**1. Introduction**

Classical swine fever virus (CSFV), a member of the genus *Pestivirus* within the family *Flaviviridae*, comprises a single positive-stranded RNA genome of approximately 12.3 kb, which encodes a polyprotein of 3898 amino acids [1]. The viral genome comprises a 5' untranslated region (5' UTR), an N-terminal protease (Npro), a capsid (C) protein, envelope (E) proteins (Erns, E1, E2, p7), non-structural

(NS) proteins (NS2, NS3, NS4A, NS4B, NS5A, NS5B), and a 3' UTR [2]. CSFV is categorized into three genotypes, 1, 2, and 3, each of which can be subdivided into three subgenotypes (1.1–1.3, 2.1–2.3, and 3.1–3.4) [3]. Classical swine fever (CSF) is a highly contagious multisystemic hemorrhagic viral disease of domestic pigs and wild boar, which can manifest as acute, subacute, chronic, or late onset disease [4]. Vaccination is used to prevent and/or reduce the number of outbreaks of CSF and, together with other control measures, was an important factor in eradicating CSF from Holland in 1985 and from the Rivas region of Nicaragua [5]. Historically, CSF vaccines have been based on attenuated strains, i.e., lapinized Chinese and tissue culture-adapted strains. Modified live vaccines (MLVs) based on several attenuated virus strains (e.g., C-strain, Thiverval, PAV-250, GPE-, and K-strains) are used most widely. The advantages and disadvantages of MLV vaccination during an outbreak have been described in previous studies [5]. Briefly, the advantages include ease of use, low cost, and an induction of life-long immunity by a single dose. However, in the 1980s and early 1990s (the time when a lapinized Chinese strain was used in Mexico), vaccinated pigs occasionally exhibited adverse reactions; some even died [5]. The CSF vaccine virus multiplies in actively replicating cells such as fetal cells and reticuloendothelial cells; some reports show that animals vaccinated with a Chinese strain virus had more respiratory infections due to a concurrent *Pasteurella multocida* infection [6]. In addition, the CSF vaccine virus may induce embryonic death and myoclonia congenita when administered to pregnant sows [5]. When herds that include pregnant sows are vaccinated for the first time, there can be a fall in sow fertility in the 6 weeks following vaccination, after which it returns to normal; therefore, researchers concluded that vaccination reduces herd productivity [5]. The OIE requirements for the safety of MLVs in young animals states that vaccination should not induce a high body temperature or leukopenia and should not allow horizontal transmission. For pregnant sows, safe MLV vaccination requires no transplacental transmission and no evidence of reversion-to-virulence after passage in piglets. The CSF MLV (LOM vaccine strain) used to vaccinate pigs in South Korea since 1974 has caused abortion in some pregnant sows [7], and prolonged virus shedding by immunosuppressed pigs after vaccination [8].

The South Korean government maintained a CSF (LOM) vaccine policy to control the disease on the mainland; however, vaccination on Jeju Island (located at the southern end of Korea) was stopped in 2000 (Jeju Island declared itself a "CSF-free region"). However, pig farms of Jeju Island have suffered continuous outbreaks of the LOM vaccine strain via various routes from the mainland, raising suspicion about the safety and reversion-to-pathogenicity of commercial LOM vaccine strains. Five outbreaks of the LOM vaccine strain infection of pigs have occurred over 19 years. The first (from 2004 to 2007) occurred through feed contaminated with the commercial LOM vaccine strain present in an animal plasma protein supplement [9]. The second was caused by the delivery of incorrect vaccine material from the mainland; in 2010, one farm accidently inoculated pigs with a vaccine mixed with the LOM vaccine strain. The third and fourth occurred on one farm (2012) and two farms (2013); the route of LOM vaccine strain exposure was not revealed, but feed contaminated with the LOM vaccine strain, or an unintentional injection of an incorrect CSF (LOM) vaccine from the mainland was suspected. In the fifth case (in 2014), pregnant sows on 20 pig farms were inoculated with a commercial swine erysipelas vaccine mixed with the LOM vaccine strain. Vaccination was stopped immediately, but a total of 111 pig farms (20 in 2014, 22 in 2015, 32 in 2016, 26 in 2017, and 11 in 2018) were exposed to the Jeju LOM strains (LOM strains isolated from Jeju pigs). Farm-to-farm transmission patterns showed mainly between high-density pig farms in the Hanlim region (located in northwestern Jeju) from 2015 to 2018 and the Daejeong region (located in southwestern Jeju) from 2016 to 2018.

Here, we investigated pig-to-pig and/or farm-to-farm transmissions, and the possibility of reversion-to-pathogenicity by comparing genetic mutations in the Jeju LOM strain viruses.

#### **2. Results**

## *2.1. Histopathological Analysis to Detect Infection by CSFV (Jeju LOM) Alone or Co-Infection with CSFV (Jeju LOM) and Other Pathogens*

Overall, 122 samples (from 100 piglets and 22 fetuses) obtained from 25 pig farms were tested; all were positive for CSFV (Jeju LOM). The Jeju LOM strain was identified in 103 of 122 samples (81 piglets and 22 fetuses). Antigens derived from the Jeju LOM strain were detected in tissue samples from 51 suckling piglets, 14 piglets were infected with Jeju LOM alone, and the remaining 37 were co-infected with Jeju LOM plus enteric pathogens (six with *Clostridium* spp, six with *E. coli*, three with PEDV, and four with rotavirus), respiratory pathogens (three with PRRSV), or *Streptococcus* (n = 7), and *Staphylococcus* spp. (n = 2) (Table 1). Samples from the 37 suckling piglets co-infected with the Jeju LOM strain and other pathogens showed evidence of interior visceral hemorrhage (18 of the kidney, 12 of the exo-endocardium, and nine of the lung), and 14 had non-purulent brain lesions (perivascular cuffing, gliosis, and neuronophagia). Fourteen piglets infected with the Jeju LOM strain only showed interior visceral hemorrhage (eight of the kidney, six of the exo-endocardium, and three of the lung), and six had non-purulent brain lesions (perivascular cuffing, gliosis, and neuronophagia). Seven of the 14 suckling piglets identified to have CSF-like specific histopathologic lesions (Figure 1). Pathogenic lesions in weaning pigs included broncho-interstitial pneumonia or fibrinous lobor pneumonia (n = 16), lung hemorrhage (n = 10), kidney hemorrhage (n = 9), peripheral lymph node hemorrhage (n = 11), exo-endocardium hemorrhage (n = 5), and non-suppurative encephalitis of brain and spinal cord (n = 5) (Table 1). Co-infection of 22 fetuses with other pathogens (i.e., PPV, ADV, EMCV, JEV, PRRSV, and PCV2) was not confirmed, and no specific pathogenic lesions were observed in their organs. Organ tissue immunohistochemistry (IHC) staining detected Jeju LOM strains in 25 of 48 suckling pigs; 17 were cases of co-infection and eight were single infections. The following organs harbored the Jeju LOM strain: tonsil (40%), spleen (22.9%), lymph node (15%), lung (14.6%), small intestine (4.3%), kidney (4.2%), and liver (2.1%) (Table 2). IHC staining of pathogenic tissues from weaning pigs detected the Jeju LOM strain in internal organs (28.6%; 6/21): lymph nodes (20%), spleen (16.7%), lung (15.0%), and tonsil (5%) (Table 2). However, no virus was detected in other internal organs (heart, liver, intestine, spinal, and brain). One of the 22 fetuses showed an infection of the kidney alone (Table 2).

**Figure 1.** Immunohistochemistry (IHC) staining to detect histopathogenic lesions. Severe multifocal hemorrhages in the renal cortex of a suckling piglet (hematoxylin & eosin (HE); mag ×100) (**A**). Perivascular cuffing in the white matter of the cerebellum of a suckling piglet (HE; mag ×400) (**B**). Brown-stained viral antigens in the cryptal epithelium of the tonsil (IHC; mag ×400) (**C**). Brown-stained viral antigens in macrophages infiltrating the spleen (IHC; mag ×400) (**D**).



*Pathogens* **2019** , *8*, 251

Suckling piglets

Weaning pigs Aborted fetuses

Infection (only Jeju LOM)

Subtotal

 Co-infection

 Infection (only Jeju LOM)

 3/11

 14/35

 1/20

 NT

> GI:

Gastrointestinal,

 CNS: Central Nervous System.

 NT

 0/8

 0/8

 0/8

 1/8

 0/8

 NT

 0/8

 1/8

 4/20

 3/18

 3/20

 0/20

 0/20

 0/19

 0/21

 0/15

 6/21

 6/40

 11/48

 6/41

 0/45

 2/48

 1/47

 2/46

 0/38 25/48

 3/10

 3/14

 2/12

 0/13

 0/14

 0/13

 0/13

 0/12

 8/14

#### *2.2. Detection of Antibodies on Pig Farms Exposed to the Jeju LOM Strain*

The average anti-CSF (Jeju LOM) antibody-positive rates on seven pig farms in the Jeju region was as follows: 87.1% ± 4.2% of sows, 77.8% ± 8.6% of suckling piglets, 24.2% ± 14.1% of weaning pigs, 21.4% ± 11.2% of growing pigs, and 38.5% ± 13.7% of finishing pigs (Table 3). There were significant differences (*p* < 0.05) in the average anti-CSF (Jeju LOM) antibody-positive rates between sows and weaning pigs/growing pigs, and between suckling piglets and growing pigs, respectively. Serum neutralizing antibody (log2) titers in pigs on antibody-positive farms were 9.42 ± 0.25 (log2) in sows, 7.04 ± 0.54 (log2) in suckling pigs, 7.2 ± 1.04 (log2) in weaning pigs, 5.52 ± 0.12 (log2) in growing pigs, and 7.44 ± 0.74 (log2) in finishing pigs (Table 3). There were significant differences in the average anti-CSF (Jeju LOM) neutralizing antibody titers between sows and growing pigs (*p* < 0.05).

**Table 3.** CSF (Jeju LOM) seropositive rates and anti-CSFV (Jeju LOM) antibody titers in pigs of seven farms on Jeju Island.


#### *2.3. Farm-Slaughterhouse-Farm Transmission*

Of the 242 samples collected from vehicles at a slaughterhouse in Jeju, 151 (62.4%) were positive for Jeju LOM antigens by qRT-PCR. The detailed results regarding infected sites were as follows: 66.2% (47/71) driver foot floor, 38% (27/71) vehicle wheels, 71.6% (53/74) pig-holding compartment, and 92.3% (24/26) "other". Among the virus-positive samples, 92 positive samples had an estimated TCID50/mL of 103.0–3.9, and 11 samples had an estimated TCID50/mL of 104.0–4.9. One sample had a TCID50/mL > 105.0 (Table 4). A peroxidase-link immunosorbent assay (PLA) virus viability test revealed strong staining of 13 samples (4 driver foot floor, 2 vehicle wheels, and 7 pig-holding compartment) and weak staining of 32 samples; 129 samples were unconfirmed due to cellular contamination by bacteria (Table 4).

**Table 4.** Contamination by the Jeju LOM strain via exposure during transport to or at a slaughterhouse.


\* TCID50: Tissue culture infective dose 50. \*\* No test: samples were inoculated PK-15 cells but IHC was not performed due to contamination by bacteria.

#### *2.4. Pig-to-Pig Transmission and Reproduction Rate (R)*

Six pigs were inoculated with the Jeju LOM strain JJ16LOM-YJK08 and, after 24 h, were placed in a pen with six non-inoculated pigs. For Group 1, we used pigs (unhealthy) with PRRSV or PCV2, and Group 2 used pigs (healthy) without specific wasting diseases. In Group 1, two non-inoculated pigs were positive for anti-CSF (Jeju LOM) antibodies on Day 21 and Day 28. However, no non-inoculated pigs in Group 2 had detectable anti-CSF (Jeju LOM) antibodies on Day 45 (Table 5). The transmission possibility estimate for group 1 was R0 = 1.22 (95% confidence interval (CI), 0.980–1.765), whereas that for Group 2 was R0 = 0.00 (95% CI, not applicable) (Table 5).


**Table 5.** Estimated transmission probability between non-inoculated pigs and pigs inoculated with the Jeju LOM strain.

R0 calculated as <sup>−</sup>In((1 <sup>−</sup> AR)/S0)/(AR <sup>−</sup> (1 <sup>−</sup> S0)), CI calculated as AR <sup>±</sup> 1.96 ARx(1 − AR)/n). <sup>a</sup> DPI, days post-infection. <sup>b</sup> R0: reproduction number; <sup>c</sup> CI: confidence interval; <sup>d</sup> NA: not applicable.

#### *2.5. Comparison of LOM Strain Genome Sequences*

The amino acid sequences of four commercial LOM vaccine strains were compared with those of five Jeju LOM strains (2004–2007), and three unique amino acid changes in the E1 (V-577-A/M) and NS4B (M-2378L and V-2383A) proteins were detected (Supplemental Table S1). Comparison of the Jeju LOM strain JJ04LOM-Tamra01 (2004) with the four other Jeju LOM strains (2005–2007) revealed six amino acid changes: Erns (D-386-N, R-480-G), E2 (L-1065-S), NS3 (K-1165-R), NS4B (A-2352-V), and NS5A (N-2816-T) (Supplemental Table S1). Comparison of the four commercial LOM vaccine strains with 12 Jeju LOM strains (2014–2018) identified unique changes in the Npro (K-57-R, L-143-Q), Erns (Y-351-H, R-476-S), E1 (I-651-T), NS3 (V-1381-I, H-1584-N, K-2006-I), NS4B (M-2348-I, T-2371-I, I-2398-M, and V-2483-A), NS5A (A-2978-T), and NS5B (N-3409-S and S-3786-N) proteins (Supplemental Table S2).

#### *2.6. Root-to-Top Divergence and Positive Selection Analyses*

Application of the heuristic residual mean squared method to all strains (the commercial LOM vaccine and Jeju LOM strains) using the TempEst program revealed a slope of 3.93 <sup>×</sup> 10−<sup>5</sup> (rate), an X-intercept (TMRCA) of 1797.24, a correlation coefficient of 0.1272, an R squared value of 1.6173 <sup>×</sup> 10−2, and a residual mean squared value of 6.9076 <sup>×</sup> 10−6. Root-to-top divergence for Jeju LOM strain JJ07LOM-JSG02 was >0.0130. Mainly, the Jeju LOM strains isolated from the field showed a high divergence value of between 0.0070 and 0.0130 (Figure 2). However, seven LOM vaccine strains (excluding 88LOM-Suri) and two Jeju LOM strains (JJ14LOM-WSH01 and JJ17LOM-LHH10) showed a low divergence value (0.0040–0.0060) (Figure 2). The omega value (*dN*/*dS*) for the commercial LOM vaccine strains and the Jeju LOM strains isolated from the field (2004–2007 and 2014–2018) showed high homology with respect to the Erns, E1, E2, NS2, NS3, NS4A, NS4B, NS5A, and NS5B proteins (Figure 3A). However, the omega value for the C and P7 proteins of commercial LOM vaccine strains was a little higher (0.34224 and 0.43781, respectively) than that of the five Jeju LOM strains isolated in 2004 to 2007. The omega values for the C, E1, and E2 structural proteins of the Jeju LOM strains isolated in the field in 2014 to 2018 were higher than for the NS proteins (Figure 3A). Jeju LOM strains isolated from pigs from 2004 to 2007 were 98.8% to 99.2% identical at the nucleotide level and 99.2% to 99.4% identical at the amino acid level, with an omega value of 0.14088. Jeju LOM strains isolated from 2014 to 2018 showed 98.2% to 99.4% identity at the nucleotide level and 99.1% to 99.6% identity at the amino acid level, with an omega value of 0.16196 (Supplemental Table S3). The Bayes empirical Bayes (BEB) analysis of Jeju LOM strains isolated from 2004 to 2007 showed the possible inclusion mutation sites at 237, 259, 577, and 2467 aa positions, but a BEB analysis of Jeju LOM strains from 2014 to 2018 showed mutations at 173, 176, 386, 564, 1337, 2676, 2988, and 3605 aa positions. A native empirical Bayers (NEB) analysis performed with Jeju LOM strains (2014–2018) revealed the mutation site of the 564 aa position (*p* > 99%; Supplemental Table S3).

**Figure 2.** Root-to-top divergence analysis of Jeju LOM strains and commercial LOM vaccine strains. Complete genomes of 29 strains (8 commercial LOM vaccines and 21 Jeju LOM strains) were analyzed using the TempEst v1.5.1 program. The commercial LOM vaccine strains are marked with a black rectangle and the Jeju LOM strains are marked with a red rectangle. The heuristic residual mean squared method used to analyze all strains revealed the following: slope (rate), 3.93 <sup>×</sup> <sup>10</sup><sup>−</sup>5; X-intercept (TMRCA), 1797.24; correlation coefficient, 0.1272; R squared value, 1.6173 <sup>×</sup> 10<sup>−</sup>2; and residual mean squared value, 6.9076 <sup>×</sup> <sup>10</sup><sup>−</sup>6.

**Figure 3.** Omega values (*dN*/*dS*), genetic p-distances, and geographic distances for Jeju LOM strains. The *dN*/*dS* value for each of the structure proteins (Npro, C, Erns, E1, E2, p7) and non-structural proteins (NS2, NS3, NS4A, NS4B, NS5A, NS5B) was calculated and compared among 8 commercial LOM vaccine strains, 5 Jeju LOM strains (2004–2007), and 16 Jeju LOM strains (2014–2018) (**A**). Light gray denotes the commercial LOM vaccine strain, dark gray denotes the Jeju LOM strains (2004–2007), and black denotes Jeju LOM strains (2014–2018) (A). Jeju LOM strains (2014–2018) from the Aeweal (AW), Hanlim (HL), and Daejeong (DJ) regions show genetic *p*-distances of 0.0018 (one strain), 0.0045–0.0139 (11 strains) and 0.0011–0.0086 (4 strains), respectively (**B**).

#### *2.7. Geographic Distance and MCC Tree Analysis*

A correlation between geographic distance and genetic *p*-distance of the Jeju LOM strains contaminating the commercial LOM vaccine strain 16LOM-KM00 from 2014 to 2018 was confirmed, as shown in Figure 3B. Jeju LOM strain JJ14LOM-WSH01, isolated at a pig farm in the Aeweal region in 2014, showed a genetic *p*-distance from commercial LOM vaccine strain 16LOM-KM00 of 0.0018. Eleven Jeju LOM strains (Hanlim region: geographic distance, 13.5–20.1 km) and four Jeju LOM strains (Daejeong region: geographic distance, 27.4–28.1 km) isolated from pig farms showed a genetic *p*-distance of 0.0045–0.0139 and 0.0011–0.0086, respectively (Figures 3B and 4). Among genotypes 1, 2, and 3 in the beast tree constructed after global analysis of the complete E2 protein of CSFV, all LOM strains (including the Jeju LOM strains) belonged to independent groups within subgenotype 1.1 (Figure 5). From the mid-1980s, LOM strains were divided into two clusters: a lower cluster comprising commercial LOM vaccine strains and an upper cluster comprising mainly Jeju LOM strains isolated from pigs (Figure 5). In the above cluster, Jeju LOM strains isolated from pigs on Jeju Island were divided according to the year of isolation (2004–2007 and 2014–2018; Figure 5). The mean tMRCA for the LOM strains was 41.466, with an ESS (effective sample size) of 2340.2436 and a 95% highest posterior density (HPD) interval of 39.018–44.3142. The clock rate for LOM strains was 5.215 <sup>×</sup> 10<sup>−</sup>4, with a 95% HPD interval of 4.1721 <sup>×</sup> <sup>10</sup><sup>−</sup>4–6.159 <sup>×</sup> <sup>10</sup><sup>−</sup>4.

**Figure 4.** Map of Jeju Island showing locations of the pig farms harboring the Jeju LOM strains (2014–2018). Pig farms harboring Jeju LOM strains (2014–2018) are marked with a red star (2014), an orange rectangle (2015), a purple rectangle (2016), a light green rectangle (2017), or a deep green rectangle (2018). Black rectangles denote pig farms without Jeju LOM strains. An old slaughterhouse (**A**) and a new slaughterhouse (**B**) built at the end of 2018 are located in the upper and middle left, respectively. The light brown and light purple curved lines denote the estimated routes by which Jeju LOM strains entered the pig farms.

**Figure 5.** A maximum clade credibility tree based on complete E2 sequences of CSFVs. Rates of nucleotide substitution per site and per year, and the most recent common ancestor (tMRCA), were estimated using a Bayesian MCMC approach. Each dataset was simulated with the following options: generation = 100,000,000, burn-in of 10%, and ESSs > 200. The confidence of the phylogentic analysis is present to numbers representing branch length (time) above the nodes. The yellow block comprises LOM strains within Genotype 1. The light green line denotes 8 commercial LOM vaccine strains. The blue line denotes 5 Jeju LOM strains (2004–2007) and the red line denotes 16 Jeju LOM strains (2014–2018).
