Infectious abortion in ruminants is a problem in animal husbandry worldwide. Its importance is related not only to economic loss in animal production but also to infectious risks posed to humans and other animals [1, 2].

The most common bacterial abortive agents involved in ruminant abortion are Brucella spp., Coxiella burnetii and Chlamydia abortus [2–4]. While Brucella spp. and C. burnetii are typically involved in bovine fertility problems as abortion, stillbirth and weak offspring [5], chlamydial infections may cause a variety of syndromes including conjunctivitis, polyarthritis, encephalomyelitis, mastitis and other urogenital tract infections [2]. In humans, brucellosis, Q fever (C. burnetii) and chlamydiosis should be considered among the most common zoonotic diseases around the world [2, 5]. Moreover, there is increasing evidence supporting the implication in bovine abortion of other Chlamydia-related bacteria [2, 4, 6]. Other bacterial agents sporadically associated with bovine abortion that can also cause serious zoonotic diseases are Salmonella spp., Campylobacter spp., Leptospira spp. and Listeria monocytogenes [2, 5]. However, many of the bacterial causes of abortion involve opportunistic pathogens and often remain undetected. These organisms are common inhabitants in the host and environment and can occasionally enter the blood stream of the dam, subsequently infecting the placenta and producing sporadic abortion [7].

Laboratories conducting abortion diagnostic examinations should perform standard tests covering the major abortive infectious diseases; however, costs dictate that these tests are limited to the most common etiologies [8]. For example, first-line routine bacteriological abortion diagnostics in cattle in Switzerland only include serology and staining for B. abortus and C. burnetii (Swiss ordinance on epizootic diseases, article 129). In general, success rates for abortion diagnoses are low. Rates ranging from 23.3 to 45.5% were reported in the United States and a diagnostic rate of 22.5% in England, Wales and Scotland [9, 10]. While positive results for the before-mentioned diseases are notifiable and gathered in a national database, this does not imply causality of an abortion and, hence, no true diagnostic rates can be reported from Switzerland. In addition, problems such as inappropriate sample collection and submission, incomplete case history, environmental contamination and/or poor condition of the fetus may further hamper achieving an etiological diagnosis [7]. The placenta is considered the most useful sample; however, after abortion, the placenta is exposed to several environmental contaminants and detection of an agent in the placenta only does not imply that it actually passed on to the fetus. Thus, fetal abomasal fluid and organs are also required for culture and molecular detection [5].

Increasing knowledge of the bacteria that are involved in cattle abortion is crucial for optimizing the diagnostic approach and revealing emerging pathogens. Culture-independent DNA technology based on sequencing of the rrs gene encoding 16S rRNA has the potential to uncover both known and novel microorganisms [11]. Thus far, no study has focused on the bacterial microbiota present in bovine abortion material.

The objective of this study was to characterize the microbiota of samples from cattle abortions using high throughput sequencing of the V3–V5 region of the 16S rRNA gene to provide new insight into the bacteria that may play a role in bovine abortion.

Studies on the bovine microbiota have gained importance in the last 5 years with the characterization of the microbiota of the rumen [21–24], the complete gastrointestinal tract [25], the nasopharynx [26], milk [27–29], the teat [30] and the vagina and/or uterus [31–34]. In contrast to previous studies that focused on implicating specific bacteria from bovine abortion material by bacterial culture and targeted molecular detection [4, 6, 35–38], we studied the total bacterial microbiota present in abortion material from cattle using a 16S rRNA amplicon sequencing approach. In addition, we used a broad-spectrum bacterial culture to compare and confirm the sequencing results.

This study revealed that the taxonomic groups present in the bovine abortion material were Proteobacteria (AC = 87.35%; PL = 72.13%), Firmicutes (AC = 10.51%; PL = 15.66%) and Bacteroidetes (AC = 1.99%; PL = 7.81%). In a previous publication, the microbiota from the abomasal mucosa of healthy dairy cattle exhibited relative abundances of Firmicutes (27.4%), Bacteroidetes (20.95%) and Proteobacteria (19.82%) [25]. The composition profiles from the placenta and abomasal content showed significant differences (p = 0.0001), with the placenta exhibiting more taxa, while some phyla were not present in the abomasal content. This is not surprising since the placenta is often contaminated with bacteria from the vagina and vulva as well as the environment [7, 39]. Therefore, the infectious process of specific abortigenic bacteria, such as Chlamydia spp., may be confined to the placenta, which makes this organ indispensable to thoroughly investigate the abortion material [39, 40].

Figure 5 shows that the microbiota varied between the different abortion cases pointing at a possible causative genus. Some cases showed one prevalent taxonomic unit (one red square present) with many reads associated with lesions in the placenta, whereas other cases had profiles that were more equally distributed (presenting several squares from light blue to red) and were devoid of histological lesions. After interpretation of results, it was possible to reveal different taxonomic profiles showing two principal components accounting for 53% of the total data variability and thereby clearly splitting autolytic samples from those with infectious bacterial causes on the X-axis (Figure 6). These observations followed the key points for establishing an etiological diagnosis and distinguishing the exogenous microbiota from a real fetal infection [40, 41]. These points include isolating the agent in a relatively pure culture and/or a large number from the fetal tissues and/or the placenta and the presence of gross and microscopic lesions consistent with a bacterial infection in the placenta. Ideally, additional tests should exclude other common abortigenic agents including viruses, fungi and parasites. Our findings revealed opportunistic abortive agents that are not detected by the routine diagnostic approach that focuses on epizootic bacterial agents. Most of the bacterial agents that cause abortion in cattle are commensal or environmental bacteria that, as opportunistic pathogens, produce sporadic abortions [5]. These opportunistic organisms are usually categorized into two groups: (1) bacteria belonging to the commensal microbiota of the mucosal surfaces and (2) common environmental bacteria [39]. Most reports, including the broad-spectrum screening of bovine abortions, are based on bacterial cultures [42–44]. Culturing may be a powerful method for isolating bacteria if the conditions are compatible with the microbial target and sample type, but it may fail to identify novel or unculturable pathogens [45, 46]. As expected, the NGS method revealed many genera that were not evident from bacterial cultures. One of the advantages of the parallel culture-based approach is the ability to assign species-level identification and overcome the restricted taxonomic result of the NGS method that amplifies only a fragment of the 16S rRNA gene [47]. The following opportunists were identified by NGS and isolated in pure culture (Table 2): Vibrio (identified as V. metschnikovii in culture), Streptococcus (identified as S. uberis in culture), Lactococcus (identified as L. lactis in culture), Trueperella (identified as T. pyogenes in culture), unidentified Enterobacteriaceae (identified as E. coli in culture), and Campylobacter (identified as C. fetus subsp. fetus in selective culture). For V. metschnikovii, S. uberis, L. lactis and E. coli, tissue was available for histopathological analysis, and we confirmed lesions in the placentas. In the cases with placental lesions in which a possible agent was only detected by NGS with a high number (≥70%) of reads, we detected unidentified Pseudomonas and unidentified Aeromonadaceae.

Vibrio metschnikovii has been identified by bacteriological culture in cattle, swine and horse abortions [48], and L. lactis, which is mainly associated with bovine mastitis [49], has also been isolated from a bovine abortion [43]. It is likely that the suspected abortive agents, unidentified Pseudomonas, S. uberis and E. coli, may spread via the amniotic-oral route to the fetus and produce infection (Table 3); however, causality between these pathogens and abortions has not been confirmed [42, 50, 51]. Interestingly, Pseudomonas was the most prevalent genus revealed by the NGS approach and it was present in all the samples analyzed (Figure 5). The implication of Pseudomonadaceae in human periodontitis was previously revealed by NGS [52]. A previously unidentified member of the family Aeromonadaceae, which was producing placental lesions, was later identified by NGS. This suggests that culturing opportunistic environmental bacteria may be difficult. To our knowledge, only two previous studies, reported in 1972 [53] and 1993 [51], identified Aeromonas hydrophila and Aeromonas spp. as etiological agents of bovine abortion.

One pure culture of T. pyogenes partially matched the NGS results since the number of reads was lower (11.28%) than those of the most abundant genus, Fusobacterium (74.04%). T. pyogenes cooperates with environmental anaerobes, such as Fusobacterium necrophorum, to produce infections, such as endometritis in cows, and increases the possibility of uterine inflammatory conditions while intensifying disease symptoms [54–56]. Fusobacterium is known to cause sporadic bovine abortion [57, 58], but the impact of mixed infections on abortions has not yet been determined. Moreover, anaerobic bacteria may be underestimated as abortifacients since anaerobic cultures are not normally part of routine diagnostic procedures [41]. The NGS approach was also able to identify multiple possible abortive taxa from the same sample in 13Ue851_Pl and 13Ue1275_Pl (Table 3). Unidentified Enterobacteriaceae (46.46%) were found together with Enterobacteriales (24.33%) and Serratia (16.14%), and S. uberis (37.24%) were found alongside Enterobacteriaceae (25.92%) and Lactococcus (21.27%). As previously suggested, mixed infections may play an underestimated role in ruminant abortions [4].

It is often difficult to assess the presence of potentially infectious bacteria in cultures, particularly in cases where rare opportunistic agents, e.g., Lactococcus, Streptococcus or Vibrio, are identified, as they tend to be overlooked as contaminants. The amplicon sequencing approach however, gives additional information through the number of reads obtained [59]. This is especially noteworthy in the case of Pseudomonas that was not apparent in the culture. Another advantage is the ability to simultaneously screen for the broadest possible bacterial variety without requiring a costly detection method, such as qPCR, DNA arrays or antibody detection [59]. Interpretation of the NGS results must be done cautiously though, keeping in mind that further analyses are needed to determine causality. Another challenge in investigating cattle abortion is that the sample site may not correspond with the location of the active infection [60]. Additionally, depending on the DNA extraction method and the amplified variable regions of the 16S rRNA, some bacteria may be over- or underrepresented [61, 62]; e.g., we did not detect Chlamydia, Leptospira or Coxiella in the sequencing results, whereas these bacteria have been reported in bovine abortions in Switzerland [4]. This might be due to a lower affinity of the primers or a lower taxonomic resolution of the amplified fragment to the specific 16S rRNA gene in these pathogens. These limitations of the amplicon sequencing approach have previously been discussed in view of its suitability for diagnostics [45].

Histopathological analysis of tissue with a high degree of autolysis is not recommended due to loss of tissue architecture and cellular detail. NGS, however, allowed us to associate a specific microbiota to the autolytic tissue. Unlike the infectious cases, the autolytic placentas did not show a dominant taxon. In these cases, a variety of bacterial genera, represented at a similar percentage of reads, appeared to be involved in the tissue spoilage (Table 3, Figure 6). Most of these bacterial genera are also causative agents of meat spoilage [63, 64]. Moreover, a high grade of contamination with commensal microbiota from the gastrointestinal tract, such as the families Rickenellaceae, Ruminococcaceae, Peptostreptococacceae, Enterobacteriaceae or the order Clostridiales [25] was detected in some samples. The presence of other environmental or/and commensal bacteria can lower the ability to accurately detect an etiological agent as has been described for the parasite Tritrichomonas foetus [65]. Our results highlight the importance of adequately preserving the samples in sterile, chilled containers and rapidly transporting them to the diagnostic laboratory to avoid autolysis and contamination.

Although amplicon sequencing has been discussed as a diagnostic approach [60], the aim of our study was not the evaluation of NGS as a diagnostic tool but as a novel research approach to gain deeper insight into the microbiota present in abortion material. Our study emphasizes the difficulties of applying 16S amplicon sequencing to abortion diagnostics, such as the nature and suitability of the sample and the presence of contaminants. NGS helped us uncover abortifacients that went undetected by traditional methods and identify possible multi-infections. Nonetheless, standardization of workflows and cut-offs for diagnostic purposes based on the interpretation of percentage of reads is difficult to achieve and analysis would not be cost-efficient, especially since identification at species level cannot be achieved by NGS only but is required for an etiological diagnosis.

The targeted fragment of the 16S rRNA gene was not amplified in the three negative controls, indicating the absence of bacterial DNA; furthermore, no growth on the different culture media was observed. In our study, the negative control samples were extracted in the necropsy hall in aseptic conditions directly from the uterus of a dead cow having a minimal contact with the environment and no contact with the commensal microbiota from the vagina. Our ideal negative control sample would be a healthy fetus and the placenta extracted through the vaginal tract from an interrupted pregnancy of a healthy cow. However, from an animal welfare point of view this experimental setup would not be commensurate to the expected gain of knowledge and thus is not approved by the Swiss Federal Animal Protection Law (455, article 19, paragraph 4). A recent study reported the existence of a microbiota associated to the bovine placenta, but the authors could not exclude a possible contamination of the biopsies during sampling [66]. Moreover, the extraction of the mentioned biopsies was done through the vaginal tract of the live cows exposing the samples to the commensal microbiota. Although the authors took care to apply antimicrobial solution before and between sampling, this does not ensure elimination of DNA which then may persist, e.g. leading to the reported presence of the phyla Planctomycetes, a slow-growing decomposer of organic matter [67] and Euyarchaeota, a methanogenic archaea group from the rumen of cattle [68] in the placenta and the amniotic fluid. A previous work showed for the first time the presence of prenatal microbiota in human placentas [69]. However, a subsequent study concluded with a new sample set that the previous one could not distinguish between placental microbiota and contamination introduced during DNA purification [70]. Our findings indicate that under timely and sterile sampling conditions, bacterial microbiota is likely absent from these tissues.

Our study underlines the potential of amplicon sequencing to identify or confirm unknown etiological agents, such as unidentified Pseudomonas, S. uberis, L. lactis, V. metschnikovii and unidentified Aeromonadaceae. These new insights encourage adaptation of the diagnostic focus and extension of the spectrum of understudied opportunistic abortive bacteria.