Cyp33 binds AU-rich RNA motifs via an extended interface that competitively disrupts the gene repressive Cyp33-MLL1 interaction in vitro

Cyp33 is an essential human cyclophilin prolyl isomerase that plays myriad roles in splicing and chromatin remodeling. In addition to a canonical cyclophilin (Cyp) domain, Cyp33 contains an RNA-recognition motif (RRM) domain, and RNA-binding triggers proline isomerase activity. One prominent role for Cyp33 is through a direct interaction with the mixed lineage leukemia protein 1 (MLL1, also known as KMT2A) complex, which is a histone methyltransferase that serves as a global regulator of human transcription. MLL activity is regulated by Cyp33, which isomerizes a key proline in the linker between the PHD3 and Bromo domains of MLL1, acting as a switch between gene activation and repression. The direct interaction between MLL1 and Cyp33 is critical, as deletion of the MLL1-PHD3 domain responsible for this interaction results in oncogenesis. The Cyp33 RRM is central to these activities, as it binds both the PHD3 domain and RNA. To better understand how RNA binding drives the action of Cyp33, we performed RNA-SELEX against full-length Cyp33 accompanied by deep sequencing. We have identified an enriched Cyp33 binding motif (AAUAAUAA) broadly represented in the cellular RNA pool as well as tightly binding RNA aptamers with affinities comparable and competitive with the Cyp33 MLL1-PHD3 interaction. RNA binding extends beyond the canonical RRM domain, but not to the Cyp domain, suggesting an indirect mechanism of interaction. NMR chemical shift mapping confirms an overlapping, but not identical, interface on Cyp33 for RNA and PHD3 binding. This finding suggests RNA can disrupt the gene repressive Cyp33-MLL1 complex providing another layer of regulation for chromatin remodeling by MLL1.


47
Proline isomerases are a superfamily of chaperones that, by catalyzing the cis to trans 48 isomerization of the ubiquitous amino acid proline, regulate the folding of proteins that then 49 influence a wide range of biological pathways (1,2). First discovered as the target of the human

65
The RRM domain of Cyp33 specifically interacts with the third plant homeodomain (PHD3) 66 finger of MLL1, contributing to Cyp33 recruitment to and association with the MLL1 complex.

67
The isomerase activity of the cyclophilin domain is involved in rearrangement of MLL1 which 68 exposes the RRM binding interface to allow stable association of Cyp33 with MLL1. (6)(7)(8)14). minutes. The Co-NTA beads were then separated to the side of the tube with a magnetic stand 180 while the supernatant was added to the binding equilibrium reaction.

182
Binding Equilibrium, Washing, and Elution for Selection: Protein at several concentrations (100 183 nM, 500 nM, or 1000 nM) was incubated in 1X SELEX buffer (conditions in Supplementary 184    318 Since most RRMs that bind RNA do so at single-stranded regions of RNA and interact with 4 to 320 6 nucleotides through each RRM (27), we used AptaSUITE (33,34) to determine whether there 321 is any significant enrichment in particular 6-mer sequences through k-mer enrichment analysis.

322
Assessing enrichment in selected sequences over low abundance sequences (n < 10), we 323 found significant enrichment of several overlapping 6-mers, revealing a preference for AU-rich 324 sequences, often seen with several repeats amongst the highly abundant sequence reads 325 (Figure 1C), and consistent with previous work that found Cyp33 preference for PolyA and

326
PolyU RNA (24) as well as modest binding to the AAUAAA sequence (7). Although the precise 327 most prevalent k-mer sequence found varied somewhat (Figure 1), no significant variation in 328 preferred sequence as a function of condition was observed. longer 90-nt initial library pool, with K D s ranging from 6.6 to ~90 µM (  half of the selected aptamer region within SO-1 as opposed to more divergence at the 3' end 384 ( Figure 2B). Likewise, the sequencing depth provides similar data for SO-3 derived mutants, 385 which shows nearly identical conservation of an AU-rich single-stranded region ( Figure 2B).

403
Comparison of these predictions suggests that the AU-rich motifs of these sequences are 404 conserved within single-stranded regions. We then tested whether this minimized single-strand 405 region was necessary and sufficient for tight binding using synthesized oligos of the SO-1 406 singled-stranded AGUGAUAAUUAA and the AAUAAUAA k-mer commonly enriched among the 407 other conditions (Figure 1). We were, however, unable to observe quantifiable binding with 408 these ligands by gel shift. This is consistent with the previously characterized weak binding of As the interaction of the Cyp33-RRM alone with RNA is not sufficient to fully explain RNA 429 binding, we employed NMR to characterize the complete FL-Cyp33 protein-RNA interaction.

430
We used solution NMR to first understand the extent the two domains interact with each other in 431 solution. Despite being 36 kDa, free FL-Cyp33 shows excellent signal-to-noise ( Figure 3A).

432
Comparison of the full-length 1 H-15 N HSQC spectrum to that of the Cyp33-RRM (Figure 3B) 433 and Cyp33-Cyp ( Figure 3C) domains reveals that the sum of the two subdomain spectra largely 434 recapitulate the FL-Cyp33 spectrum (Overlay in Figure 3D) with additional peaks from the    Cyp33-RRM domain binds short RNA and peptide elements competitively (6-8) with 561 overlapping but not identical protein surfaces ( Figure 5B)