A rRNA hybridization-based approach for rapid and accurate identification of diverse fungal pathogens

This study presents a rapid, multiplexed rRNA hybridization assay capable of accurately identifying diverse fungal pathogens from both cultured isolates and FFPE tissue within 8 hours, offering a significant improvement over current slow diagnostic methods.

The Gist

The "Fungal Fingerprint" Scanner: A New Way to Catch Invisible Invaders

Imagine your body is a bustling city. Usually, the police (your immune system) keep the peace. But sometimes, invisible invaders called fungi sneak in and start causing chaos. These aren't just the mushrooms you put on pizza; these are dangerous microscopic organisms that can cause life-threatening infections, especially in people with weakened immune systems.

The problem? Finding them is slow and hard.

Right now, doctors have to play a game of "wait and see." They take a sample, grow the fungus in a lab (like waiting for dough to rise), and then ask an expert to look at it under a microscope. This can take days or even weeks. By the time the answer arrives, the infection might have already spread.

This paper introduces a new, super-fast tool called Pan-Fungal Phirst-ID that acts like a high-tech "Fungal Fingerprint Scanner."

---

How It Works: The "Radar" Analogy

Think of every fungus as having a unique radio station playing inside it. This station broadcasts a specific signal made of rRNA (a molecule that is like the fungus's ID card).

1. The Probes (The Antennas): The scientists built a set of 91 tiny antennas (called probes). Each antenna is tuned to listen for a specific radio frequency.
   • Some antennas are super specific, listening only to one exact species (like "Candida albicans").
   • Others are broader, listening for entire families or groups (like "all Aspergillus fungi").
2. The Fingerprint: When a sample is tested, these antennas try to "lock on" to the fungus's radio signal. Because different fungi have slightly different signals, they create a unique pattern of which antennas light up and how bright they glow. This pattern is the Fungal Fingerprint.
3. The Match: A computer compares this new fingerprint against a massive library of known fingerprints. It asks, "Does this pattern look more like Aspergillus or Candida?"

The "Smart Detective" (Co-PILOT)

At first, the computer just looked at the brightest lights. But they realized a problem: sometimes the "family" antennas lit up so brightly that they drowned out the "species" antennas. It was like trying to hear a whisper in a rock concert.

So, they built a Smart Detective named Co-PILOT.

• How Co-PILOT works: Instead of guessing immediately, Co-PILOT plays a game of "20 Questions."
  • Step 1: "Is this fungus a mammal or a bird?" (It checks the broadest category).
  • Step 2: "Okay, it's a bird. Is it a hawk or an eagle?" (It narrows it down to the family).
  • Step 3: "Is it a Red-tailed Hawk or a Bald Eagle?" (It zooms in on the specific species).

By narrowing the search step-by-step, Co-PILOT ignores the "noise" and finds the exact match much more accurately.

The Results: Speed and Accuracy

The team tested this new scanner on two groups of fungi:

• The Training Class: 93 known fungi. The scanner got it right 83% of the time at the species level.
• The Final Exam: 54 new fungi the scanner had never seen before. With the help of the Smart Detective (Co-PILOT), it got it right 91% of the time!

Why is this a game-changer?

• Speed: It takes less than 8 hours from the time the sample arrives to getting an answer. Compare that to the days or weeks of traditional methods.
• No Growing Required: You don't need to wait for the fungus to grow. It works on the fungus itself, even if it's dead or stuck in old tissue.
• The "Old Photo" Test: They even tested it on FFPE tissue (tissue samples preserved in wax blocks, like old photos in an archive). Usually, these samples are useless for modern tests because the fungus is dead or the DNA is damaged. But because this test looks at RNA (which is abundant) and doesn't need to "amplify" (copy) the signal, it could identify fungi in these old samples. This is huge for patients who had surgery years ago and now have mysterious symptoms.

The Bottom Line

This new tool is like upgrading from a magnifying glass to a facial recognition camera.

• Old way: "Hmm, this looks like a fungus. Let's wait a week to see what it grows into."
• New way: "Scan complete. That is Aspergillus fumigatus. Start the antifungal medication immediately."

While there are still a few tricky "twins" (very similar fungi) that the scanner sometimes confuses, this technology offers a massive leap forward in saving lives by diagnosing fungal infections faster than ever before.

Technical Summary

1. Problem Statement

Invasive fungal infections (IFIs) are a growing global health threat, causing approximately 1.6 million deaths annually, particularly among immunocompromised populations. Current diagnostic methods suffer from significant limitations:

• Culture-based methods: Slow turnaround times (days to weeks), often delaying critical treatment.
• Serology: Limited species specificity and high rates of cross-reactivity or false positives.
• MALDI-TOF Mass Spectrometry: Requires culture growth and struggles with variable surface proteome expression across fungal life cycles.
• Sequencing (ITS/rDNA): While accurate, it often requires send-out testing with long turnaround times. Furthermore, primary clinical samples often contain scarce fungal DNA, leading to false positives from contaminants or false negatives due to PCR inhibition.
• FFPE Tissues: Formalin-fixed paraffin-embedded (FFPE) tissues, often the only specimen available for retrospective diagnosis, are difficult to culture and challenging for standard molecular workflows.

There is a critical unmet need for a rapid, sensitive, and broad-spectrum diagnostic tool that can identify diverse fungal pathogens directly from clinical samples without the need for culture or amplification.

2. Methodology

The authors extended their previously developed Phirst-ID (phylogeny-informed rRNA-based strain identification) platform, originally designed for bacteria and Candida, to create a Pan-Fungal Phirst-ID assay.

• Target Selection: The assay targets the 18S and 28S ribosomal RNA (rRNA) subunits. rRNA was chosen for its high abundance (allowing detection without amplification) and sequence conservation patterns that allow for taxonomic discrimination.
• Probeset Design:
  • A panel of 91 probes was designed to target 86 medically relevant fungal species (based on the WHO Fungal Priority Pathogen List).
  • The design includes 60 species-specific probes and 31 higher-order taxonomic probes (14 genus, 8 family, 5 order, 4 class).
  • Probes were selected to recognize regions of differential conservation, creating unique "fingerprint" patterns (Probeset Reactivity Profiles or PSRPs) for each species.
  • An outgroup of 56 environmental fungi was used during design to minimize cross-reactivity.
• Platform: The assay utilizes the NanoString nCounter system, a hybridization-based, amplification-free RNA detection platform.
• Sample Processing:
  • Cultured Isolates: Processed via crude lysis (bead-beating in Trizol) with minimal hands-on time (<30 mins).
  • FFPE Tissues: De-paraffinized and processed using RNA extraction kits, followed by bead-beating to release nucleic acids.
  • Hybridization: A modified protocol was used, including a 2-minute denaturation step at 95°C and a reduced hybridization time (1 hour vs. standard 16–24 hours) to accelerate the workflow.
• Classification Algorithm (Co-PILOT):
  • To address the issue where higher-order taxonomic probes (e.g., Class) often produced stronger signals than species-specific probes, the authors developed a hierarchical classifier named Co-PILOT (Complementarity-based Phylogeny-Informed Level-Oriented Traversal).
  • Mechanism: The algorithm starts with the full PSRP and uses Pearson correlation to identify the best-matching taxonomic class. It then iteratively narrows the search space, filtering out samples and probes not belonging to that class, and recalculating correlations at the Order, Family, Genus, and finally Species levels.

3. Key Contributions

• First Broad-Spectrum rRNA Hybridization Assay for Fungi: Successfully adapted the Phirst-ID strategy to cover 86 diverse fungal pathogens, moving beyond the previous Candida-only scope.
• Novel Hierarchical Classifier (Co-PILOT): Developed a sophisticated algorithm that corrects for signal bias in hybridization data, significantly improving species-level identification accuracy compared to simple Pearson correlation.
• Culture-Independent FFPE Pilot: Demonstrated the feasibility of identifying fungal pathogens directly from FFPE tissue, a high-value but technically difficult sample type where culture is impossible.
• Rapid Turnaround: Achieved a total assay time of <8 hours from sample to answer, with <30 minutes of hands-on time for cultured isolates.

4. Results

• Training Set Performance:
  • Tested on 93 clinical isolates spanning 32 species.
  • Simple Pearson correlation achieved 83% species-level accuracy.
  • The Co-PILOT classifier improved species-level accuracy to 89% in the training set.
  • Accuracy increased at higher taxonomic levels: 94% at the genus level and 95% at the family level.
• Validation Set Performance (Independent):
  • Tested on 54 independent clinical isolates.
  • Co-PILOT achieved 89.3% species-level accuracy, 96.2% family-level accuracy, and >98% for higher-order taxa.
  • After correcting one misidentified clinical isolate (reclassified from Mucor to Rhizopus based on assay data and confirmed by ITS sequencing), the final accuracy was 91% (species), 94% (genus), and 98% (family).
• FFPE Pilot Study:
  • Tested 27 archived FFPE samples.
  • Of 12 "high-confidence" samples (sufficient read counts), 8 were correctly identified to the species level, and 2 others were identified to the genus level or closest phylogenetic match.
  • One negative control showed a Candida albicans signature, highlighting potential contamination risks in sample processing.
• Limitations Identified:
  • Phylogenetically Close Species: The assay struggled to distinguish Cryptococcus neoformans vs. C. gattii and Scedosporium apiospermum vs. Lomentospora prolificans due to probe design issues (cross-reactivity or lack of signal).
  • Signal Bias: Higher-order probes naturally dominate signal intensity, necessitating the Co-PILOT algorithm.
  • Sample Complexity: The assay was tested primarily on monocultures; mixed infections remain a challenge.

5. Significance

This study presents a transformative approach to fungal diagnostics. By leveraging the high abundance of rRNA and a hybridization-based platform, the Pan-Fungal Phirst-ID assay offers a rapid (<8 hours), culture-independent, and amplification-free alternative to current gold standards.

• Clinical Impact: The ability to identify pathogens directly from FFPE tissues is particularly significant, as it allows for diagnosis in cases where culture is impossible (e.g., post-surgical resection of suspected fungal nodules).
• Scalability: The platform is compatible with instruments already present in many clinical pathology laboratories (NanoString), facilitating potential adoption without massive infrastructure overhaul.
• Future Directions: While currently research-use-only, the framework allows for easy expansion of the probeset. Future iterations will need to address specific probe redesigns for closely related species (like Scedosporium/Lomentospora) and optimize extraction protocols for primary clinical specimens to enable widespread clinical deployment.