In vivo multiomic Perturb-seq with enhanced nuclear gRNA capture

The authors developed in vivo multiomic Perturb-seq, a novel platform that overcomes inefficient nuclear gRNA recovery to enable high-fidelity, single-nucleus multiomic profiling, which they used to uncover cell-type-specific transcriptomic and epigenomic phenotypes of neurodevelopmental disorder risk genes in the developing cortex.

The Gist

Imagine you are trying to understand how a specific instruction manual (a gene) affects the behavior of a city's population (cells). In the past, scientists could look at the city's daily activities (gene expression) or its zoning laws (chromatin accessibility), but they struggled to do both at the same time, especially when the city was inside a living organism like a mouse brain.

The biggest problem was like trying to find a specific lost receipt in a messy attic. When scientists tried to study brain cells, they had to isolate the "nucleus" (the command center of the cell). Unfortunately, the "receipts" proving which instruction manual was being tested (the gRNA) were often left behind in the attic's outer rooms (the cytoplasm) and got lost during the cleanup. Without the receipt, they couldn't tell which cell had which instruction.

Here is what the Zheng et al. team did to solve this:

1. The "Velcro" Trick (Nuclear Anchoring)

Imagine you have a very important note (the gRNA) that you need to keep safe inside a secure vault (the nucleus). Usually, when you open the vault door, the note flies out and gets lost.

The researchers invented a molecular "Velcro" system. They attached a special hook to the note and a matching loop to the inside wall of the vault door. Now, even when they open the door to take the vault out, the note stays stuck to the wall. This ensured that the "receipt" stayed with the cell nucleus, so they never lost track of which instruction was being tested.

2. The "Super-Scanner" (Enhanced Capture)

Once they had the note stuck to the wall, they needed a scanner to read it. They upgraded their scanner (the BD Rhapsody machine) by adding custom "magnets" specifically designed to grab onto these notes. They also used a special "zoom-in" technique (nested PCR) to make sure even faint whispers of the note could be heard clearly.

3. The "Living City" Experiment

With their new, super-reliable system, they went into a developing mouse brain (the living city) to test 16 different "instruction manuals" (genes) known to be linked to neurodevelopmental disorders.

• The Setup: They injected a virus carrying these instructions into the brains of mouse embryos.
• The Result: They successfully mapped out over 84,000 individual brain cells, seeing both what genes were turned on and how the DNA was packaged, all while knowing exactly which instruction each cell received.

The Big Discovery: Context is King

The most exciting part of their findings is that one size does not fit all.

Imagine a teacher (a gene) giving a lecture.

• In a classroom of kindergarteners (one type of brain cell), the lecture might cause them to run around wildly.
• In a classroom of high schoolers (a different type of brain cell), the exact same lecture might make them sit quietly and take notes.

The researchers found that genes like Mef2c acted like the teacher who only changed the behavior of specific "classrooms" (cell types). In one type of neuron, it changed the DNA packaging and gene activity significantly. In a neighboring type of neuron, it barely made a ripple.

Why This Matters

Before this, scientists had to guess how these genes worked by looking at the whole brain as a big soup, or by testing cells in a dish (which isn't the same as a living brain).

This new method is like having a high-definition, multi-angle security camera for the living brain. It allows scientists to:

1. See the cause and effect: "We changed Gene X, and this specific cell type reacted this way."
2. Understand the "Why": They can see if the reaction happened because the DNA was unzipped (chromatin) or because the cell just started making more proteins (transcriptome).
3. Target Diseases Better: Since neurodevelopmental disorders often affect specific types of brain cells, this tool helps doctors understand exactly which cells are broken, paving the way for treatments that fix the problem without hurting the healthy cells.

In short, they built a better net to catch the "receipts" of genetic experiments, allowing them to finally read the full story of how our brains develop and what goes wrong when diseases strike.

Technical Summary

1. Problem Statement

While pooled single-cell CRISPR screens (Perturb-seq) have successfully linked genetic perturbations to transcriptomic phenotypes in vitro and in vivo, integrating epigenomic readouts (specifically chromatin accessibility via ATAC-seq) into these workflows has been a significant bottleneck.

• The Core Challenge: In vivo screens require the isolation of nuclei (nuclei-based preparations) rather than whole cells to preserve tissue integrity and enable ATAC-seq chemistry. However, standard nuclei isolation protocols deplete cytoplasmic RNA. Since guide RNAs (gRNAs) in many CRISPR systems are transcribed in the nucleus but often rely on cytoplasmic transcripts for detection, this leads to inefficient recovery of gRNA transcripts.
• Consequence: Low gRNA recovery results in poor assignment of perturbation identity to single nuclei, making it impossible to reliably link specific genetic perturbations to multiomic (transcriptomic + epigenomic) phenotypes in complex in vivo tissues.

2. Methodology: In Vivo Multiomic Perturb-seq

The authors developed a novel platform combining nuclear transcript anchoring, gRNA-specific capture, and targeted amplification to overcome the limitations of nuclei-based multiomics.

A. Vector Engineering (The "Anchoring" Strategy)

To prevent gRNA loss during nuclei isolation, the team engineered a three-part AAV-compatible vector system:

1. RNA Hairpin-Binding System: They incorporated RNA hairpins (MS2 or PP7) into the gRNA transcript. These hairpins bind to specific coat proteins (MCP or PCP) fused to a nuclear localization signal (NLS) and an outer nuclear membrane anchor. This physically tethers the gRNA transcript to the nuclear envelope, preventing its loss during lysis.
2. Polyadenylated gRNA Transcript: Similar to the CROP-seq design, the vector expresses a U6-driven gRNA for genome editing and a separate, polyadenylated transcript derived from the gRNA sequence. This ensures the transcript is stable, recoverable, and compatible with standard poly-A selection methods used in single-nucleus RNA-seq (snRNA-seq).
3. AAV Delivery: The system utilizes Adeno-Associated Virus (AAV) for in vivo delivery, which is superior to lentivirus for in vivo applications due to lower immunogenicity and better tropism.

B. Capture and Amplification Workflow

The authors adapted the BD Rhapsody single-cell multiome workflow:

• Custom Bead Modification: They modified BD Rhapsody capture beads with custom oligonucleotides complementary to the polyadenylated gRNA-derived transcript.
• Nested Target-Enrichment (TE) PCR: Instead of relying solely on in vitro transcription (IVT), they implemented an optimized nested PCR step to specifically amplify the gRNA transcripts from the captured cDNA.
• In Vivo Application: The system was tested in mouse embryos (E13.5) via intracranial injection into the lateral ventricles, followed by nuclei isolation at postnatal day 14 (P14) and P28.

3. Key Contributions

• Technical Innovation: The first successful integration of CRISPR perturbations with joint snRNA-seq and snATAC-seq in vivo.
• Enhanced Recovery: The combination of nuclear anchoring (MS2/PP7) and custom bead modification increased single-gRNA assignment rates by 3.8-fold compared to non-anchoring vectors in in vitro benchmarks.
• Scalability: Demonstrated that AAV-based delivery allows for high-throughput screening in the developing cortex, recovering tens of thousands of high-quality nuclei with reliable perturbation assignment.

4. Key Results

A. Optimization and Validation

• In Vitro Benchmarking: In HT-22 Cas9 cells, the optimal combination of 80% bead modification and nested TE PCR yielded the highest gRNA assignment rates. The engineered vectors maintained genome editing efficiency comparable to canonical vectors.
• In Vivo Performance: In the developing mouse cortex (P14), the platform achieved a 62.8% gRNA assignment rate across 11,387 nuclei, with 44.4% receiving a single-gRNA assignment. This performance matched or exceeded previous single-cell RNA-seq Perturb-seq studies, despite the added complexity of ATAC-seq.
• Specificity: The system showed low cross-gRNA correlation, confirming high specificity under low multiplicity of infection (MOI). Indel analysis directly from the transcriptome confirmed that assigned nuclei carried the expected editing signatures.

B. Biological Application: Neurodevelopmental Disorder (NDD) Risk Genes

The platform was applied to screen 16 transcription factors and chromatin regulators implicated in neurodevelopmental disorders (e.g., Sin3a, Mef2c, Myt1l, Rai1) in the developing cortex (P28).

• Dataset Scale: Across 15 animals, they recovered 84,887 high-quality nuclei spanning 12 distinct neuronal cell types (9 excitatory, 3 inhibitory).
• Cell-Type Specificity:
  • Broad vs. Restricted Effects: Perturbations of Sin3a (a chromatin coregulator) caused broad transcriptomic and epigenomic changes across multiple cell types. In contrast, Mef2c (a transcription factor) showed highly cell-type-restricted phenotypes, with massive effects in Layer 5 and Layer 6 Intratelencephalic (IT) neurons but minimal effects in Layer 6 Corticothalamic (CT) neurons, despite similar expression levels.
  • Multiomic Concordance: The study linked chromatin accessibility changes directly to gene expression changes within the same cell. For example, in Mef2c-perturbed L6 IT neurons, accessibility changes at the Slc24a3 locus correlated directly with reduced expression, suggesting a cis-regulatory mechanism.
• Validation: Results were validated against published bulk RNA-seq knockout data (e.g., Mef2c KO), showing high correlation (Pearson r = 0.816).

5. Significance

• Mechanistic Resolution: By linking perturbations to both gene expression and chromatin accessibility in the same cell, this method provides a mechanistic view of gene regulatory networks that unimodal assays cannot achieve. It allows researchers to distinguish between direct transcriptional effects and downstream chromatin remodeling.
• Disease Modeling: The ability to resolve cell-type-specific phenotypes is crucial for understanding neurodevelopmental disorders, where risk genes often have distinct roles in different neuronal subtypes.
• Generalizability: The strategy is compatible with AAV delivery and standard nuclei isolation, making it broadly applicable across diverse tissues, species, and disease models. It establishes a foundation for building predictive models of gene function and regulatory logic in vivo.

In summary, Zheng et al. have solved a critical technical barrier in functional genomics, enabling high-fidelity, multiomic CRISPR screening in living organisms and revealing previously hidden layers of cell-type-specific gene regulation in the developing brain.