One pregnant sow was purchased from a commercial farm (Farm A) and delivered to the research facilities at Lindholm, Technical University of Denmark, Kalvehave, Denmark. All 14 piglets received 0.05 ml BCG intradermally (i.d.) in the skin of the right ear base three days after delivery. One piglet was found dead on day 9, presumably crushed under the sow. After 20 weeks (day 140), the remaining 13 pigs were euthanized by captive bolt, followed by bleeding. Tissue samples were collected in sterile falcon tubes and analyzed at the International Reference Laboratory of Mycobacteriology, Statens Serum Institut, Copenhagen, Denmark, for the presence of live BCG. Tissue samples were taken from the liver, lung, kidney, bone marrow from the left femur, muscle from the shoulder, superficial dorsal cervical lymph node, retropharyngeal lateral lymph node, and the skin at the vaccine injection site.

Three pregnant sows were purchased from Farm A and delivered to the research facilities at Frederiksberg, National Veterinary Institute, Technical University of Denmark, Denmark. Two to three days after delivery, 32 piglets were randomized 1:1 within each of the three litters to receive i.d. 0.05 ml BCG (n=17) or 0.05 ml Sauton (BCG solvent used as a placebo, n=15) in the right shoulder, administered by a midwife with extensive experience in BCG vaccination of human newborns.

After BCG administration, weight was recorded on days 1, 16, 28, and 36-39, and rectal temperature on days 36-39.

On day 36, eight pigs had been lost (see the Results section for details), and the remaining pigs (n=24, BCG: 12; Sauton: 12) were allocated based on current weight into two groups, ensuring an equal distribution of BCG- and Sauton-treatments. One half (n=12, 6/6) was inoculated intranasally with Actinobacillus pleuropneumoniae 4226, serotype 2 (1 ml in either nostril as detailed below). The other half (n=12, 6/6) received 4 ml of10⁶ TCID₅₀/ml swine influenza virus H1N2 (A/swine/Denmark/101394-1/2011(H1N2) using MAD Nasal™ intranasal mucosal atomization devices (Teleflex, Morrisville, NC, cat. no. MAD100) 2 ml in either nostril). Before inoculation, the animals were anaesthetized with i.m. injection of a mixture of 12.5 mg tiletamine, 12.5 mg zolazepam, 12.5 mg xylazine, and 12.5 mg ketamine per mL at 1 mL/15 kg. The A. pleuropneumoniae group was sacrificed on day 38 (two days after bacterial infection) and the influenza group on day 39 (three days after viral inoculation). Blood was collected in PAXGene tubes (2.5ml, Qiagen) on days 0, 1, 8, and 21 and stored at -20°C until further processing.

For experiment C, 12 female pigs (Danish Landrace/Danish Yorkshire crossbreds and paternal Duroc) were delivered after weaning to the research facilities at Frederiksberg from a commercial farm (Farm B). At 5 weeks of age, the pigs were stratified by size (total weight range 5.5-11.5 kg, mean 8.9 kg) and allocated to two treatment groups, receiving either 1.5-2 vials of BCG resuspended in 0.8 ml Sauton (equaling 15-20 times a standard BCG dose, n=6) or 1.0 ml Sauton alone (n=6), applied by three adjacent injections i.d. in the right hind leg. 24 days after immunization, venous blood was collected into heparinized tubes for stimulation assays, and the pigs were sacrificed.

The tissue samples listed in the outline of Experiment A above were analyzed at the International Reference Laboratory of Mycobacteriology, Statens Serum Institut, Copenhagen, Denmark, for the presence of live Mycobacterium bovis BCG. Specimens were disintegrated with a scalpel and in a sample homogenizer (GentleMACS, Dissociator, Miltenyi Biotec, Bergisch Gladbach, Germany) and then pre-treated with MycoPrep (NALC-NaOH, Becton Dickinson, Sparks, MD). The pre-treated material was analyzed with microscopy, PCR specific for Mycobacterium tuberculosis complex (Cobas TaqMan MTB, Roche Diagnostics, Rotkreutz, Switzerland), and culture in MGIT (Becton Dickinson) and Löwenstein-Jensen (SSI Diagnostika) media. The growth of M. bovis BCG was verified with GenoType MTBC (Hain Lifescience, Nehren, Germany).

A seed lot of A. pleuropneumoniae strain 4226, serotype 2, stored at -80°C was cultured overnight at 37°C on modified pleuropneumonia-like organism (PPLO)-agar plates using E. coli as a nurse strain, suspended in 0.9% NaCl, and adjusted to the desired concentration by turbidity of McFarland standard 0.5 diluted 1:1 in BHI+0.5% NAD. Animals were inoculated with 2 ml of A. pleuropneumoniae suspension, equaling 0.9x10⁶ CFU/animal. The concentration and purity of the inoculation suspensions were verified by seeding on PPLO and blood agar plates, respectively. After necropsy, for re-isolation of the inoculation strain from tissue samples (liver, spleen, and lung), a scalpel was inserted into the affected tissue and struck on PPLO agar and cultured at 37°C; the following day, emerging colonies were enumerated.

Swine Influenza A H1N2 virus was propagated in Madin-Darby Canine Kidney Epithelial Cells (MDCK) and prepared for inoculation as previously described (9). Influenza A virus load was quantified from nasal swabs preserved in saline using RT-qPCR. Total RNA was extracted from 140 µl of the nasal swab sample by QIAamp Viral RNA Mini Kit (Qiagen) using the manufacturer’s procedure and stored at -80°C until analysis. The influenza A virus was detected by an in-house modified version of a real-time RT-qPCR assay detecting the matrix gene (10).

Necropsy was performed to characterize gross lesions in A. pleuropneumoniae-inoculated and influenza-inoculated pigs. The lungs were examined for A. pleuropneumoniae-like lesions defined as areas with hemorrhage, consolidation, necrosis of lung tissue, and fibrin exudation. Influenza lesions were defined as mucopurulent exudate in bronchi and bronchioles and lobular atelectasis of lung tissue.

In Experiment B, from serum samples collected on days 16, 36, 37, 38, and 39 (day 39 only from influenza-challenged pigs), CRP and Haptoglobin were analyzed using in-house ELISA protocols (11, 12).

In Experiment B, on day 30, a whole blood differential count was performed (ABC Vet, Horiba ABX, France) using EDTA-treated blood, yielding counts of leukocytes, red blood cells, and platelets, and concentrations of hemoglobin, mean corpuscular hemoglobin, and mean hematocrit percentages and mean corpuscular volume.

In Experiment B, on day 28, undiluted whole blood was stimulated with LPS (10 ng/ml, from Escherichia coli O55:B5, Sigma-Aldrich), Pam3CSK4 (1 µg/ml, Invivogen), Staphylococcus enterotoxin B (1 µg/ml), phytohaemagglutinin (PHA) (2µg/ml), or PPD from M. tuberculosis (10 µg/ml, Statens Serum Institut) for 22h at 37°C, 5% CO₂ humidified incubator.

In Experiment C, peripheral blood mononuclear cells (PBMC) were purified from the heparinized blood within 2 hours of venipuncture. Blood was diluted 1:1 in sterile PBS supplemented with 2% fetal calf serum (FCS) and carefully applied to a StemCell SepMate tube containing Lymphoprep density gradient medium (StemCell Technologies). After centrifugation, the PBMC-containing supernatant was retrieved and washed twice in PBS/2% FCS. To lyse red cells, the cell pellet was resuspended and incubated with red cell lysis buffer (78mM NH₄Cl, 0.05 mM EDTA and buffered with 5mM KHCO₃). The PBMCs were resuspended in RPMI-1640 Glutamax (Gibco) supplemented with 10% FBS and 1% penicillin/streptomycin, seeded out as 14x10⁶ cells/well in a 24-well Cellstar plate (Greiner Bio-One, Frickenhausen, Germany), and stimulated at 37°C, 5% CO₂ with LPS (10 ng/ml final concentration, Sigma), Pam3CSK4 (1 µg/ml, Invivogen), Staphylococcus enterotoxin B (1 µg/ml, Sigma), Purified protein derivative (PPD) from M. tuberculosis (10 µg/ml, Statens Serum Institut), or medium alone. After 7h, the plates were centrifuged, supernatants removed, and the cell pellets harvested in RLT buffer (Qiagen) with 1% v/v beta-mercaptoethanol and stored at -80°C until RNA extraction.

In Experiment B, blood was collected in PAXgene tubes, and extraction was performed according to the manufacturer’s instructions using the PAXgene Blood miRNA Kit (Qiagen). In Experiment C, RNA was extracted from PBMCs preserved in RTL buffer after in vitro stimulations using the QiaAmp RNA Blood mini protocol according to the manufacturer’s instructions (Qiagen). The quantity of the RNA was measured using a spectrophotometer (NanoDrop ND-1000, Thermo Scientific, United States). RNA integrity was measured on an Agilent 2100 Bioanalyzer (Agilent Technologies, Nærum, Denmark). RNA integrity number (RIN) values in Experiment B were in the range of 7 to 9, averaging 7.6; RIN values in Experiment C were in the range of 8.7 to 9.7, averaging 9.2.

Synthesis of cDNA was performed in duplicates from a total of 500ng RNA per reaction using QuantiTect Reverse Transcription Kit (Qiagen #205311) as previously described (13). Non-reverse transcriptase reactions were included as a control for possible genomic DNA contamination.

In Experiment B, the resulting cDNA concentration was measured on NanoDrop and diluted to 200 ng/μl in 20 μl in low EDTA TE-buffer before pre-amplification.

Fifteen cycles of pre-amplification were performed as previously described (13) using TaqMan PreAmp Master Mix (Applied Biosystems, PN 4391128), followed by Exonuclease I treatment (New England Biolabs, PN MO293L) of 30 minutes at 37°C and 15 minutes at 80°C.

qPCR was performed in 96.96 Dynamic Array Integrated Fluidic Circuits (Fluidigm, CA, USA) according to (14), combining 96 samples with 96 primer pairs ( Supplementary Table 1 ). A non-template control and the non-reverse transcriptase controls from the cDNA synthesis were included on each array chip to monitor for non-specific DNA contamination, including of genomic origin.

Expression data (Cq values) and melting curves were acquired using the Fluidigm Real-Time PCR Analysis software 3.0.2 (Fluidigm) and exported to GenEx (MultiD, Sweden) for data pre-processing including censoring for non-specific amplification or low amplification, interplate correction, correction for PCR efficiency for each primer assay individually, normalizing using GeNorm and NormFinder programs according to the expression of ACTB, B2M, RPL13A, PPIA, TBP, and YWHAE in Experiment B and ACTB, B2M, GAPDH, RPL13A, TBP, and TGPI in Experiment C, and averaging of cDNA technical repeats. Data was scaled by setting the lowest expression of each gene of interest to 1 whereby the remaining samples were expressed as the ratio relative to this lowest sample. Experiment B data was also log2 transformed.

After censoring the samples by expression data quality as described above, the analysis included the following numbers in Experiment B: day 0: BCG n=17/control n=15; day 1: n=16/14; day 8: n=13/12; day 21: n=12/11. In Experiment C, all 12 animals were included in all stimulation conditions except for one (control pig #12) missing in the LPS stimulation.

In Experiment B, IL-6 and TNF were quantified from harvested supernatants using commercial porcine ELISA kits (DuoSet ELISA, R&D Systems); IFN-γ was analyzed using an in-house ELISA protocol (15).

Data were analyzed using GraphPad Prism 6, StataMP, version 12, or SAS Enterprise Guide version 7.4.

Kruskal-Wallis was used for non-normally distributed non-paired samples and Wilcoxon sign-ranked test for non-normally distributed paired samples. In the gene expression study in Experiment B, the control and vaccinated groups were compared using unequal variances t-test. Additionally, gene response data from Experiments B and C were analyzed in multivariate models using Discriminant Analysis of Principal Components (16), in which the multivariable data consisting of numerous response variables (the combinations of stimulant and gene response) are reduced by Principal Component Analysis to fewer linear and uncorrelated variables. These canonical variables then serve as input variables in Discriminant Analysis (proc discrim in SAS). The discriminant analysis can test the probability of correct classification of vaccinated and unvaccinated individuals. Since we did not have a sample of adequate size to split into a training and test set, we used the crossvalidate and posterr options in SAS which classifies each observation by using a discriminant function computed from all the other observations (SAS). The misclassification rate of vaccination status is returned by the model for evaluation.

In Experiment C, LPS stimulation was lacking from one animal, and the LPS stimulation was therefore omitted from the Principal Component Analysis and Discriminant Analysis, as this method cannot handle missing values. In an alternative analysis, the missing values from this single animal were replaced by the median of each gene of the other pigs in the same treatment group (Sauton-treated, LPS-stimulated), whereby the LPS-stimulated samples could be included in the model.

Ethical approval for the experiments was obtained from the Danish Animal Experiments Inspectorate (case no. 2015−15−0201−00520) and the local university animal ethics committee at the Technical University of Denmark. The animal experiments took place at the research facilities of the Technical University of Denmark in 2015 and 2016.

In total, nine piglets (BCG: 6/31; Sauton: 3/15) died or were euthanized due to events considered unrelated to the immunization in the first weeks after birth in Experiments A and B. One piglet (1/14) in Experiment A was found crushed under the sow nine days after BCG vaccination. In Experiment B, three BCG-vaccinated piglets were found dead, and two BCG-vaccinated and three Sauton-treated piglets were euthanized due to severe weakness, of which seven died within the first 7 days ( Table 1 ). The causes of death or euthanasia were general weakness or mechanical conditions and did not indicate adverse events related to the BCG vaccine. No animals died prematurely in Experiment C.

No visible or palpable reactions were observed at the vaccination site after the neonatal BCG dose (Experiments A and B), whereas 3/6 BCG-vaccinated pigs in the high BCG dose Experiment (Experiment C) had palpable reactions at the injection site 24 days after vaccination. Additionally, 2/6 had mild skin redness but no palpable reaction ( Supplementary Figure 1 ). No pigs developed ulcers or scars.

In Experiment B, body weights of the pigs allocated to BCG vaccination were 15% lower than the Sauton group one day after randomization (no measure of weight available before vaccination). BCG-vaccinated pigs remained lighter throughout the study, with the relatively largest difference of 25% (p=0.04) observed on day 16 ( Supplementary Table 2 and Figure 2A ). On day 36, the BCG-vaccinated pigs were 17% (p=0.11) lighter than the Sauton-treated pigs.

Weight gain in Experiment C from randomization to 24 days after vaccination was not different between BCG and Sauton-treated pigs ( Figure 2B ).

In Experiment B on day 30, hemoglobin and hematocrit concentrations were higher in the BCG-vaccinated pigs (p=0.04 and p=0.03, respectively) ( Figure 3 ). Whole blood cell counts were not different.

In Experiment A, 20 weeks after vaccination, two of 13 pigs were culture-positive for BCG in the lateral retropharyngeal lymph node draining the vaccination site. PCR confirmed the presence of BCG. All other specimens were culture-negative and PCR-negative and there was no visible local reaction at the injection site in any of the pigs on week 20.

In Experiment B, the mRNA expression of 49 gene sequences was analyzed from whole blood before vaccination (day 0) and on days 1, 8, and 21 ( Supplementary Table 3 ). Four genes of interest (retinoic acid-inducible gene I (RIG-I), Melanoma Differentiation-Associated protein 5 (MDA5), macrophage colony-stimulating factor 1 (CSF1), and swine leukocyte antigen DR-beta 1 (SLA-DRb)), were marginally differentially regulated between BCG and Sauton treatments at one time-point after randomization. However, none of these results sustained correction for multiple testing ( Supplementary Figure 2 ).

For the remaining 44 gene transcripts tested, there was no difference between BCG-vaccinated and control. Moreover, in the multivariate analysis, using principal component analysis to reduce the data dimension, BCG-vaccinated and control animals did not segregate at any time point (data not shown). In the discriminant cross-validation analysis, the classification error rate of BCG vaccination remained high, around random annotation. Including sex as a categorizing factor together with BCG did not affect the results (data not shown).

In Experiment B, 28 days after vaccination, IFN-γ responses to overnight PPD stimulation in whole blood cultures were, on average, higher in the BCG-vaccinated pigs than in control pigs (p=0.007). However, 4/11 BCG-vaccinated pigs were low responders (less than twice the nil level) ( Figure 4A ). IFN-γ responses to the positive SEB did not differ between BCG-vaccinated and control animals (data not shown). TNF responses to SEB tended to be higher (p=0.097) in BCG-vaccinated pigs, which was not seen for TNF responses to PHA ( Figure 4B ). IL-6 responses to SEB, LPS, or Pam3CSK4 were generally low compared with the nil sample with a stimulation index (stimulation to nil ratio) ranging from 1.4 for SEB, 2.0 for LPS, and 3.0 for PamCSK4. Overall, TNF and IL-6 responses to LPS or Pam3CSK4 were not different between BCG and control animals (data not shown).

In Experiment C, 24 days after vaccination, the expression of 52 immune-related genes from in vitro stimulated PBMC was measured using qPCR ( Supplementary Table 4 ). Overall, gene expression levels in stimulations were relatively low compared to the medium-only controls, indicating suboptimal culture conditions, probably due to low stimulation doses. IFN-γ responses to PPD were higher than background (the medium only culture) in BCG-vaccinated pigs but not in control pigs, indicating specific immunogenicity of the BCG vaccination ( Figure 5 ), as also noted for whole blood stimulations in Experiment B. For a few transcripts, responses to SEB were higher in BCG-vaccinated compared to controls, significantly so for the genes encoding the cytokines IFN-γ and IL-1α ( Figure 6 ).

After principal component analysis, the 208 response variables in Experiment C (a combination of 52 gene transcripts and 4 culture stimulants including nil) were reduced to seven principal components. Next, we submitted all gene expression variables to PCA, selecting three principal components, which then served as input variables to the discriminant analysis to test if the composite gene expression data contained a signal of BCG vaccination that enabled the model to assign vaccination grouping of the 12 animals correctly. The cross-validation analysis found a total error rate of 42% (5/12) in assigning the 12 samples to either BCG or control groups, which is only slightly better than random. Including PPD response transcripts slightly improved the performance to an error rate of 33% (4/12). Imputing the missing values and including the LPS-stimulated gene responses did not improve the precision.

There was overall no difference between the BCG-vaccinated and Sauton-treated pigs in rectal temperature responses to the A. pleuropneumoniae or influenza infections ( Figure 7 ). One day after A. pleuropneumoniae inoculation, in both the BCG-vaccinated and Sauton-treated groups, there was a transient drop of approximately 1°C, whereas, after influenza inoculation, the temperature remained stable.

At necropsy after A. pleuropneumoniae inoculation, the lungs were unremarkable or showed postmortal lesions in eight pigs (BCG: n=3; Sauton: n=5). Minor lesions of consolidated lung tissue in the cranial lobes, consistent with chronic lobular bronchopneumonia, were present in three pigs (BCG: n=3). The lesions were uncharacteristic of A. pleuropneumoniae, but could be caused by other respiratory pathogens, e.g., Mycoplasma spp. One pig (BCG) had small processes of consolidated hemorrhagic lung tissue in the diaphragmatic and cranial lung lobes; though the lesions were not classic to A. pleuropneumoniae infection, this could not be ruled out on a macroscopical basis.

Among the influenza-inoculated pigs, the lungs were unremarkable or with postmortal lesions in nine pigs (BCG: n=6; Sauton: n=3). Mild chronic pleuritis was observed in one Sauton-treated pig. Two pigs (Sauton: n=2) showed minor lesions of consolidated lung tissue without exudation in the cranial lobes; though not typical for influenza infection this could not be excluded, however other microbes, e.g., Mycoplasma spp. could be responsible.

Based on RT-qPCR of the influenza matrix gene, viral load of the upper airways peaked at two days after infection but with no differences in virus titers between the BCG and Sauton groups ( Supplementary Figure 6 ). No A. pleuropneumoniae could be cultured from any of the liver, spleen, or lung samples (data not shown).

There was no difference between BCG-vaccinated and Sauton-treated pigs with respect to serum haptoglobin or CRP on the day of infectious challenge (day 36 after immunization). After A. pleuropneumoniae infection, serum haptoglobin concentrations increased in both the BCG and Sauton groups, with the highest levels measured two days after infection ( Figure 8A ), with no difference between the BCG and Sauton groups. After influenza infection, however, a haptoglobin response was only seen in the Sauton group, peaking at two days after infection, whereas BCG-vaccinated pigs did not show a haptoglobin response after the influenza challenge ( Figure 8B ). Likewise, serum CRP concentrations were increased after the influenza challenge, however, the response was attenuated or delayed in BCG-vaccinated pigs two days after inoculation ( Figure 8D ). No statistically significant difference in serum CRP concentrations was seen between BCG-vaccinated and Sauton-treated pigs after A. pleuropneumoniae challenge ( Figure 8C ).