Rapid automated antifungal susceptibility testing system for yeasts based on growth characteristics

Jinhan Yu, He Chun, Tong Wang, Ge Zhang, Jin Li, Jingjia Zhang, Wei Kang, Yingchun Xu, Ying ZhaoFrontiers in Cellular and Infection Microbiology 2023 · DOI

TL;DR

This paper introduces Droplet 48, a novel automated system for antifungal susceptibility testing (AFST) in yeasts. By monitoring fluorescence changes in real-time and fitting growth characteristics, Droplet 48 offers a faster, more automated, and potentially more affordable alternative to existing methods, crucial for combating rising antifungal resistance.

What Problem Does This Paper Address?

Invasive fungal infections pose a significant and growing threat, exacerbated by increasing antifungal resistance and a lack of rapid, sensitive diagnostic tools. Current AFST methods, while essential for guiding treatment and tracking resistance, are often labor-intensive, expensive, and not widely accessible in clinical microbiology labs. This gap hinders timely and effective patient management, especially for immunocompromised individuals or those with complex underlying diseases. The study aims to address this by developing and validating a more automated and efficient AFST system.

Methods

The core innovation lies in the Droplet 48 (D48) system, which leverages fluorescence detection to monitor yeast growth in microdilution wells containing varying concentrations of antifungal agents. Unlike traditional methods that rely on visual inspection of turbidity or color change after a fixed incubation period, D48 continuously tracks fluorescence intensity over time. This allows it to fit a characteristic growth curve for each fungal isolate and antifungal agent combination.

The system uses resazurin, a redox indicator that turns fluorescent in the presence of metabolically active cells. By analyzing changes in fluorescence intensity, D48 can infer growth inhibition or killing by the antifungal agent. The system employs sophisticated algorithms to interpret the growth curve parameters (threshold, slope, acceleration) and determine the Minimum Inhibitory Concentration (MIC). A key feature is its ability to adapt to species-specific growth characteristics, further enhancing accuracy. The D48 system was compared against the established Sensititre YeastOne (SYO) colorimetric broth method.

Datasets

The study utilized 144 clinical fungal strains, including common pathogens like Candida albicans, Candida krusei, Candida tropicalis, Candida parapsilosis, Candida lusitaniae, Candida glabrata, Candida guilliermondii, and Cryptococcus neoformans. Additionally, less common but clinically relevant isolates such as Trichosporon asahii, Exophiala dermatitidis, and other rare bloodstream infection isolates were included to provide a comprehensive evaluation. Quality control strains (Candida parapsilosis ATCC 22019 and Candida krusei ATCC 6258) were also used.

Key Insights and Ideas

  • Real-time Growth Monitoring is Key: The shift from endpoint assessment to continuous fluorescence monitoring is a significant advancement, allowing for a more dynamic understanding of antifungal drug action.
  • Growth Characteristics as a Diagnostic Signal: Fitting growth curves based on fluorescence intensity offers a more nuanced approach than simple turbidity, potentially capturing subtle effects of antifungals.
  • Automation Reduces Subjectivity and Increases Throughput: Automating the process of inoculum preparation, reagent addition, incubation, and data analysis minimizes human error and speeds up the reporting of results.
  • Species-Specific Adaptations Enhance Accuracy: Recognizing that different fungal species have distinct growth patterns and adjusting interpretation algorithms accordingly is crucial for robust performance across a diverse range of pathogens.
  • Addressing "Trailing Growth" and "Eagle Effect": The study acknowledges phenomena like "trailing growth" (partial growth in wells with higher drug concentrations) and the "Eagle Effect" (paradoxical growth at higher drug concentrations), which can complicate MIC determination. While D48 aims to mitigate these, they remain areas for further investigation.

Key Results

The Droplet 48 system demonstrated strong performance when compared to the Sensititre YeastOne (SYO) method:

  • Reproducibility: 100% reproducibility within ±2 two-fold dilutions for all tested methods.
  • Essential Agreement (EA): For eight out of nine antifungal agents, EA was >90%. Posaconazole showed a slightly lower EA of 86.62%.
  • Categorical Agreement (CA): For four antifungal agents (fluconazole, caspofungin, micafungin, and anidulafungin), CA was >90%. Voriconazole had a CA of 87.93%.
  • Discrepancies: Only two Candida albicans isolates and anidulafungin showed a major discrepancy (MD) (2.60%). No very major discrepancies (VMD) were found.
  • Reportable Ranges: The reportable ranges of D48 were deemed appropriate for clinical fungal isolates in China.

Limitations and Future Work

The authors acknowledge several limitations and suggest future directions:

  • Optimization for Specific Agents: Further research is needed to optimize the detection performance for posaconazole and voriconazole, as these showed slightly lower agreement rates.
  • Broader Clinical Validation: While the study included a diverse set of isolates, further promotion and validation of the D48 system in clinical microbiology laboratories with a larger and more varied collection of clinical isolates are recommended.
  • Addressing "Trailing Growth" and "Eagle Effect": The paper notes that phenomena like "trailing growth" and the "Eagle Effect" can still pose challenges for precise MIC determination and require further investigation and potential refinement of the technology.
  • Cost-Effectiveness and Accessibility: While D48 is presented as potentially more affordable, a detailed cost-benefit analysis and wider accessibility in diverse clinical settings would be beneficial.

Summary updated

Abstract

Fungal pathogens are a major threat to public health, as they are becoming increasingly common and resistant to treatment, with only four classes of antifungal medicines currently available and few candidates in the clinical development pipeline. Most fungal pathogens lack rapid and sensitive diagnostic techniques, and those that exist are not widely available or affordable. In this study, we introduce a novel automated antifungal susceptibility testing system, Droplet 48, which detects the fluorescence of microdilution wells in real time and fits growth characteristics using fluorescence intensity over time. We concluded that all reportable ranges of Droplet 48 were appropriate for clinical fungal isolates in China. Reproducibility within ±2 two-fold dilutions was 100%. Considering the Sensititre YeastOne Colorimetric Broth method as a comparator method, eight antifungal agents (fluconazole, itraconazole, voriconazole, caspofungin, micafungin, anidulafungin, amphotericin B, and 5-flucytosine) showed an essential agreement of >90%, except for posaconazole (86.62%). Category agreement of four antifungal agents (fluconazole, caspofungin, micafungin, and anidulafungin) was >90%, except for voriconazole (87.93% agreement). Two <i>Candida albicans</i> isolates and anidulafungin showed a major discrepancy (MD) (2.60%), and no other MD or very MD agents were found. Therefore, Droplet 48 can be considered as an optional method that is more automated and can obtain results and interpretations faster than previous methods. However, the optimization of the detection performance of posaconazole and voriconazole and promotion of Droplet 48 in clinical microbiology laboratories still require further research involving more clinical isolates in the future.
BibTeX
@inproceedings{yu2023rapid,
  author = {Jinhan Yu and He Chun and Tong Wang and Ge Zhang and Jin Li and Jingjia Zhang and Wei Kang and Yingchun Xu and Ying Zhao},
  title = {{Rapid automated antifungal susceptibility testing system for yeasts based on growth characteristics}},
  booktitle = {Frontiers in Cellular and Infection Microbiology},
  year = {2023},
  doi = {10.3389/fcimb.2023.1153544},
}
Figure