Detection of Tick-Borne Bacteria from Whole Blood Using 16S Ribosomal RNA Gene PCR Followed by Next-Generation Sequencing

Reported cases of tick-borne diseases have steadily increased for more than a decade. In the United States, a majority of tick-borne infections are caused by bacteria.

ABSTRACT

Reported cases of tick-borne diseases have steadily increased for more than a decade. In the United States, a majority of tick-borne infections are caused by bacteria. Clinical diagnosis may be challenging, as tick-borne diseases can present with similar symptoms. Laboratory diagnosis has historically relied on serologic methods, which have limited utility during the acute phase of disease. Pathogen-specific molecular methods have improved early diagnosis, but can be expensive when bundled together and may miss unexpected or novel pathogens. To address these shortcomings, we developed a 16S rRNA gene PCR with a next-generation sequencing (NGS) approach to detect tick-borne bacteria in whole blood. A workflow was optimized by comparing combinations of two extraction platforms and two primer sets, ultimately pursuing DNA extraction from blood with the MagNA Pure 96 and PCR amplification using dual-priming oligonucleotide primers specific to the V1-V3 region of the 16S rRNA gene. The amplified product underwent modified Illumina 16S metagenomics sequencing library preparation and sequencing on a MiSeq V2 Nano flow cell, with data analysis using Pathogenomix RipSeq NGS software. Results with the developed method were compared to those from a V1-V2 16S rRNA gene primer set described by the Centers for Disease Control and Prevention (CDC). The V1-V3 assay demonstrated equivalent performance to the CDC assay, with each method showing concordance with targeted PCR results in 31 of 32 samples, and detecting 22 of 23 expected organisms. These data demonstrate the potential for using a broad-range bacterial detection approach for diagnosis of tick-borne bacterial infection from blood.

INTRODUCTION

The past 2 decades have been marked by a steady rise in the number of tick-borne diseases (TBDs) caused by bacterial pathogens (1). During this time, reported cases have nearly tripled, and six new tick-borne bacterial pathogens have been identified (2). As regions where ticks are endemic expand, greater percentages of the population are at risk for TBDs (3).

Rapid diagnosis of TBDs facilitates prompt treatment, which has been linked to better patient outcomes (3). However, acute phase diagnosis of TBDs may be challenging, as clinical presentations are often nonspecific, with the differential diagnosis including both TBDs and non-TBDs. Furthermore, commonly used diagnostic methods for acute TBDs have significant limitations. Serologic assays, the historical and still most widely used method for laboratory diagnosis of most TBDs, can be negative in early disease or when antibody titers are low (4). Additionally, most widely available serologic assays (e.g., Ehrlichia chaffeensis) show species-level cross-reactivity, and thus provide only genus-level diagnosis (5). The introduction of nucleic acid amplification tests (NAATs) for detection of many TBDs has improved acute phase diagnosis, as they often afford greater sensitivity and do not rely on the generation of a host immune response. These assays are commonly designed to detect single pathogens or a group of related organisms (e.g., Borrelia burgdorferi sensu lato genospecies). To test for all potential pathogens, NAATs are commonly bundled, which can be an expensive and time-consuming practice. Additionally, targeted assays require some level of provider suspicion to facilitate ordering, and will likely miss underappreciated or novel pathogens they are not designed to detect.

With these challenges in mind, we developed a 16S rRNA gene PCR with massively parallel, next-generation sequencing (NGS) to detect and identify tick-borne bacterial pathogens from whole blood. The framework of this assay was adapted from a clinically available “broad range” bacterial PCR and Sanger-based sequencing assay performed on normally sterile, non-blood sources (6). Briefly, the former assay involved manual extraction of nucleic acids, PCR amplification of the V3-V4 region of the 16S rRNA gene, and Sanger sequencing of the resultant amplicons, and was not performed on blood. The new assay incorporated three main changes: (i) switch to automated nucleic acid extraction from manual extraction to facilitate high-volume clinical testing; (ii) replacement of the 16S rRNA gene V3-V4 primer set with primers optimized for detection of tick-borne bacteria; and (iii) transition from Sanger sequencing to NGS to facilitate greater sensitivity through depth of sequencing. The manuscript compares two extraction platforms and two 16S rRNA gene primers, developing the Mayo Clinic protocol for NGS-based detection of tick-borne bacterial pathogens from whole blood. Recognizing that the sensitivity and accuracy of 16S rRNA gene sequencing may vary by NGS protocols, we also compared the final Mayo Clinic protocol to the recently published and extensively validated CDC NGS protocol (7).

MATERIALS AND METHODS

Samples.

Patient samples consisted of residual whole blood in EDTA originally submitted to Mayo Clinic Laboratories in Rochester, MN for analyte-specific tick-borne NAATs. Samples positive for Ehrlichia spp. or Anaplasma phagocytophilum were identified with a real-time PCR assay targeting groEL (8). Samples positive for Borrelia miyamotoi or Borrelia hermsii were identified by a real-time PCR assay targeting glpQ (9). Samples positive for Borrelia burgdorferi and Borrelia mayonii were identified by a real-time PCR assay targeting oppA2 (10, 11). Samples were collected between May 2017 and October 2019 and stored at −80°C prior to testing. Use of residual patient samples was approved by the Institutional Review Board of the Mayo Clinic.

DNA extraction.

Two bead-based automated extraction systems were evaluated: the NUCLISENS easyMAG (bioMérieux, Marcy-l'Étoile, France), and the MagNA Pure 96 (Roche Life Science, Penzberg, Germany). Samples extracted on the easyMAG underwent an initial off-board prelysis where 100 μl of whole blood, 20 μl of proteinase K, and 160 μl of proteinase K buffer were added to an autoclaved lysis tube containing 20 μl of silica/zirconium beads. Digestion was performed at 60°C with shaking at 300 rpm for 60 min, followed by centrifugation at 100°C for 5 min. Tubes were spun at 10,000 rpm for 20 sec and 260 μl of supernatant was transferred to an extraction cartridge with 2 ml of lysis buffer and incubated at room temperature for 10 min. Magnetic beads were mixed with equal parts double-sterilized elution buffer and 100 μl was added to each sample in the extraction cartridge. Nucleic acid extraction was performed on the NUCLISENS easyMAG with appropriate reagents per the manufacturer’s recommendations and a final elution volume of 50 μl.

Nucleic acid extraction on the MagNA Pure 96 was performed using the universal pathogen extraction protocol 2.0 without additional sample preprocessing. The MagNA Pure 96 DNA and Viral NA small volume kit were used with an input volume of 200 μl and an elution volume of 100 μl.

PCR amplification.

PCR was performed using two primer sets targeting the V1-V3 and V1-V2 regions of the 16S rRNA gene. Alignment of the 16S rRNA gene from 15 tick-borne bacteria, including Anaplasma, Ehrlichia, Francisella, Rickettsia, and Borrelia spp., along with 18 non-tick-borne bacteria that present with similar clinical features, including Leptospira, Rickettsia, Coxiella, Orientia, and Neorickettsia spp., was performed to design the V1-V3 dual-priming oligonucleotide (DPO) primers, modified from Kommendal et al. (12). The final primers were as follows: forward 5′-AGAGTTTGATCMTGGCTCAIIIIIAACGC-3′, reverse (1) 5′-CGGCTGCTGGCAIIIAITTDGC-3′, and reverse (2) 5′-CGGCTGCTGGCAIIIAITTDGT-3′. V1-V3 amplicon PCR was performed on a LightCycler 480 (Roche) with each reaction consisting of 15 μl of the mastermix (described in Table S1 in the supplemental material) combined with 5 μl of extracted DNA. Cycling conditions were as described in Table S2. Amplification with the CDC V1-V2 primers was performed as described above using the previously described primers minus the Illumina Nextera XT index adaptor sequences (7).

Library preparation, normalization, sequencing, and data analysis.

(i) Mayo Clinic protocol. Following amplification, products generated with both the Mayo Clinic V1-V3 and CDC V1-V2 primers were initially processed using the Mayo Clinic library preparation, normalization, sequencing, and data analysis protocol as follows. Double-stranded DNA (dsDNA) concentrations of samples were first determined using the QuantiFluor ONE dsDNA kit (Promega, Madison, WI) on a Quantus fluorometer (Promega, Madison, WI). Samples were then diluted to 5 ng/μl with 1× Tris-EDTA (TE) buffer (Invitrogen, Carlsbad, CA), which was autoclaved at 121°C for 80 min, and Illumina adaptors attached to amplified DNA using the following primers: forward (V1-V2 and V1-V3 amplicons) 5′-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGAGAGTTTGATCMTGGCTCAG-3′, reverse (V1-V3 amplicons) 5′-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGCGGCTGCTGGCA-3′, reverse (V1-V2 amplicons) 5′-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGCTGCTGCCTCCCGTAGGAGT-3′. Each reaction consisted of 22.5 μl of the mastermix described in Table S3 along with 2.5 μl of each 5 ng/μl sample. Cycling conditions on a Veriti thermal cycler (Applied Biosystems, Foster City, CA) are shown in Table S4. PCR cleanup was performed using 20 μl of AMPure XP beads (Beckman Coulter, Brea, CA) with autoclaved TE buffer. Index PCR was performed following the Illumina protocol with the following modifications. Unique dual indexes from Integrated DNA Technologies (IDT, Newark, NJ) were used, with 10 μl of each unique pair added to the samples. After index PCR, cleanup was performed using 56 μl of AMPure XP beads. Library concentrations were assessed using the QuantiFluor ONE dsDNA kit and each sample was normalized to a 4 nM concentration. Libraries were pooled by combining 5 μl of each; the final concentration of the pool was determined to be acceptable if it was between 1.3 and 1.7 ng/μl. Pools were rediluted to an acceptable range when necessary. Pooled libraries were placed in an ice-water bath prior to loading flow cells. The pooled library was denatured and combined with 5% PhiX (Illumina library control) per standard Illumina protocol with paired end (2 × 250 bp) sequencing performed on an Illumina MiSeq with a 500 cycle V2 Nano kit (Illumina, San Diego, CA). Onboard Illumina processing included adapter trimming, index demultiplexing, and fastq generation. For the Mayo Clinic protocol, files were processed using Pathogenomix RipSeq NGS software resulting in reads that were Q ≥ 30 and ≥210 bases in length. Reads meeting the criteria were clustered by sequence identity and queried against the Pathogenomix Prime 16S database. Results with ≥100 reads were reported. Data analyzed and shown are from the forward reads.

(ii) CDC NGS protocol. Residual MagNA Pure 96 extracts from all 32 tested samples were also tested and evaluated using the CDC NGS library preparation, normalization, sequencing, and data analysis as previously described (7). Primary differences between the Mayo Clinic and CDC protocols included use of PCR primers with Illumina Nextera XT index adaptor sequences incorporated, use of the MiSeq V2 kit, inclusion of results with >50 reads, and data analysis using the MiniKraken database in the CDC protocol.

Data availability.

Raw sequence read data from all samples have been deposited in the Sequence Read Archive (SRA) under project accession number PRJNA685221.

RESULTS

Initial assay optimization was performed by assessing two primer sets, CDC V1-V2 and the Mayo Clinic V1-V3, and two nucleic acid extraction platforms, NUCLISENS easyMAG and MagNA Pure 96, comparing all combinations within the described Mayo Clinic NGS library preparation, normalization, sequencing and data analysis protocol. Comparisons were conducted using nine previously positive samples (5 E. chaffeensis and 4 A. phagocytophilum) and six samples negative for tick-borne bacteria (Table S5). Combinations were assessed for taxonomic identification consistent with the prior PCR result, percentage of the reads corresponding to the expected pathogen, and reads above threshold mapping to unexpected bacteria, which can include background DNA contamination from reagents or the assay itself, along with non-TBD organisms originating from the patient sample. The combination of MagNA Pure 96 plus V1-V3 and NUCLISENS easyMAG plus V1-V2 produced concordant results for 15/15 samples, while the combination of MagNA Pure 96 plus V1-V2 and NUCLISENS easyMAG plus V1-V3 was concordant in 14/15 samples (Table 1). The latter two combinations failed to detect a low positive E. chaffeensis sample (organism specific PCR crossing point [Cp] >40), while the MagNA Pure 96 plus V1-V3 and NUCLISENS easyMAG plus V1-V2 combinations identified E. chaffeensis with 29% and 11% of reads matching this organism, respectively. There was minimal variation in the mean percentage of reads corresponding to the expected organisms; however, extraction on the MagNA Pure 96 resulted in a slightly higher percentage of reads mapping to the expected tick-borne bacterium than extraction on the NUCLISENS easyMAG (Table 1). These results, in combination with workflow considerations, led to the selection of the MagNA Pure 96 as the extraction platform.

TABLE 1.

Assessment of extraction platforms and primer set combinations

Platform and primer set	Taxonomic identification consistent with targeted PCR	Mean % of reads corresponding to expected organisma	Next-generation sequencing results with nonspecific reads above thresholda
MagNA Pure 96 + V1-V2	14/15	82.6	5/9
MagNA Pure 96 + V1-V3	15/15	81.9	6/9
NUCLISENS easyMAG + V1-V2	15/15	79.1	8/9
NUCLISENS easyMAG + V1-V3	14/15	78.2	5/9

^aData from previously positive samples only.

The V1-V3 and V1-V2 primer sets were further assessed with the Mayo Clinic NGS protocol by expanding the breadth of tick-borne samples evaluated. Ehrlichia ewingii (n = 1), Ehrlichia muris subspecies eauclairensis (n = 2), E. chaffeensis (n = 1), B. miyamotoi (n = 4), B. hermsii (n = 1), B. burgdorferi (n = 2), B. mayonii (n = 2), and A. phagocytophilum (n = 1) were included as a set of previously positive samples, along with 3 previously negative samples (Table S5). The V1-V3 primer set showed concordance with prior results in 16/17 samples, failing to detect a low positive B. burgdorferi sample with a Cp >40 (Table 2). The V1-V2 primer set showed concordance with 14/17 prior results, failing to detect B. miyamotoi in two previously positive samples (Cp values of 37.24 and 30.51) in addition to the same B. burgdorferi sample. The B. burgdorferi sample was retested by the clinically available targeted PCR and was confirmed to be positive (10). Across all samples, the mean percentage of reads mapping to the expected organism was 58.9% for V1-V3 and 44.1% for V1-V2, with results from both primer sets also producing taxonomic identifications above the reporting threshold for several background bacteria (Table 2). The improved sensitivity and increased percentage of reads mapping to TBD bacteria supported the selection of the V1-V3 primers for use in the described Mayo Clinic NGS protocol.

TABLE 2.

Expanded assessment of V1-V3 and V1-V2 primer sets using the MagNA Pure 96

Primer set	Taxonomic identification consistent with targeted PCR	Mean % of reads corresponding to expected microorganisma	Next-generation sequencing results with nonspecific reads above thresholda
V1-V2	14/17	44.1	13/14
V1-V3	16/17	58.9	10/14

^aData from previously positive samples only.

As the V1-V2 primers were evaluated in prior sample sets using the Mayo Clinic NGS protocol, we sought to compare the Mayo Clinic V1-V3 assay to the full V1-V2 CDC NGS library preparation, normalization, sequencing, and data analysis protocol via testing of residual MagNA Pure 96 extracts from the previously tested 32 samples. The results from the Mayo Clinic and the CDC NGS protocols were concordant in 31/32 samples and detected 22/23 organisms from previously positive samples. The previously referenced B. burgdorferi sample not detected with the Mayo Clinic protocol produced taxonomic identification of B. burgdorferi by the CDC V1-V2 protocol. The CDC method did not detect one E. chaffeensis-positive sample that was detected by the Mayo Clinic protocol.

DISCUSSION

In this multipart study, we describe a robust 16S rRNA gene PCR and NGS protocol for detection of tick-borne bacterial pathogens from human whole blood samples and demonstrate the impact that primer sequence, extraction platform, library preparation, normalization, sequencing, and data analysis has on the sensitivity and specificity of the complete assay. Recognizing the need for automation in the clinical laboratory, we first evaluated two automated, bead-based extraction platforms for isolation and purification of target nucleic acid from blood. Performance was assessed using previously tested TBD-positive and -negative samples, allowing for comparison of results with current clinically available assays, a necessary component in evaluation noted in a recent commentary by Thoendel (13). Samples extracted on the MagNA Pure 96 yielded fewer reads mapping to unexpected organisms, indicating a lesser amount of background DNA extracts. This was seen in conjunction with a consistently higher percentage of reads aligning with the expected pathogen. In comparison, samples extracted on the NUCLISENS easyMAG showed a slightly lower percentage of reads corresponding to the expected organism. Given the similar detection of expected organisms between the two platforms and less background contamination with the MagNA Pure 96, the MagNA Pure 96 was chosen as our preferred extraction system.

We next performed studies to select a preferred primer set to use in conjunction with a standardized Mayo Clinic NGS protocol (encompassing extraction, library prep, sequencing, and data analysis). For completeness, the DNA extracts were subsequently tested using the CDC NGS protocol.

When using the Mayo Clinic NGS protocol, the V1-V2 primer set demonstrated decreased sensitivity, indicated by decreased taxonomic agreement with expected results and percentage of reads mapping to the expected organism compared with the V1-V3 primer set. This is likely due, in part, to the conventional primer design of the V1-V2 primers, allowing for amplification with minor mismatches. In contrast, DPO primers employed in the Mayo Clinic V1-V3 primer set are less forgiving of mismatches under the primer region, decreasing nonspecific amplification (12). This decrease in sensitivity was not observed when testing the same DNA extracts with the complete CDC V1-V2 NGS protocol, thus demonstrating the importance of protocol-specific differences on overall test performance. Additionally, the sensitivity of the CDC NGS protocol for detection of B. burgdorferi appears to be higher than the corresponding sensitivity with the Mayo Clinic NGS protocol and is an area for further investigation.

In addition to nonspecific bacterial reads, some samples amplified with the V1-V2 primers produced reads with “no result” when analyzed with the Mayo Clinic NGS protocol employing Pathogenomix RipSeq NGS software. These sequences were investigated with BLASTn and determined to correspond to sequences of human origin, indicating nonspecific human DNA amplification. This is particularly important, as the Mayo Clinic NGS protocol uses a smaller MiSeq V2 Nano flow cell compared to the MiSeq V2 kit in the CDC NGS protocol. The V2 Nano kit was selected for shorter run time (∼28 h versus ∼39 h) for 2 × 250 reads, cost-effectiveness for routine clinical testing, and a similar turnaround time to targeted PCRs, with the trade-off being lesser output (500 Mb versus ∼8 Gb). Using this smaller flow cell, nonbacterial reads are wasteful and costly, reducing the number of reads that can be devoted to potentially informative sequence. Thus, reduction in nonspecific amplification of human sequence is an advantage to the V1-V3 primers that can allow for more efficient, cost-effective sequencing when applied to the MiSeq V2 Nano flow cell. In contrast, the CDC NGS protocol, designed for high-throughput surveillance, benefits from the larger MiSeq V2 kit. The greater capacity allows for larger batched testing, maximizing throughput and ultimately cost-effectiveness in an application where turnaround time is less of a consideration. Dependent on the number of samples pooled, the larger flow cell may allow for greater depth of sequencing, which can provide increased sensitivity. Again, these results emphasize that all components of an NGS protocol may have important effects on overall sensitivity and specificity of the complete assay.

An advantage of “broad range” applications is that unexpected and potentially novel organisms may be detected (or even discovered). Two samples with E. muris subsp. eauclairensis served as surrogates for “novel organism detection.” At the time of testing, this organism was not part of the Pathogenomix Prime 16S database used in the Mayo Clinic protocol. Data analysis produced a high percentage of reads (50.8% and 80.0% for sample 1 and 84.1% and 72.7% for sample 2 for V1-V3 and V1-V2, respectively) mapping to low quality matches (96% ID) indicated as “result below cutoff” in the software. When these sequences were exported to BLASTn, homology with E. muris and E. muris subsp. eauclairensis was observed. With an update to the Pathogenomix Prime 16S database, reanalysis resulted in the same percentage of reads mapping to E. muris subsp. eauclairensis (99.6% ID). This shows that the software and reference database employed for sequence data analysis is an integral and potentially limiting part of accurate interpretation. To realize this benefit, clinical implementation will require analysis of all sequences with low quality matches to the database (given that such results may indicate underappreciated or novel pathogens).

Unbiased amplification of bacterial DNA has challenges. As observed in the “negative” samples, contaminant bacterial DNA is consistently detected, likely originating from a number of sources, including specimens themselves, specimen collection containers and materials, plastics and disposables used to handle/process samples, assay reagents, laboratory environments, etc. While efforts were made to mitigate contamination, complete removal is impossible and therefore recognition of background contamination is a necessary component of result analysis of both apparent negative and even positive samples, the latter to address the possibility of coinfections. In doing so, contaminant sequence “frequent fliers” can be identified and recognized. In this study, this included high-level identifications, such as Meiothermus silvanus and Ralstonia species. In addition, sequences that do not match known organisms, reported as “result below cutoff” matches, need to be assessed (as they might represent novel pathogens as illustrated by the E. muris subsp. eauclairensis situation described above). By defining likely background reads, they can be correctly disregarded as contaminants. Although not addressed here, we note that based on our clinical experience, the profile of the background contaminants varies over time, and needs therefore to be continually addressed and readdressed. Full understanding of each assay’s background is critical to enable recognition of known and emerging pathogens, as well as to not mislead clinicians by reporting contaminants in such a way that they might be considered clinically relevant. As such, it is not just accurate databases that are needed, but expertise in microbial sequence analysis.

Limitations of this initial assay development include the available bacterial diversity in previously positive samples. While the eight TBD bacterial species included in our sample sets encompass frequently encountered organisms, further development of this assay will include a wider array of pathogens, including non-TBD-causing bacteria that can have similar clinical presentation. As seen in the extensive sample assessment by the CDC, unexpected pathogenic and clinically relevant organisms can be encountered (e.g., Coxiella burnetii, Leptospira species, Rickettsia-like species), a benefit of using a broadly inclusive primer set (7). In silico analysis of the V1-V3 primer set has shown a predicted match to 785 known bacterial pathogens, suggesting potential applications beyond detection of only TBD bacteria. Additionally, as some tick vectors harbor multiple pathogens, mixed infections may be encountered. While the use of next-generation sequencing should allow for simultaneous detection of more than one pathogen, samples with more than one TBD organism were not included here. As with all assays that target the organism for detection, presence of the pathogen in the sample type at the time of collection is necessary. As the presence in blood of some TBD pathogens can be temporal and transient (e.g., B. burgdorferi), negative results do not exclude infection and clinical judgement should be exercised. Similarly, with high clinical suspicion, treatment should not be delayed while diagnostic testing is performed. No matter the method, the turnaround time of testing can often be measured in days, particularly when specific testing requires sending to a reference laboratory. The described assay is no exception, with an expected turnaround time similar to other molecular methods. Nevertheless, the ability to provide a specific diagnosis, help guide continued treatment, and potentially prevent undue additional testing, are some of the benefits of this and other diagnostic modalities.

In summary, early recognition of TBD is necessary for optimal patient care, allowing targeted treatment in the acute phase of disease. Due to the nonspecific presentation of TBDs, laboratory testing plays a key role in diagnosis. Our development of a 16S rRNA gene PCR with next-generation sequencing is a step toward filling a TBD diagnostic gap in the clinical microbiology laboratory.