Our results show that the current distribution of C. burnetii is the result of a few highly fit clones that appear to be largely confined to individual livestock species. The concept of distinct clades associated with species specific restrictions may explain the low apparent rate of clinical disease among human populations despite the high prevalence of these bacteria. Among our samples, two sequence types were highly prevalent: ST8 was exclusively found in samples from goats while ST20 dominated cow’s milk with only two examples of ST20 from goats. This pattern is consistent with other smaller studies where likely ST20 isolates (see below) were from cattle [21, 27, 28] and rarely from goats: a single ST20 sample attributed to a goat in France  and abortions in a large commercial dairy goat herd in the UK . Likewise, recent ST8 samples have been collected from sheep, goats and humans [21, 27, 30, 31]. This tendency for host restriction may be the result of a stochastic introduction into a large livestock population allowing for an increase in frequency, spread through trade, but constrained to that population through anthropogenic isolation of livestock species. However, as both genotypes show a tendency for host restriction and similar patterns are found in Europe [21, 27, 28, 30] as well as the USA, it seems more likely that these genotypes are evolutionarily adapted to certain host species.
Genotyping historical collections of C. burnetii has provided a baseline for environmental distribution of sequence types [17, 19, 20, 32]. Interestingly, contemporary sampling yields only a small subset of the known genotypes, many of which are found across multiple studies [21, 27, 28, 30] (Kersh et al., Genotypes of Coxiella burnetii strains found in the United States environment, 2006-2008, in preparation). In some cases, subtypes of the same MST genotypes were identified [27, 30, 33]. Consistent with these findings, our genotyping of milk samples revealed only three or four MST genotypes; while only two samples had unknown genotypes (and may both have the same genotype), the genotypes of all other samples are likely to be either ST20 or ST8. It is important to note that additional genotypes not detected by our sampling may be circulating at very low levels. A high proportion of recent milk, placenta, and mucus samples from goat, cow and sheep farms in Spain were ST20, but none were ST8 . Kersh et al. recently genotyped C. burnetii DNA from US environmental samples and found ST8, ST16/26, and ST20 genotypes. Samples associated with goats were ST8 and all ST20 samples came from cattle dairies (Kersh et al., Genotypes of Coxiella burnetii strains found in the United States environment, 2006-2008, in preparation). In the Netherlands, a study by Tilburg et al.  sampled ST20 from cattle and ST33 from humans, sheep and goats. Huijsmans et al.  also genotyped recent samples from the Netherlands, albeit not with MST. However, overlapping reference samples, the results from Tilburg et al.  and a comparison to the phylogenetic relationships of MST genotypes, suggests that the Huijsmans  genotypes 1, 2, 4, 6 and 8 are likely to be (or be closely related to) MST genotypes ST33, ST20, ST20, ST8 and ST18 respectively.
While likely ST8 samples have been associated with recent livestock and human clinical samples, such associations with likely ST20 samples are rare (for example see ) and it is not clear if any of the Spanish ST20 samples were from animals with clinical manifestations [21, 27, 28, 30]. From the recent outbreak in a UK dairy goat herd  and historical collections, it is clear that ST20 can cause disease in humans and livestock [19, 20]. The scarcity of ST20 among clinical samples, despite being the dominant genotype among cow milk samples, suggests that U.S. ST20 strains have a reduced ability to cause disease in humans or cause a very mild form.
Prevalence of C. burnetii on goat and cow farms has been previously assessed, but comparisons across studies are difficult due to different serological or DNA-based detection methods. Sampling individual animals, herds, or products pooled across herds also confounds comparisons although as expected, prevalence generally increases as bulk samples become inclusive of more individuals [6, 8, 13, 34–37]. Similarly, we found that milk from four of 20 sampled cows were positive while all 3 samples from the bulk milk holding tank (containing milk from 120 cows) were positive. Our milk samples from retail brands bottled in commercial processing plants likely include milk pooled from different (and much larger) dairy farms, making it impossible to know the extent and distribution of infections among cows and herds. However, our detection of C. burnetii DNA in every goat and cow milk sample from the same brands (i.e. processing plants) over time and >95% of milk samples from processing plants across the USA shows high prevalence at either or both the individual and herd levels. Indeed, the prevalence rate reported here is comparable to the high rates reported in other studies [8, 12, 13]. Notwithstanding existing immunity, infectious diseases are density dependent, leading us to suspect that the ratio of infected to uninfected cows on some farms may be greater than our single farm results. Nonetheless, while a small number of infected animals may contaminate a large quantity of milk, it is probable that a significant portion of the 9.2 million dairy cows in the USA  are infected with C. burnetii at any given time .
Across the ~2.5 year period of sample collection, there was no variation in prevalence of C. burnetii DNA in milk samples and almost no variation in genotypes detected. As processing plants receive milk from the same dairies over time, it is likely that the same herds and even the same animals were sampled multiple times. Major temporal changes in prevalence and genotypes should be detectable. Indeed, minor genotypes were detected among the goat milk samples, indicating ephemeral emergence of different types. Conversely, subtle changes may be masked by the milk pooling process and the ability of a single infected animal to contaminate large quantities of milk. Indeed, other studies suggest that there is evidence of seasonality: In cows, shedding in milk is not associated with parturition  although seroprevalence is highest in the Autumn . In goats, C. burnetii are highly abundant (up to 109 organisms/g of placental tissue) in birth tissues  and more likely to be shed after parturition . Human infections are therefore likely to be more common during livestock birthing seasons , suggesting that infection variation among goat herds might also be seasonally linked. Seasonality is often associated with a boom and bust cycle of transmission, and the lack of strong seasonal patterns may increase disease persistence.
As pathogens are dispersed across the landscape, elapsed time allows for cellular replication and opportunities for genetic mutations to accumulate, providing genetic signatures to identify the patterns and speed of dissemination. The presence of the same genotypes among samples from across the country and the world is indicative of rapid dispersal of particular gentoypes and subsequent ecological establishment across these regions. While a paucity of historical samples and sampling efforts prevents us from estimating when these STs became dominant, no ST20 isolates were collected in the U.S. before 2007 . Interestingly, the only U.S. C. burnetii samples isolated from milk with a known date were obtained from cows in California (1947) and Ohio (1958) . Both samples are ST16/26, showing that the dominant genotype among cows may have recently changed. Higher resolution genotyping will be important for discerning dissemination patterns and mechanisms of these C. burnetii genotypes as dispersal may be due to long distance aerosol spread, trade, or other anthropogenic means. For example, sexual transmission through semen  from the small stock of infected breeding bulls used to breed Holstein cows throughout the world could result in shared genotypes. However, additional resolution among ST20 and ST8 samples has been shown with MLVA  and demonstrates that dissemination speed and patterns may have allowed for the accumulation of genetic differences and thus discerning patterns, mechanisms and barriers to dispersal may be possible. Unfortunately, MLVA is not as well suited for genotyping low levels of target DNA and the little additional resolution afforded by MLVA is not likely to be sufficient for defining detailed patterns of dispersal for this pathogen. High infectivity, low virulence and ease of aerosolization coupled with the speed and global reach of modern trade has likely resulted in these complex and subtle patterns of dissemination that will be challenging to resolve. Whole genome sequencing will likely provide additional signatures that may prove to be our best hope for maximizing genetic resolution, untangling dispersal patterns and better estimating the speed and mechanisms of dispersal for C. burnetii.