Single-cell RNA sequencing (scRNA-seq) and single-nucleus RNA sequencing (snRNA-seq) both profile gene expression one cell at a time, but they start from fundamentally different material: scRNA-seq captures RNA from an intact, dissociated whole cell, while snRNA-seq captures RNA from an isolated nucleus alone, discarding the cytoplasm. That distinction sounds minor, yet it ripples through every downstream result, from which genes you detect and how many transcripts you recover per cell to which tissue types and storage formats you can work with at all. Choosing between the two is less about which is “better” and more about which suits the biological question, the tissue, and the sample you actually have in front of you.
What Each Method Actually Captures
The most important practical difference comes down to where RNA lives inside a cell. Mature, fully processed messenger RNAs are predominantly found in the cytoplasm, where they are being translated into proteins. The nucleus, by contrast, is the site of active transcription and early RNA processing. Nuclei contain the same population of pre-mRNAs (transcripts still being spliced) as the whole cell, but they hold only a small subset of the mature mRNAs that have already been exported to the cytoplasm.1PubMed Central. Differences in molecular sampling and data processing explain variation among single-cell and single-nucleus RNA-seq experiments This means snRNA-seq data are inherently enriched for unspliced, intronic reads and yield fewer total exonic read alignments per nucleus compared with a whole cell.
In practical terms, you detect fewer unique transcripts and fewer genes per nucleus than per cell. One direct comparison in human microglia found that whole cells averaged about 4,500 unique molecular identifiers (UMIs) and roughly 2,200 genes per cell, while nuclei from the same tissue averaged about 1,300 UMIs and around 880 genes per nucleus.2Cell Reports. Single-nucleus RNA sequencing identifies functional transcriptional differences in microglia and fails to detect microglial activation in human brain tissue That is a roughly threefold difference in raw sensitivity. Whether that gap matters depends on whether the biology you care about is captured by the nuclear transcriptome or requires the cytoplasmic fraction too.
Why Nuclei Work When Whole Cells Do Not
If whole cells give you more transcripts, why would anyone choose nuclei? Because for many tissues, getting intact whole cells is the bottleneck. Certain cell types resist dissociation due to their large size, fragility, or tight connections to neighboring cells.3PubMed Central. Perspectives on single-nucleus RNA sequencing in different cell types and tissues Neurons embedded in dense brain tissue, cardiomyocytes with extensive gap junctions, adipocytes that burst easily, and the interconnected cells of solid tumors all pose serious challenges for enzymatic digestion and mechanical dissociation. The nucleus, being a compact and relatively sturdy organelle, survives isolation procedures that would destroy these fragile whole cells.
Frozen tissue is the other major driver. Most clinical biobanks store samples at −80 °C, and freezing ruptures cell membranes, making whole-cell recovery impossible. snRNA-seq works on frozen specimens because nuclei tolerate freeze-thaw cycles far better than intact cells.4PLoS ONE. Single-nucleus and single-cell transcriptomes compared in matched cortical cell types This has opened decades’ worth of archived human tissue to single-cell-resolution profiling, a resource that would otherwise be completely inaccessible. For tumor research specifically, snRNA-seq has been shown to be essential for profiling frozen or hard-to-dissociate tumors that scRNA-seq simply cannot handle.5Nature Medicine. A single-cell and single-nucleus RNA-Seq toolbox for fresh and frozen human tumors
Dissociation Artifacts and Stress Responses
Even when you can dissociate a tissue into single cells, the process itself can distort the data. Enzymatic digestion, particularly with collagenase at 37 °C, triggers a conserved stress response in the cells being freed from their tissue context. One study identified a core set of 512 heat shock and stress response genes, including the immediate early genes FOS and JUN, that were artificially induced by warm collagenase digestion.6PubMed Central. Dissociation of solid tumor tissues with cold active protease for single-cell RNA-seq minimizes conserved collagenase-associated stress responses These genes then show up in scRNA-seq data as if the cells were actively stressed in the original tissue, when in reality the stress was introduced by sample preparation.
Cold-active protease digestion at 6 °C can minimize this artifact for scRNA-seq, but snRNA-seq sidesteps the problem more completely. Because nuclei are isolated mechanically or with gentler lysis buffers rather than extensive enzymatic digestion, the transcriptional snapshot they contain more closely reflects the cell’s state at the moment of tissue collection. This is particularly relevant for neuroscience, where immediate early gene expression is a meaningful biological signal that can be obscured by dissociation-induced artifacts.4PLoS ONE. Single-nucleus and single-cell transcriptomes compared in matched cortical cell types
Cell Type Representation Can Differ Dramatically
The choice between the two methods does not just affect per-cell sensitivity; it can determine whether you see certain cell types at all. A striking example comes from kidney tissue. When researchers compared scRNA-seq and snRNA-seq on the same kidney samples, scRNA-seq identified ten cell clusters but completely missed glomerular cell types, and one of the ten clusters turned out to be dominated by artifactual stress-response genes rather than genuine biology. By contrast, snRNA-seq captured cell types that were absent from the scRNA-seq data entirely, including glomerular podocytes, mesangial cells, and endothelial cells. The nuclear approach yielded roughly 20-fold more podocytes than published scRNA-seq datasets had recovered.7PubMed Central. Advantages of Single-Nucleus over Single-Cell RNA Sequencing of Adult Kidney: Rare Cell Types and Novel Cell States Revealed in Fibrosis
The reason is straightforward: podocytes and certain other glomerular cells are large, delicate, and tightly anchored. They are destroyed or lost during the mechanical and enzymatic steps required for whole-cell dissociation. Their nuclei, however, survive just fine. This kind of systematic bias means that scRNA-seq can paint a misleading picture of tissue composition, not just missing rare cells but entirely omitting whole functional compartments. For any tissue where specific cell types are known to be fragile or deeply embedded, snRNA-seq is the safer bet for representative profiling.
Quality Control Looks Different for Each Approach
Standard quality-control filters need to be adjusted depending on which method you use. One of the most common metrics is mitochondrial gene expression. In scRNA-seq, a high percentage of mitochondrial reads in a given cell is a classic marker of a damaged or dying cell whose cytoplasmic mRNA has leaked out, leaving behind the more durable mitochondrial transcripts. Researchers typically set a threshold, often discarding cells with mitochondrial reads above 20–30% of total counts, though in kidney tissue this threshold has been debated and set as high as 50% because some kidney cell populations naturally express mitochondrial genes at high levels.8Scientific Data. Meta-analysis of single-cell and single-nucleus transcriptomics reveals kidney cell type consensus signatures
For snRNA-seq, the situation flips. Isolated nuclei should contain very little mitochondrial RNA because mitochondria are cytoplasmic organelles. Typical thresholds for nuclear data are around 5% mitochondrial reads, filtering out nuclei that picked up contaminating cytoplasmic material during preparation.8Scientific Data. Meta-analysis of single-cell and single-nucleus transcriptomics reveals kidney cell type consensus signatures A direct comparison in human bone marrow confirmed this expected pattern: scRNA-seq samples showed consistently higher mitochondrial gene proportions than snRNA-seq from the same tissue.9bioRxiv. Comparative Analysis of Single-Nucleus and Single-Cell RNA Sequencing in Human Bone Marrow Mononuclear Cells: Methodological Insights and Trade-offs Applying whole-cell thresholds to nuclear data, or vice versa, would either discard good nuclei or retain contaminated ones.
Ambient RNA Contamination in Nuclear Preps
snRNA-seq has its own category of artifact that does not affect whole-cell methods in the same way. When nuclei are isolated from tissue, free-floating RNA molecules released from broken cells form a “soup” of ambient RNA in the suspension. Every captured nucleus ends up absorbing some of this ambient RNA, which then gets sequenced alongside the nucleus’s own transcripts. In brain tissue, this problem is particularly acute because neurons are the most abundant source of ambient RNA, and their transcripts contaminate non-neuronal nuclei.
Research on cortical snRNA-seq datasets found that glial nuclei, including oligodendrocytes, astrocytes, microglia, and oligodendrocyte progenitor cells, were consistently contaminated with neuronal ambient RNAs. The contamination was strong enough to potentially confound cell-type annotation, making glia appear to express neuronal genes they do not actually transcribe.10PubMed Central. Neuronal ambient RNA contamination causes misinterpreted and masked cell types in brain single-nuclei datasets Computational tools like SoupX and DecontX can partially correct for this, and sorting neuronal nuclei away from glia before capture (using markers like NeuN) reduced the problem. But the take-home message is that snRNA-seq data from heterogeneous tissues require careful ambient-RNA correction, or you risk misassigning gene expression to the wrong cell types.
RNA Velocity and Transcriptional Dynamics
One analytical approach that sits at an interesting crossroads between the two methods is RNA velocity, which uses the ratio of unspliced to spliced transcripts to infer the direction and speed of gene expression changes in individual cells. Because snRNA-seq is enriched for unspliced, intronic reads, it might seem like a natural fit. And early results are encouraging: a study comparing nucleus-derived and cell-derived RNA velocities in microglia from the same donor found very strong correlations between the two, ranging from 0.94 to 0.99, suggesting that snRNA-seq can provide precise velocity estimates.11Scientific Reports. Single-nucleus RNA velocity reveals critical synaptic and cell-cycle dysregulations in neuropathologically confirmed Alzheimer’s disease
That said, the field has not fully settled this question. The physical process of isolating nuclei distorts the balance of spliced and unspliced mRNAs in complex ways. Their relative abundances can shift depending on how quickly specific transcripts are exported from the nucleus and on residual cytoplasmic material clinging to the nuclear surface. Standard RNA velocity models assume constant degradation and nuclear export rates, and those assumptions have not been conclusively verified for nuclear data.12PubMed Central. RNA velocity—current challenges and future perspectives For tissues where whole-cell isolation is impossible, snRNA-seq velocity is the only option and has produced biologically meaningful results. But for tissues where both methods are feasible, researchers should be cautious about interpreting velocity fields from nuclear data without benchmarking against whole-cell results.
Unlocking Archival Tissue With FFPE Protocols
Frozen tissue is one thing, but the real treasure trove in clinical archives is formalin-fixed, paraffin-embedded (FFPE) tissue. FFPE is the standard format for pathology specimens worldwide, and hospitals have accumulated vast collections spanning decades. The fixation process cross-links proteins and fragments RNA, making it hostile to most sequencing methods. Yet several recent protocols have made single-nucleus profiling of FFPE tissue a reality.
One approach, called snPATHO-seq, uses serial steps of rehydration, enzyme-based tissue dissociation, and nuclei isolation before running gene expression analysis on a commercial platform.13Communications Biology. snPATHO-seq, a versatile FFPE single-nucleus RNA sequencing method to unlock pathology archives Another method, snRandom-seq, captures full-length total RNA using random primers rather than the poly-A selection that most scRNA-seq workflows rely on, which is critical because FFPE fixation degrades the poly-A tails that standard capture methods depend on.14Nature Communications. High-throughput single nucleus total RNA sequencing of formalin-fixed paraffin-embedded tissues by snRandom-seq A third strategy, snCED-seq, uses cryogenic enzymatic dissociation for rapid, high-yield nuclei extraction from FFPE samples.15Nature Communications. snCED-seq: high-fidelity cryogenic enzymatic dissociation of nuclei for single-nucleus RNA-seq of FFPE tissues Whole-cell scRNA-seq is essentially impossible on FFPE material, so this is territory that belongs entirely to nuclear approaches.
The Brain as a Flagship Tissue for snRNA-seq
Neuroscience has been the most enthusiastic adopter of snRNA-seq, for good reason. Brain tissue is dense, mechanically complex, and largely made of cells with long, fragile projections that shear during dissociation. Postmortem human brain, the primary source for studying neurological disease, is almost always frozen. Nuclear RNA sequencing from neurons stored at −80 °C was demonstrated early on, establishing that brain archives are accessible to single-cell-resolution profiling.16Nature Protocols. Using single nuclei for RNA-seq to capture the transcriptome of postmortem neurons
Since then, the scale has grown enormously. An early study sequenced over 3,200 neuronal nuclei from six cortical regions and identified previously unknown neuronal subtypes as well as regional transcriptomic differences.17PubMed Central. Neuronal subtypes and diversity revealed by single-nucleus RNA sequencing of the human brain More recent work has expanded to subcortical regions, generating molecular taxonomies across reward circuitry areas like the nucleus accumbens, amygdala, hippocampus, and prefrontal cortex, profiling over 70,000 nuclei in a single study.18PubMed Central. Single-nucleus transcriptome analysis reveals cell-type-specific molecular signatures across reward circuitry in the human brain Studies of neurological disease have pushed even further: one analysis of multiple sclerosis brain tissue profiled 632,000 nuclei from 156 tissue samples, identifying patient-specific glial signatures and revealing that most gene expression variability in MS lesions is driven by patient-level effects rather than lesion type.19Cell Press (Neuron). Single-nucleus transcriptomics of 632,000 nuclei from multiple sclerosis brain regions reveals patient-specific glial signatures None of these studies would have been feasible with whole-cell approaches.
Plants and Non-Model Organisms
The advantages of going nuclear are not limited to animal tissue. Plants present a unique challenge that makes scRNA-seq even harder than in animal systems: the cell wall. Before any plant cell can be profiled individually, researchers must generate protoplasts by enzymatically digesting the cell wall. This process is slow, tissue-dependent, and introduces acute wounding responses that contaminate the transcriptional snapshot.20PubMed. Plant Nuclei Isolation for Single-Nucleus RNA Sequencing Many plant cell types, particularly those with thick secondary cell walls like wood-forming cells, cannot be protoplasted at all.
snRNA-seq bypasses the cell wall entirely. Nuclei can be released by mechanical disruption (chopping, grinding) followed by filtration, no enzymatic digestion of cell walls required. This makes snRNA-seq applicable to a much wider range of plant tissues and species than scRNA-seq.21PubMed Central. FlsnRNA-seq: protoplasting-free full-length single-nucleus RNA profiling in plants Work on trees in the genus Populus demonstrated that nuclei isolation overcomes the specific difficulties of solid plant tissues with secondary cell walls, where prolonged enzymatic digestion not only distorts transcriptomes but can damage cellular membranes and leak cytoplasmic mRNA content.22PLoS ONE. A robust method of nuclei isolation for single-cell RNA sequencing of solid tissues from the plant genus Populus For plant genomics, the nuclear approach has shifted from a niche alternative to the default for many tissue types.
Pairing Nuclear Data With Spatial Transcriptomics
One of the most productive uses of snRNA-seq is not as a standalone method but as a reference dataset for spatial transcriptomics. Spatial methods like Visium or MERFISH map gene expression to physical locations within a tissue section, but many spatial platforms capture RNA from spots containing multiple cells or measure only a predefined panel of genes. To assign cell types to spatial spots and interpret the spatial patterns, researchers need a comprehensive reference atlas of cell types and their expression profiles, which is exactly what snRNA-seq provides.
This pairing has been applied across organisms and disease contexts. In human uterine leiomyoma, integrating spatial transcriptomics with snRNA-seq mapped the cellular architecture and gene profiles of the tumor and its surrounding pseudocapsule for the first time.23PubMed Central. Integrating Spatial Transcriptomics and Single-nucleus RNA Sequencing Reveals the Potential Therapeutic Strategies for Uterine Leiomyoma In plants, the same strategy has been used to build cell atlases of soybean nodules and roots, capturing transitional cell states during nodule maturation that neither method alone would have resolved.24PubMed Central. Integrated single-nucleus and spatial transcriptomics captures transitional states in soybean nodule maturation Because snRNA-seq works on frozen tissue, it can be run on the same sample block used for spatial sectioning, eliminating batch differences between the reference and the spatial data.
Multiomics From a Single Nucleus
The nucleus is also a natural unit for multimodal profiling that goes beyond RNA. Because the genome and its chromatin packaging live in the nucleus, protocols that jointly measure gene expression and chromatin accessibility from the same nucleus have become increasingly practical. One study of murine gonads used joint single-nucleus transcriptomics and chromatin accessibility assays to reconstruct the transcriptional regulatory networks underlying supporting cell differentiation during sex determination.25PubMed Central. Single-nucleus multiomics of murine gonads reveals transcriptional regulatory network underlying supporting lineage differentiation By reading both the RNA and the open chromatin regions from the same nucleus, researchers can link regulatory elements to gene expression changes at the single-cell level, connecting cause to effect in a way that separate measurements cannot.
Whole-cell multiomic approaches exist as well, but they face the same tissue-compatibility limitations as scRNA-seq. For frozen tissue, archived specimens, and cell types that resist dissociation, the nucleus is often the only compartment available for multimodal measurement. The growing availability of commercial platforms supporting joint ATAC and RNA profiling from single nuclei has made this increasingly routine rather than bespoke.
When to Choose Which
For fresh tissue that dissociates easily, like blood, bone marrow, or many cultured cell lines, scRNA-seq remains the more sensitive option. You detect more genes and transcripts per cell, you capture cytoplasmic RNA species (including some noncoding RNAs that are predominantly cytoplasmic), and the ambient RNA contamination profile is better characterized and easier to correct. A systematic benchmarking effort that compared seven methods across cell lines, blood, and brain tissue confirmed that the choice of method has substantial effects on data quality and that no single method dominates across all tissue types.26Nature Biotechnology. Systematic comparison of single-cell and single-nucleus RNA-sequencing methods
snRNA-seq is the better or only option when any of the following apply: the tissue is frozen or FFPE-preserved, the cell types of interest are large or fragile, the tissue resists enzymatic dissociation, you need to avoid dissociation-induced transcriptional artifacts, or you want to pair the data with spatial transcriptomics from the same tissue block. For brain, heart, adipose, kidney, and most plant tissues, the nuclear approach has become the standard. For immune cells, blood-derived samples, and fresh tumor dissociates, whole-cell methods still dominate. Many large-scale projects now run both in parallel on matched samples, using computational integration to get the benefits of each while controlling for their respective biases.