October 5, 2022 | Experiences at the Mayo Clinic integrating metagenomic next-generation sequencing (mNGS) into clinical practice was one of the featured presentations at the recent Next Generation Dx Summit in Washington, D.C. Keynote speaker Robin M. Patel, M.D., professor of individualized medicine and director of Mayo’s infectious diseases research laboratory, focused primarily on the use of the targeted-based approach where the 16S ribosomal RNA (rRNA) gene is being analyzed.
Sequencing was described prior to polymerase chain reaction (PCR) in the 1980s but wasn’t technically ready for use in clinical microbiology labs until the advent of capillary sequencers in the 1990s, says Patel. Improvements in DNA sequencing rates resulted entirely from innovations on the technology side that are “really making a difference for patient care.”
For three decades now, the Mayo Clinic has been routinely sequencing the 16S rRNA from isolates but what clinicians have always wanted is for it to be done directly from clinical specimens, Patel says. She and her team designed primers to amplify every 16S rRNA and then sequenced between the primers to identify infectious organisms, leveraging a reference database on many previously sequenced bacterial species.
“In the future we need a description of lots of novel species and whole genome sequencing from all of them that is all cleaned up,” she continues. “But right now, [16S rRNA] is the frequent flyer of sequenced microbial genes, so it makes a nice target to go after.”
Since 2017, the Mayo Clinic has offered an assay that amplifies and sequences the rRNA gene from clinical specimens, says Patel. “We began with Sanger sequencing because that’s what we had done with isolates and because I have always found with complicated assays that it is better to start simple and ... learn where things are going wrong ... [and] what you are actually finding.”
The assay was used for a two-year retrospective analysis of 566 specimens from 460 patients of the Mayo Clinic that published last year in Clinical Infectious Diseases (DOI: 10.1093/cid/ciab167). “Overall, we had positive results for 17% of cases.” The specimens underwent 16S rRNA gene real-time PCR (SYBR-Green) followed by Sanger sequencing of those with low cycle threshold (Ct) values.
Ninety percent of the positive results were associated with clinical infections, she notes, and in 5% of patients these results impacted clinic care based on changes in patient management documented in their medical records. With the 22% of positive specimens where the sequence couldn’t be sorted out, the problem was solved with mNGS.
Researchers noticed that a positive Gram stain was, somewhat unsurprisingly, associated with a 12-fold increase in positivity. The specimens subjected to the assay were normally sterile body fluids (e.g., spinal and pleural fluids) and tissues (e.g., biopsied heart valves), Patel says, among which cardiovascular specimens had the highest rate of positivity.
In a subsequent analysis of cultures in this cohort of samples, 20% were positive and 87% were associated with clinical infections, she adds. “Interestingly, musculoskeletal specimens had the highest rate for positivity.”
Also noteworthy was that 16S rRNA PCR/sequencing found a probable pathogen in 10% of culture-negative specimens, says Patel. Overall, there was concordance between PCR and culture in 77% of the samples. Sensitivity and specificity of the molecular approach was 30% and 97% and, for culture, 34% and 94%.
The data has helped inform recommendations to Mayo Clinic practitioners about how the testing should be used, Patel says. One example is in diagnosing infectious endocarditis and the causal organism or organisms, starting with blood cultures. “If they are positive, you get your answer [and] you are done. If they are negative, you go on to Coxiella burnetii serology [and], if that’s negative and the heart valve comes out, you look at the valve in the pathology lab.”
This is where the 16S rRNA gene sequencing is particularly useful, “over and above culture if you don’t have enough specimen to do both of those tests,” says Patel. “We’ve been able to take what we’ve learned and translate it into some recommendations in clinical practice.”
To deal with specimens that were clearly positive but the data from Sanger sequencing wasn’t interpretable, mNGS was used, she says. “We’re still targeting the S16 ribosomal gene. We target the V1 (variable 1) through V3 region and run an upfront PCR assay.”
The test is quick and provides “an idea of the strength of the signal we’re dealing with,” Patel says. “If the Ct value is low, then we still go to Sanger sequencing because [it is] inexpensive and fast[and] we can get our answer right away with many of those.
“If the Ct value is in between ... 32 to 34, we go straight to next-generation sequencing,” she continues. “NGS is also used when Sanger sequencing produces an uninterpretable result.” If the Ct value is high—the Mayo lab uses a cutoff of 34—the result is called a negative “unless we see a really nice melting temperature, in which case we take it through next-generation sequencing.”
In fact, many clinical specimens fall into the high Ct value category and PCR testing is a means of screening out the negatives and producing an immediate result, says Patel.
Mayo went live with the16S rRNA gene PCR/sequencing assay in 2020 and is soon to publish a new paper in Clinical Infectious Diseases where it was used to analyze 2,146 samples, 609 of which had a Ct value of less than 32, 400 that were resolved with Sanger sequencing alone, and another 209 that went to NGS and were in nearly all cases resolved and called positive, she shares.
Of the 339 samples with an “in between” Ct value that went to NGS, 147 were reported as positive, she adds. Of the small number of negatives sequenced (16), researchers were able to get a “plausible result” in all but one case.
Per these results, says Patel, it is indeed valuable to add NGS on top of Sanger sequencing. “It increased our yield of positives by 87%, both because we could resolve those that we had previously sent to Sanger sequencing which were not resolvable ... but also because we were sequencing at the little bit higher Ct values, and we were able to get usable results from some of those.” In terms of specimen type, the yield was particularly high for wound biopsies, peritoneal fluid, and abscess fluid.
Many organisms have been found, she says. “That’s the beauty of the ribosomal RNA gene.” The list includes a lot of polymicrobial infections [e.g., Streptococcus, Staphylococcus aureus] and Mycobacterium.
Overall, the targeted metagenomic assay has produced higher positivity rates than culture and this is particularly true in specimens such as heart valves, says Patel. It’s the same story with fluids, especially pleural fluid.
Mayo researchers have also looked at the effect of antibiotics, which were found to affect the positivity rate of targeted metagenomic sequencing less than cultures, she says. But the difference decreases over time as patients come off antibiotics.
In a separate analysis, they showed the likelihood of positivity decreases for both types of tests as the duration of antibiotics increases, Patel adds. “But over time you still see the sequencing-based test is more sensitive than the culture-based test.”
In addition to the 87% increase in yield in the latest study, NGS with Sanger sequencing was found to have 53% sensitivity versus 42% for culture, and there was an impact on clinical decision-making in 14% of infected cases, she reports. In the subgroup that received antibiotics until the day of sampling, the sensitivity for sequencing-based diagnostics was 63% versus 41% for culture, and 70% if combined.
“This is a very hard test to run,” she cautions. “I think it’s the hardest test I’ve ever launched into our clinical practice.”
Patel shares one instance where the targeted sequencing test was run was on bone tissue from a patient from an outside facility who had a T9 vertebral body implant and a Ct value of 36. Based on her years of experience, she could tell right off that it was a Bartonella infection after looking at all the forward and reverse reads. But it was confirmed with another test on as second sample from the same patient, this time epidural fluid from the spine, which had a Ct value of 33 and clearly met Mayo’s criteria for sequencing.
The lesson here is that “specimen collection matters,” she says. “If you collect a good specimen, you get a much better result than if you collect a poor specimen ... especially since you can sequence anything and there are microbes all around us.”
Ct values also really matter, adds Patel. At higher Ct values, “it’s just complicated. You’re down in the weeds when you’re trying to pick out things that are significant.”
She also shared the case of a 5-year-boy with back pain whose MRI showed vertebral osteomyelitis with epidural abscess at multiple levels. He also had liver and spleen abnormalities, loved to play with cats (including stray ones), and his serology for Bartonella was positive, so it was a good fit for sequencing, Patel says.
Next up was the case where a patient had a Ct value of 33, where 3,370 forward sequencing reads implicated Kingella kingae. Other organisms found were presumably “just background” noise coming from the weak positive specimen.
Another case Patel shares is that of a 1-year-old boy with knee pain who had gone to the emergency room with a cough, congestion, fever, and wheezing, and was diagnosed with an upper respiratory infection (URI) and perhaps airway disease. Three days later, when his parents bring him back, the URI symptoms are still there but now his knee hurts and it’s warm. He is again sent home but returns two days later when his knee is worse, and his leg is swelling.
The ER team now does a workup and concludes he has neutrophilic leukocytosis. The boy has an elevated erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP), and an MRI shows septic arthritis and pyomyositis of vastus intermedius. They do irrigation and debridement, pulling out cloudy but blood-tinged fluid that is analyzed with the sequencing assay to find elevated total nucleated cells and 95% neutrophils—a complete fit with Kingella kingae.
Patel presents several cases involving pleural fluid, one where the Ct value is 35 but the lab suspected an infection and most of the sequencing reads turned up Fusobacterium necrophorum. That was likewise the conclusion with an 18-year-old female who originally presented with acute pharyngitis and was diagnosed with mono but later developed a necrotizing pneumonia of the left lung with associated pleural effusion. The culprit “was not obvious to the clinician treating this patient.”
In yet another case where pleural fluid was collected, a 5-year-old boy the week prior to admission had a low-grade fever, cough, and myalgias that progressed to high-grade fevers, nonproductive cough, difficulty breathing, and abdominal pain, she says. A month prior to his admission, the boy’s father had died from COVID-19. The patient also had leukocytosis, elevated ERS and CRP, his nasopharyngeal swab was positive for influenza A, and on chest X-ray he had consolidation retrocardiac lower left lobe and within right mid lung and a moderate-sized left pleural effusion. Gram stain of his pleural fluid showed Gram-positive cocci that looked like strep, enabling the Mayo team to lock in this diagnosis.
The last case presented by Patel was findings from sequencing on left knee fluid with a high Ct value. What stood out to her, she says, were the reads for Borreliela burgdorferi (Lyme disease), prompting the lab to run a confirmatory, standalone PCR test indicating Lyme septic arthritis.
Patel next shares the story of a 59-year-old Minnesota man who presented with a rash on his face, and some peripheral neuropathy and arthritis, who had only traveled to Mexico. His treatment team found granulomas and lots of acid-fast bacillus, including around his nerves, and so they diagnosed him with leprosy. Her arm “got twisted” to run the sequencing assay on the specimen, she says, noting that she doesn’t normally test skin because of its abundance of normal microbiota.
The results show that it wasn’t Mycobacterium leprae after all but Mycobacterium lepromatosis, an organism she had never heard of and has not even been validly named yet. Sadly, she says, the two bacterium are related. The diagnosis was confirmed by the Hansen Disease Center.
The same kind of assay has been applied directly to blood, including a collection of samples that had tick-borne pathogens in them (among them E. ewingii, E. muris subspecies eauclairensis, E. chaffeensis, B. miyamotoi, B. hermsii, and phagecytophilum), she says. “We were able to pick up 16 of 17 [pathogens]; we did miss a B. burgdorferi that had been detectable using a specific PCR from blood. The Ct value ... was over 40, so it was a super weak positive.”
Shotgun metagenomics, which involved sequencing everything in a sample, is entirely different than targeted metagenomic sequencing that goes after one or a small number of genes, notes Patel. The shotgun approach is “exciting because you can find not just bacteria but parasites, fungi, viruses ... anything really that’s in there ... It is truly agnostic in terms of what you find from a microbial standpoint.” The difficulties are tied to the need to sequence deeper and have a lot of human DNA to work with.
Patel’s research in this arena looks specifically at orthopedic implant infections for which a biofilm sampling strategy has been developed, she shares. The process involves removing biofilms from explants and using ultrasound to remove the sonicate fluid. “We’ve found it to be the best specimen for finding microorganisms and we’re trying to figure out the cause for some of the culture-negative infections that we see in our clinical practice.”
In a large study using shotgun metagenomic sequencing of sonicate fluids that included 195 noninfected patients, 115 culture-positive infections, and another 98 with culture-negative infections, the approach performed comparably to culture in the first two categories (Clinical Infectious Diseases, DOI: 10.1093/cid/ciy303). “But in the culture-negative samples, in 43.9% of cases we were able to detect [the infectious] organism,” says Patel.
The pathogens detected in the culture-negative periprosthetic joint infection (PJI) cases via shotgun sequencing are ones that would typically be found in such infections, she adds, “so we think this kind of sequencing-based diagnosis could be useful.” Since they are predominantly bacterial, it also raises a question about the relative value of shotgun versus targeted metagenomic sequencing for this kind of sample.
In a study just accepted for publication in Clinical Infectious Diseases, Patel and her colleagues took sonicate fluids from the shotgun metagenomic sequencing study and subjected them to the 16S rRNA gene-based targeted sequencing assay. Of the 395 specimens, 208 had PJI and 187 were not infected, and the targeted metagenomic sequencing detected potential pathogens in 48% of culture-negative PJIs. “Very importantly, there was no difference between positive percent agreement of the targeted and shotgun metagenomic sequencing [72% and 73%, respectively].”
The targeted approach would benefit from improvements, including bacterial, DNA-free reagents. “I desperately need more of these,” she says. Other needs include faster sequencing that is accurate, lower cost of sequencing, automation, and computer-based interpretation because much of it is still done manually.