High-resolution retrospective single cell lineage tracing with mutable homopolymers

The paper introduces RETrace2, a high-resolution single-cell dual-omic method that leverages highly mutable homopolymers and sparse methylation profiling to simultaneously reconstruct detailed cell lineage trees and identify cell types with unprecedented accuracy in both in vitro and in vivo models.

The Gist

Imagine trying to figure out the family history of a massive, bustling city. You want to know who moved in first, which families stayed in the same neighborhood, and how the city grew from a single house into a sprawling metropolis.

In biology, this "city" is your body, and the "families" are the trillions of cells that make up your organs. For a long time, scientists could only guess at this history or had to tag cells with artificial "name tags" that often ran out or got confusing.

This paper introduces RETrace2, a new, super-powered detective tool that lets scientists read the natural "diaries" written inside our cells to reconstruct their entire family tree.

Here is how it works, explained through simple analogies:

1. The Problem: Reading a Faded Diary

Every time a cell divides (makes a copy of itself), it makes tiny, accidental typos in its DNA. These are called mutations. Over time, these typos accumulate, creating a unique history for every cell.

• The Old Way: Previous tools tried to read these typos, but they were like trying to read a diary written in pencil that had been left in the rain. The "typos" were too few, too hard to find, or the tools made so many mistakes reading them that the family tree looked like a mess.
• The New Tool (RETrace2): The authors realized that some parts of the DNA are like sticky notes that get messed up much faster than others. They call these homopolymers (long strings of the same letter, like "AAAAAA"). Because these "sticky notes" change so often, they act like a high-speed camera, capturing a new "snapshot" of the cell's history almost every time it divides.

2. The Upgrade: From a Flashlight to a Super-Scanner

The researchers didn't just find better "sticky notes"; they built a whole new way to read them.

• The "Sticky Note" Discovery: They found that these single-letter repeats (homopolymers) are nearly twice as informative as the old types of repeats they used before. It's like switching from reading a book with big, spaced-out letters to reading a book with dense, detailed paragraphs.
• Fixing the Noise: Homopolymers are tricky to read because they are so repetitive (like trying to count a long line of identical ants). The team invented a new protocol to stop the "reading errors." They used better enzymes (the "scissors" that cut the DNA) and a new type of sequencer (the "camera") that doesn't get confused by long strings of the same letter.
• The Result: They increased the number of "clues" they could find in a single cell by 21 times and the number of shared clues between cells by 98 times. This is the difference between trying to solve a mystery with 5 clues versus having 500.

3. The Test: The "Ground Truth" Game

To prove their tool works, they played a game with a known answer.

• The Lab Test: They took a single cell, let it multiply into a huge family in a petri dish, and then used RETrace2 to build the family tree. Because they knew exactly how the cells were related (the "ground truth"), they could check if the tool got it right.
• The Result: With the new tool, they reconstructed the family tree with 100% accuracy. Even in "normal" cells (which don't mutate as fast), they could still trace the lineage back about 60 generations. In "fast-mutating" cells (like cancer cells or mice with a specific gene defect), they could theoretically trace back fewer than 5 cell divisions. That's like being able to tell exactly which generation great-great-grandparent you are descended from, just by looking at a single cell today.

4. The Real-World Adventure: The Mouse City

Finally, they took the tool into the wild. They used a special mouse that has a "hyper-active" mutation rate (making the DNA typos happen faster, like a diary being written in fast-forward).

• The Mission: They collected cells from the mouse's brain, kidney, and liver.
• The Discovery: They built a massive family tree showing how these organs developed.
  • They found that the brain, kidney, and liver are like three different neighborhoods in the city. While some early "founding families" spread out to build all three neighborhoods, other families stayed strictly in one.
  • They even figured out what kind of cell each one was (e.g., a brain cell vs. a liver cell) by reading a second layer of data (methylation) from the same sample. It's like reading the family tree and the cell's ID card at the same time.

Why This Matters

Think of RETrace2 as a time machine for biology.

• Before: We could only see the "present day" of our cells or get a blurry, low-resolution picture of the past.
• Now: We can watch the movie of development in high definition. We can see exactly how a single fertilized egg splits and specializes into a complex human being.

This technology could help us understand:

• Development: How we grow from a baby to an adult.
• Disease: How cancer starts and spreads (clonal evolution).
• Aging: How our cells change over a lifetime.

In short, the authors turned a blurry, static photo of our cellular history into a crisp, high-definition movie, allowing us to finally see the full story of how we are built.

Technical Summary

1. Problem Statement

Retrospective lineage tracing, which infers cell history from naturally occurring somatic mutations, offers a powerful alternative to prospective genetic barcoding (e.g., CRISPR-based methods) for studying development and disease in unmodified organisms. However, existing retrospective methods face significant limitations:

• Low Marker Density & Resolution: Traditional markers like mitochondrial DNA (mtDNA) suffer from complex inheritance dynamics (heteroplasmy), while single nucleotide variations (SNVs) have mutation rates too low for high-resolution tracking without prohibitive costs.
• Technical Noise: Previous methods using microsatellites (short tandem repeats) were limited by low mutation rates of di- to hexa-nucleotide repeats and high technical errors during amplification and sequencing, particularly with homopolymers (mononucleotide repeats).
• Incomplete Readouts: Many methods fail to simultaneously capture lineage information and cell identity (cell type), requiring separate assays.

2. Methodology: RETrace2

The authors present RETrace2, a dual-omic single-cell method that simultaneously reconstructs lineage trees and identifies cell types. The workflow involves:

• Marker Selection (Homopolymers): Instead of di- to hexa-nucleotide repeats, RETrace2 targets mononucleotide microsatellites (homopolymers). These are shown to be approximately 1.9-fold more informative per locus than conventional repeats due to higher mutation rates (polymerase slippage).
• Dual-Omic Library Preparation:
  • Input: Single cells or nuclei are isolated via FACS.
  • Digestion: Genomic DNA is digested with two enzymes: MseI (enriches A/T-rich homopolymers for lineage tracing) and MspI (targets C/G-rich regions for methylation profiling).
  • Amplification: A single-primer PCR step performs linear amplification to generate high-fidelity copies before exponential amplification, reducing early-cycle errors.
  • Splitting: The library is split into two paths:
    1. Lineage Path: Hybridization capture using a custom biotinylated probe panel to enrich homopolymer targets.
    2. Identity Path: Bisulfite conversion for reduced representation bisulfite sequencing (scRRBS) to profile DNA methylation.
• Error Mitigation Strategies:
  • PCR Optimization: Reduced total PCR cycles by ~75% (from 44 to 11) and utilized SeqAMP polymerase, which demonstrated superior fidelity for homopolymers compared to Kapa HiFi or Q5.
  • Sequencing Platform: Adopted the Element Bio AVITI platform, which maintains high accuracy for long homopolymers (up to 30 bp) where Illumina platforms show significant error rates.
  • Probe Expansion: Expanded the probe set ~8-fold to include abundant short repeats (10–14 bp), increasing target capture efficiency.
• Data Analysis:
  • Lineage: Genotypes are called using HipSTR; phylogenetic trees are reconstructed using neighbor-joining algorithms based on binary genetic distances.
  • Cell Typing: Sparse methylation profiles are mapped to a reference atlas using pairwise dissimilarity scores (mean absolute difference in methylation fractions) to infer cell identity.

3. Key Contributions

• Identification of Homopolymers: Demonstrated that mononucleotide repeats are superior lineage markers, offering nearly double the information content of di- to hexa-nucleotide repeats.
• Technical Breakthroughs: Achieved a ~21-fold increase in the median number of markers captured per cell (from 1,662 in v1 to ~35,404 in v2) and a **98-fold increase** in shared markers per cell pair.
• Artifact Reduction: Systematically addressed homopolymer-specific errors (stutter, amplification bias, sequencing errors) through optimized polymerases, reduced cycle counts, and advanced sequencing hardware.
• Dual-Omic Integration: Successfully combined high-resolution lineage tracing with cell-type identification from the same single cell.

4. Key Results

• In Vitro Benchmarking (HCT116):
  • In microsatellite-unstable HCT116 cells, RETrace2 achieved 100% accuracy in reconstructing the ground-truth lineage tree using homopolymers, compared to only 64% accuracy with di- to hexa-nucleotide repeats.
  • Down-sampling analysis confirmed that homopolymers require significantly fewer loci to achieve the same resolution.
• Microsatellite-Stable Cells (3T3L1):
  • Demonstrated viability in normal tissues with low mutation rates, achieving 70.1% accuracy in reconstructing lineage trees separated by ~62 cell divisions.
  • Projected resolution of fewer than five cell divisions in microsatellite-unstable systems (e.g., cancer or MSH2-deficient models).
• In Vivo Application (Msh2-Deficient Mouse):
  • Applied to 152 single cells from the brain, kidney, and liver of a 6.5-week-old Msh2 knockout mouse.
  • Reconstructed a high-resolution phylogenetic tree revealing polyclonal organ development.
  • Clonal Analysis: Identified that ~33% of major clades were tissue-biased (restricted to one organ), while the rest were multipotent, supporting a "competition model" of early development.
  • Cell Typing: Successfully inferred cell types (e.g., GABAergic vs. Glutamatergic neurons) from sparse methylation data, mapping terminal identity onto the lineage tree.

5. Significance

• High-Resolution Developmental Biology: RETrace2 provides a scalable, retrospective tool to reconstruct continuous, high-resolution lineage trees without genetic engineering, applicable to both model organisms and potentially humans.
• Overcoming Technical Barriers: By solving the "error problem" associated with homopolymers, the method unlocks a vast theoretical reservoir of ~1 million microsatellites in the human genome, potentially allowing the resolution of every single cell division in a lifespan.
• Multi-Omic Insight: The ability to simultaneously track lineage and cell state allows researchers to correlate developmental history with differentiation trajectories, offering new insights into tissue homeostasis, disease progression, and the timing of lineage restriction events.
• Cost and Scalability: The authors project that with automation, per-cell costs can drop below $10, making organism-wide lineage studies economically feasible.

In summary, RETrace2 represents a major leap forward in retrospective lineage tracing, transforming highly mutable homopolymers from a technically challenging marker into a robust, high-resolution tool for mapping the cellular history of complex organisms.