Targeted Metagenomics Offers Insights into Potential Tick-Borne Pathogens

Tick-borne disease pathogen identification remains a diagnostic challenge due to the multiple tests necessary for diagnosis. Targeted metagenomic sequencing is an evolving method to detect multiple different pathogens with a single test. In this issue of the Journal of Clinical Microbiology, L. Kingry, S. Sheldon, S. Oatman, B. Pritt, et al. (J Clin Microbiol 58:e00147-20, 2020, https://doi.org/10.1128/JCM.00147-20) explore 16S rRNA gene-targeted metagenomics to identify bacterial pathogens in 881 of more than 13,000 specimens submitted for tick-borne pathogen testing, giving new insights into suspected tick-borne pathogens.

ABSTRACT

Tick-borne disease pathogen identification remains a diagnostic challenge due to the multiple tests necessary for diagnosis. Targeted metagenomic sequencing is an evolving method to detect multiple different pathogens with a single test. In this issue of the Journal of Clinical Microbiology, L. Kingry, S. Sheldon, S. Oatman, B. Pritt, et al. (J Clin Microbiol 58:e00147-20, 2020, https://doi.org/10.1128/JCM.00147-20) explore 16S rRNA gene-targeted metagenomics to identify bacterial pathogens in 881 of more than 13,000 specimens submitted for tick-borne pathogen testing, giving new insights into suspected tick-borne pathogens.

TEXT

The diagnosis of tick-borne diseases remains a diagnostic challenge for many clinicians. Symptoms such as headache, fever, myalgias, and arthralgias are often vague and overlap many other syndromes, let alone among other tick-borne pathogens. Other laboratory findings such as leukopenia, thrombocytopenia, and transaminase elevations suffer from the same limitations. Clinicians are then left with a myriad of diagnostic tests, including serological assays, targeted PCRs, peripheral blood smears, and culture, not to mention the “nonconventional” testing offered from some companies directly to consumers (1, 2). Some of these tests may be readily available, others may be available through reference laboratories, and yet others require working with state health departments for testing. These tests are far from perfect, with natural delays in antibody production, cross-reactivity between serological studies, low sensitivities of some molecular assays, and difficult culturing conditions all contributing to delayed diagnoses and increased costs.

Molecular diagnostic techniques have become commonplace in the clinical microbiology laboratory, allowing direct timely detection of organisms that are slow, difficult, or impossible to culture. These methods have evolved over time from single-gene- and/or -pathogen-targeted PCR to more complex assays, including multiplex PCR panels, broad-range PCRs paired with Sanger DNA sequencing, targeted metagenomics, and metagenomic shotgun sequencing. These advances have allowed for the detection of a wider range of pathogens, including multiple pathogens at a time, and the detection of species without a priori suspicion for a specific pathogen.

In the accompanying article in this issue, Kingry et al. use targeted metagenomics for the identification of tick-borne bacterial pathogens from residual specimens submitted for tick-borne pathogen detection (3). Targeted metagenomics is a method in which many genomes (i.e., metagenomics) are investigated through the analysis of a targeted sequence, in this case, the PCR products from 16S rRNA gene PCR (4). This approach has its roots from early studies exploring environmental microbial diversity by cloning 5S or 16S rRNA genes to identify bacteria that could not be cultivated. Today, this approach is now being used for the identification of pathogens directly from clinical specimens. Initially, these approaches made use of Sanger sequencing, which, in the absence of subcloning, limited the detection to typically one species per sample. If there was a mix of 16S rRNA gene sequences in a sample, the chromatogram from Sanger sequencing would be uninterpretable, making the results uninterpretable. Subcloning individual PCR products followed by separate Sanger sequencing reactions provided a method to work around this, but this is time-intensive, and the ability to detect low-frequency sequences is dependent on sequencing a proportionately large number of subclones. The incorporation of next-generation sequencing (NGS) allows for analysis of thousands to millions of PCR amplicons from a single sample, theoretically allowing for the detection of species that may be present at much lower frequencies more quickly and efficiently. This approach has become commonplace in fields such as human or environmental microbiomes and is beginning to be used for infection diagnostics, including bloodstream infections and prosthetic joint infections (5, 6).

There are, however, significant drawbacks and limitations associated with deep sequencing of samples for infection diagnostics compared to many microbiome studies. For high-biomass samples, if an organism is detected and makes up a reasonable percentage of the total reads, it is reasonable to conclude that it was present in the sample in question. However, when next-generation sequencing is applied to low-biomass, including sterile, samples, this assumption becomes problematic. Virtually any reagent or material used during processing can contain miniscule amounts of environmental DNA that contaminate the sample. This contamination is negligible in high-biomass samples but can readily be detected in sterile or low-biomass samples where there are few or no other initial 16S rRNA gene copies to amplify (7). And unlike with species-specific targeted PCRs, contaminant DNA from any bacteria can interfere with the assay, making even nonpathogenic species such as Bradyrhizobium a frequent challenge with these assays. DNA from potentially pathogenic species such as Cutibacterium acnes, Acinetobacter species, and Pseudomonas aeruginosa can also commonly be found (8), which in most circumstances means that simple subtraction of these organisms from the data prevents the detection of these species that are capable of acting as pathogens. Vector-borne pathogens are one of the rare circumstances where this overlap of contaminants and true pathogens is likely minimal, which makes it an ideal opportunity for a targeted metagenomics approach with deep sequencing to identify potential pathogens.

Kingry et al. used targeted metagenomics as a tool to better identify bacterial pathogens that were present in patients suspected of having tick-borne diseases. Their approach included the use of an initial PCR amplifying the 16S rRNA gene V1-V2 region based on preliminary bioinformatic analysis indicating that their primer selection would allow for the detection and identification of known tick-borne bacterial pathogens. They made a Herculean effort in testing >13,000 residual samples from patients suspected of having infection as well as healthy controls, with 881 specimens testing positive for 12 different bacterial pathogens, including 7 not previously known or thought to be transmitted by ticks.

There are multiple interesting aspects of this paper that deserve attention, particularly in regard to why targeted metagenomics may be more useful in this scenario than for other infections. There is presumably very little overlap between pathogens transmitted by ticks and organisms that are typically detected as contaminants in molecular assays. The authors leverage this by sequencing >1,000 samples from healthy blood donors and water samples to identify taxa that made up this contaminant background noise (with Pseudomonadales, Enterobacterales, Propionibacteriales, and Burkholderiales being the most common taxa identified). Taxa detected in controls were ignored in the clinical specimens, which allowed them to focus on other pathogens from genera including Rickettsia, Borrelia, Leptospira, Ehrlichia, and Anaplasma. This is a luxury not afforded to those trying to apply these methods to other infection types where contaminants and pathogens overlap, but Kingry et al. took advantage of this opportunity.

By using this approach, they were able to identify possible tick-borne pathogens in 881 specimens. The most common species detected were those that are well known to be transmitted by ticks, including Anaplasma, Ehrlichia, and Borrelia species, which supports the use of this approach. Interestingly, they were able to identify two bacterial species (“Candidatus Borrelia johnsonii” and Anaplasma sp.) that have previously been identified as being carried by ticks but with no prior known human infections. Two other bacteria not previously reported to cause human disease, Neorickettsia risticii and a novel Rickettsia-like organism, were also detected. They also identified species not known to be transmitted by ticks, including multiple Leptospira species, Coxiella burnetii (the cause of Q fever), and Rickettsia typhi (the cause of murine typhus transmitted by fleas).

The detection of pathogens not known or thought to be transmitted by ticks highlights the limitation of the study design for being able to link detection to actual tick transmission. Samples for the study were selected because the ordering provider had requested PCR testing for tick-borne pathogens. Ideally, these tests would have been ordered only when patients had a syndrome consistent with tick-borne diseases and plausible recent exposure to ticks. Practically, this is frequently not the case, as they will be ordered for chronic symptoms or for individuals with no plausible exposures to ticks capable of carrying the pathogens tested for. This likely accounts for some of the many negative tests. The detection of C. burnetii and multiple Leptospira species also likely reflects the method of sample selection, as there is significant syndromic overlap (fever, headache, myalgias, leukopenia, and thrombocytopenia, etc.) of these infections with the tick-borne illnesses, and why broad infectious etiology testing is often necessary.

Another limitation of the study design is that it does not produce data regarding the sensitivity of the assay for detecting tick-borne pathogens. It is unknown how many of the >13,000 samples represented true tick-borne infections, so further work will be necessary to clarify how well this assay performs in that regard. This group of authors had previously reported the results from the use of a Borrelia genus-targeting broad PCR in 7,292 (presumably overlapping) specimens, with similar findings of different Borrelia species and “Candidatus Borrelia johnsonii” being detected (9). It is reassuring that targeted metagenomic sequencing of the >13,000 samples that they had found as many or more of each species, but it is still unknown if any of the Borrelia-specific PCR-positive samples were not able to be detected by targeted metagenomics.

Borrelia burgdorferi-positive samples were underrepresented (87 of 881 positive specimens) compared to its prevalence as the most common cause of reported tick-borne infections (10). As noted by Kingry et al., B. burgdorferi is associated with a very low level of bacteremia, making the detection of bacteremia by molecular methods, whether through targeted metagenomics or pathogen-specific PCR, challenging (1). Referral bias could also contribute to the relatively low numbers, as Lyme disease is often treated based on a clinical diagnosis.

The authors were able to detect multiple tick-borne pathogens in three samples, which is an example of the advantage of using metagenomic sequencing compared to conventional 16S rRNA gene PCR paired with Sanger sequencing. Coinfections associated with tick bites are not uncommon (11). These most commonly involve B. burgdorferi with a second pathogen. The low rate of detection of B. burgdorferi from blood again is likely contributing to the limited number of coinfections detected.

When looking toward the implementation of this approach as a clinical diagnostic tool, another limitation is that it would not be an all-encompassing approach to identifying tick-borne illnesses as this assay identifies only bacterial infections, with other pathogens, such as the protozoon Babesia microti or viral pathogens such as tick-borne encephalitis virus, Powassan virus, Heartland virus, or Bourbon virus, going undetected. While viral infections are much less common and have more distinct clinical features to prompt for their testing, babesiosis shares many clinical features with bacterial pathogens, which makes it difficult to distinguish clinically. In clinical practice, the 16S NGS assay would have to be paired with a separate assay for babesiosis in areas where this disease is endemic to ensure that monoinfection or coinfection with Babesia microti is not missed. Theoretically, a panpathogen test such as metagenomic shotgun sequencing would be able to overcome this limitation. With this approach, there is no targeted amplification, and instead, all nucleic acid (DNA and/or RNA) is sequenced using NGS, allowing for the detection of bacteria, fungi, protozoa, and viruses with a single test. Indeed, this has been shown to work in principle to identify pathogens transmitted by ticks both from human specimens (12, 13) as well as by direct testing of ticks (14, 15). However, the lack of a targeted sequence requires substantially more sequencing depth, and the interpretation of data can be much more challenging.

Looking forward toward a shotgun metagenomic approach for detecting tick-borne pathogens, it is easy to envision a strategy employed similar to the ones used in the study by Kingry et al. Prefiltering human reads (a process in which next-generation sequencing reads are rapidly compared to human genome databases to identify and remove human reads prior to more complex processes such as taxonomic assignment) is often one of the earliest steps in shotgun metagenomic sequence analysis. Using an approach similar to the one in this study, by sequencing a large number of human and reagent controls, these sequences could be added to the human sequence library for prefiltering reads to quickly remove background reads to allow for quicker processing and cleaner data without the background noise commonly encountered with shotgun sequencing. Alternatively, researchers could use a strategy of ignoring taxon identifications that overlap contaminant taxa, much like the approach used in this study.

The potential of metagenomic sequencing, be it targeted or shotgun, continues to offer great promise for clinical microbiology diagnostics but is also full of many challenges unique to the methods, complicating the interpretation of results. And while extensive control sequencing to identify and eliminate contaminant sequences is not a novel approach, the study’s focus on potential tick-borne pathogens allowed for a more complete removal of background noise given the lack of overlap between background species and tick-borne species. The assay shows promise as a single assay to detect many different tick-borne pathogens, including those that mimic tick-borne illnesses; however, in most scenarios, it will need to be combined with other assays, such as Lyme serology and Babesia PCR or peripheral smear, to provide a full assessment of potential tick-borne pathogens.

The views expressed in this article do not necessarily reflect the views of the journal or of ASM.