Leapfrogging within laboratories, or the hare and the tortoise revisited
AMR surveillance information is typically generated from sequential culture-based methods of bacterial isolation, identification and susceptibility testing (figure 1). Most bacterial culture media are comprised of cheap components and other requirements are simple; modest upgrades from the time of Robert Koch.3 Superficially, this might appear to be an appropriate technology when resources are constrained given its cost, apparent simplicity and long history, but in practice this is not the case.7 Outcomes from LMIC Quality Assurance programmes indicate that, when performed, identification and susceptibility testing is associated with high error rates outside of highly controlled environments such as provided by quality assured microbiology laboratories.7 15 16 To date, establishing quality-assured laboratory networks has proven complex and costly for many resource-limited parts of the world, including many sites in Africa.4 5 This, in turn, limits diagnostic stewardship in which clinicians and healthcare workers understand recommended guidelines, perform appropriate diagnostic tests (particularly bacterial culture if indicated), and use local AMR surveillance data to provide the most optimal treatment.17
1. Shortening turn-around times
The first step in traditional bacteriology is pathogen recovery. Paradoxically, waiting for ‘fast-growing’ pathogens to multiply is the major reason why routine bacterial culture is so slow. For some pathogen-specimen combinations, ‘Gold-Standard’ culture methods only recover the pathogen in a fraction of instances, so leapfrog solutions with greater sensitivity will have global value. Cases in point include cerebrospinal fluid processing in meningitis, particularly in pretreated patients, and low-abundance blood-borne bacteria like Salmonella enterica serovariety Typhi, a key AMR priority pathogen. To address these limitations, which are not large in well-resourced settings, technology has largely used ‘tortoise’ advancements to culture-based protocols which marginally extend the sensitivity, specificity and speed attainable for identifying a huge diversity of bacteria and AMR phenotypes. Even the most innovative ‘tortoise’ advancements speed up or eliminate one or more incubation steps but typically still require some growth.18–20 More sensitive or alternate growth-detection technologies can drastically reduce turn-around times and represent longer stride tortoise strategies, often with unfortunately hefty price tags. Automated culture, identification and sensitivity testing systems such as BACTEC, BacT/Alert or VersaTREK (blood culture) and VITEK and VITEK-MS, Phoenix, Bruker MALDI Biotyper or Sensititre (identification and/or susceptibility testing), as well as DNA-leakage or early growth-detection tests18–20 accomplish this and, in some cases, are easier to quality assure than traditional protocols. Even shorter incubation times could be possible from ultrasensitive detection platforms in development.21 22 Because they still require basic specimen and isolate management and are expensive to institute and run, these generally do not constitute leapfrog technologies for LMICs although some have been deployed.
Challenging or dangerous-to-culture pathogens have provided intrinsic motivation for bypassing culture and provide leapfrog models. Analogous technologies for detecting antibacterial resistance largely do not exist but these new methods could serve as important templates. A recently described Lab-on-a-Chip, built from complementary metal-oxide-semiconductor technology coupled with loop-mediated isothermal amplification addresses common Plasmodium resistance single nucleotide polymorphisms (SNPs) in malaria23 and could feasibly do the same for bacterial resistance conferred by SNPs. While culture-free technologies offer significant promise, many alternatives to culture narrow the range of organisms that can be identified and most have the added disadvantage of not capturing the infecting organism for archiving and future research.
Antimicrobial panels for susceptibility testing are selected based on the identity of infectious pathogens and naming and subtyping are the most complicated and least generalisable protocols by traditional testing. It is sometimes possible to base identification on bacterial cell membrane composition, or its impact on the host immune system. These options open the possibility of antigen-based or antibody-based diagnostics, which, in turn, present the possibility of field-ready platforms such as microfluidic systems, simpler mobile-readable lateral-flow devices or at least protocols to remove one or more of the sequential steps needed for routine identification by culture.24 25 In a few instances, a surface-expressed resistance-conferring protein can serve as a marker for AMR but those represent the exception.26 All of these approaches could be difficult to multiplex and until that challenge is addressed, their use for multiaetiologic syndromes where bacteria predominate as pathogens will be limited.
2. Promises and pitfalls of nucleic acid amplification tests
One oft-cited leapfrog technology for AMR surveillance that offers the opportunity to bundle pathogen identification, subtyping, and in many cases resistance profiling too, is nucleic acid amplification technology (NAAT), for example by PCR, in which oligonucleotide primers bind to a complementary DNA target, permitting specific amplification and thence detection of one or more signature fragments. All nucleic acid-based tests, including NAAT and whole-genome sequencing (WGS; discussed in the next section), have the disadvantage of identifying resistance genotypes rather than phenotypes. As some genes, including many encoding resistance, are not expressed and new resistance genes can only be identified after they are known, nucleic acid-based AMR surveillance cannot completely replace phenotypic methods, which must continue to be performed at some level in each health system.
In spite of this caveat, NAAT is widely used in research, has long since superseded culture in diagnostic virology laboratories but as with culture, PCR is an easy thing to do badly. NAATs often depend on samples being appropriately taken from patients and delivered in a timely fashion to a quality assured facility, although NAAT is certainly faster than culture. The limitations of NAAT are exquisitely illustrated by the challenges seen in both LMIC and HIC settings during the first few months of the COVID-19 pandemic when real-timePCR was the only diagnostic option for the disease and could only be performed in very few diagnostic laboratories. However, the subsequent scale-up in testing in virtually every country in the world demonstrates the potential of these technologies for select applications.
When performed from first principles, NAAT tests are exceptionally prone to contamination, which is then difficult to detect, but some of this shortfall can be addressed by test format innovation, and microfluidics advances have made it possible to automate temperamental preparatory steps.27 Lab-on-chip, discs and cards can effectively bypass inadequate human resources, infrastructure and sometimes cold-chains.28 The GeneXpert MTB/RIF platform is an example of one such leapfrog technology and the BioFire FilmArray represents another. The former has been more widely deployed in LMIC settings and consists of automated PCR system initially and most commonly used for tuberculosis diagnostics and rifampicin sensitivity testing. As with BioFire, it requires less training than conventional PCR as all the specimen processing steps are completed within the cartridge and has therefore circumvented many of these problems. GeneXpert MTB/RIF has been widely deployed in a range of LMIC settings and appears to be cost-effective in at least some of them.29–31 However, even this system shares the challenge presented by low copy number bacteria, which may certainly be the case in paucibacillary mycobacterial disease. Potential solutions to insensitivity include new technologies and applications of nucleic acid-based methods such as single-cell whole genome sequence Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR)-based detection32 as well as free DNA capture as applied in fetal abnormality and cancer screening tests.33
For relatively difficult-to-culture bacteria, including mycobacteria, and sexually-transmitted bacteria, NAAT-based tests have overcome diagnostic and surveillance roadblocks in LMIC laboratories. Xpert and BioFire tests that detect priority resistant pathogens that are easier to culture, such as methicillin-resistant Staphylococcus aureus and carbapenemase-producing Enterobacteriaceae are commercially available and approved for use at the point of care.34–37 However, these are not marketed or prioritised in low-income settings even when the Xpert hardware already exists in admittedly overstretched tuberculosis programmes, where it has been applied to other diagnostics, most notably for Ebola and now COVID-19.38 Even in the best resourced settings, NAAT cannot compete with culture in terms of the diversity of bacteria and AMR types it can identify as there is a limit to the degree to which assays are multiplexed. In AMR surveillance, NAAT is comparable to the hare that raced ahead of culture, but which has yet to cross the line into routine surveillance on a large scale and may well be overtaken by other technologies.11
3. Whole-genome sequencing-based surveillance in LMICs
Beyond those bacterial pathogens that are challenging to culture, today’s exemplar leapfrog technology remains WGS, which has been revolutionised by the introduction of inexpensive, rapid and high throughput next-generation sequencing technology. Such equipment still requires significant laboratory infrastructure and advanced equipment maintenance and information technology support.39 40 Nonetheless, second-generation sequencing technologies have been used in local outbreak situations, such as the Ebola, Zika, COVID-19 and Lassa outbreaks of the current decade where portable MinION sequencers generated actionable information,41 41–43 and this technology is eminently suitable for surveillance of AMR. The diversity of sequencing options may well grow but for genomic surveillance, the present challenge is increasing their applicability both by increasing the number and types of patient specimens that reach this type of analysis as well as the scope of stakeholder epidemiologists that can access and use the information.
Sequencing requires upstream DNA preparation and manipulation steps, which currently limit its use in some settings, but progress has been made in miniaturising and automating these using microfluidic devices.27 44 While we can expect these to continue to improve and converge into combined, portable devices, field-ready WGS platforms still have the same challenges as culture and NAAT systems; a given patient must have an adequate sample taken to confer high sensitivity. Furthermore, WGS creates large data files which need analysis at source by highly skilled bioinformaticians and adequate internet connectivity. Despite these challenges, WGS combines advantages of culture and NAAT, operating in the absence of bacteriology-specific laboratory infrastructure while retaining the relative speed of NAAT, and the diverse diagnostic capability of culture WGS provides finetyping information for any species along with identification and susceptibility data at no additional cost.45
Current routine application of WGS in AMR surveillance uses sequence derived from isolates. A bigger leap can be made by sequencing direct-from-specimen (figure 1). This has been accomplished for LMIC specimens, generating in the case of meningitis, valuable information that would not otherwise have been obtained.46 It is, however, costlier because specimens contain a lot of reaction inhibitors, including host and commensal cells, all of which contain non-target DNA. Therefore, direct-from-specimen sequencing requires enriching for pathogen DNA in silico after expensively generating total metagenomic DNA or enriching for pathogen DNA in clinical specimens before sequencing, an approach that is still in development.