Heterozygosity at the A625C Polymorphic Site of the MyD88 Gene Is Associated with Mycobacterium bovis Infection in Cattle

Abstract

The study demonstrates that in cattle, animals heterozygous at the MyD88 A625C polymorphic marker have a 5-fold reduced risk for active pulmonary tuberculosis (odds ratio [OR] = 0.19; P = 6 × 10⁻¹²). The reduced risk, however, does not extend to animals with latent pulmonary tuberculosis (OR = 0.83; P = 0.40). Heterozygosity at the A625C single nucleotide polymorphism is associated with intermediate levels of tumor necrosis factor alpha, gamma interferon, and nitric oxide synthase (NOS). Accordingly, deficiency as well as overexpression of proinflammatory cytokines or NOS favor tuberculosis, while heterozygosity provides the animals with the optimal level of inflammation.

INTRODUCTION

The causative agent of bovine tuberculosis, Mycobacterium bovis, has a broad host range, which includes numerous wild and farm animal species. M. bovis is also pathogenic in humans. M. tuberculosis, the main agent of human tuberculosis, instead is nonpathogenic in cattle (1). This characteristic is attributed to differences in gene expression between the two bacterial species (2). In countries where programs for the eradication of bovine tuberculosis are operative, the periodic testing of cattle herds for tuberculosis infection, meat inspection, and milk pasteurization have reduced to <1% the cases of human tuberculosis attributable to M. bovis and confined them primarily to people infected with HIV or exposed to prolonged contact with animals (veterinarians or abattoir workers) (3). Nevertheless, bovine tuberculosis remains relevant as a zoonosis and because of the major economic losses that it causes to the cattle industry from the slaughter of infected—and often valuable—animals, quarantine of infected herds, and restrictions on animal export.

Innate and adaptive immune responses to mycobacteria rely on Toll-like receptors (TLRs), which sense several mycobacterial components. Sensing of the mycobacterial DNA requires TLR9, while heat shock protein 65 (HSP65) requires TLR4 and the lipomannan (LM), lipoarabinomannan (LAM), 19-kDa lipoprotein (19LP), and soluble tuberculosis factor (STF) require TLR2 (4). All TLRs (with the exception of TLR3) critically depend upon myeloid differentiation factor 88 (MyD88) to link bacterial recognition by TLRs with NF-κB activation and cytokine production (5). Evidence of the crucial role played by MyD88 as a signal transducer is provided by MyD88-knockout (MyD88^−/−) mice, which die within 4 weeks from the time of infection with M. tuberculosis (4, 5). MyD88 ^−/− mice infected with M. tuberculosis display reduced expression of gamma interferon (IFN-γ), tumor necrosis factor alpha (TNF-α), and nitric oxide synthase (NOS). This observation has suggested that MyD88 controls the infection by regulating the production of these mediators (6).The above-described studies and the high genetic similarity (99.95% identity at the nucleotide level) of the M. tuberculosis and M. bovis genomes (7) collectively provided biological plausibility to the hypothesis of a functional role of the MyD88 gene against bovine tuberculosis infection. The present study shows that heterozygosity at the MyD88 A625C polymorphic site is associated with resistance against active—but not latent—M. bovis infection in cattle.

MATERIALS AND METHODS

Diagnosis of pulmonary infection.

Postmortem samples were collected according to European Food Safety Authority (EFSA) recommendations (8). In the case of animals displaying macroscopic pulmonary lesions, a portion of the diseased tissue and afferent lymph node was collected. In the case of animals without visible lesions, the mediastinal, retropharyngeal, and bronchial lymph nodes were collected. Individual lung homogenates consisted of 1 g or more of pooled specimens collected from each animal. To distinguish between subjects with active tuberculosis pulmonary infection (ATI) or latent tuberculosis pulmonary infection (LTI), 10-fold dilutions (10⁻¹ to 10⁻⁸) of individual lung homogenates in sterile phosphate-buffered saline were spotted (10 μl/spot; 5 spots/dilution) on agar-Middlebrook (MB) medium and incubated at 37°C for 4 to 5 weeks. At the end of the incubation time, the numbers of CFU were counted. Negative samples were incubated for 10 days in liquid MB medium supplemented (5 μg/ml) with the mycobacterial resuscitation-promoting factor B (RpfB) (9), spotted on agar-MB medium, and incubated for 4 to 5 weeks, and the numbers of CFU were then counted. The optimal concentration of RpfB to use in the assay was found during preliminary experiments. The growth of colonies in the absence of RpfB was indicative of ATI, and the growth of colonies only in the presence of RpfB was indicative of LTI. Controls were negative by both tests.

Identification of mycobacterial species by PCR.

One colony of M. bovis was dispersed in 200 μl of distilled H₂O containing lysozyme (20 mg/ml; Sigma-Aldrich, St. Louis, MO) and incubated at 37°C for 2 h. After incubation, DNA was isolated using a DNeasy blood and tissue kit from Qiagen (Hilden, Germany). PCR was carried out as described previously (10).

Cases and controls.

The animals included in the study—both cases and controls—were from three herds declared to be infected. To exclude sex and age as potential confounders, the animals were all lactating cows between 40 and 90 months of age. This age interval was selected to represent subjects matched for age (as much as it was realistic) and, at the same time, a population sample sufficiently numerous to provide adequate power to the study. The average ages of the cases and controls were 65.4 ± 5.2 and 69.6 ± 3.9 months, respectively. To curb stratification, both cases and controls were from the same herds and the same breed (Friesian); to keep cases and controls genetically unrelated to each other, when mother and daughter were present, one of the two was excluded.

MyD88 genotyping.

The intron/exon boundaries of the bovine MyD88 gene were established by matching the published mRNA sequence of the bovine MyD88 gene (GenBank accession number NM_001014382.2) and the DNA sequence of the human MyD88 gene (GenBank accession number NC_000003.11). Alignment was carried out using DNAsis software (Hitachi Solutions America, San Francisco, CA). DNA was extracted from lung specimens with a QIAamp DNA kit (Qiagen, Hilden, Germany). The primers were 5′-TGAAGGAGTACCCCGCGC-3′ (forward) and 5′-GATGCCTGCCATGTCATT-3′ (reverse). Conditions of the PCR were 7 min at 97°C and then 45 s at 94°C, 30 s at 60°C, and 1.5 min at 72°C (35 cycles), with a final extension for 5 min at 72°C. The 1,210-bp fragments from 20 cases and 20 controls were sequenced using an ABI 3730 DNA analyzer (Applied Biosystems, Foster City, CA) and aligned by use of Chromas software (Technelysium, Queensland, Australia). The sequences were used to design primers and TaqMan probes targeting specifically the single nucleotide polymorphism (SNP) located 625 bp downstream of exon 1 (A625C). The forward and reverse unlabeled primers were 5′-GGTGGCGTGGTACTTTGC-3′ and 5′-TTTCTCCTCTACGGGCTGTCT-3′, respectively. The TaqMan VIC- and 6-carboxyfluorescein-labeled probes were 5′-TAGCAAGGGAGACATT-3′ and 5′-TAGCAAGGGCGACATT-3′, respectively, where the underlining and boldface indicate the polymorphic nucleotide. PCR conditions were 30 s at 60°C, 10 min at 95°C, and then 40 cycles each lasting 15 s at 95°C and 1 min at 60°C. Genotyping was carried out with the investigator blinded to the case or control status of the animals being tested.

TaqMan gene expression assay.

TNF-α, IFN-γ, and NOS2 mRNA levels in the lung specimens were measured using the TaqMan gene expression assay and a StepOne instrument (Applied Biosystems, Foster City, CA). Total RNA (2 μg) was reverse transcribed using a High Capacity cDNA reverse transcription kit (Applied Biosystems, Foster City, CA). The real-time quantitative PCRs were carried out following the manufacturer's protocol. The identification numbers of the probes are Bt03259155_g1 (TNF-α bovine), Bt03212722_g1 (IFN-γ), Bt03249602_g1 (NOS2), and Bt03279175_g1 (β-actin). Five animals for each genotype (AA, AC, CC) and class (control, active and latent tuberculosis) were tested in triplicate. Relative sample quantification was carried out by the comparative 2^−ΔΔCT method (where C_T represents the threshold cycle). The endogenous control gene was β-actin. The amplification efficiency of target (TNF-α, IFN-γ, and NOS) and reference (β-actin) genes was approximately the same (slope < 0.1).

Sample size calculation.

The data for 50 cases with active tuberculosis and 50 controls (odds ratio [OR], 0.3; proportion of controls with susceptible genotype, 0.46) showed that a sample of 127 cases and 127 controls would provide 96% power and a two-sided significance level of 0.01. The study enrolled 150 animals with acute tuberculosis, 150 animals with latent tuberculosis, and 300 controls.

Other methods.

Fisher's exact test and analysis of variance with the Tukey post hoc test were performed with GraphPad Prism software, version 5. Binary logistic regression was performed with the SPSS statistical package, version 18. Hardy-Weinberg equilibrium and relative risk reduction were calculated as described previously (see references 11 and 12, respectively). Conservation analysis was carried out on 11 MyD88 genomic sequences retrieved from the Nucleotide BLAST database (blast.ncbi.nlm.nih.gov/). The alignment was performed using the T-COFFEE multiple-sequence alignment server (tcoffee.crg.cat/) and map visualized with CLC Main Workbench software, version 6.8.2. The bovine intron 1 sequence was scanned for identification of overrepresented motifs by using the SCOPE (Suite for Computational identification of Promoter Elements) motif finder (http://genie.dartmouth.edu/scope/). Regulatory elements were searched for using the Encyclopedia of DNA Elements (ENCODE; http://encodeproject.org/ENCODE/).

RESULTS

Diagnosis of cases and controls.

Pulmonary tuberculosis infection can be active (ATI) or latent (LTI); the latter is characterized by the presence of dormant bacteria (viable but not culturable on usual growth media) (13). The methods commonly used to diagnose latent tuberculosis are the tuberculin skin test (TST) or the IFN-γ release assay. However, these methods do not distinguish between hosts still infected and those which successfully controlled infection (14). In the present study, grouping together different phenotypes would sensibly reduce the power of the study (15). M. tuberculosis has 5 resuscitation-promoting factor (rpf) genes coding for as many redundant proteins (RpfA to RpfE) which, in the form of recombinant proteins expressed in Escherichia coli, induce resuscitation of M. tuberculosis (16) and M. marinum (17) in vitro and ex vivo. On the basis of these findings, an in-house assay aimed at resuscitating dormant mycobacteria with the RpfB protein was developed. It was possible to recover dormant M. bovis from seven milk and seven lung specimens from animals treated with the RpfB protein but not from any of the specimens from animals untreated with RpfB when 20 of the animals included in the study were tested. The results for milk and lung specimens from the 20 animals were fully concordant. This material was used to validate the method. The test was therefore extended to all the animals, using lung specimens collected postmortem. A PCR assay discriminating between M. tuberculosis, M. bovis, or M. avium established that all cases (with ATI or LTI) were infected with M. bovis. In conclusion, the cases with ATI were subjects positive by the PCR assay and the bacteriological test in the absence of RpfB; the cases with LTI were positive by the PCR assay and the bacteriological test in the presence of RpfB; controls were subjects exposed to M. bovis infection (since they were from the same herds that also supplied the cases) but free from infection (negative by the PCR assay and the bacteriological tests in the presence or absence of RpfB) (Fig. 1).

Fig 1.

Diagnostic criteria used to classify subjects into controls or subjects with ATI or LTI. Controls, subjects with and without RpfB negative by culture and PCR tests; ATI, subjects without RpfB positive by culture and PCR tests; LTI, subjects with RpfB positive by culture and PCR tests.

Study design.

The criticism more often leveled at association studies is that they lack reproducibility (18, 19). To curb this drawback, a two-stage study was designed. The preliminary (hypothesis-generating) stage involved 50 control animals, which were separately confronted with 50 cases with ATI or 50 cases with LTI. This preliminary study displayed a significant association of the MyD88 polymorphic site A625C with ATI (P = 0.01; Table 1) but not with LTI (P = 0.84; Table 1). The A625C polymorphic site is located in intron 1 of the MyD88 gene (Fig. 2). The study also yielded the following valuable data: first, that the association is potentially robust (since it was detected using a small number of subjects) and, second, that case stratification according to the form (active or latent) of the infection would definitively provide more power to the study. Other than A625C, the SNPs shown in the reference sequence (Fig. 2) were not present in the sample population studied.

Table 1.

Heterozygosity at the A625C SNP influences active pulmonary tuberculosis infection

Study stage	TB typea	Status	No. of cows with the following genotype:				χ2b	AC vs AA		AC vs (AA + CC)
			AA	AC	CC	Total		OR (CIc)	P	OR (CI)	P
1	Active	Cases	36	10	4	50	5.2	0.30 (0.12–0.74)	0.001	0.29 (0.12–0.71)	0.01
		Controls	25	23	2	50	1.4
	Latent	Cases	28	21	1	50	1.7	0.81 (0.36–1.81)	0.68	0.85 (0.38–1.87)	0.84
		Controls	25	23	2	50	1.4
2	Active	Cases	123	23	4	150	4.4	0.19 (0.11–0.32)	6 × 10−12	0.22 (0.12–0.37)	1.8 × 10−10
		Controls	140	135	25	300	0.9
	Latent	Cases	75	60	15	150	0.3	0.83 (0.53–1.28)	0.40	0.81 (0.53–1.23)	0.36
		Controls	140	135	25	300	0.9

^aTB type, type of M. bovis pulmonary tuberculosis infection.

^bχ²_0.05 (1 degree of freedom) = 3.8.

^cCI, 95% confidence interval.

Fig 2.

Position of the A625C SNP on the MyD88 gene. The map is oriented 5′ to 3′. Source: www.ncbi.nlm.nih.gov/gene/444881.

To explore the functional role of A625C further, 11 Myd88 genomic sequences were analyzed for conservation across species (Fig. 3). The low level of conservation suggested that the A625C position is not under stringent selection. Scanning of the bovine intron 1 sequence with SCOPE highlighted 4 overrepresented nucleotide motifs, which included the polymorphic site (Table 2), evoking a possible regulatory role of the site. However, ENCODE did not find regulatory elements within intron 1 in the bovine MyD88, except for a CpG track. The analysis, repeated on intron 1 of the human MyD88, tracked the transcription factor binding site V$BACH1_01 and the chromatin immunoprecipitation (ChIP) fragment for RNA polymerase II. Interestingly, both these tracks overlap the bovine A625C site. Collectively, the data invited further investigation of the potential influence of A625C on M. bovis infection.

Fig 3.

Alignment and conservation analysis of 11 mammalian MyD88 sequences within a region surrounding the polymorphic site under study (marked with a black line).

Table 2.

Nucleotide motifs overrepresented in the bovine intron 1 sequence^a

Sequenceb	Searched consensus sequence	Counts within the Bos taurus genome
GGGTAGCAAGGGAGACA	GGGNVNVDDDSSHSACA	5
GGGAGACATTGGAGACA	GGGNVNVDDDSSHSACA	5
AGGGA	AGGGA	7
AGGGAGACAT	AGGGRBVCAT	3

^aMotifs were identified using the SCOPE motif finder (http://genie.dartmouth.edu/scope/). All strands were plus strands, and 100% coverage was achieved for all sequences.

^bThe polymorphic nucleotide is underlined.

MyD88 heterozygosity and resistance to ATI.

The study was repeated on a larger and independent sample consisting of 300 controls, 150 cases with ATI, and 150 cases with LTI. A separate experiment (with animals not included in the case-control study) showed that crosses between subjects homozygous for the A or C factor (AA × CC) yielded only heterozygous (AC) offspring. The experiment proved that A and C are transmitted as codominant alleles (data not shown). Cases with ATI were not in Hardy-Weinberg equilibrium (χ² = 4.4; Table 1). When the test was repeated on the cases with LTI, both cases and controls were in equilibrium (χ² for controls = 0.9; χ² for cases = 0.3; Table 1). The results suggested an association of the MyD88 marker with ATI but not with LTI. First, the more cogent Fisher's exact test showed that heterozygosity (the AC status) is strongly associated with resistance to ATI (OR = 0.19, P = 6.0 × 10⁻¹²; Table 1); second, the association remained strong when the homozygous classes were pooled (OR = 0.22; P = 1.8 × 10⁻¹⁰; Table 1); third, the MyD88 marker did not influence the predisposition to LTI (OR = 0.81 and 0.83; P = 0.36 and 0.40; Table 1). The binary logistic regression test supported these conclusions (Table 3). Given the frequency of the AC heterozygotes among controls (135/300 = 0.45; Table 1) and the level of protection afforded (OR = 0.19; Table 1), this genotype prevented 36% [0.45 × (1 − 0.19) = 0.36] of the potential cases of ATI in the population examined (12).

Table 3.

Heterozygosity at the A625C SNP and resistance to active pulmonary tuberculosis shown by binary logistic regression

Reference genotype	TBa	Binary logistic regression analysis result
		Wald	P	eb	H-L Pc
AA	Active	40	1.8 × 10−10	0.19	1
	Latent	0.78	0.37	0.83	1
CC	Active	0.01	0.91	1	1
	Latent	0.68	0.40	0.74	1

^aTB, M. bovis pulmonary tuberculosis infection.

^bOdds ratios estimated by the binary analyses.

^cThe nonsignificance of the Hosmer-Lemeshow (H-L) P value indicates that the model predicted by the logistic regression fits the observed data.

MyD88 heterozygosity and inflammation.

TNF-α, IFN-γ, and NOS are known to profoundly influence tuberculosis (6). It is also known that high as well as low levels of inflammation negatively impact this disease (4, 5, 20). Thus, if the MyD88 heterozygotes displayed intermediate cytokine levels compared to those of homozygotes, the association between A625C heterozygosity and resistance to M. bovis infection would gain strong biological plausibility. To test this hypothesis, the levels of TNF-α, IFN-γ, and NOS of subjects with different genotypes (AA, AC, CC) and status (controls or animals with ATI or LTI) (6 classes; 5 animals/class) were measured. The expression levels of the subjects with ATI or LTI were then compared with those of control subjects having the same genotype. Heterozygous carriers expressed levels of TNF-α, IFN-γ, and NOS significantly lower than those expressed by the AA homozygotes. Instead, carriers expressed levels only slightly higher than those expressed by the CC homozygotes; in this case, the difference did not reach statistical significance (Fig. 4). One possible explanation for this heterogeneity is that the technique used does not discriminate below a threshold level. Taken together, the data support the conclusion that an optimal inflammatory response is associated with the intermediate A625C phenotype.

Fig 4.

Proinflammatory cytokine mRNA levels measured by the TaqMan gene expression assay. Specimens were from controls or animals with active (ATI) or latent (LTI) M. bovis infection. Animals were grouped according to genotype (AA, AC, or CC) and class (controls, subjects with ATI or LTI). Each group consisted of 5 subjects. (A to C) Levels of TNF-α, IFN-γ, and NOS mRNA expression, respectively. Relative sample quantification was carried out by the comparative 2^−ΔΔCT method. The endogenous control was the β-actin gene.

DISCUSSION

The present study demonstrates that in cattle, animals heterozygous at the MyD88 A625C polymorphic marker benefit from a 5-fold reduced risk for ATI (OR = 0.19; P = 6 × 10⁻¹²; Table 1). The reduced risk, however, does not extend to the animals with LTI (OR = 0.83; P = 0.40; Table 1). Heterozygosity at the A625C SNP is associated with intermediate levels of IFN-γ, TNF-α, and NOS (Fig. 4). What is the biological advantage of an intermediate level of production of these mediators in the case of active tuberculosis? The short answer is that heterozygosity provides the optimal level of inflammation. The deficiency of IFN-γ, TNF-α, or NOS favors tuberculosis (4, 5). At the same time, some symptoms of the disease are known to be caused by the immune response of the host rather than by the mycobacterium (20). Episodes of disease reactivation and inflammatory syndrome related to preexisting M. tuberculosis (21) or M. avium (22) infection have been observed in HIV-coinfected patients after antiretroviral therapy. The study also displays differences in cytokine expression among animals of the same genotype with acute or latent tuberculosis. This difference is particularly evident in the case of the AA animals (Fig. 4). Whether the differences are caused by the mycobacterium or the host immune response, these results, though preliminary, point to increased levels of proinflammatory cytokine expression as potential markers of disease reactivation. The A625C polymorphism—located in intron 1 of the Myd88 gene—adds evidence to the notion that noncoding regions can influence gene expression. It is not surprising that this also occurs in the case of inflammation, which needs to be under fine and complex regulation.

In cattle, exposure to environmental mycobacteria, which occurs in the majority of the subjects, interferes with the diagnosis of M. bovis infection by the tuberculin skin test (TST) or the IFN-γ assay (23). Variability in the reagents, incubation time, and diagnostic cutoff levels also influence the specificity and sensitivity of these assays (24). The postmortem culture test—still the “gold standard” method (25)—was therefore preferred for the diagnosis of infection. Also, the limits of the TST and IFN-γ assays and—on the other side—the high prevalence of M. bovis infection among the enrolled animals (150 subjects with acute infection and as many with latent infection out of approximately 650 randomly tested animals) persuaded the authors that the number of false-positive and false-negative results would be better minimized by assuming that all controls were exposed subjects, rather than relying on the TST or the IFN-γ assay for exposure diagnosis. The authors do not claim that the method adopted here is superior to current methods in general; rather, they trust that it yields a better-defined disease spectrum and more reproducible results under a case-control design.

Tuberculosis is influenced by several genes interacting among themselves (26) and with the environment (15). The presence of the mycobacterium is necessary but not sufficient to acquire the disease, as shown by the control subjects, which, exposed to the pathogen, did not acquire the disease (Table 1); see also the work of Diamond (27). Environmental factors (climate, herd size, animal purchases, cattle movements) are known to promote bovine tuberculosis (2). Even strong genetic effects on M. tuberculosis can be missed when environmental effects are not taken into account (15). We claim that the unusually small OR and P values (OR = 0.19; P = 6.0 × 10⁻¹²) reported in the present study reflect how the problems confronting the genetic analysis of this complex disease were solved. Cases were made homogeneous (active and latent tuberculosis cases were analyzed separately), and the environmental confounders were either excluded (sex and breed) or randomized (age). More importantly, control subjects were from the same source population as the cases. Controls were therefore subjects that remained infection free (negative by the bacteriological and PCR tests), though they had the same opportunity as the cases to become infected. Population stratification often has been claimed to be responsible for false-positive results in association studies, yet rarely has it been demonstrated to be the culprit (28, 29). Human studies have shown that stratification might originate when different ethnicities are mixed (30). In the present study, only one breed was studied. Furthermore, replication of the association across 2 independent population samples argues against the result being a product of population stratification.

Genetic association studies are characterized by a high rate of false-positive results (29). This condition is often due to the selection of a candidate gene without a functional relation to the disease (29, 31). In the present study, MyD88 was selected on the basis of a large body of experimental data showing that—at least in mice—this gene is critical for signaling downstream the presence of mycobacterial components and inducing the production of the innate response mediators (IFN-γ, TNF-α, and NOS) against mycobacteria (4, 5). Further, the two-stage study design allowed the reproducibility of the association to be directly proved. Replication of the results at the time that they are first described is gaining consensus as an approach for reducing the number of false-positive results (28, 32). The two-stage design was also of value to define the precise phenotype (active versus latent M. bovis infection) to study (Table 1). In conclusion, the high biological relevance of the gene to study, the accurate choice of diagnostic criteria, and randomization of environmental confounders were all carefully kept in mind during the present journey in the puzzling field of association studies. However, since the association is being described for the first time, the results of this study are presented as preliminary.

Lastly, the test used here to differentiate between acute and latent disease could potentially be extended to the periodic testing of cattle for tuberculosis. The count of dormant mycobacteria awakened by RpfB in milk samples would be an easy way to know the prevalence of latent tuberculosis in the population, a parameter greatly influencing the control of the pathogen.