Triosephosphate isomerase (TPI, TIM; EC 5.3.1.1) is a cytoplasmic enzyme of the carbohydrate metabolism and catalyzes the glycolytic interconversion of the three-carbon sugars glyceraldehyde-3-phosphate and dihydroxyacetone phosphate in an extraordinarily efficient way 12. In most species, this enzyme is present as a soluble, stable dimer and appears to be highly conserved among all kingdoms. In humans, the TPI enzyme is encoded by a single gene located at chromosome 12p13 3. This gene is associated with a rare genetic disorder, triosephosphate isomerase deficiency, initially described in 1965 4, that is a unique glycolytic enzymopathy with autosomal recessive inheritance characterized by chronic haemolytic anaemia, cardiomyopathy, susceptibility to infections, severe neurological dysfunction, and, in most cases, death in early childhood 56. The disease pathology correlates with a drastic decrease in the enzymatic activity of TPI and consequently, increased concentration of the TPI substrate dihydroxyacetone phosphate (reviewed in 67). Analysis of the pathogenic mutations in a recombinant yeast model revealed that the decrease in TPI activity is not based on an per se inactive enzyme species, but likely the consequence of a misguided regulatory mechanism that might be caused by altered dimerization of the pathogenic TPI isoforms 8. These results are in good agreement with previous observations made in mice showing that a complete lack of TPI activity (homozygous TPI null alleles) causes embryonic lethality at the earliest stages 9. Accordingly, a recent study in Drosophila reveals that the pathogenesis of TPI deficiency could result from proteasomal degradation of the apparently functional enzyme 10.

Remarkably, in several human populations, unexpected high frequencies of individuals exhibiting about 50% of the normal TPI activity have been detected 1112131415, indicating heterozygously inherited null alleles. The estimated allele frequencies of these ranged from 0.002 (Europeans) to 0.02 (Afro-Americans) 1112131415, but the underlying molecular genetic defects have so far remained unknown. An association of reduced TPI activity with three allelic variants of the TPI promoter region (located at pos. -5, -8, and -24) in African-Americans 11 could not be confirmed in a follow-up study; these single nucleotide polymorphisms (SNPs) turned out to represent common haplotypes in another African-American sample 16. Keeping in mind the aforementioned lethality of homozygous null alleles, the high frequency of the heterozygous null alleles may indicate an evolutionary advantage of a TPI^+/0 genotype 8911. The nature of this potential heterozygote advantage has not yet been elucidated, however, reduced TPI activity mediates cellular resistance to conditions of oxidative stress 8. This phenotype has been studied in yeast and in C. elegans, and is based on a metabolic re-routing between glycolysis and the associated pentose phosphate pathway 17. This observation is of basic biological interest since these two metabolic pathways are not only implicated in the cellular energy supply, but are also involved in cell growth, cell death and in the ageing process. Thus, the TPI1 locus represents an interesting target for genetic analysis, especially for studies focusing on ageing or oxidative-stress processes. However, at least for European populations, comprehensive sequence information about allelic TPI variants and their distributions and frequency is still lacking. Moreover, since oxidative stress is a key player in the ageing process – and model organisms with reduced TPI activity are more resistant to it – it is also of viable interest to ascertain the distribution and frequency of allelic TPI variants in humans who have attained an exceptional life span above average. Therefore, we re-sequenced the TPI1 locus in an extended sample of 357 German long-lived individuals aged 95 to 107 years. Two of the identified frequent polymorphisms were subsequently tested for association with the longevity phenotype.

This study presents the first molecular analysis of the genetic variability of the human TPI1 gene locus in a large set of long-lived individuals. As illustrated in figure 1, the TPI1 locus is composed of 7 exons and spans a genomic region of 3.2 kb. This allowed high sequence coverage by using only 6 PCR fragments (please see Additional file 1 for the primer sequences used). We detected 17 different mutations or polymorphisms in the analyzed region of the TPI1 gene relative to the reference sequence (Table 1). Only two of them (rs2071065 and rs2071069) were previously known, whereas the other 15 were not identified in a dbSNP search. Out of these, 15 SNPs were rare, in fact, 11 variants occurred only once. Of note, one subject had three of these private mutations. None of the previously detected SNPs in African-Americans 11 were seen in the German sample.

As listed in Table 1, the two known TPI1 variants (rs2071065 and rs2071069) had a much higher allele frequency (0.270 and 0.260, respectively) compared to the others. One of these (SNP rs2071069) has been studied previously (termed TPI 2898 in 21 and TPI 2262 in 22). Although its low frequency in the random sample, this SNP was found in all subjects having inherited the most common disease allele (TPI 1591C, encoding TPI_Glu104Asp). This discovery allowed the authors to conclude that all TPI_Glu104Asp subjects are descendants of a common ancestor that lived in today's France or England more than 1000 years ago 2122.

The detected allele-frequencies of rs2071065 and rs2071069 prompted us to test whether there is a genetic association between the SNP alleles and the longevity phenotype in larger samples of LLI (n = 1422) and control individuals (n = 967). The genotype analyses revealed SNP frequencies that were quite similar to those determined by sequencing. SNP rs2071065 showed an allele frequency of 0.30 in the LLI and 0.28 in the controls, respectively; the second SNP rs2071069 had a frequency of 0.29 in the LLI and 0.27 in controls. However, neither of the two markers showed a statistically significant difference in allele or genotype frequency between LLI and control subjects, using the entire collection or subsamples stratified for gender (P > 0.05).

Fourteen of the detected mutations were located in intronic and three in exonic regions. However, none of the detected variants seem to affect the amino acid sequence of the TPI enzyme. The only SNP located within the TPI1 coding sequence (ref. pos. 81309) encodes Leu121 and is silent; and hence, none of the identified sequence variants is likely to result in a TPI1 null allele. This result seems to be inconsistent with a previous study based on a German population, in which 3000 persons were screened for reduced TPI activity; 3.7 per 1000 individuals showed reduced triosephosphate isomerase activity (in the range of 39 to 76% compared to the enzyme activity of controls) and were verified to be heterozygous carriers 14; even higher frequencies of haploid TPI deficiencies were detected in other populations 1112131415). However, there are several possible explanations for the discrepancy of these biochemical screenings with the results of our sequence analyses. First, it is feasible that the number of sequenced individuals was too small to detect the predicted null alleles. Second, the reduced enzyme activities measured in the previous studies could be, at least in part, based on factors acting is trans; e.g. mutations affecting the activity of transcription factors controlling the expression of the TPI enzyme. Indeed, in the Eber et al. study, none of the heterozygous persons expressed an electrophoretic variant of the TPI enzyme 14. Finally, it cannot be excluded from our data that inactive TPI alleles have negative effects on human life expectancy and hence, would be depleted in the analyzed LLI group. In fact, a variety of genes increasing the oxidative-stress resistances in model organisms can, dependent on environmental conditions, shorten their lifespan, and yeast as well as C. elegans with reduced TPI activity show indeed shortened lifespan when growing under standard laboratory conditions 17. This phenotype might be caused by an observed shift of the cellular redox potential versus its reduced state 17, resulting in a condition termed reductive stress. Remarkably, reductive stress has been found to be deleterious and disease-causing in mice 23. Nonetheless, our data reveals no enrichment of TPI null alleles among long-lived individuals, indicating that the potential heterozygote advantage of the TPI^+/0 genotype does not promote longevity in humans.

For a final answer to all these questions, further studies have to focus on much larger numbers of studied individuals and on the comparison of TPI allele frequencies in different age groups and populations. Fortunately, next generation sequencing technology (e.g. massively parallel pyrosequencing with 454/Roche FLX system, sequencing by synthesis with the Illumina/Solexa system or sequencing by ligation with the Applied Biosystems SOLiD system) will allow coping with the required sample quantities at affordable costs.

In this study, we analyzed and sequenced the TPI1 locus in a large sample of LLI of German descent. We identified 17 SNPs of which 15 were rare and previously unknown. None of the identified sequence variations affected the amino acid composition of the TPI protein and hence, these mutations are unlikely to impact the catalytic activity of the enzyme. Thus, genetic TPI variants occur less frequent than expected and inactive TPI alleles are not enriched in German centenarians.

The authors declare that they have no competing interests.

BT and RR conducted and designed the re-sequencing, AN and RK the genotyping, BT and AN the statistical analyses, MR, SK, HL and BT conceived and designed the study, SS conceived and organized the sampling, MR wrote the first manuscript draft, MR, SK, AN and BT the final version. All authors read and approved the final manuscript.