Interphase 3: Why UV-crosslinking?
XRNAX and its downstream applications have been optimized for UV-crosslinked cells grown in monolayers. However, we expect XRNAX to work in any system where UV-crosslinking and TRIZOL extraction has been successfully applied. Both methods have been widely applied from studies on entire organisms, such as in C.elegans, Drosophila or plant seedlings, to tissue samples, yeast or bacteria.
Although UV-crosslinking has been around since the early days of molecular biology and has been very popular in the RNA field, its underlying chemistry is still very poorly understood. For example, the exact photo-adducts between the amino acids and the ribonucleotides remain unknown. In fact, it is still very much unclear which pairs of ribonucleotide and amino acid can crosslink upon UV-irradiation. As presented in our paper and as reported previously by Kramer et al., MS analysis of peptide-ribonucleotide adducts suggests that primarily uracil crosslinks to protein. However, it remains unclear if the other ribonucleotides do not crosslink properly or just escape MS detection. For a very detailed summary of the literature on UV-crosslinking visit Kendric Smith’s and Martin Shetlar’s website, who are the true forefathers to this field. At any rate, UV-crosslinking has some features, which are conceptually important to XRNAX.
First, UV crosslinks protein to RNA only if they are in immediate contact (referred to as ‘zero-length crosslinking’). RNA-binding proteins are in close proximity to RNA and will therefore be crosslinked. Proteins in close proximity to RNA, which do not specifically interact with RNA, are expected to wiggle around too much when UV-light activates the RNA bases so that crosslinking becomes inefficient. This makes a strong case for proteins UV-crosslinked to RNA being RNA-binding. However, it leaves open the door for a whole number of controversial cases where proteins are crosslinked because they are held into place by a true RNA-binding protein (bystander crosslinking), because their concentration is very high, because they become part of a cellular compartment containing a lot of RNA… The discussion becomes here somewhat philosophical when you start asking what does ‘RNA-binding’ mean, anyway. A good borderline case to illustrate this are G-rich intrinsically disordered regions (IDRs), which have been shown to bind RNA and have been implicated in the formation of phase-separated, membrane-less organelles, some of which require RNA to form. In this case ‘RNA-binding’ means probably something different in comparison to what ‘RNA-binding’ means in the case of a classical globular RNA-binding domain such as the RRM. UV will crosslink either one of them.
Second, UV-crosslinking only crosslinks protein and RNA. The putative mechanism is that UV-light will activate RNA bases, which then react with protein in the vicinity or with water. DNA also crosslinks but only at energy doses 10-100 times higher. Why that is remains unclear, but reasons could be because of the orientation of the bases towards each other, or simply because of the lack of uracil.
Third, crosslinking is sparse, because only a low percentage of proteins bound to RNA will crosslink successfully. How high this percentage is probably depends on the protein and its specific interface with the RNA it binds to. Some in vitro studies on individual RNA-binding proteins report UV-crosslinking of more than 50 % of the RNA-binding protein in solution. For most proteins crosslinking efficiency is expected to be a single digit percentage, however, comprehensive and unbiased in vivo data on this doesn’t exist.
Our current model for protein-crosslinked RNA in XRNAX extracts arising from these basic properties of UV-crosslinking is that most of it is protein-free RNA, sporadically interrupted by protein crosslinks. Evidence for that comes from the fact that RNA in XRNAX extracts can be completely digested with RNase, meaning that there are no longer stretches of RNA protected by protein. Network formation between proteins and several pieces of RNA probably exists. However, the connectivity is low because to connect two pieces of RNA, or make a loop into the same piece of RNA, proteins need to score several successful crosslinking events, which becomes more and more unlikely the higher the connectivity. This model is supported by the observation that predigestion of XRNAX extracts improves silica enrichment (see Applications section of this website). Predigestion of weakly connected protein-RNA networks with trypsin might open up larger parts of free RNA, which interact better with silica columns and improve co-purification of protein crosslinked to RNA.