The macroecology of viral coinfection

Using the largest standardized wildlife disease surveillance dataset to date, this study reveals that while viral coinfection in wild animals is rare, it occurs more frequently than expected by chance and is significantly influenced by host age, species-specific differences, and the heightened risks associated with captive wildlife trade.

The Gist

Imagine the natural world as a massive, bustling city where every animal is a resident. In this city, it's not uncommon for a resident to catch a cold, the flu, and a stomach bug all at the same time. This is called coinfection.

For a long time, scientists studied these "double infections" mostly in sterile laboratories with just one type of animal, or by looking at a few specific cases in the wild. It was like trying to understand traffic patterns in a whole country by only watching a single intersection in a parking lot.

This paper is like finally getting a satellite view of the entire city. The authors used a massive, decade-long dataset called PREDICT, which is like a giant global health survey that tested over 65,000 wild animals (mostly bats, rodents, and birds) for viruses. They wanted to answer three big questions: How often do animals get double-infected? Who gets it most? And do certain viruses like to hang out together?

Here is what they found, explained with some everyday analogies:

1. Double Infections are Rare, but Not Random

Out of 65,000 animals, only about 223 were found to have two or more viruses at once. That's less than 1%.

• The Analogy: Imagine walking through a crowded stadium. You might see one person with a red hat and another with a blue hat. Finding someone wearing both a red and a blue hat at the same time is rare.
• The Twist: Even though it's rare, it happens more often than chance would predict. Specifically, viruses like coronaviruses, paramyxoviruses, and flu viruses seem to have a "buddy system." They show up together more often than you'd expect if they were just randomly bumping into each other.

2. The "Age" Surprise: Bats vs. Rodents

The researchers looked at how age affects getting sick.

• Rodents (The "Young and Reckless"): Young rodents were more likely to be coinfected than older ones. This makes sense; their immune systems are still learning the ropes, like a new driver who hasn't mastered the rules of the road yet.
• Bats (The "Immune Superheroes"): Here is the weird part. Young bats were less likely to be coinfected than adult bats.
• The Analogy: Think of adult bats as people who have lived in a city for 20 years. They've been exposed to so many "germ parties" that they've built up a massive library of defenses. Over time, they accumulate different viruses that just hang out in their bodies without making them sick. It's like a house that has been lived in for so long that it has a few different tenants (viruses) who have learned to share the space peacefully. Young bats haven't had time to collect these "roommates" yet.

3. The Danger of "Crowded Apartments"

The study found that animals living in captivity (like in zoos, farms, or markets) were much more likely to be coinfected than animals living in the wild.

• The Analogy: Imagine a wild animal living in a huge forest with plenty of space. Now, imagine that same animal crammed into a small apartment building with 50 other species. In the "apartment," germs spread like wildfire because everyone is breathing the same air and touching the same surfaces.
• The Risk: The study highlighted that rats and mallard ducks in captivity were the biggest offenders. This is a major red flag for human health. When different species are forced to live in close quarters, viruses can swap genetic material (like mixing paint colors) to create brand new, super-viruses that could jump to humans.

4. The "Sampling Bias" (The Camera Angle Problem)

The authors were very honest about a limitation: their data might be a bit skewed.

• The Analogy: Imagine you are trying to count how many people are wearing sunglasses. If you only take photos in a sunny park, you'll think everyone wears sunglasses. If you take photos in a dark cave, you'll think no one does.
• The Reality: The PREDICT project tested more bats than any other animal, and they tested more in Asia than in Africa. So, while they found patterns, those patterns might be partly because they looked at certain places and animals more than others.

The Big Takeaway

This paper is a wake-up call. It tells us that the natural world is a complex web of interactions. Viruses aren't just acting alone; they are interacting with each other and with their hosts in surprising ways.

• Why it matters: If we want to stop the next pandemic, we can't just look for one virus at a time. We need to understand how viruses mix and match, especially in places where humans and wildlife are forced together (like wet markets or farms).
• The Future: The authors suggest we need better "surveillance cameras" (more data) and smarter ways to look at the whole picture, not just isolated parts, to understand how these viral communities work.

In short: Nature is a crowded party where viruses are constantly bumping into each other. Sometimes, they team up. And when humans crowd the party too much, that's when things can get dangerous.

Technical Summary

1. Problem Statement

Viral coinfection (the simultaneous infection of a host by two or more viruses) is a critical driver of disease ecology, influencing host immune responses, transmission dynamics, and the emergence of novel pathogens through recombination or reassortment. However, empirical understanding of coinfection is limited by:

• Scale: Most research relies on laboratory experiments or small-scale field studies, lacking global perspective.
• Data Fragmentation: Wildlife disease data is historically siloed across surveillance, management, and research, with a lack of standardized, host-level, multi-parasite datasets.
• Knowledge Gaps: There is insufficient understanding of how host traits (e.g., age, taxonomy, captivity), environmental factors, and specific virus-virus interactions shape coinfection patterns at a macroecological scale.

2. Methodology

The authors conducted the first global-scale macroecological analysis of viral coinfection using data from the PREDICT project (2015–2019), the largest standardized wildlife disease surveillance initiative to date.

• Data Source: The study utilized the PREDICT-2 dataset, comprising 65,662 unique animals and 126,952 unique specimens (mostly bats and rodents) from West/Central Africa and South/East/Southeast Asia.
• Data Processing:
  • Integrated animal sampling data (species, sex, age, captivity status) with viral test results (consensus PCR assays).
  • Cleaned taxonomy (e.g., reclassifying Thottapalayam virus to Hantaviridae) and excluded human samples.
  • Defined "coinfection" as the detection of two or more unique viruses in a single animal.
• Statistical Analysis:
  • Prevalence Calculation: Computed infection and coinfection rates across 16 taxonomic groups.
  • Null Modeling: Used permutation testing (1,000 iterations) to compare observed coinfection rates against expected rates based on individual virus prevalence for bats, rodents, and birds.
  • Generalized Linear Models (GLMs): Modeled coinfection status (binary: single vs. multiple) as a function of predictors including geographic region, taxonomic group, sex, age class, captivity status, and testing effort (number of virus families tested).
    • All-animals model: Included bats, birds, rodents, and shrews.
    • Bats-only model: Added cave-roosting behavior as a predictor.
    • Rodents-only model: Focused on sex and age.
  • Network Analysis: Constructed coinfection matrices at the virus family and individual virus levels to identify non-random associations. Binomial tests with Bonferroni correction were used to determine if specific virus pairings occurred more frequently than expected by chance.

3. Key Results

A. Prevalence and Distribution

• Overall Rarity: Coinfection is rare in the dataset, detected in only 223 of 65,662 animals (0.34%). Among infected animals, 93.18% had single infections, while 6.45% had double infections and 0.37% had triple infections.
• Taxonomic Variation:
  • Rodents had the highest coinfection rate among infected individuals (9.11%), followed by birds (8.01%) and bats (6.22%).
  • Bats were the most sampled group (60.9% of specimens) and contributed the highest absolute number of coinfected animals (128).
• Geographic Bias: Infected animals were predominantly found in South, East, and Southeast Asia.

B. Drivers of Coinfection (GLM Findings)

• Host Taxonomy: After controlling for covariates, birds and rodents were significantly less likely to be coinfected than bats (OR = 0.24 and 0.55, respectively). This supports hypotheses regarding unique bat immune adaptations that may allow for viral tolerance and persistence.
• Age Effects (Opposing Trends):
  • Rodents: Juveniles/subadults showed a trend toward higher coinfection than adults (OR = 1.84, p = 0.066).
  • Bats: Juveniles were significantly less likely to be coinfected than adults (OR = 0.13, p = 0.006). This suggests that coinfection in bats may accumulate over a lifespan or that maternal antibodies protect young bats.
• Captivity Status: Wild animals in captivity were significantly more likely to be coinfected than free-ranging wild animals (OR = 2.12). This effect was driven largely by rats and mallards in captivity, highlighting risks associated with the wildlife trade and farming.
• Sampling Bias: Higher testing effort (more virus families tested) correlated with higher observed coinfection rates, indicating that detection probability is a major confounder.

C. Virus-Virus Associations

• Non-Random Patterns: Observed coinfection rates exceeded simulated expectations for all three major taxonomic groups.
• Key Viral Pairs: Significant non-random associations were found between:
  • Coronaviruses (CoVs) and Paramyxoviruses (PMVs)
  • CoVs and Influenza A (Orthomyxoviridae)
  • Multiple CoVs (specifically in rodents)
  • PMVs and Rhabdoviruses (in bats)
• Network Centrality: In the individual virus network, poultry-associated viruses (Influenza A, Newcastle Disease Virus) had the highest degree centrality, acting as hubs for coinfection.

4. Key Contributions

1. First Macroecological Synthesis: This is the first study to analyze viral coinfection patterns across multiple continents and diverse host taxa using a single, standardized global dataset.
2. Quantification of "Rare" Events: It establishes that while coinfection is statistically rare in surveillance data, it occurs significantly more often than random chance, particularly among specific viral families (CoVs, PMVs, Influenza).
3. Identification of Host-Pathogen Dynamics: The study reveals divergent life-history strategies, specifically the opposing age-coinfection relationships in bats (accumulation with age) versus rodents (higher in youth).
4. Risk Assessment of Anthropogenic Settings: It provides empirical evidence that captivity and the wildlife trade significantly increase coinfection risks, creating hotspots for viral recombination and spillover.

5. Significance and Implications

• Public Health: The findings underscore the importance of monitoring wildlife trade and captive settings, where high coinfection rates could facilitate the generation of recombinant or reassortant viruses with pandemic potential (e.g., novel coronaviruses or influenza strains).
• Ecological Theory: The results challenge and refine theoretical frameworks regarding parasite community ecology, suggesting that host immune history (e.g., bat tolerance) and life stage are critical determinants of coinfection outcomes.
• Future Surveillance: The authors argue for a shift in surveillance strategies:
  • Moving from single-pathogen to multiplex/metagenomic approaches to capture the full virome.
  • Implementing longitudinal sampling to distinguish between transient and persistent coinfections.
  • Standardizing data sharing to enable future meta-analyses of host-parasite networks.

Limitations: The study relies on cross-sectional data (limiting causal inference), is heavily skewed toward bats due to PREDICT's design, and is subject to detection biases inherent in PCR-based surveillance. However, it provides a foundational baseline for future targeted research.