Gold nanoparticles that cluster in response to T cell enzymes can predict cancer immunotherapy success days before tumors begin to shrink.
(Nanowerk Spotlight) When engineered immune cells attack a tumor, they leave behind a chemical signature: an enzyme called granzyme B. Detecting that signature inside a living patient without surgery has been essentially impossible. Clinicians must instead rely on indirect measures such as blood tests, biopsies, and imaging scans that reveal whether a tumor is shrinking but not whether the immune system is actively destroying it. A tumor might remain stable for weeks while T cells work unseen within, or it might appear unchanged because the therapy has failed entirely. Without a way to distinguish these scenarios early, doctors and patients are left waiting.
Adoptive T cell therapy exploits the body’s own defenses against cancer. Physicians extract a patient’s T lymphocytes, expand or engineer them to recognize tumor cells, and reinfuse them. Once inside the body, these cytotoxic T cells seek out cancer cells and kill them by releasing granzyme B and other destructive molecules. The approach has transformed treatment of certain blood cancers and is now being tested against solid tumors.
Monitoring its success, however, remains imprecise. Blood biomarkers may not reflect what is happening inside the tumor. Biopsies are invasive, prone to sampling errors, and impractical to repeat. PET and CT scans require ionizing radiation that limits scanning frequency. MRI is expensive and slow. None of these methods detect T cell killing activity directly.
Photoacoustic imaging offers a different approach. The technique fires pulsed laser light into tissue. Molecules that absorb this light heat up and expand, generating sound waves that ultrasound transducers can detect. The result combines optical sensitivity with acoustic depth and resolution, without radiation.
A study published in Advanced Science (“Protease‐Activated Plasmonic Nanosensors for Predictive Ultrasound‐Guided Photoacoustic Imaging of Tumor Responses to Adoptive T Cell Therapy”) describes a nanosensor designed to make T cell killing visible through this imaging method. Researchers at Georgia Institute of Technology and Emory University School of Medicine created gold nanoparticles coated with short peptide chains that granzyme B specifically recognizes and cuts. When cytotoxic T cells infiltrate a tumor and attack cancer cells, they release granzyme B. This enzyme cleaves the peptide coating, causing the nanoparticles to cluster together.
That clustering transforms how the particles interact with light. Individual 15 nm gold spheres absorb visible light around 520 nm because electrons on their surfaces oscillate collectively when struck by photons. When particles aggregate after peptide cleavage, their electronic fields couple, shifting absorption into the near-infrared range around 700 nm.
This shift matters because biological tissue is relatively transparent to near-infrared light, allowing deeper imaging with less background noise. The clustered nanosensors convert approximately 90% of absorbed laser energy into acoustic signals, creating an efficient “off-to-on” switch. Dispersed particles produce minimal signal at 700 nm. Aggregated particles produce strong signal.
Schematic overview of the experimental approach used in this study. a) Granzyme B (GzmB)-activated plasmonic nanosensors are composed of a gold nanosphere (GNS) core functionalized with peptides containing a GzmB-specific cleavage sequence. Upon exposure to GzmB, these peptides are cleaved, inducing nanosensor aggregation. This aggregation enhances optical absorption and PA signals within the NIR optical window due to interparticle plasmon coupling between proximate nanosensors in the aggregate, enabling specific detection of GzmB activity via US/PA imaging within the NIR optical window. b) The GzmB-mediated amplification of PA signals allows the nanosensors to noninvasively report the antitumor activity of tumor-infiltrating T lymphocytes following ACT, facilitating early assessment of therapeutic response prior to observable changes in tumor burden. (Image: Reproduced from DOI:10.1002/advs.202515111, CC BY) (click on image to enlarge)
The research team optimized particle design by varying the ratio of peptides to gold spheres. Lower peptide densities, around 0.43 per square nanometer of particle surface, yielded stronger aggregation responses. This likely occurred because granzyme B could access its target more easily without molecular crowding.
The optimized sensors detected granzyme B at concentrations as low as 1.56 nM optically and 3.13 nM via photoacoustic imaging. The slightly higher detection threshold for photoacoustic measurements likely reflects variations in acoustic coupling efficiency and pulse-to-pulse laser fluctuations inherent to the imaging technique.
Specificity proved equally important. Tumor environments contain many proteases, enzymes that cut proteins. Matrix metalloproteinases MMP-7 and MMP-9 associate with tumor progression. Caspase-3 plays a role in cell death. The nanosensors remained stable when exposed to these enzymes, generating signals roughly seven times weaker than their response to granzyme B.
Cell culture experiments confirmed the sensors could detect physiologically relevant enzyme activity. T cells stimulated with activating antibodies secreted granzyme B into surrounding medium. Nanosensors exposed to this medium produced significantly amplified signals compared to medium from unstimulated cells.
A more stringent test used T cells engineered to recognize ovalbumin, a specific antigen. When these T cells encountered tumor cells expressing ovalbumin, they released granzyme B and killed their targets. When they encountered tumor cells lacking the antigen, they did not. Nanosensors distinguished these scenarios clearly, responding only when active killing occurred.
The critical validation came in living mice. Animals bearing subcutaneous tumors received adoptive T cell therapy, followed by intravenous nanosensor injection one day later.
By day two after therapy, tumors expressing the target antigen showed significantly elevated photoacoustic signals compared to antigen-negative controls. This difference appeared before any measurable change in tumor size. Conventional volume measurements did not distinguish responding from non-responding tumors until day four, two days later.
Flow cytometry confirmed that antigen-positive tumors harbored more granzyme B-expressing T cells, validating what the imaging revealed. Statistical analysis yielded a receiver operating characteristic area under the curve of 0.93, where 1.0 represents perfect diagnostic accuracy. The nanosensors distinguished responders from non-responders with high sensitivity and specificity.
Early photoacoustic signals also predicted later outcomes. Tumors generating stronger nanosensor signals by day two were more likely to shrink subsequently, with a correlation coefficient of −0.79, indicating a strong predictive relationship. The sensors did not merely detect current immune activity. They forecast therapeutic success.
The nanosensors showed no obvious toxicity. Mice maintained normal body weight over 15 days after injection. Signals concentrated at tumor edges, reflecting both higher vascular density at the periphery and the physical barriers that limit nanoparticle penetration into tumor cores.
This platform offers something fundamentally different from existing monitoring approaches: a direct readout of immune function rather than tumor anatomy. By detecting the molecular evidence of T cell killing, clinicians could potentially identify responding patients days earlier than current methods allow. Earlier identification might enable faster adjustments to failing treatments or provide reassurance that effective therapies are working.
The peptide coating is modular. Swapping in different sequences could adapt the sensors to detect other disease-associated proteases. Translation to human patients will require answers to questions about dosing, safety, and performance in deeper or more complex tumors. One technical challenge involves spectral unmixing: variability in how nanosensors aggregate inside the body could unpredictably alter their signal spectrum, potentially complicating efforts to distinguish sensor signals from background tissue such as hemoglobin.
For authors and communications departmentsclick to open
Lay summary
Prefilled posts
ORCID information
Stanislav Y. Emelianov (Georgia Institute of Technology)
, 0000-0002-7098-133X corresponding author
Nanowerk Newsletter
Get our Nanotechnology Spotlight updates to your inbox!
Thank you!
You have successfully joined our subscriber list.
Become a Spotlight guest author! Join our large and growing group of guest contributors. Have you just published a scientific paper or have other exciting developments to share with the nanotechnology community? Here is how to publish on nanowerk.com.
