Target antigen density is the number of copies of a target antigen on the surface of a cell, usually given as molecules per cell, in the context of a surface ‘antigen’ target therapeutic.
It’s one of the strongest predictors of whether a T cell engager, immune cell engager, bispecific antibody, antibody-drug conjugate (ADC) or chimeric antigen receptor (CAR)-T therapy will act on a cell, because most of these drugs need a minimum amount of target to be present on the cell surface before they trigger a response.
In cancer, targets are often selected by virtue of their over-expression on tumour cells compared to healthy cells. This difference is utilised to provide a therapeutic window: the range of antigen densities that can be used to allow the drug to selectively target the tumour, while leaving the healthy tissue alone.
This guide covers what antigen density is, why it’s worth testing across a range of expression levels rather than at one, how to measure it, how to build target cells at defined densities, and how to use them to find activation thresholds, assess the therapeutic window, predict on-target, off-tumour activity and model antigen escape.
What is antigen density and why does it matter?
Antigen density is the number of target molecules on a cell’s surface. It matters because the response of a targeted therapy can be proportional to the amount of target: below a certain density the drug does little, above a threshold the response can climb steeply, and at some point it plateaus.
That threshold behaviour has real consequences. A chimeric antigen receptor (CAR)-T cell or T cell engager that eradicates high-expressing tumour cells in the lab can miss low-expressing ones in the patient.
There is clinical evidence of this, with relapses after CD19- and CD22-directed CAR-T cell therapy linked to tumour cells that have lost or reduced the target. This phenomenon is referred to as antigen escape and can occur by a variety of mechanisms: complete target loss, genetic alterations within the target, alternate splicing, tumour microenvironment influences and tumour heterogeneity, to name but a few.
In the other direction, a drug with a low threshold of activation may indiscriminately attack healthy cells that carry small amounts of the same antigen.
Where the threshold sits depends on the drug format, its affinity, the epitope it binds, the effector cells involved and the readout you use, so it needs to be measured for each molecule rather than assumed.
Antigen density isn’t fixed either. It varies between patients and between tumour sites in the same patient, and there is heterogeneity within the same tumour site. This is why a single ‘positive’ cell line tells you less than you’d hope.
Why test across a range of expression levels?
Testing a candidate across a range of antigen densities shows you the finer details of its effect, not just whether it works. One high-expressing cell line gives you a single point; a density series gives you the threshold where activity begins, the density where it saturates and how steep the curve is in between.
Those finer details are what you need for the decisions that matter in discovery and preclinical work:
- Ranking candidates. Two molecules with the same potency on a high-expressing line can behave very differently at lower densities. The one that stays active at low density may be the better choice for a heterogeneous tumour, or the riskier one for healthy tissue, depending on the target.
- Choosing formats and affinities. Valency, affinity and epitope all shift the threshold. A density series lets you tune them deliberately.
- Predicting safety. If healthy tissue expresses the target at a known low level, you can check whether your candidate is likely to be active at that level before an in vivo study or a clinical surprise.
- Setting potency assay conditions. Knowing where the curve is steep tells you which cell line and density will give a sensitive, reproducible cell-based potency assay later on.
Comparing activity across expression levels also protects you from a common mistake: picking the cell line that makes the drug look best.
How do you measure target molecules per cell?
The standard method is quantitative flow cytometry. You stain cells with a fluorescently labelled antibody against the target at a saturating concentration, then convert the fluorescence signal into a number of molecules using calibration beads that carry known amounts of the same fluorochrome.
In practice:
- Choose the antibody and fluorochrome carefully. Use a clone that binds the target on intact, live cells, and a fluorochrome with a defined fluorochrome-to-protein ratio. Phycoerythrin (PE) conjugates at a 1:1 ratio are the usual choice, because one PE molecule then represents one bound antibody.
- Titrate to saturation. Stain at a concentration high enough that every accessible target is occupied. Below saturation you’ll undercount.
- Run calibration beads in the same session. Bead sets with known numbers of PE molecules, or known antibody binding capacity, give you a standard curve that converts mean fluorescence intensity (MFI) into molecules per cell. Kits based on this approach are widely available.
- Report the result as antibody binding capacity (ABC) or molecules per cell, along with the clone, fluorochrome and bead set used. Different clones and methods give different absolute numbers, so those details are what makes data comparable.
Two things to watch. Bivalent antibodies can bind two targets at once, so ABC can underestimate the true number of molecules. And a population average hides heterogeneity: report the distribution as well as the median, because a line with a broad spread of expression behaves differently from a tight one at the same mean.
How do you make low, medium and high expressing target cells?
You have three main options: pick natural cell lines with different expression, engineer a panel of stable lines at different levels, or use an inducible line where one parental cell can be set to any density you choose. Each answers the question, but they don’t answer it equally.
| Approach | How it works | Strengths | Limitations |
|---|---|---|---|
| Panel of natural cell lines | Choose tumour lines known to express the target at different levels | Biologically realistic; no engineering | Every line has a different genetic background, so differences in killing may not be due to density differences alone. |
| Panel of stable transfectants | Engineer several clones of one parental line at different expression levels | Same background across the panel | Each level is a separate clone; clonal effects and drift; a limited number of levels; multiple cell lines need to be cultured and tested. |
| Inducible expression | One parental line with the target under a small-molecule inducer; density determined by inducer concentration | Same cells, same genetic background, any number of levels including zero; dose-dependent and reversible | Needs characterisation of the induction curve; the parental line must lack the target and be susceptible to the mechanism of action (MoA) under test. |
Inducible expression is the cleanest way to isolate density as the variable, because everything else about the cells stays the same.
RoukenCells’ IndEx-2 cell lines work this way. Each line carries the target under a chemically induced proximity switch, so adding more inducer produces more antigen in a dose-dependent manner, and dual-antigen lines carry two independent switches so two targets can be set separately.
The parental backgrounds are CHO-K1, which has no human antigens of its own, and A549, a human lung carcinoma line, and custom IndEx-2 lines can be built for other targets and backgrounds.
Whichever route you take, report the densities you’ve measured rather than relying on the label. ‘Low’, ‘medium’ and ‘high’ only mean something once they’re numbers.
Using IndEx-2, we have achieved target expression levels from <81 receptors per cell (the lower limit of detection) to in excess of 1,000,000 receptors per cell.
IndEx-2 lines were also used in a peer-reviewed study of CLN-978, a CD19-directed T cell engager, to test its activity against cells with low CD19 expression.
How do you find the activation threshold?
The activation threshold is the lowest antigen density at which your drug produces a measurable response, and you can determine it by running a functional assay against target cells set to a series of densities. The threshold you get depends on the readout, so decide which response you’re asking about before you start.
A typical workflow for a T cell engager or CAR-T:
- Build the density series. Include an antigen-negative control and at least five to six densities spanning the range you expect in patients, measured as above.
- Choose the effector. For a quick, donor-independent read, a Jurkat NFAT-Luc reporter line reports T cell receptor signalling as a luminescent signal within hours and can be CAR-engineered. For a more definitive answer, use primary human T cells or the primary CAR-T product itself.
- Run the functional assay at a fixed effector-to-target ratio and a fixed drug concentration, or as a full matrix of drug concentration against density if you need EC50 values at each level.
- Read out more than one response. Killing, cytokine release (IFN-γ, IL-2, TNFα), activation markers (CD69, CD25), degranulation (CD107a) and proliferation each have their own threshold. Published CAR-T studies have shown that cytotoxicity can be triggered at lower antigen densities than cytokine production — exactly the kind of functional difference that is important to understand. Watanabe et al. reported that the activation threshold for a CD20 CAR-T product was ~200 molecules per cell for cytotoxicity, while cytokine secretion was ~10x higher. This is analogous to a study with the IndEx-2 system, shown below, in which blinatumomab (a CD19-directed T cell engager) had a cytotoxicity activation threshold of ~800 molecules per cell while the cytokine threshold was ~10x higher.
- Define the threshold consistently. Fit each response against density and report the density at which the response first rises above the antigen-negative control by an agreed margin, plus the density at half-maximal response.
Run at least two or three donors for primary cell assays, because donor-to-donor variation in effector cells shifts thresholds too. RoukenBio’s poster on quantifying the activation threshold with IndEx-2 shows this workflow applied to a CD19/CD22 dual-inducible line and ‘OR’-gated CAR-T cells.
How do you assess the therapeutic window?
The therapeutic window is the gap between the antigen density that triggers your drug on tumour cells and the density found on healthy tissues that carry the same target.
To assess it in vitro, you compare the activation threshold from the density series with the densities measured on relevant healthy cells, and check for activity at those lower levels directly.
Three steps make this concrete:
- Measure the target on healthy cells. Primary cells from the tissues most likely to express the target, quantified with the same clone and bead set as your density series, tell you where the physiological ‘healthy’ level sits on your curve.
- Test at healthy-tissue density. Set the inducible line to that density and run the same functional assay. A response there is a warning sign; none is reassurance, provided the assay is sensitive enough to have detected one.
- Express the window as a ratio. The density where the tumour response saturates divided by the density where healthy-level activity begins gives a number you can use to compare candidates and formats.
HER2 is a useful example. It’s overexpressed on some breast and gastric tumours, but it’s also present at low levels on healthy epithelial cells, and on-target, off-tumour toxicity has been reported in the clinic.
A density series is the most direct way to see whether a new HER2-directed molecule respects that difference before it reaches an animal or a patient.
Which models predict on-target, off-tumour activity?
The most predictive in vitro models are ones where the only difference between ‘tumour’ and ‘healthy’ target cells is the amount of antigen they carry. Isogenic inducible lines do this directly; primary healthy cells and tissue models add the tissue context.
Used together, they cover the question from both sides:
- Isogenic inducible lines show whether the drug is active at a ‘healthy-tissue’ density with everything else held constant.
- Primary healthy cells (for example epithelial, endothelial or haematopoietic cells that express the target at low levels) show whether the real cells, with their real density and biology, are hit.
- Co-cultures and 3D models show what happens when tumour and healthy cells are present together, which matters for bystander effects and for engagers that recruit effector cells to the site.
This is also where dual-antigen lines earn their place. Logic-gated CAR-Ts and dual-targeting engagers can be designed to act only where two antigens coincide, and a line where each antigen can be set independently is the only clean way to test that the gate holds at the densities healthy tissue actually shows.
RoukenBio runs on-target, off-tumour risk profiling as part of its assay services for safety assessment, for teams that want the study designed and run for them.
How do you model antigen escape?
Antigen escape is the loss or reduction of the target on tumour cells during treatment, and it’s a leading cause of relapse after CD19 and CD22 CAR-T therapy.
You model it in vitro by exposing the drug to target cells with reduced or absent antigen, and to mixed populations, and measuring what survives.
Three experiments cover most of the ground:
- Serial density reduction. Step the inducer down and find the density at which the drug stops eliciting a biological response in the target cells. That number tells you how far the target can fall before the therapy is likely to fail.
- Mixed populations. Combine antigen-positive and antigen-negative cells at defined ratios and track the negative fraction over time. Outgrowth of the negative population is the in vitro version of antigen-negative relapse.
- Dual-antigen loss. For dual-targeting or ‘OR’-gated designs, drop one antigen while holding the other constant, then the reverse. The design should keep killing if either target is present; a dual-inducible line such as CHO-K1 IndCD19/CD22 is the direct way to prove it.
Because the switch in an inducible line is reversible, you can also model down-modulation and recovery of the target over time, rather than just presence or absence.
Where to go from here
If you’re building a density series for a specific target, the IndEx-2 collection lists the targets available off the shelf, and the IndEx-2 whitepaper goes deeper on how the platform itself was built and validated.
If you’d rather have the work run for you, RoukenBio’s cell engineering services and in vitro pharmacology team can design and run these studies as a service.
Frequently asked questions
What is antigen density in simple terms?
Antigen density is how many copies of a target protein sit on the surface of a single cell, usually reported as molecules, or receptors, per cell. Targeted therapies such as T cell engagers, CAR-Ts and antibody-drug conjugates need a certain amount of target before they act, so the number influences which cells are affected.
How many molecules per cell does a CAR-T need to work?
It depends on the CAR, the target and the response you measure, so it has to be tested for each construct. Published work on CD19 and CD22 CAR-Ts shows that killing can be triggered at lower antigen densities than cytokine release, and that tumours with reduced target density can escape. A density series with an antigen-negative control is the direct way to find the threshold for your own construct.
What is the difference between antigen density and antigen expression?
Antigen expression usually means whether a gene is translated to a protein at all. This is often measured by flow cytometry. Surface receptor antigen density is the quantified amount on the cell surface, in molecules per cell. Two cell lines can both be ‘positive’ for a target and differ in density a hundredfold, which is why the density number matters.
Can I use a Tet-On system instead of a ready-made inducible cell line?
Yes, doxycycline-inducible systems are a common way to titrate expression and work well when you build and characterise the line yourself. The trade-offs are the time to generate and validate a clone, leakiness in the off state, and the effort of running two independent switches for dual-antigen work. Ready-made inducible lines remove that setup and come with characterised induction behaviour.
What is the therapeutic window for a T cell engager?
For a T cell engager, the therapeutic window is the range of target antigen densities where the engager activates T cells against tumour cells but not against healthy cells that express the same target at lower levels. It’s assessed in vitro by finding the activation threshold on a density series and comparing it with the measured density on healthy cells.
How do I model antigen escape in vitro?
Reduce the target on your cells in steps and find the density at which your therapy stops working, then run mixed cultures of antigen-positive and antigen-negative cells and watch whether the negative cells grow out. For dual-targeting designs, remove one antigen at a time to confirm the therapy still acts on the other. Inducible cell lines let you do all of this on one genetic background.