Feature

HOW CUSTOMIZED
RNA VACCINES
MIGHT HALT CANCER
The same technology that quickly
brought the world a vaccine for COVID-19
is now showing promise as a bespoke
therapy for cancer. By Elie Dolgin

A
ngela Evatt lay face down under of the vaccine on Evatt’s recovery is difficult.
anaesthesia as surgeons removed “It’s impossible to know,” she says. “I’m just
a malignant mole from her back and happy to be cancer-free.” However, the trial
a lymph node from her left armpit. that Evatt participated in is yielding promising
The purpose of the operation was data. According to the latest number-crunch
not only to excise the cancerous tis- from the 157-person study, the combination
sue from her body, but also to begin of vaccine and checkpoint inhibitor reduces
the process of crafting a personal- the risk of disease recurrence by nearly 50%
ized vaccine that would train Evatt’s immune compared with treatment with the inhibitor
system to attack any tumour cells left behind. alone. The latest analysis also indicates that
The vaccine uses messenger RNA (mRNA), the vaccine contributes to lifespan extension.
carefully constructed to encode the unique “At the end of the day, you realize, ‘Damn!
mutant proteins, known as neoantigens, This combination seems to have activity,’”
that are found on the surface of Evatt’s mel- says lead trial investigator Jeffrey Weber, a
anoma skin cancer cells. She first received cancer immunotherapy researcher at New
this bespoke vaccine, alongside a potent York University Langone Health in New York
immune-stimulating drug known as a check- City, who presented the findings on 3 June at
point inhibitor, as part of a clinical trial in the world’s largest annual meeting of cancer
March 2020, just months before mRNA vac- biologists and oncology specialists in Chicago,
cines would become household names in the Illinois. (Weber and his colleagues published a After decades of vaccine trial setbacks, she says, ADAM GLANZMAN/BLOOMBERG/GETTY

fight against COVID-19. previous analysis of the data at the beginning “we’ve started to see the pendulum swing”.
Every three weeks, Evatt travelled from her of this year1.) Success is far from guaranteed, however,
home in Maryland to Georgetown Universi- A larger-scale study is still needed to con- and the field is thick with unresolved ques-
ty’s Lombardi Comprehensive Cancer Center firm these promising results and to support tions. Companies are trying to determine
in Washington DC to get an injection in each bringing the vaccine to market. A trial involv- which stages of cancer will see the most benefit
arm. The mRNAs enter her healthy cells and ing more than 1,000 people with melanoma from such treatments. They are also searching
then produce the neoantigens that educate kicked off last July; another for nearly 900 for improved ways to predict the most effec-
her immune system. people who have a type of lung cancer began tive neoantigens. And it is unclear whether
Despite Evatt experiencing severe flu-like a few months later. mRNA or some other vaccine technology is the
symptoms for a day or two after each injection But even as the cancer research community best way to stimulate an anticancer immune
— fever, achiness, chills — the treatment seems awaits further evidence, the early results have response.
to have been beneficial. Now in her mid-40s, injected fresh enthusiasm into the cancer vac- As scientists continue to test the treatments,
she has remained in remission for more than cine field. “It has had a big impact across all vac- all of the pieces will need to come together.
three years after completing her treatments. cine development,” says tumour immunologist “We have the first proof of concept that
As is typical of individual experiences in Nora Disis, director of the Cancer Vaccine Insti- these things can work,” says Cristina Puig-
clinical trials, determining the precise impact tute at the University of Washington in Seattle. Saus, a cancer immunologist at the University

290 | Nature | Vol 630 | 13 June 2024

 Moderna employees build RNA vaccines tailored to an individual’s tumour at a facility in Norwood, Massachusetts.

of California, Los Angeles. “Now, we just need can work in parallel. Each bay houses its own response in a vaccine recipient. A series of
to make them better.” ‘single use personalized RNA+’ machine, a artificial-intelligence algorithms make this
refrigerator-sized unit that cranks out long call, informed by an ever-growing repository
Vaccines on demand strands of mRNA encoding up to 34 specific of clinical and laboratory data from other indi-
Moderna, the company behind the vaccine cancer mutations. The mutations correspond viduals that, according to Moderna’s oncology
that Evatt received, is one of the industry lead- to different neoantigens, neatly arranged in a lead, Kyle Holen, should yield better predic-
ers working on these improvements. Capitaliz- sequence. A mixing device then encapsulates tions over time. “It’s a therapy that learns and
ing on its experience and the financial windfall the mRNA in fatty nanoparticles to enhance its can continue to improve,” he says.
from its COVID-19 response, the company stability and cellular uptake. Moderna, in partnership with the drug firm
has refined its manufacturing protocols and “That’s the magic there,” says Elizabeth Sul- Merck, which is headquartered in Rahway, New
expanded capacity to produce personalized livan, head of operations for the personalized Jersey, is currently conducting mid- to late-
medicines around the clock. vaccine programme at Moderna, as she leads stage clinical trials of its vaccine across five
In a football-pitch-sized production facil- a tour of the company’s manufacturing plant kinds of cancer.
ity in Norwood, Massachusetts — a short in mid-April. In all of these trials, the companies adminis-
drive from Moderna’s headquarters in Cam- Less visible are the innovations that go into ter their experimental vaccine to people who,
bridge — dull black lines divvy up the grey selecting which of the many possible tumour like Evatt, have had their tumours surgically
floors into 15 bays, where pairs of technicians mutations are most likely to elicit an immune removed but still face a high risk of cancer

Nature | Vol 630 | 13 June 2024 | 291

Feature
recurrence. By training the immune system’s In place of mRNA, many companies and aca- vaccine for treating pancreatic tumours, for
T cells to identify and eliminate cancer cells at demic groups rely on DNA, peptides or geneti- example, only half of the trial participants
this stage, the objective is to avert a relapse — cally engineered viruses. As Niranjan Sardesai, developed T cells that were directed against
an approach known as adjuvant therapy. founder and chief executive of Geneos Thera- any of the vaccine-encoded mutations. And
Moderna executives have also raised the peutics in Philadelphia, Pennsylvania, points among those who did, around half formed
possibility of the vaccine being used against out, each approach triggers its own type of T cells against just a single neoantigen, even
later-stage disease, when the cancer has immune response, which could affect the though the vaccines generally contained the
spread to distant sites throughout the body, success of any vaccine candidate. “How you instructions for making ten or more targets2.
a process known as metastasis. But, so far, deliver these antigens is just as crucial as which “Once in a while, the stars line up,” says
the field has had limited success in this set- antigens you deliver,” Sardesai says. Alex Rubinsteyn, a computational biologist
ting. Although initial trials have often found at the University of North Carolina–Chapel
that personalized vaccines produce antican- Hill who designs neoantigen prediction tools
cer T cells in individuals with these types of for personalized cancer vaccines. But, he says,
advanced cancer, those immune responses instances in which several of the chosen anti-

HOW YOU DELIVER

rarely lead to tumour regression or long-term gens elicit anti-tumour activity could be the
survival benefits. exception rather than the rule. Benjamin Vin-

THESE ANTIGENS IS
“It’s very hard to eradicate established cent, a tumour immunologist, also at the Uni-
tumours,” says Gal Cafri, a cancer immunol- versity of North Carolina–Chapel Hill, agrees:

JUST AS CRUCIAL AS
ogist at the Sheba Medical Center in Ramat “The field really, really wants to just say, ‘We
Gan, Israel. The types of T-cell response that can predict this from the genomics data only.’
cancer vaccines elicit are well equipped for
restraining the growth of small, residual WHICH ANTIGENS It really wants it so bad, everyone is just doing
it. But that doesn’t make it robust.”
tumours, which helps to prevent disease recur-
rence after surgery. However, these vaccines
are less effective against large, established
YOU DELIVER.” Addressing this issue head on, many
researchers are now supplementing their
computational tools with further experimen-
tumours, which have often evolved aggressive Some platforms, for example, excel at tal data. In February, for example, a joint team
tactics that involve shielding themselves from eliciting ‘killer’ T cells, which are thought at the La Jolla Institute for Immunology and the
immune attacks. to perform the bulk of tumour-cell destruc- University of California, San Diego described
Moreover, early-stage cancers tend be slow- tion. However, the real-world impact of such a platform that uses DNA sequencing and
er-growing than late-stage ones, which gives an immunological difference remains to be gene-expression analyses on tumours to first
drug developers the 1–4 months they need to seen, because only the vaccine developed by identify potential neoantigens, as many others
design, manufacture and deliver the person- Moderna and Merck has so far shown success do already. The process then goes a step fur-
alized vaccines to patients. Then, once the in a randomized trial. ther by searching in patient blood samples for
vaccine enters the body, more “time is needed A bigger differentiator, researchers say, T cells that actually recognize those antigens3.
to build up an immune response”, says Uğur could be the computational engines that help This kind of method is going to become
Şahin, co-founder and chief executive of the to determine the vaccine’s composition. Each more important and more prevalent, accord-
biotechnology company BioNTech in Mainz, engine has its own proprietary suite of tools ing to Stephen Schoenberger, a translational
Germany, which is developing a personalized that it uses to select which neoantigens to immunologist who co-led the study. “It verifies
cancer vaccine in collaboration with biotech target. rather than merely predicts which mutations
firm Genentech in South San Francisco, Cal- Most companies start with genetic sequenc- are neoantigens,” says Schoenberger, who
ifornia. ing of data from tumours and healthy tissues is also chief executive and chief scientific
According to Sahin and Ira Mellman, head of to reveal the mutations that cropped up officer of Lykeion, a company he co-founded
cancer immunology at Genentech, all of these during cancer development. T cells will not in La Jolla to develop personalized vaccines
considerations factored into the companies’ recognize all of these mutations, however, informed by this method.
joint decision to evaluate their bespoke mRNA so algorithms are used to prioritize a subset Sachet Shukla, a computational biologist
vaccine as a post-surgical treatment for peo- — Moderna uses up to 34, BioNTech up to 20 at the University of Texas MD Anderson Can-
ple with high-risk colorectal and pancreatic — that are predicted to have the most potent cer Center in Houston, is hopeful that the
cancers that are still localized and have not immune-stimulating effects. immune-stimulating potential of personalized
yet spread across the body. “When thinking Such predictions are made on the basis of cancer vaccines will improve as the research
about where is the best place to put a cancer various factors, such as the levels at which neo- community amasses more of this information.
vaccine that would give it the best chance of antigens are expressed on tumour surfaces “I think you’ll see another step jump in the
succeeding and establishing at least proof of and their anticipated binding to cellular recep- accuracy of these algorithms,” he says.
concept,” Mellman says, “all roads lead to adju- tors that aid in provoking a T-cell response. When that happens, he anticipates that can-
vant or early disease.” Machine-learning models then incorporate cer vaccines, long regarded as ineffectual, will
He even keeps a coffee mug on his desk to experimental data to improve the accuracy finally become a staple of oncology treatment:
commemorate when this strategic decision of these tools. “It’s an idea whose time has come.”
was made. It reads “2018 A.D.”, standing for The algorithms can miss their mark in pro-
‘Adjuvant Day’. voking a cancer-directed immune response, Elie Dolgin is a science journalist in
however. “Only a small percentage of the Somerville, Massachusetts.
Special delivery predicted neoantigens turn out to be immu-
Both the Moderna–Merck and BioNTech– nogenic,” notes Neeha Zaidi, an oncologist at
Genentech vaccines are formulated as mRNA. the Sidney Kimmel Comprehensive Cancer
But that is not the only way to encode neoan- Center at Johns Hopkins University in Balti-
1. Weber, J. S. et al. Lancet 403, 632–644 (2024).
tigens for processing and presentation to the more, Maryland. 2. Rojas, L. A. et al. Nature 618, 144–150 (2023).
immune system. In a small study2 of the BioNTech–Genentech 3. Miller, A. M. et al. Sci. Transl. Med. 16, eabj9905 (2024).

292 | Nature | Vol 630 | 13 June 2024