Introduction

The APOA1/C3/A4/A5 gene cluster, located in the human chromosome 11 (Fig. 1), is a key regulator of lipid metabolism (Talmud et al. 2002; Wang et al. 2004; Lai et al. 2005). These apolipoprotein genes show different patterns of tissue-specific expression even when they share cis-regulatory elements (Ogami et al. 1990; Ktistaki et al. 1994; Kardassis et al. 1996). In particular, the distal APOC3-enhancer can regulate tissue-specific expression of the three APOA1, APOA4, and APOC3 genes (Ogami et al. 1990; Ktistaki et al. 1994; Zannis et al. 2001). A recent study found that this common enhancer did not regulate in vivo APOA5 expression (Gao et al. 2005), which agrees with much lower plasma levels of apoAV compared to other apolipoproteins (O’Brien et al. 2005). For example, plasma apoAV levels (24–406 μg/l) (O’Brien et al. 2005) are 1,000–2,000-fold-lower than human apoCIII plasma concentrations (50–140 mg/l) (Marz et al. 1987; Sakurabayashi et al. 2001).

Fig. 1
figure 1

APOA1/C3/A4/A5 gene cluster. The human APOA1/C3/A4/A5 gene cluster spans about 50 kb in the chromosome 11. A common enhancer (APOC3-enhancer) regulates transcription of all the genes in the cluster but the APOA5 gene. Arrows indicate direction of the transcription

Full size image

The reason for the apparent absence of action of the APOC3-enhancer over the APOA5-promoter is not clear. The distance between the APOA5-promoter and the APOC3-enhancer is about 35 kb, and it is known that eukaryotic enhancers are able to act over promoters located up to a distance of 100 kb (Zhao and Dean 2005). A possible explanation could be related to the ability of the APOA5-promoter to effectively compete for the common enhancer. There are several possible models to explain how enhancers act over large distances (Bondarenko et al. 2003), but regardless of the mechanism the activation signal must travel from the enhancer to the target promoter. This implies that intervening elements between the enhancer and the promoter may interfere with the enhancer action. Because the APOC3-enhancer acts over the APOA4 gene (the last gene of the cluster before APOA5), any element interfering with the action over the APOA5-promoter must lie in the APOA5-APOA4 intergenic region.

In humans, the APOA5-APOA4 intergenic region spans 28 kb and contains 22 full-length Alu elements (GenBank accession NT_033899), for an average of one Alu element every 1.3 kb. This distance is almost twice as dense as the average in the whole human genome (one Alu insert every 2.5 kb) (Mighell et al. 1997). It is noteworthy that in the human chromosome 11 the density of Alu elements is one Alu insert every 2.7 kb in the whole chromosome, and one Alu insert every 3.0 kb in the intergenic regions (Grover et al. 2004). Alu elements are repetitive sequences of ≈300 nucleotides in length, and they are specific of primate genomes. Alu repeats contain a promoter for the RNA-polymerase III (Schmid and Maraia 1992), as well as regulatory elements (for example hormone response elements, HREs) for the RNA-polymerase II (Vansant and Reynolds 1995; Shankar et al. 2004). It has been shown that Alu sequences are able to regulate expression of neighboring genes, for example, having enhancer-blocking activity (Willoughby et al. 2000), regulating transcriptional interference (Willoughby et al. 2000), and showing enhancer activity (Landry et al. 2001). We hypothesized that Alu elements in the APOA5-APOA4 intergenic region may affect expression of the APOA5 gene by interfering with the ability of the APOC3-enhancer to act over the APOA5-promoter.

If Alu sequences in the APOA5-APOA4 intergenic region have functional importance they will show a slower evolutionary rate of nucleotide change as compared to other Alu elements without functional significance. By using a published sequence of the APOA1/C3/A4/A5 gene cluster of chimpanzee (Pan troglodytes) and comparing to the gene cluster of humans, we estimated substitution rates of the Alu elements in the APOA5-APOA4 intergenic region. According to our working hypothesis, significant differences in evolutionary rates will be found if some Alu sequences have functional relevance.

If Alu repeats were interfering with the action of the APOC3-enhancer on the APOA5-promoter, insulator-like elements must lie inside those Alu sequences. In vertebrates, many insulators require the binding of the zinc-finger protein CTCF to exert their action (Bell et al. 1999; Felsenfeld et al. 2004; Gaszner and Felsenfeld 2006). We scanned the APOA5-APOA4 intergenic region for the presence of potential CTCF binding sites and assessed whether those sites reside inside Alu elements.

Materials and methods

Alu sequences

Nucleotide sequences of the Alu elements in the APOA5-APOA4 intergenic region of humans were obtained from the GenBank accession NT_033899 and from the accession NW_001222312 for chimpanzee. The positions of the Alu sequences were determined using the RepeatMasker software (http://www.repeatmasker.org) (Smith et al. 1996–2004). We further retrieved the sequences of Alu elements 100-kb upstream and 100-kb downstream from the APOA5 promoter, with the aim of determining how the nucleotide substitution rates of Alu inserts within the APOA4-APOA5 intergenic region compare to the substitution rates of Alu elements in the genomic neighborhood.

Statistical analysis

Pairwise alignment of Alu sequences and estimation of evolutionary distances were performed using MEGA version 3.1 (Kumar et al. 2004). Substitution rates (R) were estimated as R = D/2T, where D is the number of substitutions per site since the time of divergence between humans and chimps. The Tamura–Nei (Tamura and Nei 1993) distance was used to estimate D because it allows for differences of substitution rates between nucleotides and does not assume equality of nucleotide frequencies. The divergence time (T) between humans and chimps was assumed to be 6 million years (Chen et al. 2001). We scanned the APOA5-APOA4 intergenic region, using the MotifScanner software of the TOUCAN package (Aerts et al. 2003), for the presence of CTCF binding sites. We used the position weight matrices (PWMs) recently derived by Xie et al. (2007).

The PROC CLUSTER in the Statistical Analysis Systems (SAS) software version 9 (SAS Institute Inc., Cary, NC) was used to identify potential groups of Alu elements with similar substitution rates. The PROC GENMOD in the SAS software was used to estimate least square means of explanatory variables on substitution rates. The explanatory variables used were: (1) distance from the transcription-start of the APOA5 gene [≤14 vs. >14 kb] and (2) type of Alu-family (oldest J-family vs. more recent S and Y families) and (3) orientation of each Alu element relative to the orientation of the APOA5 transcription.

Results

Out of the 22 full-length Alu elements in the human APOA4-APOA5 intergenic region, 21 were retrieved from the contig NW_001222312 of chimpanzee. The most proximal Alu sequence to the transcription-start of the APOA5 gene was not complete in the contig from chimpanzee; therefore, it was not included in the analysis. Figure 2 shows substitution rates (per site per 109 years) of the 21 pairs of Alu elements along the APOA5-APOA4 intergenic region. A great heterogeneity on the substitution rates was observed, from a minimum of 0.59 ± 0.41 to a maximum of 3.56 ± 0.99. It is noteworthy that Alu elements with low substitution rates tended to be located in the first 14 kb upstream of the APOA5 transcription start site. In particular, Alu sequences with the lowest substitution rates are located between 10- and 14-kb upstream from the APOA5 gene.

Fig. 2
figure 2

Substitution rates of Alu elements and predicted CTCF binding sites along the APOA5-APOA4 intergenic region. The Tamura–Nei distance was used to estimate the number of substitutions per site since time of divergence between humans and chimps. Time of divergence was assumed to be 6 million years. Note the clustering of slow-evolving Alu repeats in between 10- and 14-kb upstream of the APOA5 gene. Note the clustering of predicted CTCF binding sites in the first half of the intergenic region. Arrows indicate CTCF binding sites placed inside Alu elements

Full size image

Nine CTCF binding sites were predicted in the APOA4-APOA5 intergenic region (Fig. 2). Seven out of the nine CTCF binding sites were located in the first 14 kb of the intergenic region, and five out of those seven CTCF binding sites (indicated by arrows in Fig. 2) were placed inside Alu elements. The remaining two CTCF binding sites in the second half of the intergenic region were located outside of Alu repeats.

Cluster analysis showed two clearly defined groups of Alu elements (Table 1). The first cluster contains Alu sequences with a low substitution rate (mean 0.98 ± 0.18), while the second group includes Alu elements with a higher substitution rate (mean 2.74 ± 0.54). Compared to the low substitution rate group, the high substitution rate group had an excess of Alu sequences with the same orientation as the APOA5 gene (5 out of 11 vs. 9 out of 10, respectively, P = 0.06, exact test). The oldest Alu-J family tended to occur more frequently in the low substitution rate group (6 out of 11) compared to the high substitution rate group (2 out of 10) (P = 0.18, exact test). Finally, slow-evolving Alu sequences tended to be located in the first 14-kb upstream of the transcription-start of the APOA5 gene (8 out of 11) as compared to fast-evolving Alu sequences (3 out of 10) (P = 0.09, exact test).

Table 1 Cluster analysis of the Alu elements in the APOA5-APOA4 intergenic region
Full size table

The same trends were observed using general linear models. Table 2 shows least square means of the substitution rates by Alu-element orientation, Alu-family type, and upstream distance from the transcription-start of the APOA5 gene. In univariate models, Alu sequences with the same orientation of the APOA5 gene had a higher substitution rate as compared to Alu sequences with an opposite orientation (P = 0.02). The oldest Alu-J family showed lower substitution rates as compared to the more recent Alu-S and Alu-Y families (P = 0.04). And Alu elements in the first 14-kb upstream of the transcription-start of the APOA5 gene had lower substitution rates relative to the substitution rates of the Alu elements in the second half of the APOA5-APOA4 intergenic region (P = 0.06). To isolate independent effects of each explanatory variable, we used multivariate general linear models. After controlling each variable for the other two variables, the strongest predictor of the substitution rates was the orientation of the Alu elements (P = 0.04) followed by Alu-family type (P = 0.04). Effect of position was attenuated after adjusting for orientation and Alu-family type (P = 0.23).

Table 2 Least square means of the substitution rates of the Alu elements in the APOA5-APOA4 intergenic region
Full size table

At last, we retrieved all the Alu elements in the 200-kb region around the APOA5 promoter (100-kb downstream and 100-kb upstream from the transcription start of the APOA5 gene) to estimate the mean substitution rate of Alu inserts in the genomic neighborhood of the APOA5-APOA4 intergenic region. A total of 85 Alu elements, 21 inside and 64 outside of the APOA5-APOA4 intergenic region, were retrieved. Figure 3 shows the distribution of the substitution rates in the 200-kb genomic region that has an overall mean of 1.81 ± 0.12 substitutions per site per 109 years. This overall mean is almost twice the mean of the slow-evolving group in the APOA5-APOA4 intergenic region (0.98 ± 0.18). The lowest substitution rate in the intergenic region (0.59 ± 0.41) is observed around 12.6-kb upstream of the APOA5 transcription start, and we determined how extreme this value is in the overall distribution. We found that just 6 out of the 85 Alu elements (7.0%) in the 200 kb genomic region have substitution rates equal to or lower than 0.59 (Fig. 3).

Fig. 3
figure 3

Substitution rates of Alu elements in the 200-kb region around the APOA5 promoter. A total of 85 Alu elements were retrieved from the genomic region 100-kb downstream and 100-kb upstream of the APOA5 promoter. The Tamura–Nei distance was used to estimate the number of substitutions per site since the time of divergence between humans and chimps. Time of divergence was assumed to be 6 million years. The overall mean of the substitution rates is 1.81 substitutions per site per 109 years as compared to just 0.98 substitutions per site per 109 years in the slow-evolving Alu group in the APOA5-APOA4 intergenic region. It is noteworthy that just 6 out of the 85 Alu elements showed an equal or lower substitution rate than the lowest substitution rate in the APOA5-APOA4 intergenic region (0.59 substitutions per site per 109 years) as indicated by the asterisk

Full size image

Discussion

Even though it is known that expression of the APOA5 gene is regulated through different mechanisms (Prieur et al. 2003; Genoux et al. 2005; Nowak et al. 2005), it remains to be determined why plasma concentration of apoAV is substantially lower compared to concentrations of other apolipoproteins such as apoAI and apoCIII. A recent study showing that the APOC3-enhancer that affects expression of the APOA1, APOC3, and APOA4 genes does not up-regulate transcription of the APOA5 gene (Gao et al. 2005) provided some insight into this issue, but several points remain uncertain. For example, it is unclear why the APOC3-enhancer is not able to act upon the APOA5-promoter that is just 35 kb away, when it is known that eukaryotic enhancers may exert their effects up to a distance of 100-kb (Zhao and Dean 2005). Therefore, it is possible that the presence of intervening elements hinder the action of the APOC3-enhancer over the APOA5-promoter, when it is known that enhancer-action requires travel of the activation signal from the enhancer to the target promoter (Bondarenko et al. 2003).

We examined the APOA5-APOA4 intergenic region that extends over 28 kb in the APOA1/C3/A4/A5 gene cluster to identify evidence of cis-regulatory elements that may be interfering with the action of the APOC3-enhancer. In humans, the APOA5-APOA4 intergenic region is rich in Alu elements, with an average density of one Alu insertion every 1.3 kb as compared to a whole-genome average of one Alu insertion every 2.5 kb (Mighell et al. 1997). Because Alu elements are able to affect gene expression (Vansant and Reynolds 1995; Britten 1996; Willoughby et al. 2000), we hypothesized that some of the Alu sequences in the APOA5-APOA4 intergenic region may affect expression of the APOA5 gene, perhaps by interfering with the ability of the APOC3-enhancer to act over the APOA5-promoter. If this hypothesis were true then Alu elements with functional relevance will tend to have a smaller nucleotide substitution rate as compared to Alu elements with no function. Our data show that Alu elements with a lower substitution rate were located in the first 14 kb of the APOA5-APOA4 intergenic region, were more likely to have an opposite orientation to the APOA5-gene transcription and tended to belong to the oldest Alu-J family. Although our results do not prove the functionality of the analyzed Alu sequences, they are highly suggestive of such function. Most importantly, they identify a likely location of the Alu elements with functional relevance, namely the first 14 kb upstream from the transcription start of the APOA5 gene. This conclusion was further supported when we retrieved all the Alu sequences in the 200-kb genomic region surrounding the APOA5 promoter. First, the mean substitution rate of the slow-evolving Alu group in the APOA5-APOA4 intergenic region was just half the overall mean in the local genomic region. Also, just 6 out of the 85 Alu elements in the 200 kb genomic region showed an equal or lower substitution rate than the lowest value in the APOA5-APOA4 intergenic region.

To further locate potential regulatory regions in the APOA5-APOA4 intergenic region, we scanned this region for the presence of binding sites of the insulator protein CTCF. We found nine CTCF binding sites, and most of them (seven out of the nine) were located up to 14 kb from the transcription start of the APOA5 gene. Also, five out of the nine CTCF binding sites were placed within Alu elements. Together, these findings suggest that insulator elements might be located in the first 14 kb of the APOA5-APOA4 intergenic region. Because slow-evolving Alu elements tend to be co-located with the predicted CTCF binding sites, the present results suggest that the same selective constraints are operating in the APOA5-APOA4 intergenic region in both humans and chimpanzees. More important, these findings suggest that the mode of action of the APOC3-enhancer inside the APOA1/C3/A4/A5 gene cluster is conserved at least in great apes.

Functional experiments are required to determine whether Alu elements in the APOA5-APOA4 intergenic region do affect expression of the APOA5 gene by interfering with the action of the APOC3-enhancer as well as the possible mechanisms for the interference. A likely mechanism is that some of the Alu elements in the intergenic region act as silencers of the transcription of the APOA5 gene. This hypothesis is further supported by the presence of CTCF binding sites inside some Alu elements in the intergenic region. In fact, it has been found that Alu sequences are able to act as transcriptional silencers in the WT1 (Hewitt et al. 1995), BRCA (Sharan et al. 1999), human epsilon–globin (Wu et al. 1990), and human growth-hormone (Trujillo et al. 2006) genes among others. It is noteworthy that silencers may act over long distances, for example the silencer of the WT1 gene is located about 12 kb from the promoter (Hewitt et al. 1995), and we found that the most conserved Alu elements in the APOA5-APOA4 were located in between 10- and 14-kb upstream of the APOA5 promoter.

In summary, we found heterogeneity on the substitution rates of the Alu elements in the APOA5-APOA4 intergenic region. The presence of a group of Alu sequences with lower substitution rates (almost threefold lower as compared to the higher substitution group), as well as the occurrence of CTCF binding sites inside Alu sequences, suggests that some of the Alu elements may have functional relevance. If confirmed by further experiments, our results would have major implications in the study of the effect of the APOA5 gene on triglyceride metabolism and risk of coronary heart disease. Most of the research on the regulation of expression the APOA5 has been focused in a few kilobases upstream of the promoter. The present results suggest that elements up to 14 kb from the transcription start of the APOA5 gene may also regulate APOA5-gene expression.