Disentangling mitochondrial copy number variation and PCR amplification bias in DNA metabarcoding

This study investigates the limitations of quantitative DNA metabarcoding by demonstrating that high variation in mitochondrial copy numbers and stable PCR amplification biases prevent accurate biomass estimation, ultimately revealing fundamental constraints and the need for new methodological approaches despite the development of a mathematical model to correct for these biases.

The Gist

The Big Picture: Counting Bugs with a Flawed Magnifying Glass

Imagine you are a detective trying to figure out how many different types of bugs are in a jar. You can't count them one by one, so you decide to use a special "DNA magnifying glass" (metabarcoding). You take a tiny drop of water from the jar, zoom in on a specific genetic tag (a barcode) that every bug has, and count how many times that tag appears in your results.

The hope is: More tags = More bugs.

But this paper is a reality check. The authors found that this "DNA magnifying glass" is actually a bit broken. It doesn't just count bugs; it distorts the truth in two major ways.

---

The Two Big Problems

1. The "Battery Pack" Problem (Mitochondrial Copy Number)

Think of every bug as a house. Inside every house, there is a power plant (the mitochondria) that generates energy.

• The Issue: Some bugs have tiny power plants with just one battery. Others have massive power plants with 200 batteries.
• The Result: If you count the batteries (DNA copies) to guess how many houses (bugs) are there, you will be fooled. A bug with 200 batteries will look like 200 bugs, while a bug with 1 battery looks like just one.
• The Finding: The researchers found that even within the same species, the number of "batteries" varies wildly depending on the bug's age, health, and tissue type. You can't just assume a 1:1 ratio between DNA and actual bug weight.

2. The "Unfair Amplifier" Problem (PCR Bias)

To see the DNA, you have to copy it millions of times using a machine called a PCR (like a photocopier for genes).

• The Issue: The "photocopier" isn't perfect. It has a favorite. If the DNA of a specific bug matches the machine's settings perfectly, it copies it super fast. If the DNA has even a tiny typo (a mismatch), the machine struggles and copies it slowly.
• The Result: The bugs with the "perfect match" DNA end up with millions of copies in the final count, drowning out the bugs that were actually there in equal numbers but had a "typo" in their code.
• The Finding: This bias is specific to each species. Some bugs get amplified 100 times more than others, completely skewing the data.

---

The Failed Fix: "Just Run It Longer"

The researchers tried a clever trick to fix the "Unfair Amplifier." They thought: "If we run the photocopier for fewer cycles, maybe the bias won't have time to grow."

• The Analogy: Imagine a race where one runner is fast and one is slow. You thought that if you stopped the race early, the gap between them wouldn't be so huge.
• The Reality: They found that the bias happens in the very first two seconds of the race. Once the race starts, the fast runner is already way ahead, and running the race longer (more PCR cycles) doesn't change the ratio. The "photocopier" fixes its own mistakes after the first two rounds, locking in the bias immediately. So, changing the number of cycles didn't help.

---

The New Solution: The "Correction Factor" Cheat Sheet

Since they couldn't stop the bias, they decided to calculate exactly how biased the machine was and correct for it mathematically.

• The Analogy: Imagine you are weighing apples on a scale that is broken. The scale always adds 5 pounds to every apple. You can't fix the scale, but if you know it adds 5 pounds, you can write a rule: "Whatever the scale says, subtract 5 pounds to get the real weight."
• The Study's Method:
  1. They created "Mock Communities" (fake jars with known amounts of 5 specific bugs).
  2. They measured exactly how much each bug's DNA got amplified compared to a "Reference Bug" (Blattodea, a type of cockroach).
  3. They created a Correction Factor for each bug.
     • Example: If the machine amplifies the "Ant" DNA 30 times better than the "Cockroach," the math says: "Divide the Ant's count by 30 to get the real number."

The Result: When they applied this math, they could accurately predict how much DNA was actually in the jar, even though the machine had distorted it.

---

The Catch: Why This Isn't a Magic Bullet Yet

Even though they solved the math, the authors warn that we still can't perfectly count bugs in the wild yet. Here is why:

1. The "Battery" Problem is still there: Even if we fix the photocopier bias, we still don't know how many "batteries" (mitochondria) a wild bug has. A sick bug might have fewer batteries than a healthy one. So, we can count the DNA copies accurately, but we still can't perfectly translate that into "how many bugs" or "how much bug weight."
2. The "Unknowns" Problem: To use their math, you need to know every single species in the jar and have a correction factor for all of them. In a real forest or river, there are thousands of unknown species. If you miss even one, your math gets messy.
3. The Cost: To get these correction factors, you have to do expensive lab tests for every new species you want to study.

The Bottom Line

This paper is like a mechanic showing us exactly how a car engine is misfiring.

• Good News: We now understand why DNA counts are wrong (battery variation + photocopier bias) and we have a mathematical formula to fix the photocopier part.
• Bad News: We still can't perfectly turn those counts into a real "bug census" because nature is too variable and we don't have data for every bug on Earth yet.

In short: DNA metabarcoding is great for making a list of who is there, but using it to count how many are there is still a work in progress. We are getting the math right, but the biological reality is still a bit messy.

Technical Summary

1. Problem Statement

DNA metabarcoding is a powerful tool for biodiversity monitoring but faces a critical limitation: the inability to reliably quantify biomass or abundance from sequence read counts. The authors identify two primary sources of bias that decouple read counts from biological reality:

1. Mitochondrial DNA (mtDNA) Copy Number Variation: The number of mitochondrial marker gene copies (e.g., COI) per unit of biomass varies significantly among and within species, tissues, and life stages. This variation can be up to 200-fold, meaning sequence counts do not linearly reflect organismal biomass.
2. PCR Amplification Bias: Primer mismatches and sequence-specific differences lead to varying amplification efficiencies ($E$) during PCR. This causes certain taxa to be disproportionately over- or under-represented in the final sequencing data.

While strategies like PCR cycle calibration (varying cycle numbers to estimate initial template concentration) have been proposed to correct amplification bias, their efficacy remains controversial. Furthermore, the interaction between mtDNA copy number variation and PCR bias in complex, multi-species communities is not well understood.

2. Methodology

The study employed a rigorous experimental design using mock communities to isolate and quantify these variables:

• Mock Community Assembly:

  • Constructed 81 mock communities using five arthropod taxa spanning different orders: Acromyrmex (Hymenoptera), Blaptica dubia (Blattodea), Galleria mellonella (Lepidoptera), Gammarus (Amphipoda), and Hermetia illucens (Diptera).
  • Samples covered systematic gradients of biomass ratios (10%–90%) and various species combinations (2 to 5 species).
  • Total of 84 communities were assembled; 3 were excluded due to handling errors.

• Quantification Techniques:

  • Digital Droplet PCR (ddPCR): Used to quantify absolute mtDNA copy numbers for each species in every sample, serving as the ground truth for template abundance.
  • Metabarcoding: Performed using a two-step PCR protocol with two different primer sets targeting the COI gene:
    • Fwh2/Fwh2Rn: A degenerate primer set known for broad coverage.
    • BF3/BR2: A second primer set for validation.
  • PCR Cycle Calibration: For the Fwh2 primer, the first PCR step was run at 8 different cycle numbers (6, 8, 10, 12, 14, 16, 18, and 20 cycles) to test if read proportions shift with cycle number.

• Data Analysis:

  • Bioinformatic processing included demultiplexing, denoising, and taxonomic assignment (pooled at the order level).
  • A Representation Score was calculated to measure the deviation between relative reads and relative mtDNA copies.
  • A Mathematical Framework was developed to model non-exponential amplification bias and derive taxon-specific relative amplification efficiencies.

3. Key Contributions

The paper makes three major theoretical and methodological contributions:

1. Rejection of PCR Cycle Calibration: The study demonstrates that varying PCR cycle numbers does not alter the relative read proportions of taxa in a community. This contradicts the assumption that bias accumulates exponentially with cycles. The authors argue that primer mismatches are resolved after the first two cycles (when the original template is replaced by amplicons with perfect primer binding sites), rendering the amplification rate constant thereafter.
2. Mathematical Derivation of Non-Exponential Bias: The authors propose a mechanistic model where amplification is not purely exponential regarding bias. They derive equations to calculate relative amplification efficiency ($E_{rel}$) based on a reference taxon, allowing for the estimation of initial template copy numbers from read data without needing to vary cycle numbers.
3. Quantification of mtDNA Variation: The study provides empirical evidence of the massive variation in mtDNA copy numbers per unit of biomass across different arthropod orders, highlighting this as a fundamental barrier to biomass quantification.

4. Key Results

• mtDNA vs. Biomass: While mtDNA copy numbers scaled positively with input biomass ($\rho=0.65$), there was substantial scattering. Copy numbers varied significantly within and among species, confirming that biomass cannot be directly inferred from mtDNA counts without species-specific correction factors.
• Reads vs. mtDNA Copies: Uncorrected metabarcoding reads showed a significant but highly dispersed correlation with mtDNA copies ($\rho=0.75$).
  • Blattodea, Diptera, and Lepidoptera were relatively well-represented.
  • Amphipoda and Hymenoptera were severely under-represented (up to 2000-fold under-representation for Amphipoda).
• PCR Cycle Calibration Failure: Increasing PCR cycles from 6 to 20 did not change the relative proportions of reads. This confirms that bias is established early and remains stable, making cycle calibration an ineffective correction strategy.
• Correction Success:
  • Using the derived mathematical model, the authors calculated relative amplification efficiencies for each taxon relative to a reference (Blattodea).
  • Applying these efficiency corrections significantly improved the accuracy of predicted mtDNA copy numbers from read data.
  • For the Fwh2 primer, the correction was highly effective (Spearman $\rho=0.95$). For BF3, the improvement was also significant but less pronounced ($\rho=0.96$), as BF3 already had better inherent representation.

5. Significance and Limitations

Significance:

• Mechanistic Insight: The study clarifies that PCR bias in metabarcoding is likely non-exponential and driven by early-cycle primer-template interactions, challenging previous assumptions that bias accumulates over cycles.
• Correction Framework: It provides a viable mathematical approach to correct for amplification bias using relative efficiency factors, provided a reference taxon is present.
• Realistic Expectations: It underscores that while read counts can be corrected to estimate mtDNA copy numbers, estimating actual biomass or abundance remains highly inaccurate due to the inherent, uncorrectable biological variation in mtDNA copy numbers per cell/tissue.

Limitations & Future Directions:

• Practical Applicability: The correction method requires knowing the relative amplification efficiency of every species in a sample relative to a reference. In highly diverse environmental samples (e.g., Malaise traps with unknown species), this is currently impossible.
• Biomass Estimation: Even with perfect bias correction, the high variability of mtDNA copy numbers per unit of biomass prevents accurate biomass estimation.
• Future Needs: The authors suggest that future quantitative metabarcoding may require:
  • The use of nuclear single-copy markers (less copy number variation).
  • The development of universal spike-in standards.
  • Application to low-diversity samples (e.g., macroinvertebrate monitoring) where species lists are known.

In conclusion, the study delivers essential conceptual insights into the mechanics of PCR bias and mtDNA variation, offering a mathematical tool for correction while simultaneously highlighting the fundamental biological constraints that currently limit the quantitative application of DNA metabarcoding.