Science investigation casts doubt on genes hailed for warding off Alzheimer’s disease

Since 2023, an international research team known for finding rare genetic variants that shield against Alzheimer’s disease has dropped hints about its biggest prize yet. The scientists named it Macondo, in homage to the fictional Colombian town in One Hundred Years of Solitude, by Gabriel García Márquez, the master of magical realism who died with Alzheimer’s in 2014. Identified in the novelist’s sister, the mutation seemed to protect her into her 90s against the disease that claimed her famous brother and other family members. Its benefits were also said to be evident in a large kindred of Colombians who carry a genetic mutation causing them to develop dementia before age 50.

The Macondo findings, yet to be published, would mark the third discovery of an Alzheimer’s-protective variant in just 7 years by a married couple—ophthalmology researcher Joseph Arboleda-Velasquez of Harvard Medical School and Mass Eye and Ear and neuropsychologist Yakeel Quiroz of Boston University and Massachusetts General Hospital—and their collaborators. Both trained at the Neuroscience Group of Antioquia (GNA), the Colombian research group based in Medellín that unearthed the cohort of early-onset Alzheimer’s families. The two previous mutations they uncovered have been among the most discussed Alzheimer’s findings during the past decade. The research has also won Arboleda-Velasquez and Quiroz global prominence, professional accolades, patents, and tens of millions of dollars in research funding.

The winning streak began in 2019, when a front-page headline in The New York Times—“Why Didn’t She Get Alzheimer’s? The Answer Could Hold a Key to Fighting the Disease”—buoyed hopes for millions of patients and their loved ones. It described a Colombian woman, later identified as Aliria Rosa Piedrahita de Villegas, who carried the devastating Alzheimer’s-causing mutation seen in her large extended family. Known to the researchers as Aliria, she had escaped memory loss until her early 70s, the group reported in Nature Medicine. They posited that this was because she also carried two copies of a rare, possibly protective mutation named Christchurch, after the New Zealand city where it was first detected in a separate medical context.

Hundreds of media reports followed. Noted dementia researchers lavished the work with praise. “Very exciting,” said the University of Southern California’s Paul Aisen. An Alzheimer’s “fire … has been doused” by Christchurch, affirmed Columbia University’s Scott Small. “A rare example where the study of just one person could change the thinking of a whole research field,” gushed Fiona Carragher, chief for policy and research at the U.K.-based Alzheimer’s Society.

Aliria’s uplifting story became the engine of an emerging theory that Alzheimer’s, early- and late-onset forms alike, might someday be treated or prevented by drugs that mimic a hidden defense bestowed by a single gene. In 2023, the group announced the discovery, in the same Colombian kindred, of a second apparently protective mutation, known as Reelin-COLBOS. Then in a 2024 paper, the researchers claimed that although Aliria’s double dose of the Christchurch mutation was especially helpful, just one copy was enough to confer years of protection against genetically linked, early-onset Alzheimer’s.

“The paradigm shift is coming,” Arboleda-Velasquez wrote on LinkedIn early last year. “Are we ready to rethink Alzheimer’s therapy?”

But a Science investigation raises doubts about the basis of the new paradigm. Science reviewed more than 1000 documents including scholarly papers, patents, grant data, and communications between many of the key researchers, and interviewed more than two dozen independent experts in Alzheimer’s genetics and GNA insiders and collaborators. The investigation suggests the group embraced a dubious approach to identifying protective mutations, improperly changed patient data to tease provocative findings from Christchurch carriers, and published multiple papers containing suspect or copied images.

Forensic image analysts recruited by Science evaluated scientific figures in Alzheimer’s-related papers led by Arboleda-Velasquez, Quiroz, or close collaborators. They identified what they believe to be inappropriately copied or altered images in three of them. Those include lab studies probing Christchurch’s effects on Alzheimer’s-linked proteins and on cells that protect the brain against inflammation. Anomalies also turned up in the paper that first described Reelin-COLBOS.

“There’s clearly duplication of images” in ways that affect experimental outcomes, says Rob Howard, an Alzheimer’s expert at University College London (UCL) who reviewed findings by the forensic sleuths. Other leading neuroscientists have echoed that judgment.

Simple errors might explain duplicate images, and innocent digital artifacts sometimes resemble something more serious. But Arboleda-Velasquez, Quiroz, and others did not reply to requests from Science to supply original versions of the images that could clarify.

Independent researchers also say Quiroz and Arboleda-Velasquez ignored evidence challenging their claims and exaggerated the implications of their discoveries. At times Arboleda-Velasquez asserted that he had solved the disease. “How many cures do you think I found for Alzheimer’s? … I found two,” he boasted last year at a luncheon sponsored by a neurological disease advocacy group.

Concerns about the pending paper on the Macondo mutation have caused at least two collaborators to withdraw their names from the manuscript. After reanalyzing data from the Colombian cohort, one of them has also concluded that Reelin-COLBOS is at best protective only in combination with other, so-far-unspecified genes and believes the same will hold true for Christchurch. (Some Alzheimer’s researchers still think Christchurch alone can slow Alzheimer’s if two copies are present.)

Several longtime GNA employees or collaborators spoke on condition of anonymity for fear of losing jobs or access to data and tissues controlled by Arboleda-Velasquez and his allies. Some who have worked closely with him and Quiroz think the pair has steered the Colombian Alzheimer’s program, the life’s work of GNA’s revered late founder, neurologist Francisco Lopera, in questionable directions. Others worry the recent work might have skewed thinking and wasted resources devoted to Alzheimer’s genetics more generally. “It’s fostered this whole little cottage industry based on incredibly unstable foundations,” says Stanford University neurologist Michael Greicius, who studies genetic resilience to Alzheimer’s.

Scientists have gone to great lengths to study the Christchurch mutation in mice, and some have built entire Ph.D.s around it, Greicius says. Drug development efforts based on Christchurch are now advancing, including one led by a startup Arboleda-Velasquez co-founded, and another at Weill Cornell Medicine, where scientists recently announced funding for a human gene therapy trial.

Arboleda-Velasquez and Quiroz declined requests for interviews or to provide written replies to detailed questions. BU did not respond to a request for comment. Officials from Mass General Brigham, of which Mass Eye and Ear is a member, and Massachusetts General Hospital declined an interview request. Top officials from the two institutions said in a joint statement that the organizations conduct “robust and confidential” reviews about research concerns, and would not comment on matters that might involve patient privacy or confidential, unpublished research.

In a statement, GNA defended its work but said that if warranted, “clarifications or corrections” would be conveyed through official channels. GNA added that concerns described by Science about suspect images involve “a limited subset of publications and do not affect the authenticity of the original experimental data, the validity of the quantitative analyses, or the scientific conclusions reported in those studies.”

In 1984, while a neurology resident at the University of Antioquia, Lopera saw the first of thousands of mostly poor, rural patients with mysterious dementia that began in middle age. As Colombia suffered the ravages of drug trafficking and political violence in the decades that followed, Lopera and his research team risked their safety to locate those patients and determine the cause of their disease.

Medellín resident Lady Soto, a member of the affected family, says her grandmother, mother, aunts, and uncles participated in studies by Lopera and his colleagues, donating blood, time, and even their brains for autopsies. In return, they received groceries, diapers, and medicines when severely ill—and a measure of hope. “They’ve done so much to find a path toward a cure,” Soto says of the researchers.

In 1995, Lopera, with international collaborators, helped confirm the reason for the family’s condition: a mutation on the gene PSEN1, which is implicated in many cases of early-onset Alzheimer’s. The mutation promotes overproduction of beta amyloid, a protein that forms sticky plaques in the brain and is widely considered the linchpin of Alzheimer’s. Its abstruse official name is E280A; GNA dubbed it “paisa” after a nickname for the people of the Antioquia region of Colombia.

Lopera and his colleagues explored the genetics and natural history of the paisa form of Alzheimer’s, building sprawling genealogies of an extended family that would eventually include some 6000 members, one-fifth of them mutation carriers. Mild cognitive impairment—when obvious memory lapses appear but do not yet interfere with daily life—usually begins around age 44 in mutation carriers, and dementia, when memory loss becomes severe, starts about 5 years later, the team reported in 2011. Researchers at the time hoped families like this one, plagued by early disease, would yield genetic and biological clues to the far more common form of Alzheimer’s that occurs in older people. Lopera had discovered the largest early-onset genetic kindred in the world.

Drug developers saw a tempting opportunity to test treatments that might benefit the broader Alzheimer’s population. Paisa mutation carriers shared similar genetics, diet, socioeconomic status, and lifestyle—ideal homogeneity for trial participants. And because most suffered cognitive symptoms within a predictable age range, a medication’s possible benefits might be easier to recognize.

In 2013, Lopera’s team and scientists with the Banner Alzheimer’s Institute launched what they hoped would be a landmark clinical trial. About 250 people received either a placebo or crenezumab, an antibody drug targeting beta amyloid. Nine years later, the drug showed no benefits—a devastating setback for the paisa families. (Since 2021, larger trials have shown drugs in the same class can slow cognitive decline in people with late-onset Alzheimer’s, although their efficacy is minimal and their approvals by the U.S. Food and Drug Administration were controversial.)

But Aliria, whose exceptional resilience against dementia was discovered during the trial’s recruitment process, offered new grounds for hope. In 2022, at a party in Medellín to celebrate GNA’s 30th anniversary, Lopera raised a glass to a new project: Villa Aliria, a nearby 13-hectare research campus and hospital to be backed in part by pharma companies.

At the forefront of the high-profile work Aliria catalyzed were Lopera’s star protégés, Quiroz and Arboleda-Velasquez, Colombian-born researchers who emigrated to the United States in the early 2000s and have maintained strong ties with the group. Quiroz’s brain-imaging and biomarker studies, done in collaboration with top dementia scientists including Banner leader Eric Reiman, supported GNA’s clinical work in Colombia. After GNA clinicians flagged Aliria as an outlier, Lopera and Quiroz arranged her visit to Boston for advanced brain imaging. Arboleda-Velasquez led the probe into her genome to identify the putative cause of her 25 extra years of good health: two copies of Christchurch.

The mutation occurs on APOE3, the most common variant of the apolipoprotein E gene, whose different forms are associated with varying degrees of risk for Alzheimer’s. The protein it encodes, ApoE, transports vital fats to neurons. APOE also affects the accumulation and clearance of beta amyloid and the spread of another protein, tau, linked to the death of neurons in Alzheimer’s.

In 2023, neuroscientist Yadong Huang of the University of California (UC) San Francisco, with Quiroz, Arboleda-Velasquez, Lopera, Reiman, and others, reported that the mutation seems to counteract some negative effects of APOE4, the form of the gene known to increase Alzheimer’s risk and accelerate the spread of tau. They engineered mice and lab-grown human neurons to carry two copies of APOE4 modified to contain the Christchurch mutation. The animals and cells showed minimal tau buildup. The researchers proposed that Christchurch weakened the binding of ApoE to a type of cell-surface protein key to tau’s spread—and that the effect might be replicated by a future therapy.

That same year, Arboleda-Velasquez, Quiroz, Lopera, and their colleagues published the discovery of another genetic outlier from Colombia. A man with the paisa mutation began to have cognitive symptoms much later than expected—at age 67. They credited a single copy of a previously unknown mutation on a gene called RELN, which produces a protein, Reelin, that helps keep neurons healthy. The Reelin-COLBOS mutation, they hypothesized, acts on some of the same cell-surface proteins as Christchurch, and “may have a profound therapeutic impact” in Alzheimer’s. (The man’s sister also had one copy of Reelin-COLBOS and showed more modest resilience, not developing symptoms until age 58.)

With colleagues, Arboleda-Velasquez has raised more than $1.5 million for a startup, Epoch Biotech, to develop drugs based on the Christchurch and Reelin-COLBOS mutations. In a 2024 paper, he and collaborators described an  experimental antibody designed to mimic the effects of Christchurch by blocking the interaction of APoE with the critical cell-surface protein. The team reported that the antibody reduced tau damage in the eyes and brains of transgenic mice.

Last month at a major Alzheimer’s meeting, Epoch announced that the Christchurch-inspired antibody also reduced tau and amyloid in the cerebrospinal fluid of monkeys. And Arboleda-Velasquez described a potential therapy, still in rodent studies, that would deliver a truncated version of the Reelin protein into the brain via nasal injection.

For people at high risk of Alzheimer’s, Weill Cornell researchers are developing another approach: using a virus to deliver an APOE gene containing the Christchurch mutation into the brain. Last month, the team received an initial $3.2 million from the U.S. National Institutes of Health (NIH) to test the safety and possible benefits of the approach in a trial enrolling 15 people with two copies of the high-risk APOE4 gene. A small biotech company, Lexeo Therapeutics, is also working on a Christchurch-based gene therapy.

In recent years, Quiroz and Arboleda-Velasquez have become major rainmakers for GNA. Prior to the publication of the Aliria findings, the couple had raised nearly $11 million from NIH for Alzheimer’s-related projects, nearly all of them with GNA staff as collaborators. Following the media surge around the Christchurch and Reelin-COLBOS papers, the pair brought in almost $18 million from NIH and foundations.

Christchurch was a plausible candidate because APOE affects Alzheimer’s risk, Goate says. But with Aliria as the sole case, “it’s extremely difficult” to firmly link that mutation to her resilience, she notes. “I would have liked to have seen the systematic analysis to show me how many other genes were equally likely.”

The group’s claim that Reelin-COLBOS is protective sparked similar skepticism. “I have no idea how among … millions of variants that come up in any individual in whole genome sequencing, they honed in on [Reelin-COLBOS],” says Harvard geneticist Rudolph Tanzi. “The Alzheimer’s genetics world shares that concern and wonder.”

Arboleda-Velasquez and his colleagues flagged the two mutations via Genomizer, a free application developed in Germany that sorts through genomic data to find variants linked to physical or clinical characteristics. Such automated tools can point to potentially relevant rare mutations, but any findings need to be validated with other methods, says Cyril Pottier, a geneticist at Washington University in St. Louis (WashU). In a study last year, Pottier and his collaborators attempted to find protective variants in a U.S. man in his mid-70s who remained healthy despite a different early-onset Alzheimer’s mutation, but painstaking efforts to narrow down the possibilities still left 13, none of them Christchurch or Reelin-COLBOS.

If Aliria’s case couldn’t pin down Christchurch’s significance, heterozygotes—people carrying just one copy of that mutation—might. People with one copy, though still rare, are more common than homozygotes such as Aliria. If they also developed symptoms later than expected, albeit earlier than Aliria, that could bolster the case that the variant is protective. It might also mean a drug could be effective against the disease even without fully re-creating the mutation’s effects. “It would be a much easier drug-development case to make if you could show that heterozygotes had a middling protective effect, to show a dose effect,” Greicius says.

But studies of Christchurch heterozygotes outside Colombia didn’t support that idea. In 2020, Spanish scientists described two siblings who developed Alzheimer’s at ages 53 and 66 despite each having a single copy of Christchurch and no known early-onset genes. In 2023, researchers identified no clear protective effects among a few dozen people in the U.K. who had one copy. And a study by Greicius, posted this year as a preprint, found no evidence for a protective effect of a single copy in late-onset Alzheimer’s. (A recent analysis of four Reelin-COLBOS carriers with no early-onset mutations was also discouraging: Three became impaired in their early 70s, and one at age 59.)

A 2023 study of paisa mutation carriers was especially disheartening. Colombian and U.S. researchers led by Nicholas Cochran of the HudsonAlpha Institute for Biotechnology analyzed gene variants that might affect age of Alzheimer’s onset. The study, which included 11 Colombians who had one copy of Christchurch, named more than a dozen possibilities. Christchurch was not among them.

THEN AT AN AMSTERDAM scientific conference that summer, Quiroz turned the tables in breathtaking fashion. She displayed data from 12 paisa carriers with a single copy of Christchurch who developed symptoms 4 to 7 years later than those who lacked the variant—a modest benefit compared with Aliria, although still profound. When Cochran saw the presentation he recognized many patient cases from the GNA data he used in his own paper, but was stunned to see that for several, the ages of disease onset were different, and now suggested a benefit. In one case, Quiroz and her team reported that the patient’s symptoms started 10 years later than in the data Cochran had analyzed.

When Cochran wrote to Quiroz to ask why the data had changed, she replied that many cases in the GNA database “were not reviewed or studied using the comprehensive procedures that we now use” and some contained inconsistencies or errors. In an email Cochran provided to Science, Quiroz told him her team reopened hundreds of paisa mutation cases—not just those of Christchurch carriers—and relied on a consensus of clinicians for all changes. “Upon careful review we now have the opportunity to detect an effect that was not obvious before,” Quiroz added.

Since its inception, GNA’s methodology for estimating age at disease onset has included neuropsychological testing and extensive clinical interviews. No one had questioned GNA’s original evaluation process, until now.

A clinician who participated in the case reviews, speaking on condition of anonymity, could not recall the exact number of cases discussed, but says it was definitely not hundreds. Arboleda-Velasquez provided the lists of cases and took part in the meetings along with Quiroz, the clinician says. Invited reviewers were blinded to genetic information—although Arboleda-Velasquez and Quiroz were not—and for most cases the committee advised against revising ages of disease onset.

Science obtained relevant parts of some case files for Christchurch heterozygotes, in anonymized form, and asked a second GNA clinician, who also spoke on condition of anonymity, to review them. For the patient whose age at mild cognitive impairment had been adjusted by 10 years—from 42 to 52—the clinician saw no basis for the dramatic revision. “[Mild cognitive impairment] detected on neuropsychological evaluation at age 42,” the file says. At 52, the patient could no longer put his underwear on correctly, according to the file, and was diagnosed with dementia. Quiroz’s presentation put the dementia diagnosis at age 57.

Another patient developed mild cognitive impairment at 49 and dementia at 51, according to Quiroz. But the history in that patient’s file, taken when she was 51, showed she had dementia dating back further, according to the clinician, and mild cognitive impairment from age 47. In a third case, age at dementia onset was switched from 45 to 50. Yet the clinical narrative reveals the patient had been fully dependent on a caregiver for 2 years by age 50, noting, “She must be bathed, dressed, and even fed.”

The clinician on the review committee, looking at the same files, says the committee would not have advised the age adjustments. The clinician did not know what became of its recommendations. According to Quiroz’s email to Cochran, the updated ages were entered into the GNA database by David Aguillón, a young physician-scientist who was Lopera’s right-hand man.

In a June 2024 paper in The New England Journal of Medicine (NEJM), Quiroz and Arboleda-Velasquez—with Lopera, Reiman, Aguillón, and others as co-authors—laid out the results Quiroz had previewed in Amsterdam. Having a single copy of Christchurch delayed illness in paisa mutation carriers for a median of 5 years, they asserted. According to the paper, the team had rigorously reassessed records for 353 Christchurch-mutation carriers and noncarriers.

The study means Christchurch “is not only protective, but druggable,” Arboleda-Velasquez said in a press release, seemingly painting a bull’s-eye on the target for deep-pocketed pharma companies. But the paper only disclosed one patent among several that Arboleda-Velasquez, Lopera, Quiroz, and Reiman had filed in different countries for possible techniques and therapies based on Christchurch and Reelin-COLBOS.

Lopera had learned of Cochran’s earlier complaints to Quiroz about the heterozygote findings and had responded by mandating an internal data review, according to two longtime collaborators. It is unclear whether Lopera followed up, in part because his cancer diagnosis and rapid decline came soon after.

That summer, Aguillón assumed control of the research group as interim director, downplaying the severity of Lopera’s illness to colleagues, patients, and the public, according to GNA members and communications reviewed by Science. During the last month of Lopera’s life, when colleagues say he could no longer speak or write, GNA sent out a letter in his name to patients and families naming Aguillón as the group’s new director. University of Antioquia officials claimed in their own statement that Lopera remained “in constant contact with his team” and had been working with Aguillón to manage the transition.

In July 2024, Cochran emailed Arboleda-Velasquez to explain that Goate and Greicius had joined him in a formal letter they planned to send to NEJM challenging the finding of protection in Christchurch heterozygotes. He replied with alarm. “Dr. Lopera is gravely ill with a terminal condition and is in a very delicate situation,” Arboleda-Velasquez wrote the following month. “When Dr. Lopera was still in what I consider a good state of mind, he explicitly instructed us not to respond to your correspondence with NEJM.” He added in another message: “I am in shock. … You have caused me and a dying man a great deal of grief.”

In their letter to the journal, the three U.S. genetics researchers described the adjustments to clinical records as possibly biased. For all the cases supposedly reviewed, only a tiny fraction had been changed, they noted, except among Christchurch carriers.

“Unless and until the authors explain why the Christchurch carriers in the NEJM paper were 20 times more likely than noncarriers to have their ages at onset adjusted,” Greicius says, “I have to assume the books were cooked.”

The journal published the letter in October 2024, paired with a response signed by Arboleda-Velasquez and Lopera arguing the new analyses had been “systematically obtained by an interdisciplinary team” whose members were “unaware of genetic status,” ensuring no bias.

Lopera had died 7 weeks earlier. He would have been too sick to have weighed in on the response, according to Londoño-Castaño and other close colleagues.

As GNA’s leader, Lopera had been “extremely transparent” and had fostered a culture of openness among the group’s scores of employees and its collaborators, Londoño-Castaño says. After his death, “things were very closed off” and Quiroz, Arboleda Velasquez, and Aguillón exerted increasing control over priorities and resources. Last year, Londoño-Castaño left after 14 years with the group. Aguillón declined requests for an interview or to provide written answers to detailed questions.

Given the doubts about the NEJM paper, Science asked three leading forensic image analysts to help assess the reliability of other work supporting the claims about the protective mutations: Vanderbilt University neurologist and neuroscientist Matthew Schrag; Kevin Patrick, a nonscientist who has identified thousands of image problems, leading to numerous retractions; and microbiologist Elisabeth Bik, an expert on image-related error and misconduct. All worked without compensation.

They evaluated 12 papers reporting studies of Christchurch or Reelin-COLBOS that were supervised by Arboleda-Velasquez, Quiroz, Lopera, or close collaborators and had reviewable image data. In three of them, the experts found apparent image duplications or anomalies in images that supported the experimental hypotheses. The three contested papers describe beneficial effects of Reelin-COLBOS on the brains of people and genetically altered mice, and of Christchurch on brain organoids (lab-grown structures that mimic some brain functions) and glial cells that support nerve cells (see graphic, below).

The 12 papers included the five with reviewable images listed by Epoch Biotech as “foundational” to its work. Two were among those found to contain apparent image problems. (Epoch also described as foundational the contested NEJM study describing possible benefits from one copy of Christchurch, and the rejoinder letter to concerns raised by Cochran and others about that study.)

In addition, image anomalies appeared in a paper about Christchurch led by other scientists, to which Arboleda-Velasquez contributed. Seven other papers on related topics by GNA scientists also showed apparent image reuse or changes. The analysts did not find any anomalies in the 2023 study led by UC San Francisco’s Huang suggesting beneficial effects of Christchurch in cell cultures and animals.

In an effort to determine what happened, Science requested that Arboleda-Velasquez, Quiroz, and others share original, uncropped, high-resolution images to compare against the published versions in studies they co-authored. They did not provide any images.

Randy Schekman, a UC Berkeley molecular and cell biologist, former editor-in-chief of the Proceedings of the National Academy of Sciences, and Nobel laureate, calls the anomalies “pretty obvious,” and says they cast doubt on the researchers’ judgment. WashU neurologist David Holtzman, Tanzi, and other Alzheimer’s researchers agree that many of the flagged images appear highly problematic.

“Whether it’s evidence of misconduct or sloppy behavior, it’s an extended-enough problem that certainly harms the credibility of the research groups, and especially of the lead investigators,” Schrag says.

Others go further. Some of the apparently duplicated images seem unlikely to be simple mistakes, UCL’s Howard says. “One [problematic image] is an accident. … Two is careless. But once these things happen more than three times,” the pattern raises suspicions, he says.

YET WORK BY OTHER scientists continues to suggest intriguing biological effects of the Christchurch mutation. Holtzman and others have published evidence in mice suggesting Christchurch indeed restricts tau spread, and some studies in brain organoids echo those findings. Mayo Clinic researchers reported that expressing Christchurch in astrocytes, cells within the brain’s immune system, clears amyloid protein in mice. Chinese researchers reported that Christchurch added to liver cells tempered memory loss and brain damage in mice engineered to produce excess amyloid.

Goate and Greicius don’t dispute the validity of most such work. But the lab results will not necessarily translate to human protection against Alzheimer’s disease, they say. “We’ve seen countless experiments that look like a million bucks in mouse models, but fail miserably in humans. Some people pitch their mouse models to compensate for a lack of human genetics,” Greicius says. “That’s my problem with Christchurch.”

Holtzman thinks two copies of Christchurch might indeed delay Alzheimer’s. But, he says, “I am not convinced from the current human data that [Christchurch] in heterozygous form is protective” against any form of Alzheimer’s.

Far stronger evidence for Christchurch, experts say, would be other Alirias: carriers of paisa or other early-onset mutations who are homozygous for Christchurch and remain healthy late in life. So far no such cases have been reported. Without them, Goate says, the hypothesis about protective qualities is “kind of built on a house of cards.”

In 2022, a poster Aguillón presented at an Alzheimer’s conference in San Diego indicated his team knew about a second paisa mutation carrier with two copies of Christchurch. The case has not been mentioned since and was not represented in the 2024 NEJM paper, which contained a chart of Christchurch carriers known to the group. Correspondence reviewed by Science suggests Arboleda-Velasquez and Aguillón are carefully following that person, who is approaching the usual age of Alzheimer’s onset for a paisa carrier. If the person becomes ill within the expected age range, “it would be compelling evidence against the hypothesis,” Cochran says.

Even after the Aliria discovery alerted researchers to the importance of outliers, some fell through the cracks. Early in GNA’s research efforts, members of the Colombian kindred who staved off cognitive decline were initially assumed not to carry the paisa mutation and therefore were not tested, says pathologist Diego Sepulveda-Falla, who collaborated on seminal studies at GNA and is now at the University Medical Center Hamburg-Eppendorf. At least two who later received genetic workups and were confirmed to be paisa carriers don’t have the Christchurch, Reelin-COLBOS, or Macondo mutations, according to Sepulveda-Falla and neuroscientist Kenneth Kosik, a Lopera confidant and collaborator who is at UC Santa Barbara.

Mery Balbin, a homemaker and paisa mutation carrier who lived outside Medellín and died from a heart attack last fall at age 69, was another outlier. Although she eventually developed dementia, she only showed mild cognitive impairment at age 52, GNA records indicate. A decade later she was still living independently, according to her daughter Marta Pulgarín, who says she pleaded with Aguillón and Lopera to study her mother further. GNA ignored the case for years, finally arriving to draw blood just months before she died, Pulgarín says. Whether she had any of the mutations deemed protective is uncertain.

Arboleda-Velasquez has recently begun to tease the unveiling of the Macondo mutation on LinkedIn. “I hate memory loss,” he wrote in a post in October 2025. “I want to eradicate it. … How about a proclamation? I hereby end it! OK fine. A therapy. Maybe APOE Christchurch? How about Reelin COLBOS? [H]ow about the two combined? Or the next one that I am about to publish?”

Macondo occurs on the ciliary neurotrophic factor gene (CNTF), known to affect the health and survival of brain cells. If CNTF-Macondo helped protect against both early- and late-onset Alzheimer’s, as Arboleda-Velasquez and his colleagues claim in a draft manuscript dated early this year and described in detail to a Science reporter, it would be GNA’s most revolutionary contribution.

Its ties to García Márquez would also make it the most romantic. In a 2023 Christmas letter to his program’s staff—the last such letter he would live to write—Lopera cited Macondo, whose publication he believed imminent, as an upcoming highlight for the group. GNA leaders held meetings about Macondo with members of the García Márquez family in anticipation of a planned media rollout, say staff members involved with the effort.

But Sepulveda-Falla, who has analyzed many brain-tissue samples from paisa mutation carriers, recently raised a major concern. A handful who had Macondo developed dementia over a wide range of ages—from late 30s to early 70s—and had many other mutations of possible interest. Those findings led him to conclude that Macondo is not a reliable shield against Alzheimer’s on its own, if it protects at all.

Sepulveda-Falla withdrew from the paper and told Arboleda-Velasquez about his concerns. He says Arboleda-Velasquez responded that he would not change his conclusions. Dismayed by the draft paper’s quality, Kosik also withdrew as an author.

As a disciple of Lopera’s who has maintained ties to GNA and believes deeply in the importance of its science, Sepulveda-Falla describes the experience as painful. It set him on a path to revisit earlier Christchurch and Reelin-COLBOS papers, including ones he had worked on. Some were questioned in the dossier prepared for this investigation, such as the first Reelin-COLBOS paper, on which he is a senior author.

Sepulveda-Falla also began a systematic review of brain tissues for five paisa outlier cases, along with about 90 other samples in his possession. He found some protected individuals shared mutations with Aliria—but not Christchurch. In a recent preprint reanalyzing the Reelin-COLBOS findings, he makes the case that any protections Reelin-COLBOS may offer must occur in combination with a constellation of other mutations. He believes that conclusion applies equally to Christchurch. The preprint noted that Christchurch and Reelin-COLBOS outliers each had hundreds of other rare mutations that might have played a protective role.

Alzheimer’s experts are increasingly recognizing the importance of multigene effects in early-onset disease. They suggest a more complex, slower moving process of discovery than the single-gene revelations that enchanted funders, the media, and much of the scientific community. And although these combinations of variants, too, may someday point to therapies, they don’t hold the same immediate promise for Colombian research participants and their loved ones awaiting a cure.

Aliria’s daughter, Rocío Villegas Piedrahita, is among them. Like many others in her large extended family, she remains hopeful that the Christchurch mutation will inspire an effective treatment—and soon. Still, she took it in stride when a Science reporter summarized the concerns that have led some researchers to suggest a more nuanced scenario. “My mother was unique and special,” she says. “I’m sure she had lots of unique genes.”