*3.9. Correlations and Influence of 5*′*UTR Features on Protein Expression*

We have demonstrated a great variability in the numbers of the individual 5′UTR features among the genes of the ABCA subfamily. Considering phylogenetic relationships among ABCA genes, there is no clear pattern in these numbers. Some positive correlations among 5′UTR features were found to be statistically significant. Some of these correlations are expectable, such as the correlations among 5′UTR length and No. of uATGs, No. of 5′UTR introns and the no. of stem loops, some are interesting and need further exploration, such as the correlations between the no. of uATGs and the no. of 5′UTR introns or sATG flanking sequence context and the presence of RG4-forming sequence. Notably, the variable uATG TIS score correlated positively, however quite weakly, with uATG position (from transcription start site) and uATG flanking sequence context (from NetStart software). In relation to the canonical cap-dependent translation initiation, a possible influence of the ATG position on the ATG context

has been mentioned in several works. Rogozin et al. [23] reported a significant negative correlation between the sATG information content and 5′UTR length for several species. They also concluded that this correlation could be explained by the strong positive correlation between the number of uATGs and the length of the UTR. Lynch and coworkers [27] described a strong distance-dependent gradient of the deficit of uAUGs. Since uATG TIS score and uATG flanking sequence context should be based on the same criteria, we would expect a stronger correlation. However, this correlation was not studied by any other study for comparison. The length of a 5′UTR, which is determined mainly by stochastic events, seems to be the major factor influencing the numbers of the other regulatory features.

Although there is also great variability in the distribution and expression of the ABCA proteins, we did not find any significant correlation that would support a clear connection between the numbers of 5′UTR features and protein expression characteristics. Among the genes which have the smallest numbers of 5′UTR features are *ABCA13*, *ABCA4*, *ABCA2*, and *ABCA5*. *ABCA13* is the least-explored member of the subfamily and the information about its protein distribution and expression in physiological conditions is missing. ABCA4 protein was reported to be expressed only in the retina and the expression level is high. ABCA2 and ABCA5 were found to be expressed in many and all human tissues with mainly medium and high levels, respectively. On the other side, among the genes which show the highest numbers of 5′UTR features are *ABCA10*, *ABCA3*, *ABCA8*, and *ABCA12*. ABCA10, ABCA3, and ABCA12 proteins are all expressed in many human tissues with mainly medium levels. The ABCA8 protein was detected only in several tissues with low expression. In the set of normal human tissues; however, the protein expression levels of the ABCA proteins ranged from low to high for all members, with the exception of ABCA5 (medium to high) and ABCA10 (low to medium). Our results therefore support the view that the great variability in 5′UTR features prepares a complex playground where the other elements such as RNA binding proteins and non-coding RNAs play the major role in the fine-tuning of final protein expressions. Notably, tissue-specific translation repression by miRNAs through binding to uAUGs was demonstrated in Ajay et al. [53].
