Part 2 — The condition

5. STAG2 and the cohesin complex

Cohesin is a ring-shaped protein complex that encircles DNA. It is one of the most important machines in the cell, and it has two distinct jobs.

The ring is built from several parts. Think of it as a clamp assembled from interchangeable components:

flowchart TB
    subgraph ring["The cohesin ring"]
        direction LR
        SMC1["<b>SMC1A</b><br/>arm"]
        SMC3["<b>SMC3</b><br/>arm"]
        RAD21["<b>RAD21</b><br/>latch that<br/>closes the ring"]
        STAG["<b>STAG1 or STAG2</b><br/>the interchangeable<br/>targeting subunit"]
    end
    NIPBL["<b>NIPBL</b><br/>loader — puts the<br/>ring onto DNA"] -.->|loads| ring
    ring --> J1["<b>Job 1</b><br/>Hold duplicated chromosomes<br/>together during cell division"]
    ring --> J2["<b>Job 2</b><br/>Fold DNA into loops that<br/>control which genes turn on"]

STAG2 is one specific, swappable component of this ring.

Critically, there are two alternative versions of that component: STAG1 and STAG2 (older literature calls the proteins SA1 and SA2). Every cohesin ring contains one or the other — never both. They are partly redundant but not identical; each directs cohesin to somewhat different places on the genome.

This has a consequence that is easy to miss and strategically important:

Cells carrying the duplication don't just make more STAG2. They make more STAG2 competing for the same limited pool of ring components. Extra STAG2 may crowd out STAG1, shifting the ratio of STAG2-cohesin to STAG1-cohesin across the genome.

If the disease is driven by the ratio rather than the absolute STAG2 level, then the therapeutic target could be reframed — you might correct the balance by nudging STAG1 up rather than only pushing STAG2 down. This is, at present, a hypothesis worth asking about, not an established fact. The relevant evidence that STAG1 and STAG2 are genuinely non-equivalent (they have different loading requirements) is established; whether ratio-correction is therapeutically actionable in this disease is not.

Supporting the idea that the paralogs compensate for one another: in cells lacking functional STAG2, researchers have observed compensatory upregulation of STAG1 and even ectopic expression of the normally germ-cell-specific paralog STAG3, forming chimeric cohesin complexes. That's the loss-of-function direction, but it demonstrates that the system actively rebalances.

This is a genuinely good question to put to any laboratory working on the condition, and asking it signals real understanding: "Are you measuring STAG1 as well as STAG2 — is the relevant variable the ratio rather than the absolute level?"

Key terms:

• Cohesin — the ring-shaped complex that encircles DNA.
• Subunit — one component protein of a multi-part complex.
• STAG1 / STAG2 (SA1 / SA2) — the two mutually exclusive alternative subunits.
• Paralog — a related gene arising from duplication, with overlapping function.
• NIPBL — the "loader" that places cohesin onto DNA; mutated in Cornelia de Lange syndrome.
• RAD21, SMC1A, SMC3 — the other core cohesin subunits.
• Sister chromatid cohesion — cohesin's classical job of holding duplicated chromosomes together.

References:

• Different NIPBL requirements of cohesin-STAG1 and cohesin-STAG2 — Nature Communications 2023 — direct evidence the two paralogs are not interchangeable.
• STAG2-truncating variants reveal compensatory cohesin remodeling — iScience — documents STAG1 compensation and STAG3 ectopic expression.

6. Loop extrusion — how cohesin controls which genes turn on

This module explains the actual mechanism of the condition. It's the most conceptually demanding idea in the primer, and it's worth the effort.

The problem the cell must solve. Two metres of DNA must fit inside a nucleus a few thousandths of a millimetre across — and not merely fit, but stay organized enough that the right genes can be switched on in the right cells.

The solution: loops. Cohesin grabs the DNA and actively pulls it through its ring, extruding a growing loop — like feeding a rope through your hands to form a bight. It keeps going until it hits a boundary marker, a protein called CTCF bound at specific DNA sites. Cohesin stalls there, and a stable loop is left behind.

flowchart TB
    A["1. Cohesin lands on DNA"] --> B["2. It reels DNA through the ring,<br/>extruding a growing loop"]
    B --> C["3. It stalls at CTCF boundary markers"]
    C --> D["4. A stable loop remains —<br/>a <b>TAD</b> (topologically associating domain)"]
    D --> E["5. Genes and their enhancers are now<br/>held in the same neighborhood<br/>— or kept apart"]

Why loops determine gene activity. Genes are switched on by enhancers — short regulatory sequences that may sit very far away along the linear DNA. An enhancer can only act on a gene if the folding brings the two into physical contact. The loops therefore act as wiring: they determine which switch is connected to which light.

A useful analogy: imagine a very long extension cord, and clips that bundle it into a specific shape so that particular plugs end up beside particular sockets. Change where the clips go, and you rewire the room without changing a single wire.

So what goes wrong with extra STAG2?

More STAG2-containing cohesin means the loop-forming machinery is present in altered amount and altered composition. The pattern of loops shifts. Enhancer–gene connections that should be made aren't, and connections that shouldn't exist are formed. The consequence is many genes mis-expressed by modest amounts — not one gene catastrophically broken.

flowchart LR
    S["Extra STAG2<br/>protein"] --> R["Altered STAG2:STAG1<br/>ratio in cohesin"]
    R --> L["Shifted pattern of<br/>DNA loops / TADs"]
    L --> E["Enhancer–gene wiring<br/>changes"]
    E --> G["Hundreds of genes<br/>mildly mis-expressed"]
    G --> N["Neurons develop and<br/>function abnormally"]
    N --> P["Learning, speech,<br/>behavior differences"]

Two implications worth carrying into every scientific conversation:

1. There is no single broken gene to fix downstream. The damage is distributed across a network. This is precisely why the meaningful measurement is a transcriptomic signature — the whole pattern of gene expression — rather than any one gene. (Module 17.)

2. Timing matters. Loop-based wiring does most of its critical work during development, when neurons are forming connections. This is the honest basis for "earlier is better" — while also being the reason the reversibility question (Module 12) is not foreclosed, since gene expression is an ongoing, continuously-maintained process, not a one-time event.

Key terms:

• Loop extrusion — cohesin actively reeling DNA through itself to form loops.
• CTCF — the boundary protein that stops extrusion and defines loop endpoints.
• TAD (topologically associating domain) — a self-interacting genomic neighborhood formed by loops.
• Enhancer — a regulatory DNA element that activates a gene, often from a distance.
• Promoter — the sequence at a gene's start where transcription begins.
• 3D genome / chromatin architecture — the spatial folding of DNA and its regulatory consequences.
• Chromatin — DNA plus its associated packaging proteins.

References:

• NHGRI — Chromatin
• PubMed search: cohesin loop extrusion CTCF — the primary literature on this mechanism is large and active; this search is the entry point.

7. Why extra STAG2 harms neurons — the evidence in humans

Everything so far has been mechanism. This module covers what has actually been demonstrated in patients, which is a different and higher bar.

The foundational paper. Kumar et al., Human Molecular Genetics, 2015 — "Increased STAG2 dosage defines a novel cohesinopathy with intellectual disability and behavioral problems." This is the single most important publication for this condition, and it is worth knowing in detail.

What it did:

• Assembled 28 affected males — 18 different Xq25 microduplications plus one microtriplication, across 6 families and 13 singleton cases, plus carrier females.
• Narrowed the critical region to implicate STAG2 as the driver gene (not the neighboring XIAP).
• Showed in cells derived from affected individuals that STAG2 mRNA and protein are elevated.
• Showed that this dysregulates disease-relevant neuronal gene networks.
• Identified that STAG2 gains cause increased expression of OPHN1 — itself a known X-linked intellectual disability gene.

Why the OPHN1 finding is so useful. OPHN1 encodes a protein involved in dendritic and synaptic morphology — the physical shape of neuronal connections. So there is a plausible line from extra STAG2 → altered loop wiring → OPHN1 up → abnormal synapses → learning and behavior differences.

More practically: OPHN1 provides a candidate readout. If a drug lowers STAG2 back toward normal, OPHN1 expression should move back toward normal too. That gives a concrete, cheap, measurable thing to screen against — rather than needing a full transcriptomic experiment for every compound.

The critical caveat, and it should be held firmly. The Kumar study used cells derived from affected individuals — but these were not neurons. Deriving actual neurons from patients was not routine then. The gene networks disrupted in a blood or skin cell are only a partial proxy for what happens in a brain cell.

This is exactly the gap that patient-derived neurons can fill. Reprogramming an affected individual's skin cells into iPSCs and then into neurons (Part 4) allows the Kumar findings to be re-tested in the correct cell type. Two questions follow directly, and both are excellent things to ask of any laboratory doing this work:

1. Does the Kumar signature — including OPHN1 — replicate in patient-derived neurons?
2. If it does, is it robust enough to serve as a screening readout?

A note on what is NOT established. At the time of writing, no published iPSC-derived neuronal model of STAG2 duplication could be found in the literature. If that holds, any such line is among the first in the world. That is scientifically significant: it makes the work publishable, it makes it a magnet for collaborators, and it gives a family real standing to ask for data sharing and co-authorship arrangements.

Key terms:

• OPHN1 (oligophrenin-1) — X-linked ID gene, upregulated by STAG2 gain; a candidate biomarker.
• Biomarker — a measurable proxy for disease state or drug effect.
• Transcriptomic signature — the characteristic pattern of gene expression differences.
• Gene network — a set of functionally related, co-regulated genes.
• Driver gene — within a multi-gene CNV, the gene actually responsible for the phenotype.
• Dendrite / synapse — the receiving branches of a neuron, and the junctions between neurons.

References:

• Kumar et al. 2015 — Increased STAG2 dosage defines a novel cohesinopathy (PubMed 26443594) — read this one. DOI: 10.1093/hmg/ddv414
• Full text at Oxford Academic
• Leroy et al. — Xq25 duplication: the crucial role of the STAG2 gene (PubMed 25677961) — the paper that framed this as a cohesinopathy.

8. Two opposite diseases from one gene

STAG2 causes two distinct syndromes in opposite directions, and conflating them is the most common and most consequential error in reading this literature.

Why this matters practically:

• Most STAG2 literature is about the wrong direction. The great majority of STAG2 papers concern loss of function — in MKMS, and especially in cancer, where STAG2 is a frequently mutated tumor suppressor (bladder cancer, Ewing sarcoma, leukemias). A PubMed search for "STAG2" returns mostly cancer papers about losing it. Filter accordingly.

• Cancer literature is still useful — inverted. Cancer researchers have spent years finding compounds that exploit STAG2 loss and mapping what STAG2 does. That's a rich resource, but every finding must be mentally flipped.

• It defines the safety ceiling. MKMS is the picture of what over-treatment looks like. It is not an abstract risk; it is a described human syndrome. This is the concrete reason for the "partial reduction, not knockout" principle from Module 4.

Key terms:

• LoF (loss of function) — the gene product is absent or non-working.
• GoF / gain of dosage — too much gene product.
• MKMS (Mullegama-Klein-Martinez syndrome) — the STAG2 loss-of-function syndrome.
• Somatic vs germline — acquired in a tissue (e.g. a tumor) vs present in every cell from conception.

References:

• Mullegama et al. 2019 — Expanding the STAG2 LoF phenotype in males
• Familial STAG2 germline mutation defines a new human cohesinopathy — npj Genomic Medicine 2017

9. The cohesinopathies — the disease family

Xq25 duplication syndrome belongs to a group called the cohesinopathies — disorders caused by disruption of the cohesin complex or its regulators. Knowing the family is strategically useful: these communities are larger, better funded, and their science frequently transfers.

Why this framing is worth using deliberately:

1. Funding and advocacy. "Cohesinopathy" is a recognized disease category with existing funders and patient organizations. "Xq25 duplication syndrome" is, to most people, an unfamiliar string. The umbrella term opens doors the specific term does not.

2. Mechanistic transfer. CdLS research on how cohesin disruption alters neurodevelopment applies here even though the gene differs — the downstream machinery is shared.

3. Community. Ultra-rare families often find their most useful allies one step out, in the adjacent and better-organized community.

Key terms:

• Cohesinopathy — any disorder caused by disrupted cohesin function.
• CdLS (Cornelia de Lange syndrome) — the best-known cohesinopathy.
• Phenotypic spectrum — the range of clinical presentations from related genetic causes.

References:

• Cornelia de Lange Syndrome Foundation — a model parent-led organization, and a plausible ally.
• PubMed search: cohesinopathy neurodevelopment