Estimation of adult census size from close-kin dyads in the malaria mosquito Anopheles gambiae

This study demonstrates that the close-kin mark-recapture (CKMR) method can be successfully adapted to estimate the adult census size of the malaria mosquito *Anopheles gambiae* by incorporating specific life-history traits, such as extreme variance in reproductive success and high clutch failure rates, into the statistical model.

The Gist

Imagine you are trying to count how many people live in a massive, bustling city, but you can't just walk down the streets and ask. The people are too fast, they hide in their homes, and if you try to tag them with a sticker to track them later, the stickers fall off, or the people get scared and run away.

This is the exact problem scientists face when trying to count mosquitoes, specifically the Anopheles gambiae, the mosquito that spreads malaria in Africa. They are everywhere, they live for only a few weeks, and traditional counting methods (like catching them, tagging them, and releasing them) are a logistical nightmare.

This paper introduces a clever new way to count them using genetic detective work, which the authors call "Close-Kin Mark-Recapture" (CKMR). Here is how it works, explained simply:

1. The Old Way vs. The New Way

• The Old Way (Tag and Release): Imagine trying to count fish in a lake by catching 100, tagging them with a bright red sticker, throwing them back, and then catching another 100 later to see how many have red stickers. If you catch very few red ones, you know the lake is huge. If you catch many, the lake is small.
  • The Problem: Mosquitoes are too small, too numerous, and too short-lived for this to work well.
• The New Way (The Family Tree): Instead of tagging them, the scientists just catch a bunch of mosquitoes and take a DNA sample (like a tiny drop of blood). They then look for family members in the mix.
  • The Logic: If you catch two mosquitoes that are brothers or sisters, they must have come from the same mother. If the population is huge, the odds of catching two siblings are tiny. If the population is small, the odds are higher. By counting how many "sibling pairs" you find, you can mathematically work backward to figure out how many total mosquitoes are out there.

2. The "Broken Egg" Surprise

When the scientists applied this method to a small island in Lake Victoria, they found something weird. They found lots of sibling pairs (mosquitoes that were brothers and sisters), but they found zero parent-offspring pairs (no mothers and their babies).

In a normal animal population, you'd expect to see some moms and their kids. Why was this different?

The Analogy: Imagine a baker who bakes 1,000 loaves of bread every day.

• The Old Expectation: You'd expect to see the baker (the parent) and a few loaves (the offspring) in the shop.
• The Reality: The baker is actually baking 1,000 loaves, but 990 of them are burnt or fall on the floor before they can be sold. Only 10 loaves actually make it to the shop.
• The Result: If you walk into the shop, you see 10 loaves. You might see two loaves that came from the same batch (siblings), but you will never see the baker with a fresh loaf because the baker is too busy baking the next batch, and the "fresh" ones from yesterday are all gone.

In the mosquito world, this means that most mosquito eggs never grow up. The "clutch failure" rate was estimated at 97.6%. Most mothers lay eggs, but those eggs die due to predators, drying up ponds, or bad weather. Only a tiny fraction of mothers successfully raise a family. This creates a "sweepstakes" effect: a few lucky moms have huge families, while most have none.

3. Solving the Puzzle

Because the scientists realized that "clutch failure" was the key, they built a new computer model that accounted for this. They treated the family relationships as a "hidden variable" (a secret they had to guess based on the DNA clues) rather than just saying "Yes, these are siblings" or "No, they aren't."

The Result:

• They estimated there are about 27,000 adult female mosquitoes on that small island.
• They confirmed that the "burnt bread" theory (clutch failure) was true.
• They calculated that the "effective" population (the number of mosquitoes actually contributing to the next generation) is much smaller than the total count, because so many eggs die.

4. Why This Matters

Why do we care about counting mosquitoes?

• Fighting Malaria: To stop malaria, scientists are developing "gene drives" (genetic modifications) to wipe out mosquito populations. To do this safely, they need to know exactly how many mosquitoes are there. If they release too few, the wild mosquitoes will win. If they release too many, it's a waste of money.
• Better Tools: This study proves that you don't need to catch and tag mosquitoes to count them. You just need to catch them, take a DNA sample, and look for family ties. It's like counting a crowd by finding people wearing matching t-shirts from the same family reunion.

The Bottom Line

This paper is a breakthrough because it took a method usually used for whales and fish and successfully adapted it for tiny, short-lived insects. It taught us that mosquito populations are like a lottery: most tickets (eggs) don't win, but the few that do win big, creating huge families that we can spot with our DNA microscopes. This new counting method will help scientists design better ways to stop malaria in the future.

Technical Summary

1. Problem Statement

Accurate estimation of adult mosquito abundance (census size, $N$) is critical for designing and evaluating vector control strategies, particularly for emerging genetic control methods like gene drives. However, traditional methods like Mark-Release-Recapture (MRR) are logistically difficult for mosquitoes and often suffer from low recapture rates and behavioral biases.

Close-Kin Mark-Recapture (CKMR) offers a promising alternative by using genetic data to identify relatives (kin pairs) within a population sample, treating them as "capture-recapture" events. While CKMR has been successfully applied to long-lived vertebrates (e.g., fish, mammals), its application to short-lived, highly fecund insects like Anopheles gambiae faces two major challenges:

1. Kinship Ambiguity: Distinguishing between different close-kin categories (e.g., full-siblings vs. half-siblings vs. avuncular pairs) is difficult with standard genetic markers due to overlapping likelihood distributions.
2. Life History Complexity: An. gambiae exhibits extreme variance in reproductive success. Field data suggests that most egg clutches fail to produce any adult offspring due to environmental factors (predation, desiccation), a phenomenon known as "sweepstakes reproduction." Standard CKMR models often assume more uniform reproductive success, which could lead to biased abundance estimates.

2. Methodology

The authors developed a novel framework to apply CKMR to An. gambiae by integrating high-resolution genotyping with a modified demographic model.

• Sampling and Genotyping:
  • Study Site: Jaana Island, Lake Victoria, Uganda (a small, isolated population).
  • Sample: 714 adult mosquitoes collected indoors over 20 days.
  • Genotyping Panel: A custom 291-locus microhaplotype amplicon panel was developed using GT-Seq. These markers are highly polymorphic (mean expected heterozygosity $\approx$ 0.8 in wild populations) and distributed across all chromosome arms, excluding inversions and heterochromatic regions.
• Kinship Inference:
  • Instead of using deterministic thresholds (which fail when kin categories overlap), the authors employed a probabilistic latent-kinship approach.
  • They treated the true kinship of each pair as a latent variable, drawing 1,000 realizations of kinship categories based on log-likelihood ratios to propagate uncertainty into the population size estimation.
• Demographic Modeling:
  • The authors adapted the CKMR framework to include a clutch failure parameter ($\mu_S$).
  • The model explicitly accounts for the probability that an entire egg clutch yields zero surviving adults, distinct from daily juvenile mortality ($\mu_J$).
  • Two models were compared:
    1. Base Model (PO+FS $\mu_S=0$): Assumes no clutch failure.
    2. Extended Model (PO+FS): Includes the clutch failure parameter.
• Validation:
  • Posterior Predictive Checks: Simulated data under both models to see which matched observed kin-pair counts.
  • Individual-Based Simulations: Used the SLiM forward-time simulator to generate synthetic populations with known parameters to test if the CKMR model could accurately recover them.

3. Key Results

• Kinship Patterns:
  • Among 714 mosquitoes, the analysis detected 60 high-confidence full-sibling (FS) pairs but zero parent-offspring (PO) pairs.
  • Full-siblings showed significant temporal clustering (collected 3 days apart) and some spatial clustering within villages, though dispersal between villages (2 km apart) was detected.
• Model Selection:
  • The Base Model failed to explain the data: it overpredicted PO pairs and underpredicted FS pairs.
  • The Extended Model (incorporating clutch failure) successfully matched the observed data (0 PO pairs, ~38 FS pairs).
• Parameter Estimates (Extended Model):
  • Adult Female Census Size ($N_F$): Estimated at 26,887 (95% Credible Interval: 6,979 – 146,011).
  • Clutch Failure Probability: The daily probability of a clutch failing is 22% (95% CrI: 15–31%). Over a 15-day juvenile development period, this results in a 97.6% probability that a clutch produces no adults.
  • Reproductive Variance: The high failure rate leads to extreme variance in reproductive success.
  • Effective Population Size ($N_e$): Estimated at 5,027 (95% PPI: 2,917 – 8,922), resulting in an $N_e/N$ ratio of approximately 10%.
• Validation:
  • Individual-based simulations confirmed that the CKMR model could accurately recover true population sizes and that the observed kin-pair counts in the field were consistent with a population size of ~27,000 females under the extended model.

4. Key Contributions

1. First Empirical CKMR Application to Mosquitoes: This study demonstrates that CKMR can be successfully applied to short-lived, highly fecund insects, provided the model accounts for their specific life history.
2. Latent-Kinship Framework: The authors successfully implemented a probabilistic approach to kinship inference that handles the ambiguity between kin types (e.g., full-sibs vs. half-sibs) without relying on error-prone deterministic thresholds.
3. Quantification of Clutch Failure: The study provides the first empirical quantification of clutch failure rates in a natural An. gambiae population, revealing that ~98% of clutches fail to produce adults. This explains the absence of parent-offspring pairs in the sample despite a moderate population size.
4. Discrepancy between $N$ and $N_e$: The results highlight a massive gap between census size and effective population size due to "sweepstakes" reproduction, a critical factor for understanding the evolutionary dynamics and efficacy of genetic control strategies.

5. Significance

• Vector Control: Accurate abundance estimates are essential for calculating release ratios for sterile insect techniques (SIT) or gene drives. This study provides a non-invasive, genetic method to obtain these estimates without the logistical burden of MRR.
• Gene Drive Design: The finding that $N_e$ is much smaller than $N$ (due to high reproductive variance) suggests that gene drives might spread more easily or behave differently than models assuming uniform reproduction would predict.
• Methodological Advancement: The study establishes a blueprint for applying CKMR to other short-lived, high-fecundity species where generations overlap and reproductive success is highly variable. It emphasizes that ignoring life-history specifics (like clutch failure) leads to severe biases in demographic inference.

In conclusion, the paper validates CKMR as a robust tool for mosquito ecology, provided that models are tailored to account for extreme reproductive variance and uncertainty in kinship assignment.