Dimensions of drug delivery: Why architecture matters in nanomedicine

A chemotherapy drug first synthesized in 1957 becomes 22,000 times as potent when restructured as a spherical nucleic acid (SNA). The same drug, reorganized as a linear polymer, works via a completely different mechanism. And a single strand of RNA, folded into an origami scaffold, reprograms tumor-protecting immune cells into tumor killers by being in the right shape.

These findings demonstrate a new principle that researchers are uncovering: at the nanoscale, how you build a therapeutic is just as important as what you build it from.

For decades, nanomedicine has treated structure as a secondary consideration—an engineering detail to optimize once you’ve chosen the right cargo. And the vaccines made from messenger RNA (mRNA) and lipid nanoparticles vindicated that logic: no two particles in a batch were identical, and they still saved millions of lives.

But recent results suggest that when researchers stop treating nanoparticles as just cargo containers and start treating them as architecture, the drugs they already have can be transformed.

“I’m not here to tell you small molecules are bad,” says Chad Mirkin, a professor and chemist at Northwestern University and director of the International Institute for Nanotechnology (IIN), who invented SNAs in 1996. “But being able to take different components—small molecules and biologics—and put them into larger nanostructures where architectural control allows you to modulate therapeutic efficacy and toxicity is an incredibly powerful modality.”

Geometry becomes pharmacology

In a 2019 study, Mirkin’s group built three cancer vaccines from identical components—same adjuvant, same antigen, same dose—but arranged differently. Each started with the same spherical liposome shell bristling with cytosine-phosphate-guanine (CpG) adjuvant strands. These short DNA sequences, common in bacterial genomes, are recognized by Toll-like receptor 9 (TLR9) as an immune danger signal.

In the encapsulated version, the antigen was sealed loose inside the hollow core. In the anchored version, the antigen was chemically tethered to separate oligonucleotide strands plugged directly into the liposome’s surface. And in the hybridized version, the antigen was conjugated to a complementary strand that clasped onto the outward-facing CpG, linking antigen and adjuvant into a single double-stranded unit at the particle’s surface. A fourth group received the same components as an unstructured mixture.

One configuration was completely curative in preclinical cancer models. The unstructured mixture had no effect at all.

Why? The answer lies in timing. In the body, a dendritic cell needs to do two things at once to launch an immune response: display the antigen on its surface and express costimulatory molecules—the secondary signals that confirm a threat is real. It’s a two-key launch system. And both keys must turn at the same moment.

In the case of the hybridized SNA, where the antigen was tethered to the particle’s surface, the nanoparticle lingered in dendritic cell endosomes for 24 h—long enough for both keys to align. In the encapsulated version, the antigen washed out within 2 h.

“It’s not just adjuvant-antigen combination that matters,” Mirkin says. “It’s how you present these and how you connect them within a larger structure that dictates their efficacy.”

A study from Mirkin’s laboratory published in February in Science Advances pushed the principle further: in a human papillomavirus (HPV) cancer vaccine built on the same SNA platform, simply flipping which end of the peptide faced outward produced eightfold as strong killer-T-cell responses. Neither the peptide, adjuvant, nor dose changed.

But if architecture can transform modern biomolecules such as peptide vaccines, what can it do for the oldest drugs in oncology? 5-Fluorouracil (5-FU) has been a chemotherapy workhorse since 1957, killing cancer cells by inhibiting thymidylate synthase. It is also spectacularly wasteful: less than 5% of an administered dose is metabolized into its active form, 5-fluoro-2′-deoxyuridine monophosphate (FdUMP). The other 95% wanders around the body, damaging everything except the tumor.

In a recent ACS Nano study, Mirkin’s group incorporated FdUMP into 10-unit oligonucleotide chains and anchored roughly 150 of them to a lipid nanoparticle core—around 1,500 drug molecules restructured as an SNA. The SNA architecture made the drug selective—almost by accident of biology.

In acute myeloid leukemia, the bone marrow floods the body with immature myeloblasts. On the surface of those myeloid cells, class A scavenger receptors recognize polyanionic macromolecules such as oligonucleotides, and SNAs, with their dense shell of oligonucleotides, carry a higher density of negative charge than the same sequences delivered as free linear strands. Consequently, the myeloid cells absorb the FdUMP-loaded SNA readily; nonmyeloid cells largely ignore it.

“You can take this chemotherapeutic, which is horribly insoluble and toxic, incorporate it into the SNA, and it becomes water soluble,” Mirkin says. “It gets into myeloid cells better than anything else—and you’ve pre-phosphorylated, so it’s ready to go.”

How as well as what: Molecular structure matters

Getting a drug into the right cell is not the only way architecture can modify a medicine’s mechanism. At the Wyss Institute for Biologically Inspired Engineering at Harvard University, bioengineer Natalie Artzi has spent 2 decades showing that a nanomaterial’s therapeutic success depends not just on what it carries, but on how that cargo is attached—and what happens after the first cell swallows it.

Her group’s work on stimulator of interferon genes (STING) agonists illustrates the point. STING is an innate immune sensor—when activated, it triggers interferon signaling that recruits the immune system to attack. But free STING agonists wash out of cells too quickly to do much.

To get the molecules to stick around longer, Artzi’s team conjugated a STING agonist to a nanoparticle using a cathepsin-cleavable linker—a molecular latch that opens only inside a cell’s cytoplasm. Tethered to the particle, the drug is inert. Cleaved free, it triggers the immune cascade. Compared with the same agonist held by electrostatic attraction alone, the conjugated architecture increased type I interferon signaling while reducing the systemic inflammation and cytokine spillover that makes immunotherapy dangerous.

When treatment schedules were aligned with checkpoint blockade (immunotherapy that releases the brakes on T-cell attack) and radiotherapy, the synergy was striking. But the bigger surprise came from watching where the particles actually went.

In a solid tumor, cancer cells vastly outnumber immune cells—and to start with those cancer cells swallowed most of the dose. For a STING agonist, that’s the wrong address. But then Artzi’s team saw something no one had designed for: the cancer cells were spitting the particles back out, intact, and neighboring immune cells were picking them up.

To prove it, her group incubated cancer cells with the STING nanoparticles, washed away any that hadn’t been internalized by the cells, and exposed the preloaded cancer cells to fresh immune cells—no free particles in solution. The only way for those immune cells to activate was if the nanoparticles had traveled between cells. The immune cells activated.

Artzi calls this process the paracrine transfer effect (PTE). “PTE describes the phenomenon whereby engineered nanoparticles are not confined to the initial waypoint cell but are transferred from one cell to neighboring cells, propagating biological effects across a cellular network,” she says. “This intercellular relay amplifies or reshapes therapeutic responses beyond primary uptake, introducing a systems-level dimension to nanomedicine design.”

If a nanoparticle is architecturally robust enough to survive transit through the wrong cell, the wrong cell becomes a relay station.

The paracrine transfer effect that Artzi’s team discovered means that cancer cells could serve as intermediaries, absorbing the nanoparticles and releasing them to neighboring immune cells, where they activate the STING pathway. In a follow-up study published in 2025 in the Proceedings of the National Academy of Sciences of the United States of America, Artzi’s team designed for that intermediation from the start. Instead of delivering a finished STING agonist to cancer cells and then to immune cells, the researchers used lipid nanoparticles to deliver mRNA encoding cGAS—the enzyme that produces cGAMP, the body’s endogenous STING activator. The tumor cells then became factories, secreting the immune trigger to neighboring immune cells through the tumor cells’ own export machinery.

Combined with checkpoint blockade, the strategy enhanced antitumor immunity in mice while avoiding the inflammatory spillover that plagues free STING agonists.

In a 2025 joint perspective in ACS Nano, Mirkin and Artzi tested whether liposomal SNAs exhibit the same paracrine transfer effect as Artzi’s polymeric STING nanoparticles. They do—but differently. Endosomal enzymes broke down the liposome core during processing, and a stripped-down oligonucleotide-lipid conjugate exited, not an intact particle. The stripped-down conjugate could still enter new cells and stimulate TLR9, but less potently, proving you need the right architecture to exploit the relay.

From spheres to polymers

Mirkin and Artzi have shown that nanoarchitecture can make drugs more potent and more selective. But at Wake Forest University School of Medicine, cancer biologist William Gmeiner had been asking a different question for 2 decades: Could restructuring a molecule give it an entirely new way to kill cancer?

The story began with F10, a first-generation polymer that Gmeiner built from repeating units of FdUMP—the same active metabolite that 5-FU so rarely produces. By stringing 10 of these units into a single oligonucleotide chain, Gmeiner bypassed 5-FU’s inefficient metabolism entirely.

In 2010, when the National Cancer Institute profiled F10 against its panel of 60 cancer cell lines, its activity pattern didn’t match 5-FU’s at all. It looked more like that of camptothecin, the parent compound of topoisomerase 1 poisons such as irinotecan. The polymer had picked up a new mechanism.

Gmeiner’s second-generation version, CF10, added a cytarabine cap to resist enzymatic degradation and sharpened the effect.

“CF10 had picked up a new activity,” Gmeiner says. “But it wasn’t identical to another class of drugs either. It was something that had elements of multiple classes.”

In fact, CF10 inhibits thymidylate synthase, starving the cell of the thymidine it needs to build and repair DNA. That starvation forces the cell to incorporate more of the drug into its genome during replication. Once embedded in the DNA, CF10 triggers topoisomerase 1 poisoning, causing topoisomerase 1 to get stuck on the DNA strand, creating lethal double-strand breaks. The cell recognizes the damage and activates its repair machinery, but repair requires thymidine, which CF10 has already depleted. The fix is fatal. Gmeiner calls it “self-potentiating”: each step the cell takes to survive makes the next step more lethal.

In a 2025 study reported in Cancers, colorectal cancer cells that had evolved resistance to 5-FU remained fully sensitive to CF10.

Earlier this year, Gmeiner’s group reported what happened when they packaged CF10 inside lipid nanoparticles. It was unclear whether the dual mechanism—which depends on the polymer being processed in a specific sequence—would survive encapsulation and lysosomal release. But not only did the nanoparticles survive with their mechanism intact, but KRAS-driven cancer cells, which scavenge extracellular nutrients through macropinocytosis, engulfed the nanoparticles readily. Gmeiner calls CF10 in this form “a poisonous food.”

But the transformation isn’t universal. “It’s not a magic trick that you can do with any nucleoside analog,” Gmeiner says. Other analogs, restructured the same way, become more potent but don’t gain new mechanisms. “I think we’re in the infancy of this,” Gmeiner says “I’ve had ideas for decades. We just haven’t had the resources to move them all forward.”

Structural control through DNA origami

When thinking about forms of nanoarchitecture, one type that naturally comes to mind is DNA origami—the technique of folding a single viral genome into arbitrary 3D shapes using hundreds of short “staple” strands. DNA origami gives the nanoarchitecture field something no other platform has: nanometer precision.

Why would a few nanometers matter?

Take the work done in William Shih’s lab at Harvard, where physician-scientist Yang Claire Zeng used DNA origami to develop a vaccine platform, DoriVac, that she later spun out as DoriNano. In a study published in 2024 in Nature Nanotechnology, she arranged tumor antigens and CpG adjuvant molecules on an origami scaffold at defined spacings. She found that the distance between the adjuvant molecules determined the strength of the immune response.

Those CpG motifs are recognized by TLR9, which operates as a dimer; TLR9 can bind two CpG molecules as long as they are positioned at a precise spacing. The origami scaffold revealed something that simply mixing components in solution could never have found, because the geometry of receptor engagement was invisible without nanoscale control.

But origami isn’t limited to DNA. Hao Yan—a biomolecular engineer at Arizona State University who also pioneered several methods to design complex 2D and 3D DNA origami nanostructures—has been lately extending structural control to RNA.

Instead of folding a scaffold with hundreds of staple strands, his platform encodes the entire nanostructure in a single RNA molecule that folds through intramolecular base-pairing. One strand, produced enzymatically and self-assembled in buffer, has a property that DNA origami can lack: intrinsic immunogenicity. The double-stranded regions that form when the RNA strand folds back on itself are recognized by Toll-like receptor 3, a pattern-recognition receptor that evolved to detect viral RNA. The scaffold activates innate immunity on contact, functioning as its own adjuvant.

“DNA origami has the benefit of being able to precisely control the number of ligands you put on the structure and the spatial arrangement of those ligands,” Yan says. “But RNA origami adds a layer of biological activity that turns the structural scaffold into a therapeutic agent in its own right.”

How far can this principle be pushed? In a study published in the Journal of the American Chemical Society Au in October, Yan’s group folded a single strand of RNA into an origami structure and injected it into mice with pancreatic tumors—one of the deadliest and most immunologically hostile cancers. The RNA origami slipped into the tumor’s macrophages and reawakened them to recruit natural killer cells to attack the cancer. When administered alongside 5-FU—the same molecule Mirkin rebuilt as a sphere and Gmeiner as a line—the combination produced complete tumor regression in two of five mice.

The next era of drug design

What connects these researchers is the conviction that considering the architecture of new drugs could make medicines even more powerful. For example, an SNA vaccine has at least 11 independently tunable structural features. In 2019, Mirkin’s group synthesized 960 variants and fed the results to machine learning models that predicted immune activation from architecture alone. The researchers found that the anchoring of the oligonucleotide to the lipid core dominated all other variables.

But machine learning can optimize only within the parameters someone thought to vary. “It was very clear to me from early on,” Artzi says, “that when you take materials and they go into the body, the context matters. Hence, we design tissue-responsive materials that adapt to tissue state to maximize performance under pathological conditions.”

She has since carried that principle to the clinic. SpideRx Biotechnologies, a startup Artzi launched from the Massachusetts Institute of Technology and Brigham and Women’s Hospital, is developing a hydrogel engineered for one of the most unforgiving circumstances in oncology: a glioblastoma surgical cavity. The hydrogel adheres to the resection site, capturing residual tumor cells in a space where nearly every patient experiences relapse.

Meanwhile, seven SNA-based therapeutics have entered human clinical trials. The furthest along pairs Mirkin’s TLR9-stimulating architecture with checkpoint immunotherapy; in a small Phase 1b/2 trial in Merkel cell carcinoma, some patients achieved complete responses. Zeng’s DoriNano was selected for Memorial Sloan Kettering’s Therapeutics Accelerator, which is providing regulatory guidance and clinical trial design support for DoriVac-101, DoriNano’s vaccine for head and neck cancer, as the company prepares for a first-in-human trial.

But DoriVac may not remain a cancer-only platform for long. In a recent study reported in Nature Biomedical Engineering, Zeng’s group conjugated conserved viral peptides from SARS-CoV-2, HIV, and Ebola onto the same origami scaffold—keeping the CpG adjuvant locked at the same 3.5 nm spacing. The vaccine generated strong antibody and T-cell responses in mice, significantly outperforming the same components delivered as a simple mixture.

And when the scaffold carried the full-length SARS-CoV-2 spike protein and was compared head to head against the Moderna and Pfizer mRNA vaccines in mice, DoriVac matched their immune activation while requiring only standard refrigeration instead of deep-freeze storage.

It’s a long way from where nanomedicine started.

In 1995, the US Food and Drug Administration approved what could be considered the first nanomedicine, Doxil, which consists of doxorubicin wrapped in a liposomal shell. The packaging reduced cardiotoxicity, but in clinical trials, Doxil was no more effective than the free doxorubicin. The reason, in retrospect, is clarifying: the liposome had no instructions. It relied on the passive assumption that leaky tumor vasculature would do the targeting, a phenomenon that proved far more reliable in mice than in people.

Thirty years and millions of research dollars later, researchers are building structures whose geometry dictates where they go, when they release, which cell they recruit, and how to survive transit through the wrong one. They are encoding biological logic into geometry and structure. The molecule hasn’t changed, but the architecture has.

Down at the nanoscale, the difference between a drug that kills a tumor and one that kills a patient can come down to a few angstroms—not because of what the molecule is made of, but because of how every atom of it is arranged in space.

David Brzostowicki is a freelance science writer who covers medicine, science, and emerging technologies.