Discussion
The importance and functionality of RNP motifs in RBP-RNA interactions are well established in previous studies [9]. Since there is a high sequence similarity between NCL and hnRNP A1 RBDs, and hnRNP A1-RNA interactions are known, a preliminary prediction of putative residues that may be important in NCL-miRNA interactions could be made based on sequence conservation between NCL and hnRNP A1. The RBP-hnRNP A1 interacts with MPC and pri-mir-18 by utilizing aromatic/charged residues found on its RNP motifs as well as on non-RNP beta strands [9]. These aromatic/charged residues show strong conservation between NCL and hnRNP A1 sequences suggesting existence of a similar RNP-mediated mechanism in NCL-RNA interactions (Fig 6A ). This study confirms this prediction through detailed docking analyses in silico . Our results provide three alternative NCL-miRNA binding possibilities with common underlying conserved residues at an equivalent position as in hnRNP A1-RBD taking part in RNA binding (Fig 6C ). Aromatic residues in the RNP motifs are known to be capable of initiating interactions with multiple nucleobases at the same time through stacking interactions and promoting structural stability to the RNA-RNP binding motif [10]. We consistently predicted the involvement of these aromatic residues in all binding modes of miRNA-NCL. Additionally, we frequently predicted some charged residues on different beta strands from the same RBDs to be involved in NCL-miRNA interactions. Several arginine residues (R291, R293, and R298) were consistently predicted to interact with phosphate groups on the nucleotides via salt bridges (e.g., R293 with G16 in Fig 3, and with A34, and U45 in the figures for mir-16, mir-221, and mir-222, respectively). These residues provide additional structural stability to RNA-protein interactions similar to previous studies [9]. Our docking results also revealed that each NCL-RBD interacts with the miRNA duplex structure from opposite sides and forms a clasp around it. Linker regions between RBD3 and RBD4 were consistently predicted in several scenarios to hold the miRNA molecule from an additional third side, thus tightening the NCL-RBDs grip on miRNA (Fig 4-6 ).
Our results indicate that NCL RBDs demonstrate preference or affinity for interaction with certain types of miRNA motifs. It was previously established that NCL prefers interacting with RNA loops while driving ribosomal biogenesis [42]. Non-canonical base pairing is known to lead to kink-turns, bulges, mismatched pairs, and wobble pairs in miRNA structures and presents suitable interaction sites that would be unavailable otherwise. In canonical pairs such as G-C or A-U, amino groups of each base-pair are projected into the major groove, creating a region with positive electrostatic potential [43]. The G-U base-pair is an example of non-canonical base pair, where oxygen groups from both nucleotides face the major groove side instead of the amino groups and leading to negative electrostatic potential in the region of the major groove of the dsRNA [43]. Our results also highlighted the preference of NCL RBDs on non-canonical base pairs when interacting with miRNAs (Supporting Figure S5 ). Such regions are expected to interact with amino acids with positively charged side groups such as arginine or lysine [43]. This ties in with our results as several arginine and lysine residues from NCL RBD3-4 were predicted as some of the most frequently encountered residues in the various docking scenarios obtained.
Additionally, wobble pairs and mismatched pairs are known to be important elements of primary miRNA processing by the MPC [44]. Certain RNA motifs such as UGU/GUG are known to be enriched around the apical loop regions [45] and the preference of NCL to interact with regions close to the apical loops has been demonstrated in previous studies investigating NCL-rRNA [42] and NCL-mRNA [26] interactions. This consensus sequence is also important in pri-miRNA processing by the MPC as DGCR8 is thought to recognize and interact with this consensus sequence [46]. Our results revealed that NCL RBDs were able to recognize this UGU/GUG motif for all miRNA molecules tested in this study (Fig 7 ). In the cases of mir-15a and mir-103a, these regions were identified adjacent to apical loop structures. It is also interesting to note that, in binding mode 2 (observed in mir15a and mir103a), RBD4 by itself seems sufficient to drive NCL-miRNA interactions (Fig 7A ). In the cases of mir-21 and mir-16-1, NCL interacts with regions distant from apical loops, but closer to bulge regions towards the middle of the miRNA molecule. (Fig 7B ). Since DGCR8 also recognizes the same motif for interactions, our results suggest that NCL potentially binds adjacent to DGCR8 or could even replace it in as a possible scenario of miRNA – NCL interactions. Alternatively, in the cases of mir-221 and mir-222, NCL-miRNA interactions were predicted to be localized slightly distant to the apical loop region and closer to the basal stem region containing a mismatched GHG/CUC motif (Fig 7C ). This manifests as a mismatched bulge, which is a common element in most pri-miRNA structures [46]. Drosha cleaves its miRNA targets around the basal stem close to this motif [46]. This cooperation model suggests that NCL RBDs binds closer to Drosha. In both scenarios, these motifs likely serve as anchoring points for NCL RBDs when they interact with miRNA in cooperation with MPC.
Similarly, multiple other RBPs including hnRNP A1, Lin28B, RBFOX3, and HuR interact with either the terminal loop or the stem regions of the pri-miRNA molecules to either promote or suppress pri-miRNA processing by the MPC proteins depending on the location of the interactions and the targeted miRNA [47]. For example, both Lin28B and hnRNP A1 interact with the terminal loop of a subset of pri-miRNA structures. hnRNP A1 is predicted to help with the processing of pri-miR18-a by binding to the terminal loop and causing a relaxation of the miRNA structure and therefore making it easier for MPC to interact with miRNA [48]. However, when hnRNP A1 interacts with let-7 pri-miRNA terminal loop, it outcompetes binding of another RBP, KH-type slicing regulatory protein (KSRP) known to promote biogenesis of let-7 [48] and decreases pri-miRNA processing by Drosha. Lin28B, on the other hand, always negatively regulates this process by interacting with the terminal loop and inhibiting Drosha from interacting with miRNA [49]. Both RBFOX3 and HuR bind to the basal stem and inhibit miRNA processing by blocking catalytic activity of Drosha [50,51]. Based on these studies, it is clear that the relationship of RBPs with the MPC is both location and context dependent.
A recent structural study investigating interactions of the MPC with pri-miR16-2 revealed that 2 DGCR8 proteins interact with nucleotides adjacent to the apical loop region of the pri-miRNA molecule [52]. Canonical function of DGCR8 is to interact with pri-miRNA using its double stranded RNA binding domains (dsRBDs) and present the pri-miRNA molecules to Drosha for processing [52]. This is illustrated inFig 8A. Many RBPs can interact with regions of pri-miRNA molecules that are also regions where DGCR8 proteins bind. We speculate that NCL-RBDs promotes pri-miRNA processing of certain miRNAs by a similar mechanism as observed in hnRNP A1-mir18 interactions. As a RBP capable of binding double stranded RNA molecules, NCL could potentially replace the pri-miRNA presenting functions of one or both DGCR8 proteins. Our results suggest that for certain miRNA, NCL can wrap around the double stranded miRNA molecule with two RBDs like DGCR8 (Fig 8B ). Since the lengths of oncogenic miRNA transcripts are variable, we speculate that NCL may act as a bridging agent between DGCR8 and Drosha when processing longer transcripts (Fig 8C ). Our findings present a snapshot of NCL-miRNA that give an initial insight into these interactions. We envision future studies to elaborate the NCL-miRNA interaction dynamics over longer time-scales to get a finer grained picture of the underlying molecular mechanisms and their downstream effects.
Drosha and DGCR8 interact with different RNA types including precursor-mRNA (pre-mRNA) and non-coding RNA and are also involved in double stranded DNA break repair mechanisms [53]. Since NCL and MPC proteins can interact in a non-miRNA context [21], it is quite likely that they cooperate in other biological pathways as well. Studies testing NCL- MPC cooperation in pre-mRNA processing may be a natural extension of this study since NCL is also known to interact with mRNA UTR molecules to manipulate their expression levels in human cancers [20,25,26].
NCL cellular localization is highly complex and context-dependent. Pathophysiological functional implications of NCL are even more enigmatic. Despite this multidimensional behavior, NCL is a promising target for cancer therapeutics. Therapeutics such as the immuno-agents 4LB5-HP-RNase [54], G-rich DNA oligonucleotide, aptamer AS1411 [55], and antagonist pseudopeptides, N6L [56] and HB19 [57], target either surface or cytoplasmic NCL to regulate its functions in miRNA synthesis, RNA metabolism, cell proliferation, angiogenesis and metastasis, in a variety of cancer types including breast cancer.
In summary, we predict two putative binding modes where NCL RBD3-4 specifically drive NCL-miRNA interactions and an alternative binding mode where a small contribution from RBD12 is also involved. As we had hypothesized RNP motifs on RBD3-4 play a significant role in these interactions; additionally, we report novel residues important for the interaction that have not been identified in any other previous study. Importantly, our data support an idea that the exclusive presence of NCL RBD3-4 in animals might be evolutionarily relevant for NCL functions as drivers for microRNA processing. Our data delineate critical residues in NCL and recognition motifs in the miRNA important for NCL-miRNA interactions. Future studies designed to validate critical residues in these motifs that are important for interactions, using site-directed mutagenesis in cellular conditions will be required. Once confirmed experimentally, NCL-miRNA interacting interface provide a valuable drug targets for the development of cancer therapies to control specific gene expression.