When a patient comes into the hospital with sepsis, doctors begin treatment with a broad-spectrum antibiotic before knowing the microorganism at play and before knowing which antibiotics can kill it. There just isn’t time to wait.

This scenario plays out again and again with bacterial infections, particularly in life-threatening situations. Once doctors identify the bacteria causing the infection and the most effective antibiotics to treat it, they then tweak the drug type and its dosage, if needed. But before then, the use of broad-spectrum antibiotics, or the wrong antibiotic, can fuel the development of antibiotic-resistant bacteria.

The standard way to determine the best antibiotic for a particular bacterial infection is a growth-based antimicrobial susceptibility test (AST) — also called a broth microdilution test — where clinical microbiologists test a series of antibiotics at different concentrations against a culture of bacteria and look for which antibiotics at which concentrations inhibit bacterial growth.

When I started in the lab, everything was very manual. The impact on patient care is mind blowing to me.” 
- Andrea Prinzi, bioMérieux 

For sepsis cases where the bacterial density in blood is low, clinical microbiologists must first confirm the presence of bacteria in blood. They do this using a blood culture to increase bacteria numbers until they reach a detectable level. Unlike other types of bacterial cultures, blood cultures contain anticoagulants due to the high amount of blood present and use sensors to detect carbon dioxide as a surrogate for bacterial growth. Generally, it can take a few hours to days for a blood culture to give a positive result, depending on the organism and its initial density in the blood. Then, microbiologists use the positive blood culture to further grow the bacteria overnight on plates (18-24 hours). Once grown, the isolated microorganisms undergo AST using broth microdilution (16-20 hours incubation). From here, clinical microbiologists can determine the lowest antimicrobial concentration that can kill the microorganism, the minimum inhibitory concentration (MIC).

While clinical microbiologists have performed this gold-standard broth microdilution method for decades, “it’s very technically challenging to do,” said Thao Truong, a clinical microbiologist at the University of Washington. Plus, using this method means that it can take days before a clinician knows which antibiotics will actually work against a particular infection.

As an alternative, scientists are developing a myriad of new techniques to determine antimicrobial susceptibility faster and with less hands-on time for clinical microbiologists, ultimately leading to better care for people with serious bacterial infections.

Automation: from detecting turbidity to volatile organic compounds

One of the first ways researchers made ASTs faster and easier was through automation. The first of these systems, developed by researchers at Pfizer and Abbott Laboratories, came on the market in the 1970s, shortly followed by the AutoMicrobic System, the predecessor to the now widely-used VITEK® systems from bioMérieux (1).

Now, many of the clinical microbiology labs use automated broth microdilutions with manufactured panels, said Truong. “The big difference is the setup [and the readout] are more automated. You’re not visually inspecting these by eye,” she said. “That’s a lot more practical and much more feasible in a busy hospital lab setting.”

An updated version of the VITEK® system, called VITEK® 2, tests several different concentrations of antibiotics and measures turbidity, an indicator of bacterial growth, every 15 minutes to generate growth curves. Because of these additional measurements, it’s possible for the instrument to detect growth earlier compared to waiting to take one reading after an overnight culture. The device compares these curves to thousands of known microbial growth patterns. From there, the instrument determines the MIC. This system requires the same positive blood culture and overnight culture as the gold-standard method, but it generally provides a MIC after 5-18 hours of growth — depending on the organism or antibiotic tested — compared to the standard 16-20 hours broth microdilution takes.

The Food and Drug Administration has cleared VITEK® 2 for testing bacteria and yeast, but a newer variation called VITEK® REVEAL™ just gained approval earlier this year for identifying gram-negative bacteria directly from a positive blood culture. This means skipping the overnight culturing step required for the VITEK® 2 and saving 18-24 hours. Instead of relying on turbidity measurements, VITEK® REVEAL™ measures the volatile organic compounds (VOCs) microorganisms emit using small molecule sensors that sit over each well.

Measuring bacterial nanomotions

Though automating broth microdilution assays allows clinical microbiologists to process more samples faster, many are still hindered by the need to wait for bacteria to grow before running the assay. Research groups are developing new methods to shorten, or even eliminate, the overnight culturing step by measuring other unique characteristics of the bacteria.

Resistell, a biotech company developing a rapid AST device called Phenotech AST, is one of these teams. They use atomic force microscopy and machine learning to measure small nanomotions from bacteria to determine antibiotic susceptibility (2). Nanomotions are “a ubiquitous phenomenon that all cells do due to metabolic activity and cellular activities,” said Alexander Sturm, Chief Scientific Officer at Resistell (3). 

Their Phenotech AST device consists of a cantilever that acts like “a little diving board,” said Sturm. One side can swing while the other side is fixed. When bacteria attached to the cantilever move, they cause deflections in the cantilever that a detector monitors. Bacteria that are sensitive to the antibiotic would cause fewer deflections as they become more metabolically inactive and slowly die. In contrast, bacteria that are resistant to the antibiotic would increase the deflections as they respond to added stress from the drug. 

To classify whether bacteria are resistant or susceptible to a particular antibiotic, Phenotech AST uses a machine learning algorithm to find commonalities in deflection characteristics from strains that are susceptible as well as commonalities between the resistant strains. 

The team tested this method with Escherichia coli and Klebsiella pneumoniae isolates against four clinically important antibiotics from two different classes. They found that the algorithm correctly predicted susceptibility and resistance between 89.5 and 98.9 percent of the time, in line with standard clinical microbiology methods.

“VOCs are a metabolic byproduct that precedes cell division, so you can actually get an estimate of growth faster,” said Andrea Prinzi, the Field Medical Director at bioMérieux. The VITEK® REVEAL™ produces results in 5.5 hours on average directly from a positive blood culture.

Unlike traditional methods, this process only requires around 500 bacteria. “That’s one advantage. You [need] much less inoculum,” Sturm said. This means clinical microbiologists can start the test after a six-hour blood culture instead of waiting for additional overnight growth. Once cultured, strain identification happens in 30 minutes using MALDI-TOF mass spectrometry, and then the Phenotech AST determines antimicrobial susceptibility profiles in two hours. “Two hours is actually the magic number because then you can do it still in a [hospital] shift,” said Sturm.

Resistell is adapting their device to assess multiple bacterial samples at once so that hospitals can test different drugs at different concentrations.

Microfluidics monitors responses from individual cells

To eliminate the need for culturing altogether, Johan Elf, a biophysicist from Uppsala University, is developing a microfluidics technique to monitor cell growth on the single-cell level. These microfluidic chips contain thousands of cell “traps” that act like a sieve to capture single bacterial cells in long, narrow channels. When scientists expose a portion of the traps to antibiotics, they can use microscopy to monitor growth rates of individual cells.

Elf and his team first tested this system in 2017 and found that it correctly classified 49 clinical E. coli isolates from urinary tract infections (UTIs) in less than 30 minutes (4). Since then, they’ve expanded their assay to include species identification using fluorescence in situ hybridization after AST (5). They designed these initial tests for UTIs, and the tests were the focus of Astrego, a company Elf co-founded in 2017. When Sysmex Corporation acquired the company in 2022, they brought the test to market.

However, blood samples are quite a bit different from UTI samples: While there are generally some blood cells along with bacteria in a urine sample positive for a UTI, there are many more blood cells compared to bacterial cells in a blood sample. To overcome this, Elf introduced an additional centrifugation step to separate bacterial cells from blood cells. They could then lyse any remaining blood cells without affecting the bacteria. In a study, which has not yet been peer reviewed, they used the microfluidic traps to detect low densities of E. coli (nine colony forming units (CFU) per ml), K. pneumoniae (seven CFU/ml), and Enterococcus  faecalis (32 CFU/ml) from blood (6). This contrasts with bacterial densities after a positive blood culture which can reach over one million CFU per ml (7). Sample sources with higher bacterial cell densities load faster onto the chip than samples with lower bacterial cell densities, but in terms of sample density required for the platform, “there is no lower limit,” Elf said.

Aside from the speed advantage of microfluidics-based AST, Elf described many other advantages of working at the single-cell level, such as revealing small fractions of the population that are resistant and co-infections with different species.

The microfluidic approach may also accelerate diagnosis of slow-growing bacteria such as Mycobacterium tuberculosis, which have a doubling time between 16 and 22 hours, compared to 20 minutes for E. coli.

“It would be a month or two months to get the results from classical AST,” said Sturm.

Elf’s team at Uppsala University applied their microfluidics platform to M. tuberculosis and enabled rapid AST in just one day in a study that has not yet been peer reviewed (8). The researchers at Resistell are also working on applying the Phenotech AST to hasten antimicrobial susceptibility profiling of M. tuberculosis.

Introducing rapid AST to the clinic

While the VITEK® systems are already in the clinic, other rapid ASTs are emerging there too. “There has been a real explosion in [the development of] rapid tests,” said Elf.

Troung added, “Something that's becoming more and more standard across clinical labs across the country are molecular methods direct from blood [cultures].” Methods like polymerase chain reaction (PCR) can help identify the organism and their resistance genes in just a couple of hours, and next-generation sequencing methods are becoming more accessible.

“There’s still a question of when the phenotypic and the genotypic tests will be best. I think that remains to be thought about and when they complement each other,” said Elf. Prinzi also noted that researchers can’t get a MIC from a genotypic test. “[Fast] phenotypic approaches are very exciting to see because it gives us so much information and it aligns with how [clinical microbiology laboratories] have always run,” she added.

Despite these developments, the path for rapid ASTs to enter the clinic can be challenging. “Diagnostics need to be cheap,” said Sturm. “This is difficult for new technologies to be developed.” Aside from cost, he noted that the ability to automate, multiplex, and turn around samples quickly is key for diagnostics used in clinical microbiology laboratories.

But it’s not just advances in technology that can propel rapid AST forward. Clinicians also need to consider how rapid tests would integrate into existing systems and practices.

“The challenging part for UTI is that [AST] diagnostics are currently not used. It would be a change of the whole market sector to require a diagnostic before prescribing an antibiotic — although it would be good for the fight against antibiotic resistance,” said Elf.

Troung said that although rapid antibiotic susceptibility testing “was shortening the time to result from three days to, let's say, less than one day, it didn't make as much of a difference clinically, unless clinicians actually acted on it too.” To see change, communication workflows between clinical microbiology labs and clinicians would also need an overhaul. “The biggest roadblock for us is being able to interface with providers to make sure that the results are actionable in a prompt manner. Logistically, that’s challenging,” she added.

Even so, ASTs have already come a long way since the early days spent culturing bacteria and inspecting culture tubes manually. Prinzi, who was a clinical microbiologist for 13 years before joining bioMérieux, noted how far the field has come in that short amount of time. “When I started in the lab, everything was very manual. The impact on patient care is mind blowing to me.”

References

1. Wheat, P.F. History and development of antimicrobial susceptibility testing methodology. J Antimicrob Chemother  48, 1–4 (2001).
2. Sturm, A. et al. Accurate and rapid antibiotic susceptibility testing using a machine learning-assisted nanomotion technology platform. Nat Commun  15, 2037 (2024).
3. Kasas, S. et al. Detecting nanoscale vibrations as signature of life. Proc Natl Acad Sci USA   112, 378-81 (2015).
4. Baltekin, Ö. et al. Antibiotic susceptibility testing in less than 30 min using direct single-cell imaging. Proc Natl Acad Sci USA  114, 9170-9175 (2017).
5. Kandavalli, V. et al. Rapid antibiotic susceptibility testing and species identification for mixed samples. Nature Commun  13, 6215 (2022).
6. Miguélez, M. et al. Culture-free Rapid Isolation and Detection of Bacteria from Whole Blood at Clinically Relevant Concentrations. bioRxiv (2024).
7. Chandrasekaran, S. et al. Direct-from-Blood-Culture Disk Diffusion To Determine Antimicrobial Susceptibility of Gram-Negative Bacteria: Preliminary Report from the Clinical and Laboratory Standards Institute Methods Development and Standardization Working Group. J Clin Microbiol  56, e01678-17 (2018).
8. Tran, B. et al. One-day phenotypic drug susceptibility testing for Mycobacterium tuberculosis variant bovis BCG using single-cell imaging and a deep neural network. bioRxiv (2024).