Supplementary Materials and Methods

Supplementary Materials and Methods Cloning and Target Selection Full length human 3’-UTRs of CNTFR, PDCD4, SPRY1, SPRY2, BTG2, PTEN, TMEM2, THBS1 and MTF2 were amplified by PCR and directionally cloned (BamHI/SalI or AgeI/EcoRI restriction sites) downstream of EGFP, which was driven by a CMV promoter in dual-fluorescent reporter plasmids. Plasmids additionally contained the gene for tdTomato driven by a second CMV promoter, which served as passive transfection control. Seed and supplementary pairing mutations were further introduced using site-directed mutagenesis with mutation-containing PCR primers. All PCR products were confirmed by sequencing after cloning. For competitive binding assays generic miR-21 and let-7 dual-fluorescent reporters were generated by primer annealing and ligation downstream of EGFP. The let-7 reporter contained two 8-mer target sites (Fig. S2A), while the miR-21 reporter contained one cleavable complementary target site, which was attenuated by the introduction of a single mismatch opposite nucleotide 11 of the microRNA. For miR-21 this was due to weak changes for regular 8-mer target sites (Fig. S4A), while a fully complementary site was completely silenced without detectable GFP fluorescence. We confirmed the validity of the attenuated reporter to measure miR-21 activity using antimiR (Fig. S2B) and mimic (not shown) transfections. Inhibitor plasmids contained the gene for a far red fluorescent protein (iRFP(1)) driven by a CMV promoter. The 3’-UTRs used in dual-fluorescent assays were directionally cloned downstream of iRFP (BamHI/SalI restriction sites). Same plasmid without a 3’-UTR served as “no inhibitor” control for normalization. All primer sequences are provided in Table S3. Target selection of CNTFR and let-7 targets was based on a search of 23-species 3’-UTR alignments from TargetScan (V6) for conserved 6-mer seed matches (to nucleotides 2-7 of the miRNA) and conserved 5-mer supplementary matches to miRNA nucleotides 12-16 or 13-17. Pairing was counted as conserved if all nucleotides were conserved in the majority of the species including mouse and human. The supplementary pairing search was limited to a region 3 to 13 nucleotides upstream of the seed match, following previously established parameters for such sites(2). For miR-21 CNTFR was our top hit with extensive conserved seed and supplementary pairing. For let-7 we further reduced the number of candidates based on a) presence of only one let-7 binding site, b) strong overall Ago2 association in HeLa cells(3) and c) strong Ago2-CLIP peaks overlapping the binding site in HeLa cells(4), resulting in the selection of TMEM2, THBS1 and MTF2.

1

Transfection for reporter assays NIH3T3 cells were cultured in 10% FCS in DMEM with 1x Penicillin/Streptomycin. For dual-fluorescent reporter assays 20,000 NIH3T3 cells were seeded per well in 96-well optical square-bottom black plates (“µ-Plate”, ibidi). On the next day cells were transfected with a mix of reporter plasmid (200 ng/well), LNA-antimiR (50nM) and Lipofectamine 2000 (ThermoFisher, 0,5 µl/well) in a total volume of 100 µl/well, at 5% FCS and without antibiotics. Reporter plasmids were thereby mixed with appropriate LNA-antimiRs before Lipofectamine addition to increase co-transfection efficiency. At least three wells per plate were transfected for each condition. After 48 h cells were carefully washed once with 37 °C PBS, fixed with 4% PFA in PBS for 6 minutes at room temperature and washed again with PBS. As the last step all PBS was removed and 100 µl of 50% glycerol in PBS containing 0.5 µg/ml DAPI was added to each well. Plates were thereby handled in a dark room to avoid bleaching the fluorophores. Plates were then stored at 4 °C overnight and images were acquired with fluorescence microscopy on the next morning. For the competitive binding assay generic miR-21 and let-7 dual-fluorescent reporters were premixed with the different inhibitor-vectors at a reporter:inhibitor ratio of 1:1 for miR-21 and 1:10 for let-7 targets and a total of 200 ng/well were co-transfected using 0.5 µl/well of Lipofectamine. The remaining steps were carried out as above. Assays were analyzed using automated microscopic acquisition and morphological analysis as described below. Competitive binding influences target binding following the equation: !" =

[!]& ["] " + () (1 +

, ) (-

(Where M: miRNA, T: target, I: inhibitor; with K being the respective dissociation constants.) We therefore expect that differences in apparent KD for preferential targets vs. regular targets in vivo will result in stronger competitive potency.

Automated image acquisition Plates were measured in an automated way on a Zeiss Observer.Z1 microscope. Each well was imaged at four predefined positions. Images were acquired at each position with different filters for each of the three detected fluorophores. The filter configuration (excitation, emission, splitter) was: GFP (target signal): ET470/40x, ET525/50m, T495LPXR; RFP (tdTomato/normalization): ET560/40x,ET630/75m,T585lp; DAPI (nuclear staining): D350/50x,ET460/50m,T400LP. We confirmed for each fluorophore that no crosstalk was present between the channels. Images were acquired in 12-bit depth using a Retiga-4000dc

2

camera (Qimaging). Infrared RFP in competitive binding assays was not measured and did not show crosstalk into the other channels.

Morphological image analysis We performed individual cell detection and GFP and RFP value extraction for the fluorescent images to obtain similar data to a FACS measurement. The automated microscopic acquisition and analysis has the advantage of scalability, allowing us to interrogate a large number of conditions and replicates per plate. Processing was done in MetaMorph (Molecular Devices) using a custom workflow. Images were first corrected for background fluorescence using top-hat transform and cell borders were then recognized using the watershed algorithm on inverted and smoothed (morphologic open filter) RFP images. DAPI images were thresholded and objects in RFP without overlap to a DAPI signal of predefined intensity were excluded. Further filtering based on object area, intensity, length, breadth and shape factor were applied to select valid objects (cells) in the RFP channel in the linear exposure range of the camera. Finally, red and green corrected fluorescence intensities were reported for each detected cell. Similar to the analysis previously performed for FACS-derived data(5, 6), we then binned cells based on RFP intensity (which corresponds to the transfection strength) and calculated median GFP/RFP ratios per bin. As previously described(5, 6) we observed that the extent of target regulation typically decreased in higher bins (Fig. S2A), suggesting that strong overexpression perturbs microRNA function. We therefore report the GFP/RFP values for the lowest bin (RFP(arbitrary fluorescence units)=100-200) in Fig. 2, with the exception of Fig. 2C left panel (miR-21 competitive binding), where the highest bin (AFU=400-800) is reported, as it showed the strongest reporter de-repression for all groups.

Ago2 ribonucleoprotein immunoprecipitation (RIP) Ago2-RIP reflects the amount of mRNA bound to Ago2. However, multiple binding sites for different miRNAs and possibly miRNA-independent Ago-interactions (e.g. by co-precipitation of higher order RBP complexes) can be present for a specific RNA. We therefore assessed changes in RIP-fractions induced by target site mutation or miRNA inhibition as a measure of contribution of specific interactions to the overall RIP fraction. We expect for this change (Δ) that: Δ

RIP-fraction total

∝ ["!] = ["] 0

[TM] T +[TM]

(1)

(where T: target, M: miRNA)

3

For two targets (1 and 2) of one miRNA with Δ

RIP-fraction1 RIP-fraction2 >Δ total1 total2

replaced with: [)C] )

=

[C] DE

, it therefore follows that KD18 for all samples) and a ribosome depletion was performed using a RiboMinus Eukaryote Kit v2 (ThermoFisher) following manufacturer’s protocol. For both input and IP fractions the RNA was then fragmented in 1x magnesium RNA fragmentation buffer (NEB) for 3.5 min at 94 °C and extracted using phenol-chloroform. First strand cDNA synthesis was then performed using Superscript III (ThermoFisher), followed 5

by second strand synthesis with DNA polymerase I (final concentration 1 U/µl) in the presence of 0.04 U/µl RNase H (ThermoFisher). The resulting double-stranded DNA was then purified using AMPure XP paramagnetic beads (Beckman Coulter). End repair and further preparation steps were performed using NEBNext Ultra II DNA library prep kit for Illumina (NEB) as per manufacturer’s protocol. Libraries were multiplexed and all libraries pertaining to one experiment (6x input, 6x IP) were sequenced together in an Illumina HiSeq 4000 machine using a paired read chip at 100 bp read length. For microRNA quantification NIH3T3 cells were transfected with a CMV-tdTomato plasmid and sorted based on fluorescence intensity to match the conditions for Ago2-RIP assays above. Total RNA was then extracted using RNApure reagent and small RNA libraries were prepared using NEBNext small RNA library prep set for Illumina (NEB) following manufacturer’s protocol and sequenced as above.

Bioinformatic analysis of deep sequencing data For RNA-Seq experiments each biological replicate was sequenced in two technical replicates. Reads obtained from library sequencing were processed using flexbar(10) to remove adaptor readthroughs and trim low quality bases. Only reads retaining a length of ≥50bp in both read pairs were kept. Reads were aligned to mouse genome (mm10) using TopHat2(11). For each protein coding gene we selected a representative mRNA transcript annotation (20,974 total entries) based on Gencode version M3 and 3P-seq tags from TargetScan Mouse(V7.1)(12). The number of uniquely aligned reads overlapping each mRNA were quantified using htseq-count(13). For each biological replicate counts for each mRNA in technical replicates were summed up (+1 to eliminate 0 values) and normalized per million of total mRNA read counts in the sample (FPM). We then used the formula in Fig. 3A to calculate mRNA fold changes (dExpr) and Ago2-RIP fold changes (dEnr) for each mRNA in each sample. The values of the biological replicates were then averaged. For further analysis we retained those mRNAs with either ≥10 reads in each of the LNA-Ctrl input samples and an FPKM≥1 or ≥50 reads in LNA-Ctrl input samples. To select top candidates (Fig. 5, red curves) for each mRNA with a) ∆Expr ≥ 0.7, b) ∆Enr ≥ 0.7, we applied following rational criteria: c) ∆EnrLNA-Mix (in LNA-Mix) ≥ 0.35, d) ∆EnrLNA-16