TL;DR
Perturb-DBiT is a powerful new spatial multi-omics platform that simultaneously maps CRISPR-induced genetic perturbations and the whole-transcriptome—including non-coding RNAs—within intact tissue sections. By co-sequencing single-guide RNAs (sgRNAs) and total RNA base-by-base, it allows researchers to observe how specific gene knockouts reshape cellular behavior, metabolism, and the surrounding tumor microenvironment in situ.
What Problem Does This Paper Address?
Functional genomics has long relied on CRISPR screens to identify gene functions, but these are typically performed in "bulk" or single-cell suspensions. While effective, these methods destroy the spatial architecture of the tissue. In complex environments like tumors, the "where" matters as much as the "what." A cell’s behavior is dictated not just by its own mutations, but by its neighbors and its location within the tissue (e.g., near a blood vessel vs. a lymph node).
Existing spatial CRISPR technologies suffer from several bottlenecks: they are often limited to small panels of target genes, can only detect protein-coding mRNAs, or require "proxy" measurements from serial tissue sections rather than the exact same cells. Perturb-DBiT aims to provide a "panoramic" view, capturing the full complexity of the transcriptome and large-scale CRISPR libraries on a single tissue slice.
Methods
Perturb-DBiT builds upon the DBiT-seq (Deterministic Barcoding in Tissue) framework, which uses microfluidic channels to deliver DNA barcodes in a grid-like fashion directly onto a tissue slide. The authors introduced two distinct workflows:
- Perturb-DBiT PAC (Polyadenylated Capture): Uses in situ enzymatic polyadenylation to add "tails" to all RNA species (mRNA, lncRNA, tRNA, miRNA, and sgRNAs). This allows for an unbiased, "total RNA" capture.
- Perturb-DBiT DC (Direct Capture): Uses specific primers designed to capture the constant scaffold region of sgRNAs alongside standard mRNA capture.
The core innovation is the base-by-base next-generation sequencing (NGS) of the sgRNAs. Unlike imaging-based methods (like FISH) that are limited by the number of fluorescent "colors" available, NGS allows Perturb-DBiT to scale to genome-wide libraries containing tens of thousands of different guides.
Datasets
The researchers validated the platform across several challenging biological models:
- Human Cancer Metastasis: HT-29 colorectal cancer cells in a lung colonization model.
- Immune-Competent Mouse Model: E0771 breast cancer cells in a syngeneic lung model to study the tumor immune microenvironment (TME).
- Liver Cancer GEMMs: Genetically engineered mouse models (mTSG library) to study autochthonous tumor initiation.
- CRISPR Libraries: Ranged from small (2–86 sgRNAs) to medium (288 sgRNAs) to genome-scale (Brunello and Brie libraries, ~80,000 sgRNAs).
Key Insights and Ideas
- The Non-Coding Connection: One of the most striking findings was the impact of perturbations on the "dark matter" of the genome. For example, knocking out the RNA-editing gene ADARB1 led to a global downregulation of tRNAs, which the authors linked to a failure of these cells to integrate into tumor colonies.
- Metabolic Reprogramming: The study revealed that specific knockouts (like sgFGL2) trigger distinct shifts in tRNA isotypes (e.g., serine tRNA upregulation), suggesting that genetic perturbations drive localized metabolic adaptations to the tissue environment.
- Clonal Cooperation: By mapping sgRNAs spatially, the authors observed "co-localization" of specific guide pairs (e.g., sgS100A4 and sgSLC22A16). This suggests that different mutant clones might cooperate or provide a collective selective advantage to drive tumor expansion.
- Spatial Immune Modulation: In the immune-competent model, knocking out Zc3h12a (a negative regulator of inflammation) created localized "hot spots" of immune infiltration near lymph nodes, a detail that would be lost in non-spatial single-cell sequencing.
Key Results
- High Fidelity: In a medium-sized screen, Perturb-DBiT captured 95% of the target genes and outperformed traditional bulk sequencing (MIPS) in sensitivity.
- Scalability: The authors successfully detected over 5,700 unique sgRNA UMIs from a genome-scale library in a single tissue region.
- Data Quality: The PAC method yielded an average of 3,371 genes and 6,429 UMIs per spot for exonic reads, with significantly higher counts when including intronic sequences, providing a deep look into the transcriptional state.
- Validation: Top hits identified as tumor suppressors (e.g., SEMA6B, MT1E, ASAH2) were validated using in vitro clonogenic and wound-healing assays.
Limitations and Future Work
- Resolution: While the 20-µm spot size is near-cellular, it is not yet true single-cell resolution. The authors used a computational framework called iStar to help impute single-cell-level data, but physical single-cell resolution remains a goal.
- Tissue Coverage: The current microfluidic chips cover a limited area (e.g., 5mm x 5mm). Scaling this to cover entire whole-organ sections would require further engineering.
- Sequencing Depth: Genome-wide spatial screens are currently limited by the immense sequencing depth required to capture every guide across every spot. The authors suggest that future iterations will benefit from increased multiplexing and deeper sequencing strategies.
Summary updated