Female iPSC X-chromosome inactivation (XCI) erosion and its transcriptomic effects during CRISPR gene editing and neural differentiation

This study reveals that while X-chromosome inactivation (XCI) erosion varies in CRISPR-edited female iPSCs but remains largely preserved in their neural derivatives, XIST-mediated epigenetic silencing still drives significant allelic imbalance in both X-linked and autosomal genes, thereby confounding transcriptomic analyses of differentially expressed genes in neurodevelopmental disease modeling.

The Gist

The Big Picture: A Library with a Broken "Do Not Read" Sign

Imagine your body's cells are like a massive library containing two copies of every book (one from mom, one from dad). For most books, you read both copies. But for the X-chromosome (which is like a massive, important volume in the library), female cells have a special rule: they must close one copy and only read the other. This ensures that women don't get "double the dose" of instructions compared to men, who only have one X-chromosome.

The "Do Not Read" sign that closes that second book is called XIST. It's a molecular sticky note that says, "Silence this section."

The Problem: Scientists use human stem cells (iPSCs) to study brain diseases. They often edit these cells using CRISPR (a molecular pair of scissors) to fix or break specific genes. However, the researchers in this paper discovered that during this editing process, the "Do Not Read" sign (XIST) sometimes falls off or gets erased. This is called XCI Erosion.

When the sign falls off, the cell starts reading both copies of the X-chromosome books again. This messes up the cell's instructions, potentially making the results of the experiment look wrong.

What the Scientists Did

The team took hundreds of female stem cells, edited them with CRISPR to study genes related to brain disorders (like autism or schizophrenia), and then turned them into neurons (brain cells). They wanted to know:

1. Does the editing process knock off the "Do Not Read" sign?
2. Does the sign stay off once the cell becomes a brain cell?
3. Does this mess up our data?

The Key Findings (The Story)

1. The Sign is Fickle in the Lab

They found that in the stem cell "waiting room," the XIST sign is very unstable. Some cells kept the sign firmly in place, while others lost it completely. Interestingly, the loss of the sign wasn't caused by the CRISPR scissors themselves; it was just a random thing that happened to the cells in the lab culture. It's like a sticky note that randomly peels off the book depending on how the cell is feeling that day.

2. The Sign Stays Off (or On) When Cells Grow Up

When they turned these stem cells into neurons, the status of the sign mostly stayed the same.

• If a stem cell had lost the sign, the resulting brain cell usually didn't have it either.
• If a stem cell had the sign, the brain cell usually kept it.
• Analogy: Imagine a student who forgets their homework assignment in high school. When they get to college, they usually still haven't done the homework. The habit (or lack thereof) persists.

3. The "Double Dose" Effect

When the sign (XIST) was missing, the cells started reading both copies of the X-chromosome. This caused a "double dose" of certain genes.

• X-Linked Genes: These were the most affected. It's like turning the volume up on a radio station that was supposed to be muted.
• Autosomal Genes (The other 22 pairs of chromosomes): Surprisingly, the sign didn't affect these much in the bulk of the cells. However, when they looked at single neurons (one cell at a time), they found that the sign did have a subtle, quiet effect on these other genes too. It's like a whisper that you can only hear if you are standing right next to the speaker.

4. The "Hot Spot"

The researchers found a specific neighborhood on the X-chromosome (between 70 and 80 million letters down the line) where the sign's absence caused a lot of confusion. This area contains genes crucial for brain development. It's like a specific street in a city where, if the traffic light breaks, everything gets gridlocked.

5. The Danger to Science (The Confounding Effect)

This is the most important takeaway for other scientists. When researchers compare edited cells to unedited cells to find "differences," they often assume the only difference is the gene they edited.

• The Trap: If one group of cells accidentally lost the XIST sign and the other didn't, the "double dose" of X-chromosome genes will show up as a difference.
• The Result: Scientists might think they found a new brain disease mechanism, but they actually just found that one group of cells forgot to close its X-chromosome book. It's a "ghost in the machine" that can lead to false conclusions.

The Takeaway for Everyone

This paper is a warning label for the scientific community. When using female stem cells to study the brain, you have to check if the "Do Not Read" sign (XIST) is still attached.

If you don't check, you might be blaming a broken gene for a problem that is actually just caused by the cell reading too many books. By accounting for this "erosion," scientists can make their models of brain disorders more accurate and their discoveries more reliable.

In short: Before you trust the results of a female stem cell experiment, make sure the cell hasn't lost its "mute button" for the X-chromosome, or you might be listening to a song that's playing way too loud.

Technical Summary

1. Problem Statement

Human induced pluripotent stem cells (hiPSCs) and their differentiated derivatives (neurons) are critical models for studying neurodevelopmental and psychiatric disorders, often utilizing CRISPR/Cas9 gene editing. A fundamental biological process in female cells is X-chromosome inactivation (XCI), mediated by the long non-coding RNA XIST, which silences one X chromosome to ensure dosage compensation.

However, in cultured hiPSCs, XIST expression is often unstable, leading to XCI erosion (loss of silencing). While XCI erosion is known to affect gene expression in pluripotent cells, its dynamics during the lengthy CRISPR editing process and subsequent neural differentiation remain poorly understood. Key gaps include:

• Does CRISPR editing or neural differentiation stabilize or exacerbate XCI erosion?
• How does XCI erosion affect the expression of both X-linked and autosomal genes in neurons?
• Does variable XIST status confound differential gene expression (DEG) analyses in disease modeling, potentially leading to false positives or misinterpretation of CRISPR editing effects?

2. Methodology

The study leveraged a large-scale resource from the SSPsyGene consortium, utilizing CRISPR-based cytosine base editors (CBEs) to introduce loss-of-function (LoF) mutations in 66 neurodevelopmental disorder (NPD)-associated genes.

• Cohort: 297 CRISPR-edited and non-targeting control (NTC) iPSC lines derived from 6 donor lines (4 female, 2 male).
• Differentiation: A subset of these lines (n=234) was differentiated into excitatory and inhibitory neurons, co-cultured with mouse astrocytes.
• Sequencing:
  • Bulk RNA-seq: Performed on 297 iPSC lines to assess global transcriptomic changes and XIST status.
  • Single-cell RNA-seq (scRNA-seq): Performed on ~190,000 neurons (121k excitatory, 69k inhibitory) to resolve heterogeneity at the single-cell level.
• Analytical Approaches:
  • XIST Stratification: Samples were classified as XIST+ (expressing XIST) or XIST- (eroded) based on expression thresholds.
  • Allele-Specific Expression (ASE): Used to detect allelic imbalance at heterozygous SNPs, providing a sensitive measure of epigenetic regulation distinct from average expression levels.
  • Differential Expression (DEG): Compared DEG results with and without XIST expression as a covariate to assess confounding effects.
  • Gene Ontology (GO) & Positional Analysis: Identified conserved genomic loci and biological pathways affected by XCI erosion.

3. Key Contributions

• Systematic Characterization of XCI Dynamics: Provided the first extensive evaluation of XCI erosion across a large cohort of CRISPR-edited iPSCs and their derived neurons, demonstrating that erosion status is largely preserved from iPSCs to neurons.
• Mechanistic Insight into ASE: Demonstrated that XIST erosion causes a shift from monoallelic (ASE) to biallelic expression, with a significantly stronger effect on X-linked genes than autosomal genes.
• Identification of Conserved "Hot Spots": Mapped specific genomic regions (e.g., 70–80 Mb on the X chromosome) where XIST-induced allelic imbalance is conserved across different genetic backgrounds and cell types.
• Methodological Warning: Quantified the confounding effect of variable XIST expression on standard DEG analyses, highlighting the necessity of accounting for XCI status in female iPSC studies.

4. Key Results

A. XCI Erosion is Widespread and Persistent

• In iPSCs: XCI erosion was highly variable across CRISPR-edited lines from the same donor, attributed to heterogeneity in the source population rather than the editing process itself or passage number (passages 19–26).
• In Neurons: The XIST status of the parent iPSC line was largely preserved in differentiated neurons. While most neurons retained the parental XIST status (XIST+ or XIST-), stochastic reactivation or loss of XIST was observed in a minority of cases.

B. Transcriptomic Effects: X-linked vs. Autosomal

• X-linked Genes: XIST- female lines showed significantly higher expression of X-linked genes compared to XIST+ lines and male controls. This effect was robust in both bulk iPSCs and single neurons.
• Autosomal Genes: Unlike previous studies in naive pluripotent cells, this study found no significant global change in average autosomal gene expression between XIST+ and XIST- lines in bulk RNA-seq. However, single-cell analysis revealed a modest negative correlation between XIST levels and autosomal gene expression in neurons.
• Mitochondrial Genes: XIST loss in iPSCs led to increased expression of mitochondrial (MT) genes, bringing female levels closer to male levels. This effect was not observed in differentiated neurons.

C. Allele-Specific Expression (ASE) Analysis

• X-linked Bias: XIST+ cells exhibited strong monoallelic expression (ASE bias) for X-linked genes. XIST- cells showed a shift toward biallelic expression.
• Autosomal Bias: While weaker than on the X chromosome, XIST erosion also induced subtle ASE shifts in autosomal genes.
• Conserved Loci: The study identified specific loci with conserved ASE bias across all female donor lines.
  • X-linked: A "hot spot" at 70–80 Mb on the X chromosome (containing XIST, FTX, JPX, and neurodevelopmental genes like TTC3P1) showed consistent allelic imbalance.
  • Autosomal: 339 autosomal genes shared across all lines were influenced by XIST status, enriched for immune signaling and translation processes.

D. Confounding Effects on Differential Expression (DEG)

• When analyzing LoF mutations in neurons, including XIST expression as a covariate in the linear model significantly altered the list of differentially expressed genes (DEGs).
• Approximately 20% of LoF genes showed a Jaccard similarity index < 0.75 between models with and without XIST correction.
• Non-overlapping DEGs were predominantly X-linked and showed larger fold-change differences, indicating that failing to control for XCI erosion can lead to false attribution of gene expression changes to the CRISPR edit rather than epigenetic drift.

5. Significance and Implications

• Disease Modeling Validity: The study underscores that XCI erosion is a persistent confounder in female hiPSC models. Researchers must screen for XIST status and include it as a covariate in statistical models to avoid misinterpreting data in neurodevelopmental disorder research.
• Epigenetic Regulation: The findings suggest that XIST exerts a conserved, locus-specific repressive effect on both X-linked and specific autosomal genes, even in differentiated neurons, challenging the view that XCI effects are limited to the pluripotent state.
• Gene Discovery: The identification of conserved "hot spots" and specific neurodevelopmental genes (e.g., PAK3, DCX, ALG13) affected by XCI erosion provides new targets for investigating the molecular mechanisms of sex-biased neurological disorders.
• Experimental Design: The paper provides a framework for handling female iPSC data, recommending that future studies utilizing CRISPR editing and differentiation in female lines explicitly account for XCI erosion to ensure data reproducibility and accuracy.