal genetic background conferring them a selective advantage over the commensal strains (Chalmers et al., 2008; Hibberd et al., 2011; Lepp et al., 2013; Lacey et al., 2016), but that these strains would also harbor many plasmid-borne virulence genes organized in pathogenicity loci (Lepp et al., 2010; Lacey et al., 2016). Within these loci, some genes encoding toxins, antibiotic resistance, and adhesion factors might preferentially contribute to the ability of the virulent strains to persist over time and to colonize the birds (Lacey et al., 2016; Prescott et al., 2016; Zhou et al., 2017). Indeed, some genes found in these strains could be associated with a better adaptation to an environment where the disease is raging and where they will find the conditions to undergo positive selection and multiplication (Sawires and Songer, 2006; Prescott et al., 2016). The increase in toxin gene carriage observed in the C. perfringens population found in NE-affected farms of our study could also be partially explained by the chromosomal background of these virulent strains, providing them with a heightened capacity of exchanging and carrying plasmids. Among the relevant virulence genes, netB, which plays a critical role in NE pathogenesis, and cpb2 are located on those plasmids found in the virulent strains (Lepp et al., 2010; Parreira et al., 2012; Prescott et al., 2016). Surprisingly, in healthy farms, the proportion of strains carrying the netB gene at the beginning and at the end of the study was considerable. Though this situation has already been reported (Abildgaard et al., 2010), we would have expected to see this higher prevalence of strains harboring these genes in diseased farms. This result highlights the fact that the sole presence of netB is not sufficient to cause clinical disease in a broiler flock (Zhou et al., 2017). The absence of clinical NE in these healthy farms during the study could explain why we did not see any increase in the number of C. perfringens strains harboring these two genes at the end of the experiment. Interestingly, in diseased farms, the proportion of C. perfringens strains carrying cpb2 and netB genes underwent a significant increase during the field study. Two hypothesis can be proposed to explain this increased frequency in netB and cpb2 harboring strains in the diseased farms. The selective multiplication and amplification over time of dominant clones harboring these genes may have contributed to the observed increase, as well as to the reduced genotypic richness noted for NE-affected farms (Figure 2A). Since cpb2 and netB genes reside on mobile genetic elements, we cannot exclude the possibility of an increased horizontal gene transfer between the different variants of C. perfringens colonizing the gut of NE-affected birds, a well-recognized ability of C. perfringens strains in laboratory conditions (Bannam et al., 2011; Stecher et al., 2012; Lacey et al., 2016). Our results are consistent with those of previous studies reporting the important role of plasmid DNA conjugative transfer in antibiotic resistance and virulence gene transfer in influencing C. perfringens adaptation (Wisniewski et al., 2015). However, conducive conditions and proper chromosomal background seem to be essential to promote an optimal horizontal transfer in these NE-causing strains. Since we did not see this upward trend in the frequency of strains harboring more virulence genes, such virulent strains might not have been present or may not have encountered the necessarily favorable conditions for NE development in healthy farms (Lepp et al., 2010; Lacey et al., 2016). Moreover, even if both cpb-2 and netB toxin genes were present at relatively high frequencies in healthy farms at the start of the experiment, no clinical necrotic enteritis outbreak was observed throughout the duration of the study in these farms. The increase observed in the final proportion of strains harboring virulence genes only in NE-affected farms further demonstrates a significant positive selection of virulent C. perfringens clones in diseased farms. Since all participating farms were standardized and because the drug-free program was applied in a stringent and uniform way between farms, we can conclude that other contributing factors might have been involved (Gholamiandekhordi et al., 2006; Keyburn et al., 2008; Martin and Smyth, 2009; Brady et al., 2010). The different C. perfringens populations living in these farms probably had a role to play, not only in the occurrence of NE at the beginning of the field study, but also in the recurrent nature of these outbreaks (Barbara et al., 2008; Cooper and Songer, 2009; Lepp et al., 2010; Shojadoost et al., 2012).

Our findings show that commercial broiler chicken farms participating in a drug-free program can show significantly different profiles according to clinical NE. These findings can be explained in multiple ways. Among these, comparative genomics realized between virulent and commensal strains of C. perfringens recovered from the participating farms could help explain why clinical necrotic enteritis was not observed in healthy farms, despite the presence of strains harboring toxin genes (Lepp et al., 2010, 2013; Parreira et al., 2012). The presence of a specific and beneficial intestinal microflora in birds could also have contributed to this healthy profile and would deserve a more comprehensive assessment (Torok et al., 2011; Singh et al., 2014; Stanley et al., 2014). Ultimately, a resampling of the C. perfringens populations living in healthy and diseased participating farms after the reintroduction of antimicrobials could help us to better assess the evolution of these populations under different selective environments.

Conception or design of the work; or the acquisition, analysis, or interpretation of data for the work: M-LG, JA, GP, SQ, AL, and MB. Drafting the work or revising it critically for important intellectual content: M-LG, GP, JA, SQ, MB, and AL. Final approval of the version to be published: SQ, JA, GP, AL, MB, and M-LG. Agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved: M-LG, JA, GP, SQ, AL, and MB.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.