Scalable genotyping in fixed transcriptomes resolves clonal heterogeneity via single-cell sequencing

The authors introduce Genotyping in Fixed Transcriptomes (GIFT), a scalable single-cell assay that simultaneously profiles whole transcriptomes and targeted somatic mutations with high accuracy, enabling the resolution of genotype-to-phenotype relationships and clonal heterogeneity in large cohorts and FFPE tissues.

The Gist

Imagine you are trying to understand a massive, chaotic city. You have two ways to look at it:

1. The "Who" (Genotype): You have a list of every person's DNA (their unique ID card).
2. The "What" (Phenotype): You have a live feed of what everyone is doing, saying, and thinking (their transcriptome).

For a long time, scientists could only look at these two things separately. They could take a census of the city's DNA, or they could watch the live feed of the people, but they couldn't easily link a specific person's ID card to their specific actions at the same time, especially in a crowd of millions.

This paper introduces a new super-tool called GIFT (Genotyping In Fixed Transcriptomes) that solves this problem. Here is how it works, using some everyday analogies.

The Problem: The "Blurred Photo" and the "Broken Library"

Previously, trying to link a person's ID to their actions had three big flaws:

1. Low Resolution: Old methods could only check 1 or 2 ID cards at a time. If you wanted to track a criminal gang (a cancer clone) that had 50 different ID markers, you were stuck.
2. The "End of the Book" Rule: Old methods could only read the ID if it was written at the very end of a sentence. But mutations (ID changes) can happen anywhere in the sentence.
3. The "Moldy Library": Most hospitals keep patient records in old, formalin-fixed jars (FFPE tissue). The DNA and RNA in these jars are like books that have been glued together and turned to mush. Old tech couldn't read them.

The Solution: The "Gap-Filling" Trick

The authors invented GIFT, which is like a clever detective technique to read these messy, glued-up books.

The Metaphor: The Bridge Builders
Imagine the RNA (the book) is a long road.

• The Old Way: You tried to read the whole road at once, but the road was broken.
• The GIFT Way: You send out two small construction crews (probes) to stand on either side of a specific pothole (the mutation). They leave a tiny gap between them.
• The Magic Step: A special "bridge builder" enzyme (a polymerase) walks across that gap, reading the road exactly as it is, and builds a bridge.
• The Result: Because the bridge is built only if the road matches a specific pattern, the system knows exactly who is standing there. If the road has a pothole (mutation), the bridge is built differently.

This "gap-filling" trick allows them to:

• Read hundreds of ID cards at once (scalability).
• Read IDs anywhere on the road, not just the end.
• Read books that were glued together (FFPE tissue).

The Big Experiments: What Did They Find?

1. The "Cell Line Mix-Up" (Testing the Tool)

They mixed three different types of leukemia cells (like mixing red, blue, and green marbles) and used GIFT to sort them.

• Result: GIFT was incredibly accurate (>99%). It could tell which marble was which, even though they were all jumbled together, and it could check 600 different ID markers at once.

2. The "Moldy Library" (FFPE Tissue)

They took old, preserved brain tumor samples (glioblastoma) from a hospital archive. These samples are usually useless for this kind of deep analysis.

• The Fix: They added a "de-gluing" step (heating the tissue) to make the RNA readable again.
• Result: They successfully reconstructed the family tree of the tumor. They could see which cells were related to each other and how the tumor evolved, even from samples that were years old.

3. The "Blood Cancer City" (MPN Cohort)

This was the big one. They looked at blood samples from 35 patients with a blood cancer called Myeloproliferative Neoplasms (MPN). They analyzed 712,000 cells in total.

• The Discovery: They found that the famous JAK2 mutation (a common driver of this cancer) doesn't just turn on a switch; it changes the volume of the cell's voice.
  • The Analogy: Imagine a radio station.
    • No Mutation: The radio is off.
    • One Copy of Mutation: The radio is at low volume, playing "Inflammation" music.
    • Two Copies of Mutation: The radio is blasting at max volume.
  • They discovered that the more copies of the mutation a cell had, the louder the "Inflammation" signal was. This "volume knob" effect was something previous tools were too blurry to see.

4. The "Family Tree" (Lineage Tracing)

In one patient who was getting sicker (turning from blood cancer to leukemia), GIFT acted like a time machine.

• It found a tiny, rare mutation that was hiding in a small group of cells.
• Later, as the patient got worse, that exact group of cells took over the whole system.
• The Lesson: GIFT can spot the "bad apple" before the whole barrel rots, helping doctors understand how cancer evolves before it becomes aggressive.

Why Does This Matter?

Think of GIFT as the Google Maps for Cancer.

• Before, we had a blurry satellite photo (bulk sequencing) or a street view of one house (single cell, but no ID).
• Now, GIFT gives us a high-definition, real-time map where we can see exactly who lives in every house and what they are doing, even if the map is drawn on old, crumpled paper.

This allows doctors and scientists to:

1. Track Evolution: See how cancer clones grow and change over time.
2. Personalize Treatment: Understand why some patients get sicker than others based on their specific "family trees."
3. Use Old Data: Unlock the treasure trove of millions of old hospital samples that were previously impossible to study this way.

In short, GIFT is a new lens that lets us see the invisible connections between our DNA and our health, turning a blurry mess into a clear, actionable picture.

Technical Summary

1. The Problem

Single-cell transcriptomics has revolutionized the understanding of cellular heterogeneity, but a critical gap remains: the inability to link somatic mutations (genotype) directly to transcriptional states (phenotype) within the same cell at a large scale.

• Limitations of Existing Methods:
  • Scalability: Current RNA-based genotyping methods (e.g., GoT) typically target only 1–3 loci per cell, insufficient for resolving complex clonal structures that require tens of mutations.
  • Positional Bias: Many methods rely on poly(A) priming, restricting targets to the 3' ends of transcripts. Somatic mutations can occur anywhere, often in the middle of transcripts.
  • Sample Compatibility: Existing methods generally require fresh, viable cells with intact RNA. They are incompatible with Formalin-Fixed Paraffin-Embedded (FFPE) tissues, which constitute the vast majority of clinically annotated human samples but suffer from RNA degradation and crosslinking.
  • Accuracy: Genotype inference is highly sensitive to errors; rare technical artifacts can distort clonal lineage reconstruction.

2. Methodology: GIFT (Genotyping In Fixed Transcriptomes)

The authors introduce GIFT, a novel assay that integrates targeted mutation detection with whole-transcriptome analysis (WTA) in single cells, compatible with both fresh and fixed (FFPE) samples.

• Core Innovation: Gapfilling Reaction
  • GIFT builds upon the 10x Genomics Flex workflow, which uses pairs of ssDNA probes to hybridize to RNA.
  • Instead of ligating adjacent probes directly, GIFT designs probe pairs with a deliberate "gap" (2–10 nucleotides) spanning the mutation site.
  • A strand-displacement-deficient (SDD) DNA polymerase (specifically Sulfolobus DNA polymerase IV) is used to fill this gap using the native RNA template as a guide.
  • This "gapfill" product is then ligated, barcoded, and sequenced. The sequence of the filled gap reveals the native genotype (wildtype vs. mutant).
• Key Technical Optimizations:
  • Enzyme Selection: The use of a polymerase lacking 3'→5' and 5'→3' exonuclease activity and strand displacement activity prevents the displacement of the 3' probe, ensuring accurate gapfilling.
  • Chemical Enhancements: The addition of betaine increases genotyping yield by ~60% by reducing secondary structures and melting temperature disparities.
  • FFPE Adaptation: A high-temperature decrosslinking step (80°C with Proteinase K) was introduced to reverse formalin-induced crosslinks, restoring transcript accessibility for probe hybridization in archival tissues.
  • Probabilistic Genotyping Framework: To handle noise (PCR template switching, sequencing errors), the authors developed a computational pipeline (giftwrap) that:
    1. Estimates the probability of PCR-mediated template switching per UMI.
    2. Defines a constrained feature set of plausible alleles.
    3. Uses an edit-distance model to assign unmatched gapfill sequences to the most likely allele.
    4. Aggregates evidence across UMIs to output genotype probabilities rather than discrete calls, allowing for flexible downstream filtering.

3. Key Contributions

1. Unprecedented Scalability: GIFT enables the simultaneous profiling of hundreds of targeted mutations (up to 611 loci tested) in a single cell, a ~100-fold increase over previous transcriptome-based genotyping methods.
2. FFPE Compatibility: It is the first method to successfully perform joint single-cell genotyping and transcriptomics on archival FFPE tissues, unlocking a massive repository of clinical samples.
3. Positional Flexibility: Unlike 3'-biased methods, GIFT can target mutations anywhere within a transcript, including the middle of genes.
4. High Accuracy: The method achieves >99% genotyping accuracy, even in noisy FFPE samples, by leveraging a probabilistic error model.

4. Key Results

The authors validated GIFT across three distinct biological contexts:

• Validation in Cell Lines:

  • In a mixture of three leukemia cell lines (HEL, K-562, SET-2), GIFT accurately genotyped 30 variable loci and 65 control sites.
  • It achieved 99.3% accuracy and successfully distinguished heterozygous vs. homozygous states.
  • Comparison with "dual probe" methods showed GIFT had significantly higher accuracy (>99% vs. ~90% for dual probes) without compromising gene expression data quality.

• Application to Archival FFPE (Glioblastoma):

  • Applied to FFPE-preserved Glioblastoma (GBM) tissue. Without decrosslinking, yields were low; with the optimized protocol, gene expression recovery increased ~10-fold.
  • Successfully reconstructed clonal lineage trees from a single FFPE sample, distinguishing between sequential and parallel evolution of mutations (e.g., EGFR and AXL mutations) that bulk sequencing could not resolve.
  • Reduced the space of possible lineage trees from 73 (based on bulk data) to 6 based on single-cell co-occurrence data.

• Cohort-Scale Profiling in Myeloproliferative Neoplasms (MPN):

  • Profiled 712,664 cells from 35 MPN patients and 2 healthy controls.
  • Genotyping Depth: Captured up to 55 loci per cell (median 164), a massive improvement over prior studies (3–8 loci).
  • Phenotypic Impact of JAK2 V617F:
    • Revealed that JAK2-mutant cells are enriched in neutrophils and show impaired apoptosis.
    • Identified a dosage-dependent gradient of interferon-γ (IFN-γ) response: Wildtype < Heterozygous < Homozygous mutant cells showed increasing IFN-γ signatures in monocytes.
    • Demonstrated that JAK2 mutation drives specific transcriptional programs in Hematopoietic Stem Cells (HSCs), including upregulation of coagulation genes and downregulation of proliferation markers (E2F/MYC) in fibrotic stages.
  • Lineage Tracing in AML Transformation:
    • In a patient progressing from MPN to Acute Myeloid Leukemia (AML), GIFT reconstructed the phylogeny, identifying a low-frequency NRAS mutation in a specific clone that later expanded during transformation.
    • Linked specific clones (defined by CALR, TP53, EZH2, ZRSR2) to distinct transcriptional programs and differentiation trajectories within the same patient.

5. Significance

• Bridging Genotype and Phenotype: GIFT provides a rigorous framework to infer the functional impact of somatic mutations by directly comparing mutated and wildtype cells within the same individual and cellular context, eliminating confounding inter-patient variability.
• Unlocking Clinical Archives: By enabling multi-omic analysis of FFPE tissues, GIFT allows researchers to study somatic evolution in the vast, clinically annotated archives of human disease that were previously inaccessible to single-cell resolution.
• Precision Medicine: The ability to resolve clonal heterogeneity and link specific mutations to drug resistance or disease progression (e.g., MPN to AML) offers new insights for therapeutic targeting and understanding disease evolution.
• Scalability: The platform's ability to process hundreds of thousands of cells with hundreds of targets makes it feasible to study complex somatic evolution across large patient cohorts, moving beyond case studies to population-level insights.

In summary, GIFT represents a paradigm shift in single-cell multi-omics, overcoming the scalability, accuracy, and sample-type limitations of previous methods to provide a comprehensive view of how somatic mutations drive cellular diversity and disease progression.