Substrate specificity of human MCPIP1 endoribonuclease

MCPIP1, also known as Regnase-1, is a ribonuclease crucial for regulation of stability of transcripts related to inflammatory processes. Here, we report that MCPIP1 acts as an endonuclease by degrading several stem-loop RNA structures and single-stranded RNAs. Our studies revealed cleavage sites present in the stem-loops derived from the 3′ untranslated region of the interleukin-6 transcript. Furthermore, MCPIP1 induced endonuclease cleavage at the loop motif of stem-loop structures. Additionally, we observed that MCPIP1 could cleave single-stranded RNA fragments. However, RNA substrates shorter than 6 nucleotides were not further affected by MCPIP1 nucleolytic activity. In this study, we also determined the dissociation constants of full-length MCPIP1D141N and its ribonuclease domain PIN D141N with twelve oligonucleotides substrates. The equilibrium binding constants (Kd) for MCPIP1D141N and the RNA targets were approximately 10 nM. Interestingly, we observed that the presence of a zinc finger in the PIN domain increases the affinity of this protein fragment to 25-nucleotide-long stem-loop RNA but not to shorter ones. Furthermore, size exclusion chromatography of the MCPIP1 and PIN proteins suggested that MCPIP1 undergoes homooligomerization during interaction with RNA substrates. Our results provide insight into the mechanism of MCPIP1 substrate recognition and its affinity towards various oligonucleotides.

Because the activity of MCPIP1 is dependent on the presence of Mg 2+ or Mn 2+ metal ions, all degradation assays were performed in buffer with the divalent cation Mg 2+ . Additionally, to decrease non-specific electrostatic  Table 1. Nt sequences of fluorescently modified oligonucleotides used for the RNase assays and affinity determination assays. Nts that form loop fragments of stem-loop structures are underlined. Sequences with numbered residues are part of the 3′UTR of transcripts from mouse or human IL-6. These sequences were numbered such that the first nt after the stop codon of the coding sequence is marked as 0. RS -reverse stem modification of mIL-6 82-106 (altered nts are in bold). YR -purine and pyrimidine residue modification of mIL-6 82-106 (altered nts are in bold) .
SCientiFiC RepoRts | (2018) 8 interactions between MCPIP1 and nucleic acids, the cleavage studies were performed at physiological salt concentration (150 mM NaCl). The observed MCPIP1 ribonuclease activity products were reproducibly consistent for proteins obtained from different batches. In each case, the RNA cleavage assay was carried out for 30, 60, 120, 180 and 240 minutes. We initially performed an RNase assay of the mIL-6 82-106 stem-loop structure. We observed that MCPIP1 WT induced degradation starting from the 3′ end of the mIL-6 82-106 5′FAM, and MCPIP1 WT cleaved the 25 th single nt as the first one. Then, the 24 th nt was cleaved (Fig. 1A). Simultaneously, mIL-6 82-106 5′FAM stem-loop cleavage occurred at the loop site, between the C10 and U11 nts (Fig. 1A). In the consequence, a 10 nt single-stranded RNA fragment was generated from the 5′ end of the mIL-6 82-106 5′FAM stem-loop structure. Next, additional processive degradation of the nascent 10-nt-long ssRNA was observed (Fig. 1A). However, MCPIP1-induced degradation was not observed for ssRNA fragments consisting of 6 nt (Fig. 1A).
Next, to verify the stereospecificity of MCPIP1-induced cleavage, we reversed the sequence of the mIL-6 82-106 at the stem site of this stem-loop. Surprisingly, after reversing the stem sequences (mIL-6 82-106 RS oligonucleotide), we observed a single nt product induced by MCPIP1 WT activity, indicating that enzymatic hydrolysis occurred between first (A) nt and the second (C) nt (Fig. 1B). Thus, the reverse stem sequence (mIL-6 82-106 RS) was cleaved between the same nts as the basic mIL-6 82-106 stem-loop, in which MCPIP WT triggered enzymatic hydrolysis between A 25 and C 24 nt in the vicinity of the 3′ end of the oligonucleotide (Fig. 1A,B). Moreover, accumulation of the 7-nt-long degradation product of the mIL-6 82-106 RS oligonucleotide indicated that MCPIP1 WT introduced cleavage between the A 7 and A 8 nts (Fig. 1B).
We also examined whether the analysis of oligonucleotide degradation was affected by potential E. coli contaminants remaining from the protein purification procedure. Therefore, we analyzed oligonucleotide cleavage induced by MCPIP1 D141N with a substitution of the conserved aspartate at the catalytic center of the PIN domain. No ribonucleases activity of the MCPIP1 D141N was observed for the mIL-6 82-106 RS stem-loop oligonucleotide (Fig. 1B,E). However, low-efficacy nuclease activity of MCPIP1 D141N was observed for mIL-6 82-106 5′FAM, as shown in Fig. 1A. MCPIP1 D141N induced cleavage occurred only at 3′ end of this oligonucleotide. Thus, the D141N mutation does not completely abolish in vitro enzymatic activity of MCPIP1.
The characteristic feature of the unmodified mIL-6 82-106 stem-loop is a high presence of pyrimidine residues at the 5′ site of the stem. Therefore, to assess the role of this characteristic pattern, we modified the stem sequence to achieve balanced distribution of the purine and pyrimidine residues at the stem site of this stem-loop. Nts 6-9 and 16-19 were changed in mIL-6 82-106 YR (Table 1 and Fig. 1C). Our results showed that MCPIP1 WT -induced degradation of the mIL-6 82-106 YR occurs at the same time at the loop site of the stem-loop structure or at the 3′ end of the stem-loop structure (Fig. 1C). Furthermore, we determined that after destabilization of the mIL-6 82-106 YR stem loop structure through loop cleavage induced by MCPIP1 WT , the 10-nt-long ssRNA was increased and subsequently processively degraded (Fig. 1C). Additionally, we observed that degradation of mIL-6 82-106 YR stops at the fragment consisting 6 nt, similar to the degradation of the unmodified mIL-6 82-106 5′FAM oligonucleotide (Fig. 1A,C). Alteration of purine with pyrimidines (mIL-6 82-106 YR) did not change the cleavage sites in the stem loop structure, and degradation was triggered as in the case of mIL-6 82-106 5′FAM (Fig. 1A,C). Thus, we concluded that MCPIP1 WT -induced in vitro degradation is not dependent on stem sequence of the stem-loop.
To avoid negative results due to diminished cleavage susceptibility of sites where nts are modified by fluorescent labeling, we labeled the mIL-6 82-106 stem-loop structure at the 5′ end or at 3′ end (mIL-6 82-106 5′FAM, mIL-6 82-106 3′FAM, respectively). For the mIL-6 82-106 3′FAM sequence, we observed a 1-nt-long degradation product; therefore, the first cleavage induced by MCPIP1 WT occurs between C 24 and A 25 nts as shown in Fig. 1D. Thus, cleavage between C 24 and A 25 was observed for both the mIL-6 82-106 5′FAM and mIL-6 82-106 3′FAM sequences. However, for the 3′FAM-labeled oligonucleotide, the degradation was less efficient. Comparison of Fig. 1A,D suggests that the presence of fluorescent dye on a cleaved nt does not significantly affect the MCPIP1 activity.
Kinetics of oligonucleotide degradation triggered by MCPIP1 depends on many factors. In the next step, we examined whether the stem-loop structure, oligonucleotide length or nucleotide sequence affected MCPIP1 nucleolytic efficiency. To verify the impact of different stem-loop structures and sequences on MCPIP1-triggered cleavage, we used a set of short stem-loops consisting of 17 or 18 nts (mIL-6 85-101 , hIL-6 82-99 , consensus stem-loop) ( Table 1). We observed that MCPIP1 WT induced cleavage of two nts from the 3′ end of these oligonucleotides and also accumulation of bands that are 10, 9 and 7 nt long ( Fig. 2A). These findings indicated that MCPIP1 WT introduces endonucleolytic cleavage in the loop region of those short stem-loops (mIL-6 85-101 , hIL-6 82-99 , consensus stem-loop). Initial MCPIP1 WT -induced enzymatic hydrolysis occurs simultaneously for four phosphodiester bonds between 9-12 nt at the loop motif of the mIL-6 85-101 sequence ( Fig. 2A). We determined that the pattern of loop cleavage of the mIL-6 85-101 stem-loop is different than that for mIL-6 82-106 5′FAM, which was cut between the C10 and U11 nts (Fig. 1A). Therefore, the stem length of the stem-loop affects the cleavage sites recognized by MCPIP1.
To verify the influence of size of high-order RNA backbone structures on the oligonucleotide cleavage rates, we performed kinetic analysis. The kinetics of oligonucleotide degradation are shown as the level of uncleaved oligonucleotides obtained from densitometric analysis of the results from oligonucleotide degradation assays. In the subsequent time points of the RNase assay, uncleaved mIL-6 85-101 oligonucleotide levels were significantly decreased compared to uncleaved levels of the mIL-6 82-106 oligonucleotides (Fig. 2C). We observed that MCPIP1 WT -triggered cleavage of the mIL-6 85-101 stem-loop was relatively faster than that of the hIL-6 82-99 stem-loop, which possesses longer stems ( Fig. 2A). These results showed that the kinetics of degradation of RNA stem-loop structures containing short stems is faster than that of stem-loops possessing longer stems. We concluded that unwinding of shorter stems from the stem-loop structures results in a more efficient degradation (Figs 1A and 2A,C). To determine the importance of loop fragments in MCPIP1-triggered stem-loop cleavage, we compared stem-loops that contain a 3, 4 or 6 nt long loop motif. However, we did not observe major differences in MCPIP1-induced degradation of these oligonucleotides ( Fig. 2A).
Subsequently, we assessed whether MCPIP1 WT degrades unstructured ssRNA. After MCPIP1 WT -triggered destabilization of stem-loop structures, a subsequent cleavage occurred in the nascent ssRNA. Thus, we examined 12-nt-long ssRNA oligonucleotides from the mIL-6 82-93 and 7-nt-long mIL-6 82-88 ssRNA (Fig. 2B). Using the RNA folding software mFOLD 26 , we confirmed that the mIL-6 82-93 and mIL-6 82-88 sequences did not show base pairing interactions at room temperature; thus, they do not fold into stable secondary structures. We noticed that for the unstructured ssRNA, the rate of MCPIP1 WT -induced degradation was increased compared to cleavage of the stem-loop sequences (Fig. 2D). Degradation of either mIL-6 82-93 or mIL-6 82-88 indicated that ssRNAs shorter than 6 ribonucleotides were not efficiently cleaved by MCPIP1 WT (Fig. 2B). The levels of shortened oligonucleotides formed as a result of the MCPIP1 WT induced cleavage of the mIL-6 82-106 5′FAM indicated high increase of the 6 nt long truncated oligonucleotide (Fig. 1F). Furthermore, we observed 11-fold increase of the level of 6 nt product of the MCPIP1 WT induced cleavage of the mIL-6 82-93 (Fig. 2F). Moreover, there was marginal catalytic activity of MCPIP1 D141N for ssRNA, which presented as cleavage of two nts from the 3′ end of the mIL-6 82-93 oligonucleotide (Fig. 2B,E). To verify the sequence specificity of ssRNA cleavage, we performed degradation assays using poly-U sequences. However, it appeared that MCPIP1 WT processively cleaved the poly-U homopolymer, and the oligonucleotide degradation stopped when the fragment consisted of 6 nt (Fig. 2B). These results indicated that MCPIP1 WT cleaves unstructured ssRNA in a sequence-independent manner.
We next investigated whether MCPIP1 exhibits RNA substrate specificity. Therefore, in RNase cleavage assays, we used single-stranded and double-stranded DNA (ssDNA and dsDNA) as a substrate. These DNA sequences were similar to RNA sequences consisting of 12 nts present in the mIL-6 82-93 oligonucleotide. We observed that MCPIP1 WT cleaves both ssDNA and dsDNA (Fig. 3A). Degradation of these sequences occurred from the 3′ end; however, the kinetics of these processes was lower compared with the cleavage of mIL-6 82-93 ssRNA (Figs 2B,E and 3A,B). Degradation of the mIL-6 82-93 ssDNA had approximately equal efficiency using either MCPIP1 WT or MCPIP1 D141N (Fig. 3A,B). Therefore, the aspartate 141 residue of MCPIP1 is crucial for RNA cleavage but not for DNA processing (Figs 1A,B, 2B and 3A,B).
We showed that MCPIP1 D141N does not possess activity against mIL-6 82-106 RS (Fig. 1B). However, we have observed that MCPIP1 D141N possesses low nuclease activity in some of the investigated systems (Figs 1A, 2B and 3A). Therefore, to further confirmation that presented RNA cleavage assay is not affected by contaminations from E. coli extract we used another control which is MCPIP1 438-599 protein deprived of PIN nuclease domain. Applying MCPIP1 438-599 to RNase assay we did not observe degradation of investigated oligonucleotides ( Supplementary Fig. S2E). Thus, our results are not affected by contaminations and we conclude that single mutation D141N of MCPIP1 is not sufficient to completely abolish in vitro MCPIP1 nuclease activity. All identified cleavage sites observed in degradation assays are listed in Supplementary Table S1. To verify the sequence specificity of MCPIP1-triggered degradation of RNA, we presented the identified sites of cleavage as a consensus logo ( Supplementary Fig. S3). For logotype preparation, we used the sequence logo generator software WebLogo 27 . We figured out that cleavage sites lacking G nts in the immediate vicinity of the cut site were preferable for MCPIP1-induced cleavage ( Supplementary Fig. S3).
To further confirmation of our observation about in vitro nonspecific cleavage of RNA oligonucleotides by MCPIP1 WT we performed additional experiments. We checked whether MCPIP1 WT might cleave the template which were previously reported at in vivo studies as not degraded by MCPIP1 nuclease activity. The fragment comprising 1-81 nt from the mIL-6 3′UTR is not regulated through MCPIP1 activity in cells studies 5  part of the mIL-6 3′UTR contains the mIL-6 82-106 3′UTR stem loop which is the putative element responsible for IL-6 transcripts destabilization through MCPIP1 nuclease activity. Due to limitations of synthesis methodology we used 45 nts long sequence from the mIL-6 1-45 3′UTR (Table 1 and Supplementary Fig. S2F). Using RNA folding software mFOLD 26 we showed that the mIL-6 1-45 oligonucleotide possibly forms two stem-loops structures as shown at Supplementary Fig. S2F. Degradation of the mIL-6 1-45 3′UTR clearly indicates endonuclease activity of MCPIP1 WT . The fragment of the mIL-6 1-45 tends to form two stem-loop secondary structures, thus observed cleavage induced by MCPIP1 WT should be introduced at loop site of these stem loops. Indeed we observed endonuclease cleavage of the mIL-6 1-45 at both loop sites ( Supplementary Fig. S2F). However, due to obtained low electrophoresis resolution we could not precisely describe the exact nucleotides between which cleavage takes place ( Supplementary Fig. S2F).

Dissociation constants of the MCPIP1 complex with oligonucleotides.
Our results from oligonucleotide degradation assays did not reveal a strong structural or sequence preference of in vitro RNA cleavage by recombinant MCPIP1 WT . However, we determined that single-stranded RNA or 17-nt-long stem-loops were cleaved with a faster rate than 25-nt-long stem loops. For that reason, we investigated whether there were any differences in MCPIP1 D141N affinity for the tested oligonucleotides. We used FAM-labeled oligonucleotides to develop a method for determination of the MCPIP1 D141N affinity to oligonucleotides. Previously, we used electrophoretic mobility shift assays (EMSAs) to show that MCPIP1 D141N has the potential to form stable complexes with 3′UTR fragments of the C/EBPβ transcript and obtained complexes possessing two distinct quaternary structures 21 . Observed shifts at EMSA assay indicated that the marginal nuclease activity of MCPIP1 D141N did not repress formation of the nucleoprotein complex. Estimated binding affinities of the complexes of MCPIP1 D141N with RNA based on our results published previously by Lipert et al. were between 640-1580 nM (Supplementary  Table S2) 21 . The obtained Kd varies from previously used the 3′UTR sequence fragments of the C/EBPβ transcript. However, in our previous EMSA experiments, we were not able to determine the equilibrium dissociation constants of the achieved complexes. Herein, we determined the apparent equilibrium dissociation constants of the human MCPIP1 D141N complexes with different types of oligonucleotides: stem-loop RNA, ssRNA, ssDNA and dsDNA (Fig. 4B,  Supplementary Fig. S4 and Table 2). The slopes of dose-response curves were very steep for protein concentration values between 10 nM and 100 nM. Amplitudes of the fluorescence signals were changed approximately 2 times depending on the sequence. Fluorescence polarization assay which is commonly used for affinity determination might be affected by high fluorescence intensity changes observed in our measurements. Thus, we decided to analyze only fluorescence intensity signal which also gives us better residuals of obtained fits. The two-phase course of fluorescence intensity changes, observed in all investigated cases, prompted us to model the affinity data with a complex double equilibrium binding equation where N + P + P  NP + P  NPP (N -oligonucleotide, P -protein) ( Fig. 4B and Supplementary Fig. S4). These results showed that two protein molecules sequentially bind to a single oligonucleotide. The best model describing our data was a sequential binding model with equal equilibrium-binding dissociation constants, K d1 = K d2 . We also analyzed the model characterized by K d1 ≠K d2 , which was rejected because it inconsiderably improved the residual distribution, however, the standard deviations of the calculated Kd were higher than those for the model with Kd 1 = Kd 2 . A third analyzed model with a single equilibrium constant, N + P + P  NPP, was rejected because it had the highest residuals of curves that were fitted to the measured data points. The final selected model (sequential binding analysis, K d1 = K d2 ) reflects the measured data well, and the obtained K d are shown in Table 2. This model was characterized by the lowest standard deviation of the obtained dissociation constant values and low residuals of the fitted curves.
We observed that for the set of investigated oligonucleotides comprising stem-loop structures, ssRNA, and ssDNA, we did not find major differences between dissociation constants of the complexes with MCPIP1 D141N (Table 2). Therefore, MCPIP1 D141N can efficiently bind diverse oligonucleotide sequences. Minor differences in the MCPIP1 D141N affinity to stem-loop structures or single-stranded oligonucleotides suggest that the nucleic acid double-stranded helical structure is not necessary to interact with MCPIP1. Additionally, we observed that MCPIP1 has lower affinity to dsDNA comparing to other investigated nts ( Table 2). We showed that the affinity of full-length MCPIP1 D141N to oligonucleotides is significantly higher than that for fragments of this protein represented only by the nuclease domain (PIN D141N ) (Fig. 4A,B, Supplementary Fig. S4 and Table 2). Moreover, we noticed that the zinc finger domain increased the affinity of the PIN D141N subunit to 25-nt-long oligonucleotides but not to shorter oligonucleotides (Fig. 4A,B, Supplementary Fig. S4 and Table 2).
Binding assays using free FAM dye did not showed significant changes of fluorescence intensity at investigated systems (Fig. 4C). MCPIP1 D141N and buffer condition did not affect fluorescence emission of the free FAM label. The unlabeled hIL-6 81-98 RNA oligonucleotide did not affect fluorescence emission of free FAM label in the presence of MCPIP1 D141N (Fig. 4C).Thus, we assume that described interactions are the effect of the assembly of the MCPIP D141N complex with oligonucleotides. The shape of fluorescence spectra of the FAM labeled oligonucleotides were consistent for all examined MCPIP D141N concentrations ( Supplementary Fig. S4A). Fluorescence intensity of the FAM labeled oligonucleotides were changed due to MCPIP1 D141N nucleoprotein complex formation which affected FAM fluorescence probe (Supplementary Fig. S4A).
Observed dissociation constants for MCPIP1 D141N complexes with mIL-6 82-106 5′FAM were substantially weaker for EMSA system than in fluorescence based assay (Table 2, Supplementary Fig. S5 and Supplementary  Table S2). We suppose that differences in dissociations constants are the results of the complex binding kinetics of MCPIP1 interaction with RNA that possibly is characterized by relatively fast k off rates. In case of high k off the EMSA as a non-equilibrium method will give higher dissociation constants compared to equilibrium techniques. The EMSA shift for mIL-6 82-93 ssRNA and mIL-6 82-93 ssDNA were observable at a relatively low concentration of MCPIP D141N (400 nM) (Supplementary Fig. S5A) although, at higher concentration of the MCPIP1 D141N the oligonucleotides were not completely bounded in nucleoprotein complex. Therefore, we didn't calculate the Kd   Fig. S5A).

Homooligomerization of MCPIP1.
Interestingly, the two-phase course of fluorescence intensity changes was observed in the obtained affinity assay graphs during oligonucleotide-binding processes (Fig. 4B, Supplementary Fig. S4). Thus, two protein molecules sequentially bind to a single RNA molecule. According to other studies, PIN domain superfamily proteins are frequently described as oligomers: dimers or tetramers 28 . Therefore, to obtain precise data of the MCPIP1 protein oligomerization state, we analyzed protein size exclusion chromatography results. In both cases, single Gaussian peaks were observed, indicating the monodispersity of the analyzed protein fragments. Analysis of protein size based on the retention volume (Fig. 5A,C) indicates that both the PIN and PIN-ZF domains were in a monomeric state. The mouse PIN domain was previously suggested to be a dimer 19 . In contrast, for full-length MCPIP1 WT and MCPIP1 D141N , we observed wide elution peaks that shifted in favor of possible oligomeric forms (Fig. 5A). To assess these elution profiles, we performed multiple Gaussian peak fit analyses (Fig. 5B). Comparing the obtained maxima of fitted peaks with the column calibration curve, we observed that the calculated molecular masses of the fractions corresponded to tetrameric, dimeric and monomeric forms of MCPIP1 WT (Fig. 5C,D). Therefore, the full-length MCPIP1 WT or MCPIP1 D141N is most frequently present as a dimer in native conditions, however, tetrameric and monomeric fractions were also present in solution (Fig. 5A-D). Buffer supplementation with 1.6 M urea resulted in narrowing peaks in size exclusion chromatography, indicating the additional increase of the MCPIP1 dimers fraction in the sample (Fig. 5A-D). Native PAGE electrophoresis of MCPIP1 revealed two heterogeneous bands for MCPIP1 WT (Fig. 5E). These results confirmed that MCPIP1 homooligomerization occurs in native conditions. Equilibrium between the dimeric and tetrameric states of MCPIP1 could be influenced by changing the buffer composition. After addition of urea to the native PAGE buffer, the equilibrium was shifted in favor of the MCPIP1 tetrameric fraction (Fig. 5D). Urea as a chaotropic agent alternates hydrogen bonding between water molecules and proteins, and therefore affect hydrophobic interactions 29 . The small concentration of urea or other commonly used denaturation agent: guanidine hydrochloride increases stability of some proteins and might increase reactions rates 30 . Urea concentration (1.6 M) used in our experiment has a much lower concentration than typically is used for denaturation of proteins, however, the concentration of this osmolyte is sufficient to significantly change the content of bulk water 29 . Urea can substantially influence the polypeptides solvation, increasing protein solvent accessible area which might consequently lead to conformation changes of MCPIP1 and probably its might affect oligomerization 30 . Although, this indirect mechanism appears to be the most likely in our case, there are reports indicating a possible alternative mechanism, in which the urea molecules directly interacts with protein molecules in a divalent manner 31 . The two distinct quaternary structure of MCPIP1 D141N nucleoprotein complexes were also observed in EMSA, using long UTRs fragments as well as single stem loop of the mIL-6 82-106 , as shown by Lipert at al. in Supplementary Fig. 5 21 . Together, the size exclusion chromatography and native PAGE results indicate that the both dimeric and tetrameric forms of MCPIP1 homooligomers were found in the investigated conditions.

Discussion
Previous studies indicated that MCPIP1 is a selective ribonuclease that cleaves translationally active mRNA at the 3′UTR 5 . To determine how MCPIP1 recognizes the unique molecular targets that were reported in biological systems, we applied in vitro analysis using recombinant MCPIP1. We identified MCPIP1 as an endoribonuclease that degrades diverse sets of RNA stem-loop structures. Collectively, our data did not indicate a strong structural or sequence preference during in vitro cleavage of the RNA stem-loops, as all investigated sequences were affected by MCPIP1. However, our results revealed that unstructured single-stranded RNA is highly prone to cleavage by MCPIP1 WT . We observed that MCPIP1 WT -induced degradation of the mIL-6 82-106 stem-loop starts from the 3′ end of the sequence. Simultaneously, cleavage occurs at the loop site of the stem loop. As a consequence of loop cleavage, the stem-loop structure is destabilized, and ssRNA fragments are generated, which are further processively degraded in the next step (Fig. 6A). Surprisingly, we observed that 6-nt-long ssRNA was not rapidly cleaved by MCPIP1 WT . A possible explanation for this process might be that the 6 nt RNA substrate is too short to reach the nuclease site. We hypothesized that the region of MCPIP1 that is crucial for RNA binding must be proximal to the catalytic cleft in the PIN domain since the 7-nt-long mIL-6 82-88 substrate was still bound with high affinity to the PIN D141N domain. We hypothesized that the positively charged region that is present in the structure of the MCPIP1 PIN domain is essential in the RNA recognition results showed that PIN and PIN-ZF were monomeric and suggest that full-length MCPIP1 most frequently occurs as a dimer in native condition. Stoichiometry of the MCPIP1 -RNA interaction was based on the size exclusion chromatography results and the results from affinity determination assays where the sequential binding model were used. Thus, for full-length MCPIP1, we proposed a sequential binding model: oligo + MCPIP1 Dimer + MCPIP1 dimer  oligo-MCPIP1 dimer + MCPIP1 dimer  oligo-MCPIP1 tetramer . The presented dissociations constants of the complexes were estimated based on the affinity determination assays shown in Table 2. process and protects bound fragments of short ssRNA from further cleavage. Preferential cleavage of oligoribonucleotides triggered by MCPIP1 was observed for sequences lacking a G nt at positions −1 and +1 of the cleavage site, however, this observation might be affected by the low complexity of the analyzed sequences.
Degradation of 3′FAM-labeled oligonucleotides indicated that introduction of the fluorescent label did not disable the recognition of the cleavage sites by MCPIP1 WT . However, for 3′FAM-labeled oligonucleotides, we observed a decrease in the nucleolytic efficiency of MCPIP1 WT . Moreover, we observed that MCPIP1 cleaved poly-U ssRNA oligonucleotides in a processive manner. These findings may suggest 3′ to 5′ exonuclease activity of MCPIP1 against ssRNA. Additionally, previous reports revealed that the MCPIP1 PIN domain shares high structural homology with the T5 D15 5′-exonuclease 11,32 . Nevertheless, the successive exonuclease degradation of single-stranded RNA by MCPIP1 is not relevant in vivo due to the low rate of observed 3′ to 5′ exonuclease cleavage activity. Moreover, in vitro endonuclease activity of recombinant MCPIP1 had a strong background as shown in the results of cleavage of the loop sites of the investigated stem-loops. Additionally, previous results have also shown degradation of longer transcripts, such as IL-6, IL-8 or CEBPβ, which indicated that preferable sites of endonucleolytic cleavage are present in these transcripts 5,6,21 . The endonuclease activity of recombinant MCPIP1 was also confirmed from in vitro degradation of circular RNA fragments 5 .
We hypothesized that the marginal in vitro activity observed here of either MCPIP1 WT or MCPIP1 D141N towards DNA molecules is irrelevant in vivo, since MCPIP1 has a primary cytoplasmic localization and should be considered as a ribonuclease. In contrast, the EndoV nuclease efficiently cleaves both RNA and DNA substrates 33 . There are also evidences of nuclear localization of the MCPIP1 for which essential is nuclear localization signal (RKKP) that is present in amino acid sequence of the MCPIP1 34 . However, observed here DNA cleavage was inefficient thus we estimate DNAse activity of MCPIP1 as biologically insignificant.
Ribonucleases possessing PIN domains usually lack strong sequence specificity in in vitro studies with recombinant proteins 13 . However, protein engineering can modify the specificity of these RNases. One example is the engineered PIN-PUF nuclease that possesses a high sequence specificity of RNA degradation 35 . Most likely, modification of the PIN domain from MCPIP1 will enhance its specificity and will be beneficial for the development of a highly sequence-specific molecular tool.
To the best of our knowledge, the equilibrium dissociation constants of the complex of MCPIP1 with oligonucleotides have not been previously described. To investigate oligonucleotides, we determined the K d values of the complex with MCPIP1 D141N , PIN-ZF D141N and PIN D141N . The dissociation constant studies revealed a high affinity of MCPIP1 D141N to oligonucleotides, however, they did not show a major difference in affinity parameters using different oligonucleotides. We observed that the affinity of MCPIP1 D141N and its fragments towards ssRNA, ssDNA and dsDNA substrates is lower than that for oligonucleotides forming stem loops. Moreover, we did not observe significant differences between the affinity of MCPIP1 D141N or PIN-ZF D141N to the investigated oligonucleotides. Therefore, we hypothesized that the PIN-ZF fragment is crucial for maintaining the complex with oligonucleotides. We also observed that the zinc-finger domain significantly increased the affinity of the PIN D141N domain to 25-nt-long oligonucleotides. Interestingly, for shorter oligonucleotides, we did not observe significant differences in the affinity for PIN D141N or PIN-ZF D141N . We hypothesized that the zinc finger does not reach short substrates, which were localized in proximity to the catalytic pocket of MCPIP1. A zinc finger tethered in the vicinity of the PIN catalytic domain might enhance the re-association of the substrate and facilitate subsequent cleavage. In contrast, previous data suggested that long RNA fragments derived from the C/EBPβ 3′UTR mRNA interact with full-length MCPIP D141N but not with PIN-ZF D141N 21 . We hypothesized that our previous results might be affected by the non-equilibrium conditions of the EMSA method. It is also possible that for maintaining of PIN-ZF D141N interaction with long mRNA fragments, additional domains of MCPIP1 are crucial for the stability of the ternary complex.
Size exclusion chromatography results indicate that MCPIP1 exists in equilibrium between the dimeric and tetrameric state, we proposed a stoichiometric model of the MCPIP1 -RNA interaction. Two molecules of the MCPIP1 dimer interact with a single stem-loop structure (Fig. 6B). Thus, for the full-length MCPIP1, we proposed a sequential binding model: oligo + MCPIP1 dimer + MCPIP1 dimer  oligo-MCPIP1 dimer + MCPIP1 dimer  oligo-MCPIP1 tetramer . We hypothesized that the binding of oligonucleotide substrates induces homooligomerization of the MCPIP1. Tetrameric oligomerization of the PIN domain was previously shown for Nob1p and PAE2754, where the PIN domains of these proteins form a ring structure with a central hole that is wide enough to accommodate ssRNA or ssDNA but not double-stranded oligonucleotides 36,37 . Nevertheless, further validation of this model of MCPIP1-RNA complex stoichiometry should be performed. Our model is based on size exclusion chromatography results and double binding equilibrium observed in the affinity curves. Resolving the quaternary structure of MCPIP1 with RNA oligonucleotides is crucial for understanding its detailed mechanism of RNA regulation. To date, there is no resolved holoenzyme structure of any of the PIN domains; thus, further studies of MCPIP1 complexes are highly interesting.
We observed increased rates of degradation of hairpins with short stems, which are consistent with results published by Mino and coworkers from HITS-CLIP analyses where small stem-loops that contained a 3-ntor 4-nt-long loop motif were predominantly identified in complexes with MCPIP1 D141N 5 . Recent studies have revealed the importance of UPF1 for cytoplasmic mRNA decay catalyzed by MCPIP1 5 . UPF1 is an RNA helicase that participates in degradation of mRNAs with premature termination codons that are crucial for nonsense-mediated mRNA decay (NMD) 38 . SMG6, one of the key proteins for NMD, also contains a PIN domain at the C-terminus that is responsible for ribonucleolytic activity and mRNA turnover 39 . For in vitro RNA substrate cleavage induced by recombinant SMG6, the presence of additional proteins with helicase activity is not necessary. We showed that MCPIP1 alone is sufficient to unwind and degrade substrates with stem-loop secondary structures in vitro. Nevertheless, UPF1 might enhance unwinding of MCPIP1 substrates as an RNA helicase since we observed that degradation of more stable (with low Gibbs free energy) stem-loops was less efficient. Interaction with UPF1 and other proteins may increase the rate of degradation of selected RNA targets and broaden the recognition potential of the MCPIP1 complex.
Our biochemical studies revealed numerous cleavage sites introduced by recombinant MCPIP1 in the investigated sequences. We found that MCPIP1 induced endonuclease cleavage in the loop motif of stem-loop structures. We hypothesized that the presence of strong and non-sequence-specific interactions with RNA would enable MCPIP1 to efficiently search for the stem-loop elements in transcripts, and identification of a stem-loop would result in endonucleolytic cleavage and transcript destabilization. Nevertheless, MCPIP1 has been identified as selective ribonuclease that cleaves translationally active mRNAs at the 3′UTR. This raises the possibility that additional proteins that are elements of a ternary complex consisting of transcripts and MCPIP1 might determine the final MCPIP1 specificity.

Methods
Cloning and protein purification. The human ZC3H12A gene that encodes MCPIP1 was optimized for efficient expression in E. coli strains and ordered as a synthetic gene from GenScript (USA). The cloning, expression and purification of the full-length MCPIP1 WT protein and its mutant form MCPIP1 D141N were previously described 6 . The procedures for purification of the N-terminus His 6 -tagged proteins PIN-ZF WT and PIN-ZF D141N (134-327 residues) were described previously 21 . The same methods were used to purify PIN WT and PIN D141N (134-297 residues), which were also tagged with His 6 at the N-terminus. Briefly, E. coli BL21-CodonPlus-RIL cultures were grown at 37 °C in LB medium until reaching an OD 600 of 0.5. Protein expression was induced with addition of 0.5 mM IPTG. All proteins were expressed for 3 hours at 37 °C. Full-length MCPIP proteins were purified using ion-exchange chromatography (TMAE) in denaturing conditions. PIN WT , PIN D141N , PIN-ZF WT and PIN-ZF D141N were purified using Ni-NTA affinity chromatography in denaturing conditions. Finally, all proteins were dialyzed and purified using a gel filtration Superdex 200 prep grade 10/300 (GE Healthcare) column in a buffer comprised of 25 mM Tris, pH 7.9, 300 mM NaCl, 10% (w/v) glycerol, 1 mM DTT, and 0.5 mM EDTA. Chromatography was performed using an Äkta FPLC purification system (Amersham Pharmacia). Table 1 were purchased from Sigma-Aldrich. These oligonucleotides were fluorescently labeled using 6-carboxyfluorescein (6-FAM). The 5′ ends labeling of oligonucleotides were made by attaching 6-FAM to phosphate group of the 5′ terminal nucleotides. The mIL-6 82-106 3′FAM labeling was done by coupling 6-FAM to phosphate group of the 3′ terminal nucleotide. Labeled and purified with high-performance liquid chromatography oligonucleotides were purchased from Sigma-Aldrich. The double-stranded mIL-6 82-93 dsDNA was prepared by mixing mIL-6 82-93 ssDNA and the complementary oligonucleotide in a 1:1.2 ratio. Subsequently, for dsDNA, oligonucleotides were annealed by heating to 95 °C for 5 minutes and cooled at room temperature. Analysis of RNA secondary structure of the investigated oligonucleotides was performed using the Vienna RNA web server 40 .

RNase assays.
In vitro cleavage assays of FAM-labeled oligonucleotides were performed in buffer containing 25 mM Tris-HCl, pH 7.9, 150 mM NaCl, 10% (w/v) glycerol, 2.5 mM MgCl 2 , 1 mM DTT, 0.5 mM EDTA and 0.05 mM ZnCl 2 . Labeled oligonucleotides and MCPIP1 protein concentrations were 7.5 µM and 2 µM, respectively. Samples were incubated at 37 °C, and reactions at different time points were stopped by freezing in dry ice. After addition of twofold excess of concentrated loading dye consisting of 95% (w/v) formamide, 0.5 mM EDTA, 0.025% (w/v) xylene cyanol, and 0.025% (w/v) bromophenol blue, reactions products were denatured at 95 °C for 1 minute. An alkaline hydrolysis RNA ladder for each oligonucleotide was generated through denaturation at 95 °C for 25 minutes in alkaline buffer containing 50 mM sodium bicarbonate, pH 9.5, and 1 mM EDTA. Samples were resolved in denaturing gel electrophoresis in TBE (Tris/borate/EDTA) buffer. Denaturing gels contained 20% polyacrylamide and 7.5 M urea. Fluorescence signals were detected using ChemiDoc gel imaging device with ImageLab 5.2 software (BioRad Laboratories). Signal acquisition times 0.5 sec were the same for each of the gels.
Affinity determination assays. The concentration of FAM-labeled oligonucleotides was 2 nM in a system with MCPIP1 D141N and 20 nM in a system containing PIN D141N or PIN-ZF D141N proteins. Free FAM label (6-Carboxyfluorescein, C0662 Sigma-Aldrich) was used as a control of affinity determination assay. Unlabeled and HPLC purified hIL-6 81-98 RNA oligonucleotide was purchased from Sigma-Aldrich. Protein concentrations were determined by measuring the absorbance at 280 nm using a NanoDrop 2000 spectrophotometer (Thermo Scientific). Proteins absorption coefficients were calculated on the basis of amino acid sequence. Samples were prepared using the twofold serial dilution method; thus, in each sample, the concentration of protein gradually changed. The reaction buffer for detection of sample fluorescence contained 25 mM Tris-HCl pH 7.9, 150 mM NaCl, 5% (w/v) glycerol, 2.5 mM MgCl 2 , 1 mM DTT, 0.5 mM EDTA and 0.05 mM ZnCl 2 . Fluorescence signals were collected using the FluoroLog FL3-12 spectrofluorometer (Horiba Jobin Yvon). Excitation and emission wavelengths were 495 nm and 514 nm, respectively. Measurements of fluorescence was performed at 25 °C using a temperature controlled cuvette holder. The dimensions of the quartz cuvette were 3 × 3 mm (Hellma). Dissociation constants (K d ) were determined using DynaFit software (version 4.07.111, BioKin) 41 . Determination of the binding model was based on residual distribution of fitted curves and standard deviation of determinated dissociation constants. For calculation of the dissociation constants, a sequential binding model was used: + + +   N P P NP P NPP (N -oligonucleotide, P -protein), where K d1 and K d2 were equal dissociation constants. Additionally, two binding models were tested. The first one was characterized by K d1 ≠ K d2 , and the second one was simplified to the single equation +  N P NP. The graph errors represent standard deviations from 3 independent experiments. For statistical analysis of differences between calculated dissociation constants for oligonucleotide complexes with MCPIP D141N , PINZF D141N and PIN D141N one-way ANOVA followed by Tukey's multiple comparison test was used.

Gel filtration assays. Analytical size exclusion chromatography was performed using a Superdex 200
Increase 10/300 GL column (GE Healthcare) that was calibrated with the following protein standards: myoglobin, α-chymotrypsinogen, β-lactoglobulin, ovalbumin, bovine serum albumin, apoferritin and thyroglobulin. The apparent molecular weight of MCPIP1 proteins was determined based on the column calibration curve. For determination of homooligomerization of the analyzed samples, multiple Gaussian peak fits were performed for chromatogram data using OriginPro 2017 software (OriginLab).
Native polyacrylamide gel electrophoresis. Protein samples for native electrophoresis were prepared with the addition of twofold excess of concentrated loading dye that comprised 62.5 mM Tris-HCl, pH 6.8, 25% glycerol, and 1% (w/v) bromophenol blue. The gels contained 300 mM Tris-HCl, pH 8.8, and 6% polyacrylamide (concentrations of acrylamide/bis-acrylamide were 30%/1% w/v). Electrophoresis was performed at 80 V using running buffer containing 25 mM Tris and 192 mM glycine. Gels were stained with Coomassie Brilliant Blue G-250 solution.
Data availability. The datasets analyzed during the current study are available from the corresponding author on reasonable request.