The use of embryonic chicken eggs as an alternative model to evaluate the virulence of Salmonella enterica serovar Gallinarum

Salmonella enterica serovar Gallinarum (S. Gallinarum) can cause fowl typhoid, a severe systemic disease responsible for considerable economic losses. Chicken pathogenicity test is the traditional method for assessing the virulence of S. Gallinarum. However, this method is limited by several factors, including ethical considerations, costs, and the need for specialized facilities. Hence, we established a chicken embryo lethality assay (ELA) model to determine the virulence of S. Gallinarum. Three virulent and three avirulent representative strains, which were confirmed by the chicken pathogenicity test, were used to perform the ELA. The most significant difference between the virulent and avirulent strains could be observed when 13-day-old embryos were inoculated via the AC route and incubated for 5 days. Based on a 50% embryo lethal dose (ELD50), isolates considered to be virulent had a Log10ELD50 of ≤ 4.0, moderately virulent strains had a Log10ELD50 of 4.0−6.1, and avirulent isolates had a Log10ELD50 of ≥ 6.1. Different abilities to invade the liver of embryos were found between the virulent and avirulent strains by a growth curve experiment in vitro. The maximum colony-forming units (CFU) of the virulent strain was about 10,000 times higher than that of the avirulent strain in the liver at 5 days post infection. The ELA results of 42 field strains showed that thirty-two strains (76.2%) were virulent, nine were moderately virulent (21.4%), and one strain was avirulent (2.4%). In conclusion, these results suggest that the ELA can be used as an alternative method to assess the virulence of S. Gallinarum, which will contribute to the study of virulence genes, virulence evolution, pathogenic mechanisms and vaccine development.


Introduction
Salmonella enterica serovar Gallinarum (S. Gallinarum) is the causative agent of fowl typhoid (FT), a serious systemic disease that causes huge economic losses to the commercial poultry industry [1]. This disease occurs at all ages in chickens and is characterized by severe anorexia,

Bacterial strains and growth conditions
Strains of S. Gallinarum 287/91 (NCTC 13346), A17-DW-005, and A18-GCVP-014 (field strains used in the present study) were selected as representative virulent strains and the vaccine Nobilis SG9R (Intervet International, Boxmeer, The Netherlands), A17-DW-005ΔspiC (spiC is a virulence factor encoded within the Salmonella pathogenicity island 2) and A17-DW-005ΔwaaJΔspiC were selected as representative avirulent strains [24,25]. The above mutant strains were constructed in our lab. Forty-two field strains were isolated from clinical samples of chickens (layer, broiler, and Korean native chicken) at necropsy from 2016 to 2019 in South Korea. All strains were stored in Luria-Bertani (LB) medium containing 20% glycerol at −70˚C, and then cultured on LB agar at 37˚C for 24 h. Before each experiment, a single colony was picked and inoculated in LB broth and incubated overnight at 37˚C with a shaking speed of 180 rpm, and the culture broth was then used for the assay.

Effect of chicken embryo inoculation age and route on virulence
Specific pathogen-free (SPF) chicken embryos (Hy-Vac Laboratory, Redfield, Iowa, USA) were incubated until use in a constant temperature incubator at 37˚C with 40 to 60% humidity. To determine the optimal inoculation age, 6, 10, 13, and 16-day-old embryos were selected. To determine the optimal inoculation route, allantoic cavity (AC), chorioallantoic membrane (CAM) and yolk sac (YS) were selected. Six-day-old embryos were inoculated only via the YS. Ten-day-old, 13-day-old, and 16-day-old embryos were inoculated via the AC and CAM. Three virulent strains 287/91, A17-DW-005, A18-GCVP-014 and one avirulent strain SG9R were used in this experiment. Bacteria were grown to a density of 1.1 by OD 590 (approximately 1×10 9 CFU/ml) and different doses were obtained by serial 10-fold dilution. Fifteen chicken embryos were used per dilution. Vitality was assessed every 12 h by candling for 5 days and deaths were recorded. Embryos that died within 24 h of inoculation were assumed to have suffered lethal trauma during the inoculation and were removed from the experiment. Different doses were used to determine the 50% embryo lethal dose (ELD 50 ) of each strain [26]. Dead embryos and embryos that survived the experiment were chilled for at least 4 h at 4˚C followed by necropsy. The following scoring systems were used to assess the gross lesions of the embryonic body and liver. Embryonic body: 1 for normal, 2 for medium body size, 3 for small body size and 4 for small body size plus hemorrhage. Embryonic liver: 1 for normal, 2 for swelling, 3 for few necrotic foci, 4 for multiple necrotic foci. Bacterial re-isolation was performed from the AC fluid and the embryonic liver.

Chicken pathogenicity test
The eleven selected S. Gallinarum strains were used to challenge chickens to confirm their virulence. Ten Salmonella-free 4-day-old Hy-Line brown layers in each group were used to evaluate the virulence of the S. Gallinarum strains in vivo. In order to guarantee the best environmental conditions, the isolator conditions (temperature, humidity, ventilation) were constantly monitored. The whole staff taking care of or handling the laboratory animals was well trained. Each of the S. Gallinarum strains was orally inoculated into the chickens at a dose of 10 8 colony-forming units (CFU) [27]. After inoculation, the chickens were checked twice a day for 14 days. The clinical signs were scored as 0 for being normal, 1 for being depression and ruffled feathers, 2 for depression, ruffled feathers, respiratory distress and 3 for the above mentioned clinical signs plus anorexia, emaciation and green-yellowish diarrhea, and 4 for death. When birds showed the clinical score of 3 (humane endpoint), the chickens were humanely sacrificed by cervical dislocation performed by trained veterinarians immediately [28]. Despite our efforts, some of the chickens used in the present study died before euthanasia (natural death by fowl typhoid). All dead animals including natural death and euthanasia were necropsied in order to evaluate the presence of S. Gallinarum. Surviving chickens were euthanized at 14 days post-inoculation (dpi), and bacterial re-isolation was conducted.

In ovo growth curve
To compare the growth curves of virulent strains and avirulent strains in ovo, embryos were inoculated with three virulent strains 287/91, A17-DW-005, A18-MRA-014, and one avirulent strain SG9R. Twenty 13-day-old eggs inoculated with 10 4 CFU per strain via the AC were sacrificed at 0, 3, 6, 12, 18, and 24 h and daily thereafter for 4 days (a total of 5 days). Three eggs were used for each time point. A group inoculated with sterile PBS was also assessed. The AC fluids and livers were collected aseptically. The viable bacterial counts were estimated by plating dilutions of the AC fluid and homogenized liver onto MacConkey agar plates.

Virulence assessment of the field strains by ELA
To evaluate the virulence of field strains, 42 field strains that originated from clinical cases were tested. ELD 50 values were determined by inoculating different dilutions of the strains into the AC of 13-day-old chicken embryos according to the method described above. Fifteen eggs were inoculated per dilution.

Statistical analyses
Statistical analysis was performed using SPSS version 21.0 (SPSS Inc., Chicago, IL, USA). ELD 50 values and 95% confidence intervals (CIs) were calculated by using probit analysis of Bliss [29]. The one-way analysis of variance (ANOVA) was used for the analysis of significant differences between ELD 50 values of virulent and avirulent strains using different age/route combinations. One-way ANOVA was also used for the analysis of significant differences in the CFU of different strains at the indicated time points from the AC fluid and embryonic liver, respectively. Differences were considered statistically significant at: � P <0.05, �� P <0.01, ��� P <0.001.

Effect of chicken embryo inoculation age and route on virulence
To determine the optimal inoculation age and route, different embryo ages and inoculation routes were selected. As shown in Table 1, in terms of the age of the chicken embryo, ELD 50 values obtained from 13-day-old chicken embryos were higher than those obtained from 10-day-old chicken embryos. Similarly, ELD 50 values obtained from 16-day-old chicken embryos were higher than those obtained from 13-day-old chicken embryos.
Regarding the inoculation route, all 6-day-old chicken embryos died when inoculated by YS, which indicated that virulent and avirulent strains could not be distinguished through the YS route. Based on the results of statistical analysis, 13 day/AC (P = 0.000001), 13 day/CAM (P = 0.001371), 16 day/AC (P = 0.000004), and 16 day/CAM (P = 0.006511) (S2 Table) could be used to distinguish between virulent and avirulent strains. When the 13-day-old chicken embryos were inoculated by the AC, the most significant difference between the ELD 50 values of the virulent and avirulent strains was manifested. Therefore, 13 day/AC was the optimal inoculation combination through which the virulent and avirulent strains could be distinguished to the greatest extent.

Determination of virulence of representative strains by ELA
As shown in Table 2, the Log 10 ELD 50 values of the virulent strains were 4.0 or less (4.0 was the highest value within the 95% CIs among the virulent strains), whereas the Log 10 ELD 50 values of the avirulent strains were 6.1 or greater (6.1 was the lowest value within the 95% CIs among the avirulent strains). Therefore, the Log 10 ELD 50 value range of the moderately virulent strains could be defined as 4.0−6.1. None of the embryos inoculated with PBS died during the 5 days.
Based on the scoring system shown in S1A Fig, gross lesions of the embryonic body were noted and recorded. There were no statistical differences among the three strains (A17-DW-005, A18-MRA-014, and SG9R) when 10 6 CFU was inoculated into chicken embryos by the AC (P > 0.05). When 10 4 CFU was inoculated, there was no statistical difference between A17-DW-005 and A18-MRA-014 (P > 0.05). However, when 10 2 CFU was inoculated, there was a significant statistical difference among the three strains (P < 0.05) (S1B Fig).
Based on the scoring system of S2A Fig, the gross lesions of the embryonic liver were scored. There were no statistical differences among the three strains (A17-DW-005, A18-MRA-014, and SG9R) when 10 6 and 10 4 CFU were inoculated, respectively (P > 0.05). However, when 10 2 CFU was inoculated, there was a significant statistical difference among the three strains (P < 0.05) (S2B Fig). Hence, when 10 2 CFU was inoculated into chicken

PLOS ONE
Virulence of S. Gallinarum in embryonic chicken eggs

PLOS ONE
Virulence of S. Gallinarum in embryonic chicken eggs embryos by the AC, virulent, moderately virulent and avirulent strains could also be distinguished based on the scoring results of the gross lesions of the embryonic body and embryonic liver.

Correlation between growth in ovo and virulence
In ovo growth curves were determined in both the AC fluid and the embryonic livers to investigate the correlation between replication (or invasion capability) of S. Gallinarum and strain virulence (Fig 1). By comparing the bacterial loads in the AC fluid, it could be observed within the first 24 hours that the four strains multiplied at almost the same rate and reached about 10 8 CFU/ml. The bacterial counts of the virulent and moderately virulent strains were maintained

PLOS ONE
Virulence of S. Gallinarum in embryonic chicken eggs at 10 7 to 10 8 CFU/ml at 2−5 dpi. However, the bacterial count of the avirulent strain decreased slightly at 2 dpi, and was maintained at 10 6 to 10 7 CFU/ml up to 5 dpi. By comparing the bacterial counts in the livers, the numbers of virulent, moderately virulent and avirulent strains were significantly different at 4 and 5 dpi (P < 0.001). At 5 dpi, the bacterial numbers of the virulent and moderately virulent strains in the liver were about 10,000 and 100 times higher than that of the avirulent strain, respectively.

Virulence evaluation of field strains by ELA
Forty-two field strains (shown in Table 3) isolated from clinical cases, including the eight S. Gallinarum strains challenged in chickens, were inoculated into 13-day-old embryos via the AC to confirm their virulence (Fig 2). The ELA results of the 42 field strains showed that thirty-two strains (76.2%) were virulent, nine were moderately virulent (21.4%), and one strain was avirulent (2.4%).

Discussion
The chicken pathogenicity test is the traditional method used for assessing the virulence of S. Gallinarum. However, the use of chicken models is limited by several factors, including ethical considerations, costs, and the requirements for specialized facilities. For decades, embryonic eggs have been used as a convenient alternative infection model for studying viruses and evaluating bacterial virulence [9]. However, to our knowledge, embryos have never been used as a model to study the virulence of S. Gallinarum. Studies have shown that for S. Pullorum and S. Typhimurium, the higher the virulence of the strain in chickens, the higher the mortality rate of the strain in chicken embryos, suggesting that ELA could be used to assess the virulence of S. Pullorum and S. Typhimurium strains. Hence, in the present study, we analyzed whether chicken embryos could be used as a model for studying the virulence of S. Gallinarum. According to the published literature, both chicken embryo age and inoculation route have a marked effect on the lethality of chicken embryos. The susceptibility of different chicken embryo ages (8 day, 13 day, 16 day) to virulent, partially virulent and avirulent strains of E. coli has been studied, which showed that 16-day-old chicken embryos were the most resistant [11]. The effect of the embryos age on the susceptibility to lethal infection with R. anatipestifer was also reported: 7 day > 10 day > 13 day > 15 day [18]. In addition, similar experimental results have been reported for fungal chicken embryo infection tests [8,30].
Consistent with these studies, our results also showed that with increasing age, the embryo became more resistant (Table 1). Except for the inoculation route of YS, as the embryo age increased, the sensitivity of the embryo decreased, and the strain ELD 50 increased accordingly. It has been suggested that the increased resistance of older embryos to pathogen infection may reflect the maturation of the embryonic immune system. The enhanced immunity can reduce the burden of bacteria on the host and prevent the spread of pathogens, thereby improving the survival rate of the host.
For the embryo inoculation route, we chose the three most commonly used routes, AC, CAM, and YS. First of all, in terms of the YS route, our results showed that 6-day-old embryos were so susceptible that both virulent and avirulent strains caused 100% mortality. Similar results have been reported for E. coli and Enterococcus faecalis (E. faecalis). The YS route could not distinguish the virulence of E. coli isolates because both virulent and non-virulent strains had 100% mortality [12]. The YS inoculation route produced high embryo mortality, which made it difficult to distinguish the virulence of E. faecalis strains and to estimate the LD 50 [31]. Thus, inoculation via the YS is not appropriate for use in the ELA to estimate the virulence of S. Gallinarum strains.
Regarding the AC and CAM inoculation routes, our results showed that, compared with the CAM route, the AC route was better for distinguishing virulent and avirulent strains. The most significant difference (P < 0.000001) (S2 Table) was demonstrated in 13-day-old embryos inoculated via the AC route, although differences could also be seen among other age/route combinations (Table 1) [18]. In conclusion, our experimental results emphasized the importance of the age and route of inoculation for the determination of the virulence of S. Gallinarum in the ELA.
A comparison of the growth curves of representative virulent, moderately virulent and avirulent strains in the AC showed that the difference in virulence was not related to the proliferation ability of bacteria in the AC. Because the AC fluid itself supports excellent bacterial growth, the proliferation rates of the strains were almost the same in the first 24 hours, with all of the bacterial counts reaching about 10 8 CFU/ml (Fig 1). Although the number of bacteria in the avirulent strains had decreased over the next few days, it still maintained a bacterial concentration of about 10 6 CFU/ml. However, by comparing the colonization levels of bacteria in the liver, it could be seen that the higher the virulence of the strain, the stronger its invasiveness to the liver. By comparing the bacterial counts in the livers, it could be seen that the higher the virulence of the strain, the higher the level of bacterial colonization. On 4 and 5 dpi, the bacterial load of the virulent strain in the liver was 10 3 −10 4 times higher than that of the avirulent strain. Also, we observed that about 80−90% of embryos inoculated with virulent strains 287/91 and A17-DW-005 died on 4 and 5 dpi, but embryos inoculated with avirulent strain SG9R did not. Therefore, we speculate that virulent strains have a strong ability to invade the liver, leading to the proliferation of a large number of bacteria, which may be the main reason for embryo death.
Based on the optimal age/route combination of 13 day/AC, we determined the ELD 50 values and 95% CIs of multiple virulent and avirulent representative strains. By comparing the 95% CIs, we have defined the criteria for distinguishing the virulence of S. Gallinarum strains in ELA. Regarding the criteria for virulence classification, the selection of
Secondly, how to classify the virulence is the second issue to be considered. In the present study, we clearly observed a correlation between the inoculation dose and embryo mortality. Although no reports on ELD 50 have been found in the Salmonella-related literature, ELD 50 has been used to distinguish virulence in other bacteria. It was reported that there was at least a 6-log difference between the least virulent (ELD 50 , > 3.3×10 8 ) and most virulent (ELD 50 , 2.2×10 2 ) Campylobacter jejuni strains [17]. In addition, for Neisseria meningitidis, it was reported that low virulence strains had an ELD 50 of 10 3 CFU or greater, and high virulence strains had an ELD 50 of approximately 10 1 or less [27]. In our results, there was up to 3.5-log difference between the ELD 50 values of the virulent and avirulent strains, which indicated that the virulence of S. Gallinarum strains could be accurately distinguished by ELD 50 (Table 1). In addition, as shown in Table 2, there was a clear correspondence between the chicken pathogenicity test and ELA. The virulence of S. Gallinarum on chickens was reflected in the ELA. The results of these studies indicated that ELA could be used as an alternative method to evaluate the virulence of S. Gallinarum.
According to the criteria defined in the present study, in the 42 field strains of S. Gallinarum, most of the strains (76.2%) isolated were virulent, a small part (21.4%) were moderately virulent, and only one (2.4%) was avirulent. In the present study, most of the strains were virulent or moderately virulent. This may be because the majority of the strains were isolated from sick and dead chickens. In addition, the successful isolation of an avirulent strain indicated that ELA could be used to screen for natural attenuated vaccine candidates. These results verified that the ELA can be applied to evaluate the virulence of field strains of S. Gallinarum.
Supporting information S1 Table. The