Two experimental antibody shuttles reached the same average concentration in mouse brain one day after injection. By day seven, one measured 1.8 nM and the other 2.7 nM. A team choosing between them on day one would have missed a difference that mattered six days later.
That finding from a University of Michigan brain-shuttle study came to mind when I read Roche’s November 16 programme at the Clinical Trials on Alzheimer’s Disease meeting. Roche featured two kinds of PET imaging for its investigational Alzheimer’s antibody, trontinemab. In one, investigators followed a radioactive label attached to the antibody. In the other, a separate tracer showed how amyloid plaque burden changed.
For a team developing an Alzheimer’s antibody, those measurements answer different questions. Does the molecule recognize the intended form of amyloid? Does its shuttle bind the receptor in the species being tested? And how much antibody reaches brain tissue long enough to act?
Would your animal model recognize the shuttle?
Trontinemab uses transferrin receptor 1 as its route across the blood–brain barrier. The Michigan group studied a different shuttle target, CD98hc, and encountered a problem worth catching early. Their starting antibody recognized human and cynomolgus monkey CD98hc. Binding to the mouse version was undetectable.
In their peer-reviewed study in Nature Biomedical Engineering, Wen-Hao Wu, Neelan Sivaneri and colleagues used Nicoya Alto to compare four antibodies against CD98hc from all three species. Surface plasmon resonance (SPR) let them follow each interaction as it formed and dissociated. They identified variants that recognized all three versions of the receptor and then evaluated brain exposure in separate animal experiments.
That’s a comparison I’d want before interpreting a disappointing mouse study. If you’ve only tested the human receptor, low brain uptake leaves two explanations open. The shuttle may have bound the mouse receptor and travelled poorly, or it may barely have recognized the receptor in the first place.
What I’d put on the binding panel before imaging
For a plaque-directed programme, I’d compare the parent antibody and each shuttle-equipped candidate against the same preparation of fibrillar amyloid. I’d also run each candidate against human TfR1 and TfR1 from the study species. With a parent and three new designs, that’s already 12 antibody–target pairs before references and repeats.
The amyloid preparation deserves as much attention as the antibody. In a Nature Communications study of p3, a shorter amyloid peptide, Tian and colleagues used an Applied Photophysics Chirascan to follow the change from a largely disordered peptide to β-sheet-rich fibrils. Electron microscopy showed the assemblies that had formed. I’d record the aggregation state and split one preparation across the antibody panel. Otherwise, a difference between curves could come from what was in the amyloid tube.
For the binding assay itself, I’d borrow a control from Carstens and colleagues’ antibody study. They used Alto to test binding to human claudin-18.2 across five concentrations, with a blank and an unrelated antibody as controls. That was an oncology study, but the control earns its place in an amyloid assay too. If the amyloid preparation also sticks to an unrelated antibody, I’d investigate that signal before using the curve to rank a candidate.
This is where the 16-channel Alto makes a practical difference. It prepares serial dilutions on the cartridge, and Nicoya specifies a minimum sample volume of 2 µL. For a lab running smaller panels, the four- or eight-channel Revo offers automated serial dilutions with the same minimum volume. When the first purification yields less antibody than expected, I’d use that flexibility to keep the species comparison and repeat the borderline result. Those are the measurements I’d most regret cutting later.
What day seven revealed
In the Michigan study, the researchers used radiolabelled antibodies to measure brain concentrations after a single dose in wild-type mice. Two shuttles in Figure 8 illustrate why sampling time belongs in any discussion of brain delivery.
| CD98hc shuttle | Day 1 brain concentration | Day 7 brain concentration |
|---|---|---|
| IgG/D4-4 | 3.1 ± 0.2 nM | 1.8 ± 0.1 nM |
| IgG/3.5D19 | 3.1 ± 0.2 nM | 2.7 ± 0.1 nM |
Mean ± standard deviation, with three to four mice per time point.
The binding experiment established that both candidates could engage the mouse receptor. The tissue measurements showed how their exposure changed over time. To answer the next question for a plaque-directed antibody, I’d also want to see where it localized relative to blood vessels and plaques. A concentration averaged across a piece of brain leaves that detail unresolved.
What changed in the plaques?
In its phase Ib/IIa CTAD presentation, Roche reported a decline of about 70 Centiloids on amyloid PET after six trontinemab doses of 1.2 mg/kg given four weeks apart. Centiloids express plaque burden on a standardized scale. For patients, the next question is whether treatment helps preserve memory and daily function. That is what the phase III TRONTIER programme is designed to assess.
If I were reviewing an underperforming candidate, I’d want the next experiment to change a specific decision. If amyloid recognition weakens when the shuttle is added, revisit the antibody format. If the shuttle misses the study species’ receptor, revisit the model. If both binding results hold up, measure tissue exposure over time before spending another design cycle chasing tighter affinity.
Desmond Gilmour
Email: dgilmour@nicoyalife.com
Phone: +1 519-546-9876
I work at Nicoya Lifesciences. The views expressed here are my own and do not necessarily reflect those of Nicoya. This is a personal publication, not an official company communication.
