Gene regulation at the TP73 locus is complex. The mRNA transcripts are initiated from two alternative promoters, leading to the production of two different protein isoforms. "Full-length" p73 mRNA, initiating from the P1 promoter, encodes a protein with an N-terminal transcriptional activation domain, while the shorter "delta-N" (ΔNp73) isoform, initiating from the major downstream P3 promoter, lacks this domain. The full-length p73 protein is pro-apoptotic, while the ΔNp73 acts as a natural dominant-negative, and protects cells from apoptosis [4]. The fact that the ΔNp73 isoform is preferentially expressed in neurons explains the net anti-apoptotic function of p73 in the peripheral nervous system, and probably in cortical neurons of the brain.

We therefore searched for additional sequence variants in TP73. As a screening panel, we chose genomic DNA from 12 people (24 chromosomes), including several individuals from each of the three ethnic groups. This analysis would not be expected to detect rare variants, but would detect common alleles. Sequencing of the exons encoding the DNA binding domain of p73 did not reveal any coding variants, although the BanI SNP, which is coding-neutral, was observed with a high frequency and served as a positive control. However, we did find a common SNP in the P3 promoter of TP73. This variant was a G→A transversion at position minus-386 relative to the transcriptional start site of the ΔN-p73 isoform. A transcription factor binding site database search [15]http://www.cbil.upenn.edu/tess/ showed that this SNP converted a variant consensus progesterone receptor binding site overlapping the -386A allele to a non-consensus sequence in the -386G allele (Fig. ​(Fig.1).1). This SNP also overlapped with a p53 consensus binding half-site, but no second nearby half-site was found, suggesting that p53 binding to the P3 promoter [16] is not likely to be affected by the -386G/A SNP.

To ask whether the -386G/A SNP might be associated with AD risk, we genotyped the series of cases and controls. These individuals were also genotyped for coding variants at the APOE locus, which are known modifiers of AD risk [17,18]. The results are shown in Table ​Table4.4. The -386G/A allelic system was in Hardy-Weinberg equilibrium in the controls, and the -386G allele was significantly associated with AD in the group as a whole, both when the data were analyzed as unadjusted odds ratios, and after multivariate logistic regression analysis correcting for age and education.

After stratifying by APOE genotype, the AD association of the -386G TP73 allele was seen in the groups with one or zero APOE ε4 alleles, but was not seen in the group of APOE ε4/ε4 homozygotes (Table ​(Table5).5). Since homozygosity for APOE ε4 in itself conveys a high risk of AD, this result is not surprising.

We next asked whether allele frequencies of the -386G/A SNP might differ by ethnicity. As shown in Table ​Table6,6, there was indeed a difference among the three ethnic groups, with the highest frequency of the -386G allele in African-Americans and the lowest frequency in European Caucasians. Consistent with the known history of ethnic admixture in the Caribbean, Hispanics showed an intermediate allele frequency. When we reexamined the AD associations stratifying by ethnicity, we found a significant association of homozygosity for the -386G allele with AD in the Hispanics, and a similar direction of the effect, but failure to achieve significance, in the smaller number of African-Americans (Table ​(Table7).7). There were too few Caucasians with the -386G allele to allow conclusions.

Although multiple interpretations of these results are possible, one hypothesis is that the -386G/A SNP, or closely linked polymorphisms, might influence AD risk via a functional effect on ΔNp73 expression in the brain. Small variations in the net expression of the ΔNp73 mRNA isoform in human brains would be hard to quantify, but examining allele-specific expression, in which one allele serves as an internal control for the other, is expected to be more sensitive. We therefore screened a series of 34 brains for heterozygosity at the TP73 exon 5 BanI polymorphism, which would allow mRNA from the two alleles to be distinguished by RFLP and direct sequencing. Since the collection of autopsy brains available to us was primarily, though not exclusively, from Caucasians, the allele frequency at this marker was low. Nonetheless, we found 4 brains, 1 fetal mid-gestation and 3 adult, which were heterozygous. Two of the adult brains were affected by AD and one was normal.

Among these 4 brains, 3 manifested approximately equal biallelic mRNA expression of TP73 mRNA, as indicated by comparison with genomic PCR controls, while one adult brain (the normal brain) showed a strong allelic expression bias (Figure ​(Figure2).2). This bias was reproducible in multiple independent RT-PCR reactions (Fig. ​(Fig.2),2), and was also reproduced starting from two separate RNA preparations from this same brain. We utilized RFLP analysis since this method gives a simple read-out of the two alleles. Heteroduplex formation can cause artifacts in this method when the cDNA is over-amplified, so we kept the PCR cycle number to a minimum (35 cycles). Genomic controls were run in parallel using the same PCR conditions, and these gave highly reproducible RFLP results, which were taken as the standard for biallelic representation. Lastly, the results of direct sequencing, which is not sensitive to heteroduplex formation, were consistent with the RFLP results (Fig. ​(Fig.2).2). From this analysis, we conclude that some human brains manifest an allelic bias for TP73 mRNA expression. Lastly, we established the genotype at the -386G/A P3 SNP in the 4 brains (Table ​(Table8).8). The adult brain with strongly biased expression was indeed heterozygous for this marker, but heterozygosity was also found in the fetal brain, which showed approximately equal biallelic mRNA expression.

Lastly, we tested for major effects of the -386G/A SNP on P3 promoter activity by transfecting promoter-luciferase reporter plasmids into human SK-NSH-N neuroblastoma cells. No significant difference in basal promoter activity was seen in this assay (Fig. ​(Fig.33).

Our brain expression data add to the findings of one prior study, in which 3 of 4 informative fetal brains showed virtually monoallelic expression of TP73 mRNA, with one fetal brain showing equal biallelic expression [9]. While our current results are in a broad sense consistent with those findings, our experimental approach was different in two respects, which make the two sets of data complementary. First, we examined adult brains; and second, we utilized the exon 5 BanI RFLP as a marker, while the previous study utilized a different polymorphism (StyI RFLP) in exon 2, upstream of the P3 promoter. This second difference means that we have examined TP73 mRNA derived predominantly from the P3 promoter (the major isoform in neural tissues), while the prior study examined mRNA derived exclusively from the upstream P1/P2 promoters.

Explanations for an allelic expression bias include genomic imprinting, which seems unlikely at the TP73 locus based on the negative data from mice but has not been strictly excluded in humans, or the presence of functional sequence polymorphisms in the promoters/enhancers of this gene. The -386G/A SNP itself is one candidate. The adult hippocampus that showed an allelic expression bias in our study was indeed heterozygous for this SNP, but in the fetal cerebral cortex that we examined, heterozygosity for this SNP evidently did not result in biased expression. So either the -386G/A SNP is not the critical sequence variant, or alternatively the effect of this variant is manifested in the adult hippocampus but not in whole fetal cerebral cortex. Our promoter-reporter assays in SK-NSH-N neuroblastoma cells did not indicate a strong difference in basal P3 promoter activity in comparing the -386G and -386A alleles. However, this polymorphism affects a predicted progesterone receptor binding site and progesterone-dependent transcriptional activity remains to be tested.