A survey of an eight-breed cross swine heterogeneous population demonstrates the associations among immune cell traits, lung lesions and Mycoplasma abundances

Background: Developments of pulmonary diseases, often accompanied by infections of bacteria, severely affect meat production and welfare of pigs. This study investigated the association of lung lesions and Mycoplasma levels inferred from 16S rRNA sequencing of bronchoalveolar lavage uid with 57 immune cells and 25 hematological traits in 307 pigs at age of 240 days from an eight-breed heterogeneous cross. Result: At a false discovery rate threshold of 0.05, we found that the greater severity of lung lesions were signicantly associated with higher CD8+ to CD3+ cell ratio (CD8+/CD3+), neutrophil to lymphocyte ratio (NLR), and standard deviation of red blood cell volume distribution width (RDW-SD), and lower CD4-CD8-/CD3+, CD3+CD4-CD8-/PBMCs, CD14-CD16-/PBMCs, mean corpuscular hemoglobin concentration (MCHC), lymphocyte count (LYM) and lymphocyte count percentage (LYMR), reected an status of inammation, immune suppression and hypoxia of the pigs accompanying the development of the lung lesion. The Mycoplasma abundance showed positive correlations with neutrophil count (NEU), neutrophil count percentage (NEUR), neutrophil-to-lymphocyte ratio (NLR), monocyte count (MON), RDW-CV, and RDW-SD, and negative correlations with MCHC, LYM, and LYMR, these correlations are largely consistent with those of lung lesions, supporting the comorbidity of lung lesions and Mycoplasma infection. We also observed a nonlinear association that the sharp increases in NEU and NEUR occurred only when Mycoplasma abundance raised to a level above the population-average. Conclusion: This study showed that the pigs from an eight-breed cross heterogeneous population reared under standardized housing conditions suffered lesion averagely covered 40% of lung, and the lung lesions were signicantly linked to load of Mycoplasma in bronchoalveolar lavage uid.


Background
Respiratory disease is common in pig production industry, often associated with multiple microbial infection, and causes great treatment burden and economic loss (1). Enzootic pneumonia in pigs is a type of respiratory disease links to infection of pathogenic bacteria such as Mycoplasma pneumoniae and accompanies with development of lung lesions (2). The lung tissue of pigs is accessible at slaughter, which allows for direct evaluation of lung lesions and quanti cation of lung microbiota or viruses from bronchoalveolar lavage through serological tests, PCR, or sequencing technology (3,4). For example, a study of 125 farrow pigs found that Mycoplasma hyopneumoniae, porcine reproductive and respiratory syndrome virus, and swine in uenza viruses are the major pathogens involved in the development of pneumonia-like lung lesions (4). We measured microbial communities from 20 lung samples with extreme lung lesion phenotypes from a pig heterogeneous population derived from eight founder breeds, and found that Mycoplasma is the major pathogen associated with the severity of lung lesions and signi cantly affect the growth performance of pigs (3).
The host immune and hematological systems have essential functions in resisting pathogen and maintaining animal health (5,6). Clarifying the associations between immune and hematological cells with the lung lesion and pathogenic bacteria would provide knowledge that is useful to diagnose, prevent, and treat respiratory diseases. It has been reported that the ratio of CD4 + T cells, NK cells, and B cells to white blood cells was lower in Landrace pigs that were selected to be resistant to Mycoplasma than the control group (7). Despite progresses have been made, studies aimed to reveal the links between immune and hematological responses of pigs to respiratory disease and bacterial infection remain limited.
In this study, we investigated the correlation of lung lesions and Mycoplasma levels with 57 immune and 25 hematological traits measured in 307 generation 7 (G7) pigs at age of 240 ± 10 days from the pig heterogeneous population. We expected that the sample size of more than 300 individuals together with broad genetic and phenotypic variations of the population would offer reasonable power to found biologically meaningful associations.

Experiment design
This work is a cross section association study largely takes advantage from the design of the pig heterogeneous population. The pig heterogeneous population was created by crossing eight founder pig breeds consisting of 4 western commercial breeds (Large White, Duroc, Landrace and Pietrain) and 4 Chinese indigenous breeds (Bamaxiang, Tibetan, Erhualian and Laiwu), with major objectives to investigate genetic basis of various complex traits of pig. The eight founder breeds conferred greater genetic and phenotypic diversity in the heterogeneous population than a normal F 2 or three-cross populations. Moreover, as all the pigs were reared in uniform and standardized conditions and slaughtered at approximately the same age, we expected that the environmental confounding effect on the phenotype-phenotype association results are well controlled. Another unique feature of this work is that each individual were simultaneously measured with multiple phenotypes including the lung lesion, microbiome composition from bronchoalveolar lavage uid through 16S rRNA sequencing, immune cell using ow cytometry and hematological traits from routine blood test, which is rarely reported in pigs, therefore, an integrative analysis on these traits could provide novel information on the host response to pulmonary bacterial infection and lesion in pigs (Fig. 1A).

Animals
All animals investigated in this study are from generation 7 individuals of the heterogeneous population.
Each individual is expected to contain ancestry from the eight founder breeds. The population was kept under uniform indoor conditions with ad libitum water supply at the experimental farm in Jiangxi Agricultural University (Nanchang, China), and fed with corn-soybean feed (Jiangxi Huada Group) containing 3100 kJ digestive energy, 15.5% crude protein, 0.78% lysine, 0.6% calcium, 0.35% phosphorus and 0.55% salt. Each pig was vaccinated with seven vaccines including Classical Swine Fever Vaccine,

Measurement of the lung lesions and microorganism levels
Detailed measurement of lung lesion score and microorganism were described in our previous studies with a few adjustments (3). Brie y, the whole lung tissue was harvested during the slaughter and photographed from both anterior and posterior sides. The lung lesions were scored from the photographs; lungs were divided into seven sections, including apical lobe, cardiac lobe, and diaphragmatic lobe at both the left and right side, and the intermediate lobe. Then, each section was assigned a score of 0-5 corresponding to a lesion proportion of 0%, 0-20%, 20-50%, 50-75%, 75-90% and > 90%. We assigned the weights to the seven sections of lung according to their volume, which is 5%, 6%, 7%, 8%, 32%, 36% and 5% for left and right apical lobe, left and right cardiac lobe, left and right diaphragmatic lobe and intermediate lobe, respectively. Finally, the lesion score of the lungs was calculated as a weighted sum of lesion scores for the different sections of lung from both the anterior and posterior sides.
The bronchoalveolar lavage samples were collected within 20 min poster slaughter. After photographing, we infused about 100 ml sterile PBS buffer (137 mM sodium chloride, 2.7 mM potassium chloride, 10 mM phosphate, and 2 mM potassium phosphate) into each lung with the trachea cut off, and gently pressed the lung sample for 2 min. About 50 ml lung lavage uid was collected. The lavage uid sample was centrifuged at 4000 g for 30 min at 4 , the pellet was transferred to a 2 ml sterile centrifugation tube, and stored at -80 for subsequent DNA extraction. DNA was extracted from the lavage samples using QIAamp DNA Stool Mini Kit (Qiagen, Hilden, Germany). The V3-V4 region of 16S rRNA gene was ampli ed using barcode-indexed universal bacterial primers. The amplicons were then puri ed and sequenced on an Illumina MiSeq sequencing platform (Illumina, SanDiego, CA, USA). The raw sequence reads were ltered as previously described (3). The Operational taxonomic units (OTUs) were clustered using UCLUST cluster algorithm in USEARCH v7.01090 (8). The taxonomy of each 16S rRNA gene was analyzed using RDP v2.2 (9) based on Green-Gene v13.5 reference database (10). The genera level of Mycoplasma used in this study was determined in previous study (3).

Flow cytometry analysis
Peripheral blood samples from the neck arteries were collected during slaughter and stored in lithiumheparin tubes at room temperature for further analysis. We used ow cytometry (BD Accuri C6) to measure the absolute and relative number of immune cells using six antibodies by three combinations/panels: 1) CD3 FITC (BD Biosciences), CD4 Alexa Fluor 647 (BD Biosciences), and CD8 PE (BD Biosciences); 2) CD14 (Bio-Rad) combined with Alexa Fluor 647 -conjugated A niPure Goat Anti-Mouse IgG (Jackson ImmunoResearch) and CD16 RPE (Bio-Rad); and 3) CD163 FITC (Bio-Rad). For each panel of antibodies, 100 µl peripheral blood was incubated for 30 min at 4℃ with speci c antibodies. Red cells were lysed for 15 min at room temperature (20-25℃) by erythrocyte lysis buffer (BD Biosciences). The cells were washed by 1 ml PBS and centrifuged at 300 × g for 10 min. The supernatants were removed and the pellets were re-suspended in 500 µl PBS (pH = 7.4), and analyzed on an Accuri C6 ow cytometer. Finally, we obtained immune cell traits for 307 G7 pigs including 176 males and 131 females (Additional le 1: Table. S1).

Statistical analysis
The summary statistics of phenotypic traits were calculated by R program (12). The repeatability of a trait was calculated by , and are between and within sample variance components, respectively, that were computed using ANOVA in the R program (12). The phenotypic records that deviated from mean ± 4 standard deviations of population were set to be missing. The phenotype values of the immune and hematological traits were rst log2 transformed, and then corrected for potential covariates including age, sex, batch of experiment, and PBMCs concentration (the latter only for immune cell traits). Two statistical approaches, Pearson correlation and quadratic regression, were employed to investigate the association of lung lesions and Mycoplasma levels with immune cell and hematological traits. The correlation between a pair of traits was computed using cor function in R. The quadratic regression was performed by lm function in R based on following formula: Where y is a vector of corrected phenotypic values of lung lesion or microorganism, and x represents a vector of corrected phenotypic values of immune or hematological trait. The signi cance thresholds of the Pearson correlations were determined by the false discovery rate (FDR) calculated from 1000 permutation tests, in which we randomly shu ed the labels of phenotype data to calculate the null distribution of association statistics.

Descriptions of phenotypes
The summary statistics of traits investigated in this study are detailed in Additional le 1: Table. S1. The average lung lesion score was 1.96, indicating on average, 40% (1.96/5) of an individual pig's lung developed lesions. We determined the composition of bacteria at genera level, and investigated their correlation with lung lesion score across 625 G7 individuals. The results showed that Mycoplasma displayed much more signi cant correlation (P value = 3.3 × 10 − 15 ) with lung lesion than the other bacteria genera (Fig. 1B). In light of this result, we focused on the Mycoplasma in the subsequent analyses. The coe cient of variation for the Mycoplasma levels is 1.16, which is greater than the coe cient of variations for the immune cell and hematological traits measured in the same population (Additional le 1: Table. S1), re ecting a large variability of Mycoplasma levels in the lung of heterogeneous pigs that were reared in uniform housing and feeding environments. We pro led 57 immune and 25 hematological traits in arterial blood samples of 138-307 G7 individuals (Additional le 1: Table. S1). The average repeatability of the 57 immune cell traits estimated from 8-27 pairs of replicated samples was 0.895, ranging from 0.629-0.996, suggesting measurement of immune cell phenotypes was reliable.
Both of the immune cell and routine blood test parameters were measured from the arterial blood, which directly come from lung, therefore, it could re ect the situation of lung more accurately than venous blood (13). Out of 307 individuals investigated in this study, we also measured immune cells from venous blood of 58 pigs at age of 120 days. Interestingly, the immune cell counts measured from the venous blood samples of 120 days pigs were highly correlated with those measured in the arterial blood of the same individuals at 240 days of age (Fig. 2), indicating that the arterial blood can also re ect inter-individual difference in blood cell counts in venous blood, thus the results generated in this study are potentially comparable with those studies using venous blood.

Evidence of a nonlinear relationship between NEU and Mycoplasma abundances
Next, we use quadratic regression to assess the nonlinear relationship among traits. Notably, we obtained considerable enhancement in association signi cance between Mycoplasma levels and two hematological traits, NEU (P value = 2.6 × 10 − 7 ) and NEUR (P value = 1.7 × 10 − 8 ) based on quadratic regression compared to those of Pearson correlation analysis (Fig. 4A), suggesting existence of a nonlinear relationship between Mycoplasma levels with NEU and NEUR. In these two cases, we observed sharp increases of NEU and NEUR were occurred when the abundance of Mycoplasma in lung raised above the population average level (Fig. 4B-E).

Discussion
Respiratory diseases such as enzootic pneumonia cause great economic loss in pig production industry. The pathogenesis of some respiratory pathogens in pigs has been studied, the host immune response to infection and tissue lesions remain largely unclear and worth to be surveyed. In this study, we examined the association of various immune cell traits and hematological parameters with lung lesions and Mycoplasma levels. The identi cation of signi cant associations in this study would provide valuable information on the immune and hematological response of host to Mycoplasma infection and related lung lesion.

Immune and hematological traits positively correlated with lung lesion
We observed that pigs with more severe lung lesions had higher CD8+/CD3+, NEU, NEUR, NLR, RDW-CV, and RDW-SD. CD8 + T cells are a class of cytotoxic T cells which kill infected cells by secreting cytokines (TNF-α and IFN-γ), cytotoxic molecules (perforin and granzymes), and interacting with Fas/FasL (14). Increased CD8 + T cells in peripheral blood, lung tissue were reportedly associated with lung lesions in humans and pigs (15,16). Further, a study in mice lacking the CD8 gene exhibited a reduced in ammatory response and did not develop lung disease after prolonged exposure to cigarette smoking (17). The excessive increase in the proportion of CD8 + in CD3 + T cells observed in the study, re ecting an imbalance of CD8 + T cell differentiation and severe in ammatory response, could contribute to development of lung lesion.
Neutrophils are thought to be the rst line of defense against invading pathogens, which can trigger an in ammatory response which further prompts neutrophils to in ltrate tissues. There is evidence that higher levels of neutrophils are associated with lung lesions in pigs (18); this agrees with the positive correlation observed between lung lesions and neutrophil-related traits identi ed in this study. In humans, NLR is considered to be the biomarker of in ammation, which was positively correlated with the severity of COPD and predictor of mortality (6). The percentage of peripheral blood neutrophils was reported to be signi cantly increased in people with lung disease (19). Moreover, the lung cancer patients with higher NLR had poor overall survival (20).
RDW-CV (or RDW) and RDW-SD are measures of variation in red blood cell volume. The positive correlations of RDW-CV and RDW-SD with lung lesion observed here, suggested that the progression of lung lesion is accompanied by the increase in variation of red blood cell size. To our knowledge, no study in pigs has reported the association between RDW and lung lesions. However, the association of RDW with lung disease has been identi ed in humans. A study has found negative correlation of RDW with lung function (21). The evaluated RDW is reported to be related to chronic in ammation, which can alter erythropoiesis and promote erythrocyte membrane deformation (22). Higher RDW in individuals with severe lung lesions may promote the in ammation and further exacerbate the disease.

Immune and hematological traits negatively correlated with lung lesion
In this study, we found that CD4-CD8-/CD3+, CD3 + CD4-CD8-/PBMCs, CD14-CD16-/PBMCs, LYM, LYMR, and MCHC were negatively associated with lung lesions. As far as we know, this study is the rst report of several immune associations in pigs. CD4-CD8-T cells also exist in rodents and humans, where they constitute about 1-5% of the CD3 + T cell population (23). CD4-CD8-CD3 + cells were known as regulatory T cell, as they were found to be involved in autoimmune diabetes and intracellular infection in humans (24). The CD4-CD8-T cell population was reported to be signi cantly reduced in severe pulmonary tuberculosis patients (25). CD14 and CD16 are monocyte markers in both pigs and humans. We reported a signi cant association of lung lesion with CD14-CD16-cell population, which encompassed different cells including B cells, T cells, NKs, DCs, further investigations are required to verify the involvement of these cells in lung lesion.
Among the parameters from routine blood test, lymphocytes are an important part of the immune system; reduced lymphocytes levels is usually associated with immunosuppression (26). An in vitro study showed that lipid-associated membrane proteins of Mycoplasma induced the apoptosis of lymphocytes and monocytes from PBMCs (27). We therefore hypothesize that the invasion of Mycoplasma may lead to the apoptosis of lymphocytes and immunosuppression of host.
MCHC, a measure of mean corpuscular hemoglobin concentration, was found to be negatively correlated with lung lesion. The decreased levels of MCHC could be linked to the deformation of red blood cells as indicated by the evaluated RDW-CV. A recent study in humans showed that MCHC levels were lower in a control (healthy) group than in a newly diagnosed COPD group (28). The process of hypoxia, induced by lower MCHC, may contribute to the lung tumor cell metastasis and poor prognosis (29). We postulate that the decline of MCHC may have contributed to the development of the disease by putting the animals in an anemic hypoxic state.

Immune and hematological traits associated with lung Mycoplasma levels
Previously, we found that Mycoplasma was the most abundant and only bacteria signi cantly associated with lung lesions in pigs (3). Similarly, we found that the pattern of correlations of Mycoplasma with six hematological parameters including RDW-SD, NLR, NEUR, LYMR, MCHC, and LYM was similar to the lung lesion correlations. These results agree with the nding that the percentage of neutrophils and lung damage appeared to be increased in pigs infected with Mycoplasma compared to the control group (30). Mycoplasma pneumoniae was also reported to be associated with the pathogenesis of asthma in humans (31); asthmatic children with Mycoplasma infection demonstrated a marked decline in T lymphocytes (32).

Conclusion
This study showed that the pigs from eight breed cross heterogeneous population reared under standardized housing conditions suffered lesion averagely covered 40% of lung, and the lung lesions were signi cantly linked to load of Mycoplasma in bronchoalveolar lavage uid. Much more signi cant correlation of Mycoplasma with lung lesion than other genera of bacteria observed hereby supporting Mycoplasma is a main genera of bacteria that could lead to exacerbation of lung lesion in pigs. Through association analysis with immune cell and hematological traits, we showed that evaluated load of Mycoplasma and deterioration of lung lesions consistently induced signi cant changes of hematological status including the increase of RDW-SD, NLR, and NEUR, and decrease of LYMR, MCHC, and LYM in pigs, processes indicating the in ammation, immune suppression and hypoxia status of host, these results further support the comorbidity of Mycoplasma proliferation and lung lesion. Moreover, we found that lung lesions were positively associated with CD8+/CD3 + cell ratio, and negatively associated with CD4-CD8-/CD3 + cell ratio, indicating the balanced differentiation of CD4 and CD8 T cells is important to present or alleviate lung lesion. This study provides useful information that may be helpful to the diagnoses of relevant respiratory disease in pigs.

Competing interest
The authors declare that they have no competing interest.