It is not yet clear what the findings mean for research in this area. Roses notes methodological differences between his work and Goate’s that he said may explain the discrepancies. The issue will likely not be settled until data come in from even larger studies. For example, the Alzheimer's Disease Genetics Consortium is currently running a Tomm40 study with several thousand participants, said co-director of data analysis Lindsay Farrer at Boston University, Massachusetts. Meanwhile, Roses’ company, Zinfandel Pharmaceuticals, Durham, North Carolina, has a collaboration in the works with Takeda Pharmaceutical Company Limited, Osaka, Japan, that will seek to validate Tomm40’s status as a biomarker for individual AD risk, as well as test the ability of Takeda’s diabetes drug pioglitazone to delay AD progression (see Takeda press release). The trial is currently seeking FDA approval, Roses said, and will likely start in late 2012 with around 4,000 participants.

In 2009, Roses and colleagues first stirred up the Tomm40 buzz with their finding that the length of a variable poly-T repeat in an intron of the Tomm40 gene helped predict the age at which certain people will develop AD symptoms (see ARF related news story and Roses et al., 2009). The researchers repeated the finding in a small prospective cohort (see ARF related news story), showing that, in people who carry the ApoE3 allele, those with at least one short poly-T repeat develop AD about seven years later than those with two copies of long poly-T repeats. In other words, the short repeat seems to be protective, while longer repeats confer more risk.

Efforts to validate the finding have been mixed so far. Some research has supported the link; for example, scientists at the University of Wisconsin, Madison, found an association between Tomm40 variants and gray matter volume and cognition in middle-aged ApoE3/3 carriers that is consistent with Roses’ results (see Johnson et al., 2011). Likewise, researchers at New York University also turned up data suggesting that the short poly-T repeat is protective against AD (see Bruno et al., 2011). However, scientists at Carnegie Mellon University could not replicate Tomm40’s effect on onset age in a cohort of about 900 people (see Chu et al., 2011), while a study from the University of Navarra, Pamplona, Spain, showed inconsistent results (see Cervantes et al., 2011). The AlzGene database lists Tomm40 polymorphisms linked to AD.

To try to clarify the picture, Goate and colleagues undertook a large case-control study. First author Carlos Cruchaga combined data from two cohorts, for a total of around 1,600 AD patients and 1,200 controls. Roughly half the participants came from the Knight Alzheimer's Disease Research Center at WashU, the other half from the National Institute on Aging LOAD Family Study. Almost half of the participants had the crucial ApoE3/3 genotype. The authors defined onset as the age when clinical symptoms first appeared, as recorded in the clinical assessment, Cruchaga told ARF, but noted it can sometimes be difficult to be sure of the exact age of onset.

When they analyzed the data, Cruchaga and colleagues found a small but significant association suggesting that very long poly-T repeats confer a lower risk of AD in an ApoE3/3 background. This contradicts previously published results suggesting long repeats are high risk (but see below for Roses’ comments on unpublished data). Cruchaga and colleagues found no other significant associations, including no effect of Tomm40 variants on the AD age of onset. Separately, the authors examined about 120 postmortem brains from the WashU Neuropathology Core and found no links between Tomm40 variants and the expression of either ApoE or Tomm40 proteins in the brain. They also looked at cerebrospinal fluid biomarker levels in samples from WashU and from the Alzheimer’s Disease Neuroimaging Initiative (ADNI), and again found no correlations with Tomm40 polymorphisms. Collectively, the data suggest that Tomm40 may be involved in disease pathways, but the mechanism is unlikely to involve expression levels or β amyloid or tau pathways, the authors note.

“It is still possible that this variant affects risk of the disease, but the effect is going to be very small,” Cruchaga suggested. “We do not think [the Tomm40 variant] is going to be useful to separate individuals with ApoE3 genotypes, based on our results.”

Farrer said the study by Cruchaga and colleagues was well designed and sufficiently powered to detect effects of the size seen in previous Tomm40 papers. Given the larger sample size in the new paper, Farrer suggested, “Unless there is some ascertainment bias in one of the studies, [the data] would have to favor the results of this study.” Farrer noted that the mitochondrial Tomm40 protein may still be an attractive candidate for scientists studying disease pathways, since several cellular investigations show a role for mitochondrial dysfunction in AD. Nonetheless, based solely on the results of this study and prior published studies, Farrer said, “To me it appears, at least in terms of the genetic variance of the disease, that [Tomm40] does not account for the genetic risk.”

Roses takes a different view. “I believe there are valid reasons why there are differences between their results and ours,” he wrote to ARF (see full comment below). One key factor is how the age of onset is determined. When onset ages are taken from a retrospective case-control study such as this one, inconsistencies in how ages are reported may lead to inaccurate data, Roses said. The best way to get accurate onset ages, he suggested, is to prospectively follow a cohort of cognitively normal people using neuropsychiatric tests to see when the first Alzheimer’s-like cognitive problems develop. He cited the work of Richard Caselli at the Mayo Clinic in Scottsdale, Arizona, as an example of this approach. Caselli presented preliminary data at the 2010 International Conference on Alzheimer’s Disease that supported Roses’ findings (see ARF related news story). In an update at the 2011 AAIC meeting, Caselli discussed more extensive results from a cohort of about 900 people that again replicate the association between Tomm40 variants and age of disease onset, Roses said. Cruchaga agrees that a case-control study does not provide as accurate an age of onset as a prospective study, but he notes that the decrease in accuracy is corrected for by the large sample size in his study. His sample had enough statistical power to detect an eight-year difference in disease onset, even if the ages of onset were misdiagnosed by as much as two years, Cruchaga told ARF.

In regard to the link between very long poly-T repeats and late ages of onset seen by Cruchaga and colleagues, Roses noted that his group has seen the same effect, which they described at the 2011 Alzheimer’s Association International Conference, held 16-21 July. The very long/very long genotype has a biphasic presentation, Roses said. Most people who develop late-onset AD before the age of 60 carry two copies of the very long poly-T repeats, indicating high risk associated with the genotype. However, these people make up a minority of those with the very long/very long genotype. The majority of people with this genotype instead have a late disease onset, typically over age 80. It is not yet clear why this is, Roses said, and he is currently investigating the phenomenon. He suggested there may be another genetic factor that discriminates between these groups. Roses noted that the planned Takeda-Zinfandel trial will be restricted to people between the ages of 62 and 83. Since people with the high-risk very long genotype typically develop symptoms before age 62 and will therefore not be included in the study, the very long genotype will be treated as protective in the trial design, Roses said.

Meanwhile, other scientists propose different interpretations for the Tomm40 results—for example, that variants in the ApoE promoter region, which affect ApoE expression, are responsible for the observed association. “The Tomm40 observations are really a reworking of the observation, mainly by Chartier-Harlin and colleagues, that the ApoE haplotype influenced disease risk,” wrote John Hardy at University College London, U.K., in an e-mail to ARF (see also ARF related news story and Lambert et al., 2002). However, Cruchaga noted that they did some fine mapping of the ApoE region and did not see any effect from common genetic variants on risk or age of onset, making this idea unlikely to explain their findings.—Madolyn Bowman Rogers.

Reference:
Cruchaga C, Nowotny P, Kauwe JS, Ridge PG, Mayo K, Bertelsen S, Hinrichs A, Fagan AM, Holtzman DM, Morris JC, Goate AM; for the Alzheimer’s Disease Neuroimaging Initiative. Association and expression analyses with single-nucleotide polymorphisms in Tomm40 in Alzheimer disease. Arch Neurol. 2011 Aug;68(8):1013-19. Abstract