Learn/Tracer studio/Example 1

What can a PET scan tell us?

A PET image is a map of radioactivity. Turning it into a statement about the brain takes two further steps — and each adds assumptions. This first example follows one tracer, [¹¹C]UCB-J, from what the scanner records to what it can, and cannot, say about synapses.

[¹¹C]UCB-JSV2AIllustrative simulationAbout 10 minutes

00 The tracer

One tracer, one target

[¹¹C]UCB-J is a PET tracer: a small molecule that binds tightly and selectively to SV2A (synaptic vesicle glycoprotein 2A), labelled with radioactive carbon-11 so that the scanner can follow it.

SV2A sits in the membrane of synaptic vesicles — the small packages that store neurotransmitter inside nerve terminals — and is found in terminals throughout the brain. That is why [¹¹C]UCB-J binding is widely used as an in vivo index of synaptic density.

SV2A is also where the anti-seizure drug levetiracetam binds. That gives a simple test of whether a signal really reflects SV2A: give levetiracetam first and see whether the signal falls. You can try it in layer 2.

Schematic, not to scale. [¹¹C]UCB-J reaches nerve terminals from the blood and binds SV2A on synaptic vesicles; some tracer is always present unbound.

Every PET result passes through three layers. Keeping them apart is the difference between a measurement and a conclusion.

1Measured radioactivityWhat the scanner records
2Target bindingWhat a model infers from it
3Biological interpretationWhat it means — and what it doesn't

1 Layer one

Measured radioactivity

What the scanner records

When a carbon-11 atom decays, it emits a positron. Within a short distance it meets an electron, and the two annihilate into a pair of gamma photons travelling in opposite directions. The scanner records these pairs in coincidence, and image reconstruction turns millions of them into maps of radioactivity concentration.

A dynamic scan repeats this in time frames — short ones just after injection, when the signal changes fast, and longer ones later. For each brain region, that gives a time–activity curve.

At this layer, every carbon-11 atom counts the same wherever it is: tracer bound to SV2A, tracer free in tissue or stuck to other structures, and tracer still in the blood vessels. The image alone cannot tell them apart.

Carbon-11 also decays quickly — its half-life is about 20 minutes — so the raw signal fades during the scan. Values are corrected for decay, but late frames are built from fewer counts and are noisier. Switch between the two views to see both effects.

Can tell us

Where the radioactivity was, and how it changed over time.

Cannot tell us

How much of it is bound to SV2A.

One brain region · 90-minute dynamic scanIllustrative simulation

Show values as a table

2 Layer two

Target binding

What a model infers from the measurement

To get from radioactivity to binding, we use what is known about how a tracer moves: it is delivered by the blood, crosses into brain tissue, binds its target and comes off again. A kinetic model describes these steps with a few rate constants and is fitted to the measured curve.

The model also needs an input: the concentration of intact tracer in arterial blood, or a reference region that lacks the target. Neither is trivial for [¹¹C]UCB-J.

In our rat study, the tracer was broken down quickly: twenty minutes after injection, only about a third of the radioactivity in plasma was still intact [¹¹C]UCB-J, so the blood curve had to be corrected for metabolites. And in our minipig study, levetiracetam also lowered the signal in white matter — so white matter could not serve as an SV2A-free reference region there.

The main outcome is the total volume of distribution, V_T: the ratio of tracer in tissue to tracer in plasma at equilibrium. It includes both specifically bound and non-displaceable tracer. The binding potential, BP_ND, compares the two directly and is proportional to the density of available SV2A sites — but it requires knowing V_ND, which can only be estimated indirectly.

Our [¹¹C]UCB-J studies have reported V_T from a one-tissue compartment model. In rats, levetiracetam doses of 10, 40 and 100 mg/kg occupied about 33%, 78% and 82% of the sites, and V_ND was estimated at 2.5 mL/cm³ — well below regional V_T values of 32–46 mL/cm³. Most of the signal was specific.

The simulation is built from separate free and bound compartments so that you can see them. Its baseline values sit in the same range as the rat data.

V_T = V_ND (1 + BP_ND) BP_ND ∝ B_avail / K_D

Can tell us

How much tracer is specifically bound — if the model's assumptions hold.

Cannot tell us

What that binding means biologically.

Assumptions worth knowing. The model assumes that the tracer is given in trace amounts, that the brain is in a steady state during the scan, and that the blood input is measured correctly. In animal studies, anaesthesia and scan conditions must also be matched between groups.

What the model sees inside the same curveIllustrative simulation

SV2A sites available 100%

V_T · total

–

V_ND · non-displaceable

–

unchanged by binding

BP_ND · binding potential

–

Show baseline values as a table

Figure from Binda et al. 2025. Panel a: coronal rat brain slices; columns show MRI, [11C]UCB-J V_T maps at baseline, and V_T maps after taVNS, with frontal cortex, striatum and midbrain outlined; colour scale V_T 0 to 30. Panel b: time–activity curves in the left striatum, radioactivity in kBq/cc over 90 minutes, baseline in black and after taVNS in red. — **[¹¹C]UCB-J PET in the rat.** (b) Radioactivity in the left striatum over 90 minutes, at baseline and after 30 minutes of transcutaneous auricular vagus nerve stimulation (taVNS). (a) V_T maps from kinetic modelling, overlaid on MRI. Binda KH, Real CC, Simonsen MT, Grove EK, Bender D, Gjedde A, Brooks DJ, Landau AM. *Psychophysiology* 2025;62:e14709 · doi:10.1111/psyp.14709 · CC BY-NC-ND 4.0, reproduced unmodified.

In a real study

The same three layers, in one published figure

1
Measured radioactivity
Panel (b): the time–activity curve in the striatum — what the scanner recorded, before and after stimulation.
2
Target binding
Panel (a): V_T maps. A one-tissue model, with an input curve taken from the heart in the images, turned the curves into a binding measure for every voxel.
3
Interpretation
SV2A binding was lower after stimulation, while [¹⁸F]FDG PET in the same study showed no clear change in glucose metabolism. The authors note that ex vivo experiments are needed to corroborate the in vivo findings.

3 Layer three

Biological interpretation

What it means — and what it doesn't

Suppose a study finds lower [¹¹C]UCB-J binding in one brain region in a disease model than in healthy controls. Which of these statements can the data support? Decide for each, then check.

Can we conclude…
“SV2A binding is lower in this region in the model than in controls.”

Yes — this is what was measured
This is the direct result, provided both groups were scanned under the same conditions, the model fitted well and the groups were large enough to separate a real difference from noise.
Can we conclude…
“There are fewer synapses in this region.”

Consistent with — not proof
Because SV2A is found in nerve terminals throughout the brain, lower binding is consistent with fewer synapses, and this is how it is usually read. But it could also mean fewer vesicles per terminal, fewer SV2A copies per vesicle, or fewer available sites. Measurements in brain tissue from the same animals can test the synaptic reading.
Can we conclude…
“The remaining synapses are working less well.”

Not from this measurement
SV2A binding reflects how much of the protein is present, not how active the synapses are. Questions about activity need other measures — for example glucose metabolism with [¹⁸F]FDG PET, or electrophysiology.
Can we conclude…
“Excitatory synapses have been lost.”

Not from PET alone
SV2A is present in both excitatory and inhibitory nerve terminals, and [¹¹C]UCB-J does not tell them apart. Tissue analyses with markers for particular synapse types can.
Can we conclude…
“The lower signal is not simply because the region is smaller.”

Rule this out first
PET images are blurred over a few millimetres. If a region has shrunk, its signal mixes with its surroundings — the partial-volume effect — and can look lower even if SV2A per volume of tissue is unchanged. Structural MRI helps to check and correct for this.
And after treatment…
“A treatment that raises SV2A binding has built new synapses.”

Not from PET alone
A rise is consistent with more SV2A-containing terminals, but also with more SV2A per terminal. Showing that new synapses formed needs direct evidence from tissue — and attention to when the scan was done relative to treatment.

Six statements, four kinds of answer.

Three pairs of line drawings joined by double arrows: a neuron with few versus many dendritic spines; a nerve terminal with few versus many synaptic vesicles; a synaptic vesicle with two versus six SV2A proteins in its membrane. — **What can "altered SV2A binding" represent?** A change in the number of synapses, in the number of vesicles per terminal, or in the number of SV2A molecules per vesicle — each would move the PET signal. Rossi R, Arjmand S, Bærentzen SL, Gjedde A, Landau AM. *Frontiers in Neuroscience* 2022;16:864514 · doi:10.3389/fnins.2022.864514 · CC BY 4.0; panel labels added.

[¹¹C]UCB-J PET measures SV2A binding in the living brain. Synaptic density is an interpretation — often a good one — and it is strongest when tested against other measurements.

That is why we combine imaging with analyses of brain tissue: to understand what imaging signals represent biologically, and how they change with disease and treatment. How our methods fit together →

04 In our work

Where this example comes from

We use imaging of synaptic vesicle glycoprotein 2A (SV2A), including PET with [¹¹C]UCB-J, to investigate changes associated with disease and treatment. By combining imaging with analyses of brain tissue, we examine how changes in SV2A binding relate to synaptic biology and synaptic density.

We characterised [¹¹C]UCB-J PET in rats and in Göttingen minipigs, testing it with levetiracetam blocking and comparing it with [³H]UCB-J autoradiography in brain tissue. The papers below are a starting point.

2021
In vivo imaging of synaptic SV2A protein density in healthy and striatal-lesioned rats with [¹¹C]UCB-J PETPET
Thomsen MB, Jacobsen J, Lillethorup TP, … Brooks DJ, Landau AM · Journal of Cerebral Blood Flow & Metabolism
2020
Preclinical PET studies of [¹¹C]UCB-J binding in minipig brainPET
Thomsen MB, Schacht AC, Alstrup AKO, … Brooks DJ, Landau AM · Molecular Imaging and Biology
2022
Synaptic vesicle glycoprotein 2A: features and functionsReview
Rossi R, Arjmand S, Bærentzen SL, Gjedde A, Landau AM · Frontiers in Neuroscience
2025
Acute transcutaneous auricular vagus nerve stimulation modulates presynaptic SV2A density in healthy rat brain: an in vivo microPET studyPET
Binda KH, Real CC, Simonsen MT, … Brooks DJ, Landau AM · Psychophysiology
2021
Exercise protects synaptic density in a rat model of Parkinson's diseaseAutoradiography
Binda K, Lillethorup T, Real C, … Chacur M, Landau A · Experimental Neurology

05 Keep going

From physics to psychiatry

▶ Watch · 15 min

Teaching · PET in psychiatry

Seeing Chemistry

A narrated lecture for medical students, with five pause-and-think questions. These chapters pair with this page:

Next in the studio

Dopamine and [¹¹C]raclopride

A second example — how a dopamine D2/D3 receptor tracer responds to changes in dopamine, and what that can tell us — is in preparation.

About the simulation. The curves on this page come from a two-tissue compartment model with illustrative parameters (K₁ = 0.70 mL·cm⁻³·min⁻¹; k₂ = 0.28, k₃ = 1.60, k₄ = 0.10 min⁻¹; 5% blood volume; 80% occupancy in the levetiracetam example), chosen so that V_ND (2.5 mL/cm³) and V_T (about 42 mL/cm³) fall within the ranges reported in our rat study, a typical frame schedule, and noise that follows counting statistics. They mimic the general behaviour of [¹¹C]UCB-J but are not fitted to any data set and are not results from our lab.

Questions for review

Is the level right — mainly for students and interested visitors, with the formulas as optional depth?
Would you like the example tied to one of our published [¹¹C]UCB-J studies rather than a generic "disease model"?
Any statement in layer 3 you would phrase differently?

MethodsFrom signal to evidenceHow PET, autoradiography, tissue analyses and behaviour complement each other.Read → ResearchDopamine signalling and GPR6Our work on a receptor of the striatum — and the groundwork for imaging it.Read →