# Nuclear Architecture and Membraneless Compartments

> The nucleus has no membrane-bound compartments inside it, yet its chromosomes keep to their own regions and its nucleoplasm holds bodies with different jobs. This lesson describes chromosome territories and the two Hi-C compartments, the nucleolus and its smaller neighbours, and the proposal that many such bodies form by phase separation. Afterwards you can state what phase separation predicts, compare an ideal one-protein phase with a fixed set of binding sites by asking where extra protein goes, work a mass-balance test on concentration data, and say why round shapes, FRAP and hexanediol do not settle the mechanism.

Course: Cell Biology (Membranes and compartments). Level: core university (third year). About 26 minutes.
Lesson page (link this one when citing): https://stemhop.com/lessons/nuclear-architecture-condensates/

Builds on: Organelles as Distinct Environments · Experiments That Distinguish Mechanisms · The Nuclear Envelope and Nuclear Transport

## What you'll learn

- Describe how interphase chromosomes occupy territories that intermingle at their edges, and how Hi-C sorts chromatin into open and closed compartments.
- Identify the nucleolus and its three parts, nuclear speckles, Cajal bodies, PML bodies and paraspeckles by what each holds.
- Explain the phase-separation hypothesis and its ideal one-protein predictions at equilibrium under fixed conditions (a threshold, fixed concentrations inside and outside the bodies, exchange across a boundary), and compare it with binding to fixed sites and polymer bridging by asking where extra protein goes, a comparison that supports a model rather than proving it.
- Separate what is well supported about nuclear bodies from what is still contested, and design an experiment that tests the mechanism in a living cell.

**A human nucleus is only several micrometres across and holds roughly two metres of DNA, but it has no membrane-bound compartments inside it, and it is still not a uniform soup.**

The short answer to the hook: chromosomes keep to territories, bodies gather molecules by binding them to scaffolds, and some bodies may form by phase separation. Any membraneless body that concentrates proteins and nucleic acids is a **biomolecular condensate**, whatever made it. The first film’s droplets are one proposed way to make one, **liquid-liquid phase separation**: above a certain concentration a solution demixes into a **dense phase** (the droplets) and a **dilute phase** (here the **nucleoplasm**, the nuclear interior around them). For proteins it is thought to be driven by **multivalency**, many weak, reversible binding sites per molecule (repeated domains, disordered regions, RNA; [that lesson](https://stemhop.com/lessons/water-weak-bonds/)). The first film’s sites are **clustered sites**: many copies of a DNA or RNA sequence in one place, bound with no demixing.

> **The one idea to carry through this lesson.** A bright spot shows that something is concentrated, not how. Add more protein and ask where the extra goes. For one protein demixing on its own, at equilibrium under fixed conditions, the extra adds dense-phase volume (bigger drops or more of them) while the inside and outside concentrations stay fixed. On a fixed set of sites that does not grow, the extra raises occupancy: spots brighten until the sites fill, and only then does the outside climb. Each pattern supports its model; neither proves it alone.

The first film showed that comparison; it leaves out what the imaging checks really measure and why the graphs have their shapes. The second film tours the bodies and chromatin a real nucleus holds, then examines liquid-phase tests in a frog nucleolus and in heterochromatin. The last section (Where the test is weaker) takes up the first film’s closing complications.

## Why looking harder does not help

Why did the first film’s usual checks pass both kinds of spot? Look at what each one measures. **Roundness** alone shows neither softness nor mechanism, because clusters below the resolution of the microscope look round too. Real **fusion**, two bodies merging and relaxing into one round body, is what drops do, but clusters that drift together can look merged unless that relaxation is seen. **[FRAP](https://stemhop.com/methods/frap/)**, fluorescence recovery after photobleaching ([that lesson](https://stemhop.com/lessons/membrane-organization-topology/)), measures turnover, which binding and unbinding also give.

**Predict.** 1,6-Hexanediol, an alcohol thought to weaken hydrophobic contacts, dissolves a nuclear body. What would you want to know before calling the body a liquid?

> **Check your prediction.** What else the alcohol does. Itoh and colleagues found that 5% hexanediol for 5 minutes strongly suppressed chromatin motion in live HeLa cells, and foci of one nuclear body, the Cajal body (below), fell in number in the same treatment. A reagent that changes chromatin this much cannot tell you which mechanism held a body together.

**Half-body FRAP** has more power: bleach half of a large body and follow both halves and the nucleoplasm beside it. A droplet has a boundary, so molecules are expected to mix inside it faster than they cross its edge. That pattern favours a phase without proving one: the weak interfaces of real condensates may slow crossing only a little, and molecules that rebind sites inside a dense cluster can also mix internally.

## Add more protein: the numbers

The first film’s picture was the simple case; here is the fuller one: why each graph has its shape, and where it stops. **Total** (along the bottom) is averaged over the whole nucleus; **outside** (up the side) is measured between the spots.

- **A phase of one protein.** Below the **saturation concentration**, csat, there are no bodies and outside equals total: a 45° line. Above it the line goes flat: like salt past its solubility, extra protein forms more dense phase, whose own concentration, cdense, also stays fixed. That holds at equilibrium and fixed conditions, until the dense phase fills the nucleus (φ = 1, below).
- **Sites.** While they are far from full, each molecule has the same chance of being bound at any moment, so a fixed fraction stays outside: a straight line through zero, below the 45° line. Once every parking space is taken, extra protein stays outside and the line bends up. This assumes fixed sites; if added protein brought in new sites or reshaped the chromatin that carries them, the spots could grow too.

A flat outside line alone is weaker evidence than it looks: cooperative binding can flatten it too (Where the test is weaker).

For a phase, the amounts must add up: the protein in the dilute phase plus the protein in the dense phase is all of it.

> **Equation card: What fraction of the nucleus the dense bodies fill, once the protein is above its saturation concentration.**
>
>     φ = (c_total − c_sat) / (c_dense − c_sat)
>
> In words: The bodies' share of the nucleus is the excess over saturation, divided by the gap between the dense and dilute concentrations.
>
> - φ: the fraction of nuclear volume taken up by the dense phase (0 to 1)
> - c_total: the protein's average concentration over the whole nucleus (μM)
> - c_sat: the saturation concentration, where the dilute phase stays (μM)
> - c_dense: the concentration inside the bodies (μM)
>
> Sanity check: At c_total = c_sat, φ = 0: the coexistence boundary, with a nonzero dense fraction just above it. At c_total = c_dense, φ = 1: the dense phase fills the nucleus. In between, at equilibrium and fixed conditions, c_sat and c_dense stay fixed, so only φ grows.
> Where it comes from optional Step 1: c_total = c_sat × (1 − φ) + c_dense × φ (the dilute volume times its concentration, plus the dense volume times its). Step 2: multiply out: c_total = c_sat + φ × (c_dense − c_sat). Step 3: solve for φ.

**Predict.** A protein has csat = 1 μM and cdense = 50 μM. Its total rises fourfold, from 2 to 8 μM. Do its bodies double, quadruple or more than quadruple in volume?

> **Check your prediction.** More: sevenfold. φ goes from 1/49 = 2.0% of the nucleus at 2 μM, through 3/49 = 6.1% at 4 μM, to 7/49 = 14.3% at 8 μM. Volume tracks the excess over csat, which goes from 1 to 7, not the total, which goes from 2 to 8. The nucleoplasm stays at 1 μM.

**Worked problem (made-up data).** Two proteins, each tagged at its own gene, are set to total nuclear concentrations of 2, 4 and 8 μM.

1. **Protein A:** outside 1.0, 1.0 and 1.1 μM, while the total body volume grows at the same brightness. Bigger, not brighter: consistent with the simple phase model.
2. **Protein B:** outside 1.5, 2.8 and 5.9 μM, or 1.5/2 = 0.75, 2.8/4 = 0.70 and 5.9/8 = 0.74 of the total, while its foci brighten in proportion. A constant fraction outside is the straight line through zero: sites far from full.

*Interactive on the lesson page:* Condensate lever rule: how bodies grow as the protein goes up

Below 1 μM, the outside follows the total and φ stays zero; above 1 μM, watch φ grow while the outside stays at 1 μM.

Those are made-up proteins. The second film picks up here: it tours the bodies and chromatin a real nucleus holds, then examines two systems tested for liquid-like behaviour.

**Film transcript: Real Bodies, Real Tests**

### Film transcript: Real Bodies, Real Tests

Chapter one ended on a clue: add protein and see whether a spot grows bigger or brighter. Now meet the bodies and chromatin a real nucleus holds. Which have been tested as liquids, and what did those tests show? Some order needs no spot at all. After a cell divides, each chromosome partly unpacks but keeps a region of its own: a chromosome territory.

Picture balls of yarn in a box, with loose loops poking across where two meet. A sequencing method called Hi-C maps which stretches of chromatin touch. Its maps reveal two broad compartments. A is open and mostly active. B is closed and mostly quiet, often against a mesh under the nuclear envelope. Standard maps average over millions of cells, not one nucleus. Now the bodies. The nucleolus forms around repeated genes for ribosomal RNA.

Mammalian nucleoli have three parts along an inside-out assembly line. Fibrillar centres hold the genes and copying machinery. New RNA emerges at their borders, where processing begins. The outer granular component assembles ribosomal subunits. Smaller bodies sit around it, each known by what it holds. Speckles hold splicing factors; Cajal bodies, the marker protein coilin; PML bodies, the PML protein. Paraspeckles are built on a long RNA called NEAT1. These bodies are real. They exchange molecules with the nucleoplasm, and several help with a job. But how does each one form? Only a few have had their material properties tested closely. The strongest liquid-phase evidence here comes from the nucleolus in the egg-forming cells of a frog. Its dense fibrillar and granular components behave like two liquids that coexist without mixing.

What should purified nucleolar proteins do when mixed with RNA in a tube? They reproduce the layering: drops with one phase inside the other. This is strong evidence that the frog nucleolus contains coexisting liquid phases. Heterochromatin, densely packed and mostly silent chromatin, adds a third mechanism. A protein can bridge distant stretches until the fibre collapses into a compact ball. Physicists count that as a phase transition of the chromatin. But it makes a ball of chromatin, not a liquid drop of protein. The verdict splits by system. In early fruit fly embryos, foci of HP1a looked liquid-like. In mouse fibroblasts, half-body FRAP found no preferential mixing inside heterochromatin clusters, favouring collapsed chromatin.

Neither study speaks for all heterochromatin. Whether phase separation builds a given body in a given living cell is still contested. The text below shows where the test itself grows weaker.

From the film

- Chromosomes keep to territories, and Hi-C reveals open A and closed B compartments.
- Mammalian nucleoli have three parts that build ribosomal subunits, read from the inside out.
- Speckles, Cajal bodies, PML bodies and paraspeckles are each known by what they hold.
- The frog oocyte nucleolus provides strong evidence for coexisting liquid phases.
- Chromatin bridging is a third mechanism, and for heterochromatin the verdict splits by system.

## What the nucleus holds

The second film toured the territories and bodies; here is the detail behind them.

### Chromosomes keep to territories, mostly

During mitosis each chromosome is a compact rod. Afterwards it partly unpacks, but it does not spread through the whole nucleus. Labelled with fluorescent DNA probes, a method called [chromosome painting](https://stemhop.com/methods/karyotyping-and-chromosome-fish/), each one fills a region of its own, a **chromosome territory**. The DNA is packed as **chromatin**, DNA wound on histone proteins, and how tightly it is packed affects whether its genes are used ([that lesson](https://stemhop.com/lessons/eukaryotic-regulation/)). Picture balls of yarn in a box: each keeps its own space, but where two press together, loose loops of one poke into the other.

**Predict.** Why can painting a whole nucleus not show whether neighbouring territories mix at their edges?

> **Check your prediction.** A conventional light microscope blurs everything below a few hundred nanometres ([that lesson](https://stemhop.com/lessons/microscopy-and-images/)), and in a whole nucleus light from above and below blurs every edge. In 2006 Miguel Branco and Ana Pombo painted thin frozen sections of fixed human lymphocytes, which removes most of that blur. Neighbouring territories overlapped at their edges, and how much two chromosomes overlapped correlated with how often that pair takes part in translocations (exchanges of pieces between chromosomes) measured in human lymphocytes after ex vivo irradiation. That is a correlation, not proof that overlap causes any given translocation.

*Figure (AI-generated illustration).* Each chromosome fills a territory, but where two meet, loops of one poke into the other. A frozen section about 150 nm thick removes most of the blur, so the overlap can be seen. Image: a nucleus with three painted chromosome territories whose edges interleave, cut by a thin frozen section, with an inset contrasting a blurred whole-nucleus edge against a sharp thin-section edge.

**[Hi-C](https://stemhop.com/methods/chip-seq-and-chromatin-profiling/)**, a sequencing method, gives a second view: chromatin is cross-linked in the cell, cut and re-joined, so segments lying close in space end up in one DNA molecule, and sequencing millions of these junctions maps which parts of the genome touch. The 2009 maps of human cells, at a resolution of one megabase, showed that open and closed chromatin segregate into two genome-wide compartments, A (open, gene-rich, mostly active) and B (closed, mostly quiet). Closed chromatin often lies against the nuclear lamina, the protein mesh under the nuclear envelope ([that lesson](https://stemhop.com/lessons/nuclear-transport/)). The maps average over millions of cells: they report how often two regions are near each other, not where they sit in one nucleus. Later single-cell versions map one nucleus at a time.

*Figure (AI-generated illustration).* In Hi-C, chromatin is cross-linked, cut and re-joined so that regions close in space end up in one molecule; each sequenced junction records two regions that were near each other. Standard Hi-C data average over millions of cells. Image: four steps of Hi-C, from cross-linking two chromatin stretches that lie close in space, to cutting, joining their ends into one piece, and sequencing that junction.

### Bodies without membranes

A **nuclear body** is a region of the nucleoplasm where particular molecules are concentrated, with no membrane around it. The largest, the **nucleolus**, carries out ribosome biogenesis. It forms around the repeated genes for ribosomal RNA (the rDNA) and, in mammalian cells, has three parts, like an assembly line read from the inside out. The **fibrillar centre** holds the rDNA and the machinery of RNA polymerase I, the enzyme that transcribes it; active transcription occurs at or near its border with the **dense fibrillar component**. In that component, which is rich in fibrillarin, the newly made ribosomal RNA precursor (pre-rRNA) begins to be processed. The outer **granular component**, rich in nucleophosmin (NPM1), is where ribosomal subunits are assembled before export.

The smaller bodies, with one key experiment each:

| Body | What it holds | One key experiment |
| --- | --- | --- |
| **Nuclear speckles** | Pre-mRNA splicing factors, which exchange with the nucleoplasm ([splicing](https://stemhop.com/lessons/rna-processing/)) | Mouse embryonic stem cells: genes near speckles showed more co-transcriptional splicing (splicing while the RNA is still being made) than genes farther away. Human HEK293T cells: tethering a reporter pre-mRNA to speckles increased its splicing |
| **Cajal bodies** | Coilin, their marker, and components of the small nuclear ribonucleoproteins (snRNPs) that splice pre-mRNA | Engineered mammalian cells: tethering Cajal-body components to chromatin started apparently functional Cajal bodies from scratch |
| **PML bodies** | The PML protein that builds them, SUMO (a small protein tag) and partner proteins | Cultured mammalian cells and mice: oxidation drove PML to multimerize and nucleate the bodies, which then enhanced stress-induced SUMO modification of partners inside |
| **Paraspeckles** | Their scaffold, the long noncoding RNA NEAT1, and its bound proteins | Mice lacking Neat1 have no paraspeckles yet were reported viable and fertile in the laboratory; later work found defects in pregnancy and lactation |

These facts are well supported: the bodies exist, they exchange molecules with the nucleoplasm, and several of them help with a job. How each forms is the second film’s question, and few have had their material properties tested closely.

*Figure (AI-generated illustration).* The nucleus holds bodies with no membrane, each known by what it concentrates. The nucleolus has three parts, read from the inside out: fibrillar centre, dense fibrillar component, granular component. Image: a cross-section of a nucleus with no membrane-bound compartments holding a large nucleolus and smaller speckles, Cajal bodies, PML bodies and a paraspeckle, each labelled by what it holds, with an inset showing the nucleolus as three nested layers.

## Real bodies: one strong case, one split verdict

The strongest liquid-phase evidence here comes from the nucleolus in oocytes of the frog Xenopus laevis. There the fibrillarin-rich and NPM1-rich regions behave as two liquids that coexist without mixing, and purified nucleolar proteins mixed with RNA reproduce the layering (see Marina Feric’s card). This is strong evidence that these frog nucleoli contain coexisting liquid phases, not that every nucleolus is a single liquid phase.

**Heterochromatin**, densely packed, mostly silent chromatin, adds a third mechanism. A protein can **bridge** chromatin, cross-linking distant segments so the fibre collapses into a compact globule. Physicists count this collapse as a phase transition of the chromatin polymer, but the result is a globule of chromatin, not a liquid drop of the protein.

**Predict.** A protein makes a focus by bridging one stretch of chromatin. If you double the protein, must the focus act like the first film’s sites?

> **Check your prediction.** Not necessarily. If the chromatin stretch, its compaction and its sites are fixed and nearly full, extra protein mostly raises occupancy, like sites. But more bridges can pull in and compact more chromatin, changing the size of the focus. Bridging is its own mechanism, which the comparison does not sort automatically.

The verdict on heterochromatin splits by system. Purified Drosophila HP1a demixed in the test tube, and HP1a foci in early Drosophila embryos looked liquid-like, with mature domains proposed to contain liquid and stable parts. In mouse fibroblasts, by contrast, chromocentres, dense clusters of heterochromatin, behaved like collapsed chromatin globules rather than HP1 liquid pools, and half-body [FRAP](https://stemhop.com/methods/frap/) found no preferential mixing inside them. Test-tube numbers for mouse HP1α do not settle it either (Going deeper). Neither study speaks for all heterochromatin.

## Where the test is weaker

The first film ended on three complications. Here is what each changes, after a look-alike neither chapter showed.

**A flat outside line can mislead.** If many molecules must bind together before a site holds any (cooperative binding), the sites stay nearly empty at low concentration and then fill over a narrow range, like a threshold, so the outside stays nearly flat while they fill.

**Predict.** A protein’s outside concentration creeps only from 1.6 to 2.2 μM while its total triples, from 3 to 9 μM. Is it a phase?

> **Check your prediction.** Not yet. A made-up model of cooperative sites gives just these numbers; its outside climbs (to 4 μM at a total of 14 μM) only once the sites are full. Ask where the extra went: more body volume at the same brightness supports a phase, brighter foci of the same size support sites.

**Mixtures.** A protein that fails the test may still phase separate as one part of a mixture of several components, which need not buffer each one. In human cells, the dilute-phase concentration of the nucleolar protein NPM1 rose with its total expression, which was explained by its interactions with other nucleolar components (Riback and colleagues, 2020).

**Seeds.** In engineered Corelet experiments in living cells, local capture of a protein on a small engineered seed triggered phase separation below the global concentration needed for it. If chromatin acts as a seed in the same way, a body could form on it where the average level looks too low.

**Tiny spots.** Below the blur of the microscope, a few hundred nanometres, size and brightness cannot be told apart.

**Function.** A protein that passes has not been shown to need the body for any function. Speckle tethering shows that location matters, not that phase separation is the reason, and that mice without paraspeckles survive in the laboratory does not mean the bodies do nothing.

What is contested, then, is whether phase separation is the mechanism for a given body in a given living cell, and whether its material state matters for its job.

### Design problem

A laboratory reports that protein X forms round nuclear foci that fuse, recover after [photobleaching](https://stemhop.com/methods/frap/) and dissolve in 1,6-hexanediol. It concludes that X drives a phase separation needed for transcription. What is wrong with the argument, and what would you do instead?

> **One way to do it.** None of the three observations is diagnostic: roundness, apparent fusion and FRAP recovery also fit binding to clustered sites or chromatin bridging, and hexanediol also suppresses chromatin motion.
>
> Instead, tag X at its own gene (its endogenous locus) and check that the tag does not change its amount or behaviour. Raise and lower its level, measure concentrations inside and outside the foci, and ask where the extra goes: into more focus volume at a steady outside concentration, or into brighter foci of fixed size (support for a model, not proof). Use half-body FRAP to ask whether X mixes inside a focus faster than it crosses the edge. Make a mutant that keeps DNA binding but loses its multivalent region, and see whether foci and transcription change together. Finally, test transcription separately from assembly, with a control that moves the same genes to or from the foci by another route, much as Bhat and colleagues tethered a reporter pre-mRNA to speckles to test splicing. Only then can you ask whether the foci themselves, rather than protein X alone, are needed.

## Going deeper

### More on territories, compartments and bodies

Branco and Pombo (2006) activated their human lymphocytes with phytohemagglutinin before fixing and sectioning them. Deeper [Hi-C](https://stemhop.com/methods/chip-seq-and-chromatin-profiling/) maps in 2014, at kilobase resolution, also found smaller contact domains and thousands of chromatin loops whose anchors usually bind the protein CTCF. [Immunofluorescence](https://stemhop.com/methods/immunohistochemistry/) usually shows about 20 to 50 nuclear speckles per nucleus in mammalian cells. PML bodies take their name from promyelocytic leukaemia, in which a fusion of PML with a retinoic acid receptor is associated with disrupted bodies.

### The HP1α turbidity numbers

Purified mouse HP1α mixed with DNA turned cloudy (turbid), a sign that droplets may be forming, reaching half-maximal turbidity at 45 μM untagged and 23 μM with a GFP tag. These are in vitro values: a number that a tag alone can roughly halve, measured in a tube, cannot be read as the protein’s saturation concentration in a nucleus.

---

From STEMhop (https://stemhop.com), free lessons by Christopher Deppmann. Lesson text © the authors. Practice questions: https://stemhop.com/lessons/nuclear-architecture-condensates/
