Study area

The study was carried out in DNP, located in south-western Spain (37°0' N, 6°30' W) and covering 54,000 Ha. This is a flat region of sandy soils bordering the Atlantic Ocean, with a maximum elevation of 47 m. The climate is Mediterranean sub-humid with marked seasons. In the wet season (winter and spring), most of the marshlands are flooded and wildlife and cattle tend to graze in the more elevated scrublands [37]. In summer, the wetter and more productive ecotone between the scrublands and the marshes supports aggregations of wild and domestic ungulates. Human access is restricted and management is carried out by Park authorities. Limited traditional exploitation of some natural resources, such as logging, and cattle and horse rising are allowed. After 1994, when bTB in wildlife was first diagnosed in DNP a Government-sponsored program was initiated to eradicate bTB-positive cattle. Ungulate populations have been culled by shooting (between 200 and 500 individuals/year, the majority of them wild boar, or about 10-20% of the wild ungulate population estimated at 3,500 individuals).

Animal sampling

From April 2006 to April 2007, 124 European wild boar, 95 red deer, and 100 fallow deer were sampled within the park by shooting. The culling of wild ungulates was approved by the Research Commission of Doñana National Park in accordance with management rules established by the Autonomous Government of Andalucía. For each animal we recorded the exact position with GPS. Sex and age, based on tooth eruption patterns (animals less than 12 months old were classified as juveniles, those between 12 and 24 months as yearlings, and those more than 2 years old as adults; [38]), were recorded in the field. A necropsy was performed on site and the presence of tuberculosis-like lesions recorded by macroscopic inspection of lymph nodes and abdominal and thoracic organs [6]. This protocol included the examination of the lungs for the presence of TB-compatible macroscopic lesions during field inspection and a sample was collected. A tonsil and a head lymph node sample from each individual were collected for culture (Figure 1; Table 1). In wild boar, one piece of the tonsils and one from both mandibular lymph nodes were submitted for microbiological studies. In red deer and fallow deer, one piece of the tonsils and head lymphnode samples, always containing at least half left and half right medial retropharyngeal lymph node, were submitted for culture. Due to logistic and budget constraints, no thoracic or abdominal lymphoid tissues were cultured except when TB-compatible macroscopic lesions were evidenced.

			Mycobacteria Other Than Tuberculosis (MOTT)					Mycobacterium bovis
Host	Site	n	M. scr.	M. int.	M. xen.	M. int.	Total MOTT	A1	A3	B2	B5	C1	D4	E1	F1	Total M. bovis
Wild boar	CR	14						12							1	13
	SO	18	3				3	8						2		10
	EB	31	2	6		3	11	5		2						7
	PU	29	1			5	6	7		12						19
	MA	32						5		7	1					13
	Total	124	6	6		8	20	37		21	1			2	1	62
Red deer	CR	35						8	1				1			10
	SO	35	6			1	7	8				1				9
	EB	12				1	1	2		1						3
	PU	3						1		1						2
	MA	10				1	1
	Total	95	6			3	9	19	1	2		1	1			24
Fallow deer	CR	36	2				2	7						1		8
	SO	35	9		1		10	8					2			10
	EB	9	3			1	4	2								2
	PU	5	2				2	1								1
	MA	15
	Total	100	16		1	1	18	18						3		21
	TOTAL	319	28	6	1	12	47	74	1	23	1	1	1	5	1	107

M. scr. = Mycobacterium scrofulaceum; M. int. = Mycobacterium interjectum, M. xen. = Mycobacterium xenopi, M. int. = Mycobacterium intracellulare

	MOTT pos		MOTT neg
Host	M. bovispos	M. bovisneg	M. bovispos	M. bovisneg
Red deer	1	8	26	60
Fallow deer	3	15	19	63
Wild boar	4	16	57	47

Social groups were defined as animals sampled the same day at the same site, and with characteristics that were compatible with forming a stable (e.g. female-yearling) or seasonal (e.g. rut mixed) group. Only part of the individuals belonging to a given social group was sampled. Sampling was performed according to European (86/609) and Spanish laws (RD 223/1988; RD 1021/2005), and current guidelines for ethical use of animals in research (ASAB, 2006) and UCLM animal experimentation committee.

Microbiological procedures

Culture: The specimens were liquefied and decontaminated with an equal volume of N-acetyl-L-cysteine and 2% NaOH, mixed by centrifugal swirling, and then incubated for 15 min (24). The reaction was neutralized by adding 0.0067M phosphate-buffered saline (pH 6.8), to a final volume of 50 mL. The specimens were concentrated by centrifugation at 3,000 × g for 15 min. The supernatant was discarded, and the sediment was re-suspended in 0.5 mL of sterile water. The sediment was used to inoculate two Löwestein-Jensen with pyruvate solid medium. Lowëstein-Jenssen slants were incubated at 37°C for 6 weeks and inspected weekly for growth. When growth was detected, a smear was prepared to confirm the presence of acid-fast bacilli from suspect colonies by Ziehl-Neelsen staining.

Identification

We identified M. bovis and MOTT to the species level and characterized M. bovis strains with spoligotyping and MIRU-VNTR typing.

Macroscopic morphology of the colonies and pigment production was recorded. Identification at species level was performed with the GenoType^®MTBC (Haim lifescience GmbH, Germany) for the Mycobacterium complex strains that allows the differentiation of M. africanum I, M. bovis BCG, M. bovis ssp. bovis, M. bovis ssp. caprae and M. tuberculosis/M. africanum II/M. canettii. MOTT strains were identified by the GenoType^® Mycobacterium CM and Genotype^® Mycobacterium AS MTBC (Haim lifescience GmbH, Germany). The GenoType assays were performed according to the manufacturer's instructions: DNA extraction by the DNA SSS method (REAL, DURVIZ, Valencia, Spain) was followed by PCR amplification of a trait of the 23S rRNA gene, as recommended. Reverse hybridization and detection were carried out on a shaking water bath (TwinCubator; Hain lifescience GmbH, Germany). The final identification was obtained by comparison of line probe patterns with the provided evaluation sheet [39].

Typing

The M. bovis isolates were further characterized by spoligotyping [40]. The amplified product was detected by hybridization of the biotin-labelled PCR product onto spoligotyping membrane (Isogen Bioscience BV, Maarssen, The Netherlands). Purified sterile water and chromosomal DNA of M. tuberculosis H37Rv and M. bovis BCG P3 were included as controls in each batch of tests. The patterns were allocated a number in the M. bovis spoligotyping database. The results were recorded in SB (spoligotype bovis) code, followed by a field of 4 digits as defined on the M. bovis Spoligotype Database website (http://www.mbovis.org).

All wildlife isolates (n = 107) were also subjected to MIRU-VNTR analysis (Table 1). Extensive documentation (online, Adobe PDF manual, and Flash tutorials) on the service and the genotyping methods is available at the MIRU-VNTRplus website (http://www.miru-vntrplus.org).

The analysis was carried out as described [41, 42], and primers previously reported were used to amplify the ETR-A and ETR-B loci (formally VNTR 2165 and VNTR 2461) [43], MIRU 4, 10, 16, 23, 26, 31 and 40 [14] using the recommended PCR conditions. Amplicon sizes were estimated by electrophoresis on a 1.5% agarose gel at 45 V during 2 h, using 100-bp ladder (Biotools B&M). Figure 2 presents the spoligotyping patterns, VNTR allelic profiles and typing pattern (TP) codes defined for this study.

Statistics

Considering the animals in which any mycobacterial infection was diagnosed, three generalized linear mixed models (GLMM, SAS 9.0 software, GLIMMIX procedure) were explored to test different explanatory variables that affect the presence of a mycobacterial type or group. The most common mycobacterial groups were: (i) M. bovis (ii) M bovis A1 and (iii) M. scrofulaceum. The presence or absence of infection in a mycobacterial group was considered as a binary variable. The model was fitted using a logit link function. The model considered social group as a random effect. The model included host species (wild boar, fallow deer and red deer), the study area and age (juvenile: less than 2 years, adult: older than 2 years) as categorical explanatory variables. The distance to the water (log₁₀-trasnformed) was included as a continuous predictor. To compare the spatial associations of infection by specific mycobacterial type and hosts, we included as explanatory continuous variable the ratio (log₁₀-transformed) between the nearest neighbor distance from host to a different host species with the same type of mycobacteria relative to the nearest distance to a con-specific host with the same type of mycobacteria (calculated using ArcGis version 9.2, ESRI, Redlands, CA). A ratio >1 indicates that the nearest distance to a host with the same spoligotype is higher for a different host species. All the aforementioned explanatory variables we also included in the models interacting with the host species. Due to over-parameterization of the models and zero inflated data, no interactions were included in the M. bovis A1 and M. scrofulaceum models. P-value was set as ≤ 0.05. We estimated exact confidence limits for prevalence (proportions) using Sterne's exact method.

Mycobacteria species and molecular types

We obtained a total of 154 mycobacterial isolates from DNP wildlife. This included 107 M. bovis isolates belonging to 8 different typing patterns (spoligotyping pattern + VNTR profile, TP), and 47 isolates belonging to four MOTT (Table 1). M. bovis TPs and MOTT species were isolated from wild boar (n = 82 isolates), red deer (n = 33 isolates), and fallow deer (n = 39 isolates) (Figure 3). Wild boar and red deer had 5 M. bovis TPs each, whereas fallow deer presented only 2 TPs. The number of different isolates per host (MOTT and M. bovis TPs combined) was 8 in wild boar, 7 in red deer and 5 in fallow deer (Table 1).

Regarding M. bovis, we identified 6 different spoligotyping patterns and 5 different VNTR allelic profiles (Figure 2). One spoligotyping pattern was new according to the M. bovis database, and was therefore introduced with code SB1610. Co-infection of a single host by two M. bovis TPs occurred in all three wild ungulate species. One adult male red deer was infected with TPs A1 and B2, one adult male and one adult female fallow deer were co-infected with TPs A1 and E1, and two wild boar (weaner and juvenile) were co-infected with TPs A1 and B2.

MOTT species found in wildlife hosts included M. scrofulaceum (28 isolates) and M. intracellulare (12 isolates), both found in all host species, M. interjectum (6 isolates, only in wild boar), and M. xenopi (1 isolate in a fallow deer; Table 1). In four deer and four wild boar, M. bovis and MOTT were isolated concurrently (6 M. scrofulaceum, 1 M. interjectum and 1 M. intracellulare). In a single wild boar, both types of mycobacteria were simultaneously isolated from the two tissue samples collected and cultured, while in the remaining cases M. bovis was isolated from either lymph nodes or tonsils and the MOTT from the tissue where M. bovis was absent. We recorded no cases of co-infection by different MOTT.

Changes in mycobacterial typing patterns over time in DNP

Spatial structure

Regarding the MOTT (Table 1, Figures 4 and 5), M. interjectum was only found in wild boar from EB, in the central part of DNP. In contrast, M. scrofulaceum was found in all three wildlife hosts (but not in cattle) in CR (2 isolates), SO (18), EB (5), and PU (3). The only MOTT found in cattle (one M. intracellulare isolate) was isolated from a cow raised in PU. M. intracellulare was often isolated from wild boar in PU and EB, and also from one fallow deer in EB and two red deer in SO and MA, respectively.

Table 5 shows the Czechanovsky similarities between the mycobacteria isolates in different sites and host species in DNP. For example, in column and row 1 from Table 5, the similarity indices of the CR mycobacterial community (in the north of DNP) decrease towards the south of the Park (MA; 20%; see also Figure 6). The highest similarity indices were observed between neighboring sites such as between EB and PU (89%) and MA and PU (75%). All hosts had their highest similarities with mycobacterial communities from the central sites of DNP.

	WBr	RDr	FDr	RDFDcr	WBFDcr	WBRDcr
WBcr	22			29
RDcr		25			29
FDcr			75			29

Factors affecting the presence of M. bovisTPs and MOTTs

Statistics concerning the GLMMs to test the factors affecting the presence of a given mycobacterial type or group are shown in Table 9. Concerning the M. bovis vs MOTT GLMM, the distance to water was statistically higher in MOTT infected individuals than in M. bovis ones (MOTT mean distance to water = 1989 ± 245 m; M. bovis mean distance to water ± SD = 1513 ± 164 m). The ratio of the minimum distances to similarly infected hosts (which in average were always higher than 1 for the three host species and analyzed mycobacterial groups) statistically interacted with the host. The ratios (log₁₀-trasnformed) were similar for MOTT and M. bovis in both deer species (2.13 ± 0.36 and 2.11 ± 0.32 for MOTT and M. bovis in red deer; 2.01 ± 0.11 and 1.95 ± 0.35 m for MOTT and M. bovis in fallow deer), whereas they were higher for M. bovis than MOTT in the wild boar (2.71 ± 0.36 and 3.55 ± 0.20 m for MOTT and M. bovis). This would indicate that in wild boar the intraspecific spatial aggregation of M. bovis is higher than for MOTT.

When attending to specific mycobacterial types, there were statistical differences between zones for bovis TP A1, so that it was dominant in wild ungulates from the north of DNP (Table 1, Figure 6). There were statistical differences in the probability of infection by M. scrofulaceum relative to other types among host species (wild boar = 7.3%; Red deer = 18.2% and fallow deer = 41.0%; Table 1). M. scrofulaceum presented a lower intraspecific spatial aggregation than the remaining mycobacterial types (2.19 ± 0.20 and 2.90 ± 0.15 m ratios for M. scrofulaceum and the remaining types, respectively).

Mycobacterial species and typing patterns

Contrary to most previous studies in wildlife, where single TPs tend to dominate in each geographical region [e.g. [19, 20, 45]] we detected a high richness of both MOTT and M. bovis TPs in DNP. Whereas single TPs are indicative of single introduction events of M. bovis, in our case the high identified TP richness is probably a consequence of (i) historical cattle breeding and consequent exchanges with breeders from outside the park, (ii) variable conditions provided by high environmental diversity, and (iii) the diversity and abundance of suitable wildlife hosts.

Multiple infection of a wildlife host with several M. bovis TPs had recently been found in one wild boar from this study area [32]. This observation is rare in wildlife M. bovis hosts [46]. To the best of our knowledge, this is the first study reporting co-infection of red deer and fallow deer with several M. bovis TPs. Moreover, the efficiency of isolating mycobacteria could have been improved with the inclusion of liquid media, suggesting that we detected only part of the true co-infections. The relevance of these findings is that they demonstrate that M. bovis infected wildlife hosts may become infected more than once under natural conditions, at least in areas of high infection pressure such as DNP. These results also suggest that cross-protection between different M. bovis strains is limited, further underlining the importance of genetic factors rather than immune responses in controlling mycobacterial infections in wildlife [11, 47, 48]. Additionally, the infection exclusion reported for closely related genotypes of other intracellular bacteria of the genus Anaplasma [49] did not appear to occur for M. bovis TPs.

Co-existence of members of the M. tuberculosis complex and MOTT, such as M. intracellulare, had already been reported in human patients [50]. As previously discussed, the fact that we found several M. bovis - MOTT co-infections suggests that infection by one organism does not impede infection by the other in these wildlife host species. However, in all three wildlife hosts, isolation of one group of mycobacteria occurred more frequently in individuals not infected by the other group, suggesting that either some competition between mycobacteria or some laboratory bias towards the first identifiable growth may exist. This is also suggested by the finding that in all cases but in one, M. bovis was isolated from either lymph nodes or tonsils and the MOTT from the tissue where M. bovis was absent. In humans, it has been suggested that BCG vaccination protects children against cervical lymph node infection by MOTT [27].

Several authors have reported infection of wild boar with M. scrofulaceum, M. interjectum, M. xenopi and M. intracellulare [51, 52]. All four MOTT recorded in this study had also already been reported in other wildlife species [18, 53]. However, this is the first report of M. xenopi in deer.

Changes over time in DNP

Apparently, the community of M. bovis in domestic cattle lost strain richness from time one (1998-2003) to time two (2006-2007), which may result from the application of the official test and slaughter program. However, the alternative hypothesis of some rare strains going undetected at any sampling period cannot be completely excluded. Part of the new TPs isolated from wildlife had been reported in cattle in the earlier survey (D4, F1). This suggests cases of spill-over from cattle to wild ungulates, and subsequent maintenance of these TPs in wildlife reservoir hosts. Other TPs had been detected neither in DNP cattle nor in wildlife, but are widespread in Spain (e.g. F1, SB0120). This would suggest a recent introduction, possibly via infected cattle. However, TP E1 is of particular interest. This TP had never been detected, but is similar to the dominant TP A1 except for one spacer. More sampling and long term studies are needed in order to test whether pathogen evolution resulted in higher TP richness in wildlife species when compared to cattle [32].

Spatial structure

Our finding that different wildlife species were infected with the same types at a very local scale suggests that transmission is likely to occur between the species. Fallow deer differed from red deer and wild boar in showing more homogeneity in their mycobacterial isolates, regardless of the sampling area. This may be due to a higher rate of movement of fallow deer between areas and therefore relates to specific territorial and aggregation behaviors as commented above. This in turn would be relevant for disease control, suggesting a higher capacity of this host for spreading pathogens over long distances. The different distribution patterns of M. bovis TPs may be due to historical introduction of different TPs, presumably by infected cattle, in different parts of DNP or, alternatively, if environmental survival of mycobacteria plays a role, to a better adaptation of certain TPs to the varying habitat characteristics of northern and southern DNP.

Factors affecting the presence of M. bovisTPs and MOTTs

In a previous paper we found that infection risk in wild boar was dependent on wild boar M. bovis prevalence in the buffer area containing interacting individuals. However, this was not evidenced for deer [21]. In this study, using the identification of social groups as an alternative tool, we found that red deer and wild boar were more prone to be infected with M. bovis if they were part of a group with at least one infected deer, while as commented above, this was not the case for fallow deer. High intraspecific transmission rates at early ages within wild boar social groups have been suggested in wild boar from Spain [6], and this probably relates to close interaction when foraging or routing. Animal behavior is an important aspect of disease/host dynamics that as yet has not been well documented but may play an important role in the transmission in free-ranging wildlife populations [33]. Owing to higher contact rates and common environmental risk factors, bTB transmission should occur more frequently within certain social groups. Recently, [1] used host population genetics to show that contact within family groups probably was a significant mechanism of M. bovis transmission among white-tailed deer (Odocoileus virginianus) in Michigan (USA). In DNP, modelling suggested that wild boar infection probability depends on wild boar bTB prevalence in a buffer zone of interacting individuals, while no such effect was observed in deer [21].

The fallow deer was the only species whose mycobacterial community showed more intra-species similarity throughout DNP than site similarity. Although fallow deer displayed the lowest prevalence (which is probably related to a lower natural host susceptibility, [21]), its highly gregarious behavior and subsequent increased transmission risk (at least during seasonal rutting) may cause mycobacterial strains to be shared by many social groups after social disruption. This is consistent with the finding that fallow deer displayed the lowest M. bovis prevalence, but a disproportionally high social group prevalence (i.e. spread across population subunits) as compared to that of red deer. That is, the findings that fallow deer belonging to groups with infected individuals were only rarely infected, or that most infected fallow deer groups had only one infected animal, strongly suggest that either the intra-specific intra-group transmission rate or the susceptibility of fallow deer to bTB is lower than in red deer. However, the alternative explanation that culturing from head lymphoid tissue only missed to detect infection disproportionally more in fallow deer than in red deer or wild boar cannot be excluded.

Confirming the above discussed, a spatial structuring in the mycobacterial isolates was evidenced for M. bovis A1 type, so that it was dominant in wild ungulates from the north of DNP while B2 was dominant in the south (Table 1, Figure 6). When we assessed the spatial associations (measured as nearest distances to similar and other host species) of M. bovis TPs, MOTT, and M. scrofulaceum, our findings were consistent with spatial aggregation of the host species with the same types.

The spatial distribution of M. bovis versus MOTT was probably linked with the water, since wildlife hosts infected with M. bovis were most often sampled closer to the marshland than MOTT. Environmental water sources could act not only as environmental sources of mycobacteria but also by favoring closer contact between the species [7], and this could promote more the transmission of M. bovis by close contact than indirect transmission of MOTT, which one would expect to be more dependent on external factors.

There were statistical differences in the probability of infection by M. scrofulaceum relative to other types among host species. M scrofulaceum is a slow-growing atypical mycobacteria that is found in environmental water sources. Nonetheless, no association was evidenced with distance to marshland. We speculate that the rooting behavior of wild boar may relate to increased exposure to this mycobacteria than other hosts. Nonetheless, our study does not discard that advanced host species-pathogen interactions may also result in different relative occurrences of mycobacterial types across the studied host species.