Skip to main content
Premium Trial:

Request an Annual Quote

ECCMID Talks Highlight Promise of Whole-Genome Sequencing for Clinical Infectious Disease Work

Premium

AMSTERDAM (GenomeWeb) – For molecular diagnosis of infectious disease, PCR remains the undisputed king. However, next-generation sequencing-based whole genome analysis may be next in succession, as evidenced by the burgeoning number of scientific presentations on the technology at this year's European Congress of Clinical Microbiology and Infectious Diseases, held here this week.

Specifically, researchers discussed using WGS in a clinical setting to help predict antibacterial resistance, to inform influenza vaccine development and develop new hypotheses about flu disease severity in individual patients, and to diagnose and build drug resistance profiles for tuberculosis.

Optimism about WGS as a clinical tool for infectious disease diagnosis was tempered, however, by the reality that the technology is still not always cheap, fast, or accurate enough for routine use — although it appears to be getting there.

Combating drug resistance

Tim Read of Emory University in Atlanta, in a presentation on Saturday, discussed his group's efforts to establish WGS as a routine clinical tool for detecting antibiotic resistance.

Read noted that slow and laborious culture-based methods had long been the dominant technology for antibiotic resistance profiling until molecular methods like nucleic acid amplification tests emerged. And while these assays are fast, sensitive, and ideal for use with both cultured and direct clinical samples, they frequently suffer from specificity issues and only provide limited information.

WGS, on the other hand, is unbiased; comprehensive in that it samples 99 percent of the genetic makeup of an organism; and archivable, meaning that it is easy to store data and go back and examine it when needed. In addition, one genome sequence replaces many individual genetic tests such as assays for toxins, antibiotic resistance, and genotypes; and it also allows passive detection of traits or "stuff you weren't looking for," Read said.

Read and his Emory colleagues have been investigating WGS as a routine tool to detect vancomycin-intermediate Staphylococcus aureus (VISA), a "poster child" for complex antibacterial resistance in that these strains evolve in response to prolonged vancomycin therapy and contain multiple chromosomal mutations.

More specifically, the researchers have been trying to develop a genomic predictor for VISA by sequencing genotype-matched strains and conducting genome-wide association studies. They hypothesized that because the VISA phenotype is rare, the mutations that cause it must be rare in the S. aureus population.

In their study, they focused on candidate resistance genes and used the available database of 3,277 public S. aureus genomes that were available at the time. To build their classifier, they used both known and rare mutations in informative phenotypes, homed in on 26 candidate SNPs, and applied their classifier to a 28-strain training set, achieving about 89 percent accuracy in predicting VISA.

To improve the test, Read said the group now needs to sequence enough lab-isolated mutants to saturate the most common non-synonymous mutations. "Really we must just sequence as many of these as possible, as sequencing becomes cheaper," he said.

Summarizing, Read noted that routine clinical strain genome sequencing is likely to become a workhorse in clinical microbiology, given a competitive technology platform and the blessing of regulatory agencies. "We'll then have things like prediction of unusual antibiotic resistance, and large strain sets for classifier development," he said. "Many antibiotic resistance phenotypes can already be called with higher accuracy based on genomic data. Some will take intensive efforts to build predictors. We may be eventually able to predict certain [antibiotic resistance] phenotypes only using sequencing-based tests."

However, Read conceded that WGS is not going to be widely used for this purpose until it is on par with current molecular tests, such as PCR, for a number of parameters. Specifically, he said, it needs to be cheap (on the order of $10 per genome), fast (an hour or less turnaround time), versatile (it can run one or hundreds of genomes), and accurate, which it is already proving to be.

Sequencing platforms such as Pacific Biosciences' RS II, Illumina's MiSeq, and Oxford Nanopore's MinIon all have shown great promise thus far, but are still not there, he said. "You can argue about this, but none of these techs really fit the requirements for routine use in clinical labs today," Read said.

"But we're moving quickly to the point where we can envisage this being possible, and need to think about how we're going to manage all this data," he added. "Even though the common wisdom is that there's too much data, and we're drowning in it, in my opinion, we don't have enough."

Keeping up with influenza

In another presentation, Maria Zambon, director of reference microbiology at Public Health England, discussed how WGS is becoming a useful tool for picking apart the molecular epidemiology of the flu and developing vaccines for the influenza virus.

Zambon said that in the last 30 years or so that she has been researching influenza, viral detection tools "have changed beyond recognition from low-sensitivity fluorescence to gold standard qPCR," which is now rapidly being coupled with or replaced by WGS. This, she noted, has caused a shift in virology surveillance workflows. Previously, most labs were dependent on culture methods, which changed to PCR and culture in the 1990s to 2005. Then, from about 2005 to 2014, this changed to PCR first, followed by some WGS; and after 2014, WGS began being used for all PCR-positive samples and to select variants for culture.

Finally, she predicted that most labs will eventually move to a WGS workflow first in order to "pick out things from sequencing that seem different enough to try and culture the virus."

Her lab at PHE has been conducting Sanger-based WGS for several years now, a method that proved particularly useful in a few notable flu outbreaks. First, there was a wave of fatal pediatric cases in 2003 and 2004, all of which PHE analyzed by WGS, discovering surprising information. For example, the viruses were distributed along all genetic lineages, fatal cases were genetically similar to mild control cases, and no genetic changes were exclusively associated with a fatal outcome.

Then, PHE applied WGS to two different pandemic flu waves in the UK in 2009 and 2010, and discerned no genetic differences in the fatal cases examined in each wave.

Taken together, both of these findings, along with a few similar WGS studies, led to the conclusion that there may be patient-specific genetic factors at play that determine susceptibility to disease, making this an important future area of research.

Zambon said that her lab is now shifting from Sanger-based WGS to NGS technology. "In Sanger, you have relatively low coverage of genomes, while in NGS, you have huge coverage of genomes [and] coupled synthesis and detection, but desynchronization is a major source of error," Zambon noted, adding that "the creation of NGS pipelines is not that different in many respects from working with Sanger."

Zambon also highlighted all the major NGS platforms currently on the market for WGS, and noted that different platforms have different systematic errors that stem from the chemistry used. "The propensity is to move toward systems with longer reads and lower error rates," she said, adding that "at this point, there are no more- or less-superior technologies. One needs to think about the specific application."

Currently, her lab is using WGS data to provide information for vaccine recommendations, noting that it offers another helpful layer of information alongside the typical antigenic profiling. However, "we need to [create] workflows that will work if we just want a quick [hemagglutinin] sequence analysis, or if we want a whole genome, and make sure we can get those to construct whole datasets," she said.

For this application, WGS has already begun to prove its worth, Zambon said, but it has also presented challenges. For instance, in February of the 2014/2015 flu season, antigenic analysis suggested that the vaccines for that season were well matched to circulating strains, but vaccine effectiveness proved to be quite low. WGS showed that there was a mismatch between phenotypic and genetic data, and this year's vaccine is proving to be similar.

With WGS, she said, "we can get comprehensive pictures, and we are actually changing our focus from single gene segments to whole genomes, but there are quite a few challenges in achieving a paradigm shift, one of which is always linking phenotypic and genomic data."

"Now, we are moving to an unbiased approach to analysis," she added, but "data is now a bottleneck."

Eradicating TB?

Finally, two researchers from the UK presented their visions for using NGS-based WGS to diagnose, resistance-type, and potentially eradicate Mycobacterium tuberculosis from the general population.

In one presentation, Grace Smith, a deputy director and clinical lead for Public Health England, presented preliminary data on a prospective study undertaken at PHE to compare WGS for TB diagnosis and drug susceptibility to current routine tests.

Smith noted that the TB burden in the UK is relatively high for a resource-rich country: In 2014, there were 6,520 cases reported, and the incidence rate was 12 per 100,000, which is four times the rate in the US and one of highest rates in Western Europe.

She said that PHE recently implemented a new TB strategy with 10 specific areas for action, and that WGS technology can satisfy criteria for six of those areas.

As such, the agency has initiated a real-time equivalence study at the Public Health Laboratory for the Midlands in Birmingham, UK, comparing WGS with currently used tests such as a Hain Life Sciences' PCR-based assay, Ziehl-Neelsen staining, and the tuberculin skin test.

So far, over 11 months, the lab has processed isolates from 1,700 patients, of whom 600 were infected with TB complex, using WGS and routine diagnostic methods in parallel and in real time. Smith did not disclose what sequencing platform the group is using, but noted that WGS failed to give a result in 3.6 percent of cases. "We've been very cautious in calling [results] at this point, and if it wasn't a clear call, it was regarded as a 'fail'," a point that will be important to monitor in the ongoing study, Smith said.

Overall, the group achieved good sensitivity and specificity, and 99 percent and 97.2 percent concordance for identification of TB complex and common nontuberculous mycobacteria, respectively, demonstrating that WGS "is not inferior to current TB diagnostic tests," Smith said.

The group is now investigating discordant results, timing of tests, cost effectiveness, and obtaining accreditation through lab standards organizations, with the eventual goal to implement WGS as part of England's national TB reference service. Should they succeed, Smith said, she believes the lab will become one of the first public health services to broadly implement WGS for such a purpose.

In another TB-focused talk, Timothy Walker, a clinical medicine graduate student in the Nuffield Department of Medicine at the University of Oxford, gave what he admitted was a more hubristic presentation, entitled "How can whole-genome sequencing help to eradicate tuberculosis?"

Walker, who was presented with ECCMID's young investigator award for his work, noted that the World Health Organization has actually set a target to eradicate TB in the next 20 years, which means reducing worldwide cases to fewer than 10 per 100,000 individuals.

It's an uphill battle, to be sure, littered with numerous obstacles, such as 2 billion people currently harboring latent infections; an insufficiently effective vaccine; an inability to predict which latent patients will develop disease; relatively basic diagnostic methods; and long, toxic, and expensive treatment courses.

Walker noted that Cepheid's GeneXpert tuberculosis test has helped the situation, especially in high-burden countries, but in general, TB outcomes are still poor. An ideal diagnostic, he noted, must be able to identify the Mycobacterium species, determine all drug susceptibilities, identify predictors of virulence, and link the infecting organism to the source of acquisition. It must also be fast, usable at the point of care, and inexpensive. WGS, he said, has the potential to satisfy all of these criteria, though the physical technology characteristics still have a ways to go.

Recapping work that he and his colleagues published in December, Walker said that as a first step, they have now sequenced every Mycobacterium species and will soon make the sequences publicly available.

Last year, the group published an algorithm in The Lancet Infectious Diseases that they developed to use in combination with WGS on Illumina's MiSeq to retrospectively diagnose Mycobacterium infection and predict drug susceptibility. Because the MTB genome is so large, they took a candidate gene approach, first examining 23 genes already known to be associated with resistance. "We made no assumptions, and tried to create an algorithm to ID both known mutations and new mutations," Walker said. This involved dividing samples into a training set and a validation set, and using the former to test the latter.

One of the reasons their algorithm worked, Walker said, was a "serendipitous finding that most samples that are phenotypically [drug]-susceptible have no mutations in the genes being considered, but resistant phenotypes mostly had one mutation." Examining about 1,500 new patient samples, only 10 percent had a mutation that they'd never seen before, and using their algorithm, they were able to predict resistance with a sensitivity of 92 percent and specificity of 98 percent, "better than any existing test has achieved so far," Walker said.

Then, in work that was published in December in The Lancet Respiratory Medicine, the researchers put their method to the test in an eight-site pilot study, showing that it was comparable in terms of performance to standard diagnostic methods, but with faster turnaround time and potentially better cost effectiveness.

The researchers are now exploring implementing the test in routine clinical diagnosis. This is looking promising, Walker said, but there are still some major obstacles to overcome. First, they would prefer a test that works directly with primary samples, and not cultured samples, which is not a trivial undertaking and something they're partnering with other labs to solve.

"Second, can you work with a portable sequencer?" Walker said. The Illumina MiSeq, he noted, sits on a bench and needs a high-precision laboratory environment. As such, they've begun doing some sequencing with the Oxford Nanopore MinIon to see whether it is feasible for use in the field or at the point of care.

Finally, researchers need to be able to analyze their data, ideally locally, Walker said. To that end, colleagues in the laboratory of Zamin Iqbal at the Wellcome Trust Centre for Genetics in the UK have recently developed the Mykrobe predictor software, which can be used to analyze microbial genomes on a desktop or laptop computer, as opposed to needing cloud connectivity.

"Once you've got something that's portable, and doesn't need Internet, then you just need a power source," Walker said. He pointed out that one of WHO's 20-year TB action plan stipulations was to have a next-generation diagnostic test in place by 2025, so "we have 10 years to work on this aspect, and I think we're moving in the right direction."