A compound that directly and selectively stalls PCSK9 translation

Proprotein Convertase Subtilisin/Kexin Type 9 (PCSK9) plays a key role in regulating the levels of plasma low density lipoprotein cholesterol (LDL-C). Here we demonstrate that the compound PF-06446846 inhibits translation of PCSK9 by inducing the ribosome to stall around codon 34, mediated by the sequence of the nascent chain within the exit tunnel. We further show that PF-06446846 reduces plasma PCSK9 and total cholesterol levels in rats following oral dosing. Using ribosome profiling, we demonstrate that PF-06446846 is highly selective for the inhibition of PCSK9 translation. The mechanism of action employed by PF-06446846 reveals a previously unexpected tunability of the human ribosome, which allows small molecules to specifically block translation of individual transcripts. One Sentence Summary A small-molecule PCSK9 inhibitor targets the human ribosome and selectively prevents PCSK9 synthesis.

demonstrates in vivo activity. We show that PF-06446846 induces the 80S ribosome to stall 66 while translating PCSK9. We further demonstrate using ribosome profiling that despite acting 67 through protein translation, a core cellular process, PF-06446846 is exceptionally specific, 68 affecting very few proteins. The PF-06446846 mechanism of action reveals a previously 69 unexpected potential to therapeutically modulate the human ribosome with small molecules as a 70 means to target previously "undruggable" proteins.  Table S1). The synthesis and physiochemical characterization of only inhibited by 20% (Fig. S1D). Translation of the protein fusion constructs was driven by the 98 EMCV-IRES, indicating that PF-06446846 is unlikely to target PCSK9 translation initiation 99 directly. When all the codons of PCSK9(1-33) were mutated to either common or rare 100 synonymous codons (Fig. S1E), PF-06446846 still inhibited translation of PCSK9 (1-33)-101 luciferase (Fig. 1C), ruling out a role of the mRNA sequence. Conversely, PF-06446846 did not 102 inhibit translation of a PCSK9(1-33) construct with two compensatory frameshifts that result in a 103 near endogenous mRNA sequence but a non-endogenous amino acid sequence (Figs. 1C & S1E).

104
These data indicate that PF-06446846 sensitivity is primarily dependent on the amino acid 105 Lintner et al.
6 sequence of PCSK9. To further define the sequence requirements, we tested the activity of PF-106 06446846 against sets of N-terminal deletions, C-terminal deletions and alanine scanning 107 mutations of PCSK9 (1-33). The most important regions in PCSK9(1-33) that confer sensitivity 108 to PF-06446846 are Leu15-Leu20, residues 9-11 which include two tryptophan amino acids, and 109 residues 31-33 ( Fig. S10A-B). However, most mutations partially reduced the activity of PF-110 06446846, suggesting that multiple amino acid features of PCSK9(1-35) make contributions to 111 its sensitivity to PF-06446846.  133 To explore the safety of the compound and to gain insight into the in vivo activity of PF-134 06446846, male rats were orally administered PF-06446846 at doses of 5, 15 and 50 mg/kg daily 135 for 14 days. Plasma PF-06446846 (Table S8) and PCSK9 concentrations were measured at 1, 3,  During the two-week dosing period, PF-06446846 was tolerated. There was a small 150 decrease (11-13% relative to vehicle) in food consumption at 50 mg/kg PF-06446846 that was

11
To identify and quantify the sensitivity of individual proteins to PF-06446846, we 220 adopted a computational approach to identify transcripts that could potentially have PF-221 06446846-induced stalls, estimate the position of the stalls, and use differential expression 222 analysis using only reads aligning 3' to the stall site (Fig. 5A). To estimate a 3' bound for the 223 positions of potential PF-06446846-induced pauses, we adapted the approach previously 224 reported (21). For each gene, we plotted the percentage of the total reads aligning at or 5' to each 225 codon to generate cumulative fractional read (CFR) plots ( Fig. 5B-C). If PF-06446846 induced a 226 stall, the CFR plot should increase rapidly 5' to the stall and level off 3' to the stall. We define 227 the maximum divergence between the PF-06446846 and vehicle CFR plots as the D max and the 228 codon at which D max occurs as the D max position, which is analogous to the KS position 229 previously described (21). For genes with one or more PF-06446846-induced stalls the D max 230 position occurs 3' to the last stall site. For genes with a high D max (Z-score > 2) we used reads 231 mapping 3' to the position of D max for differential expression analysis (Fig. 5D, see methods for 232 details). Using this approach, we identified 22 PF-06446846-sensitive genes at the 60 minute 233 timepoint (Figs. 5 E-I and S14, Table S9) and 44 genes at the 10-minute timepoint (Dmax Z-234 score > 2, DeSeq FDR > 10%). With the exception of CDH1 and IFI30, all PF-06446846-235 sensitive proteins identified at the one hour timepoint were also identified at the 10-minute 236 timepoint. To test the robustness of our approach we also analyzed the data from the second 237 ribosome profiling experiment with the same pipeline. Despite most stall peaks being smaller in 238 the second study, we identified all of the same PF-06446846-sensitive sequences as for the data 239 from study 1, demonstrating the advantage of using information from the entire transcript instead 240 of a single codon. when data were analyzed using the same criteria as for the secretome (Fig. S9E-H). With a 246 reduction in stringency (proteins with 3 unique peptides accepted), a 2-to 3-fold reduction in  To validate our approach, we tested the translational inhibitory activity of PF-06446846 259 for a set of the targets identified in the 1-hour datasets in HeLa-derived cell free translation 260 reactions. In the all but two cases, translation of constructs consisting of the predicted stall site 261 fused to luciferase was inhibited by PF-06446846 (Fig. 5J). For the other two proteins, Midikine 262 and BCAP31, the translation of luciferase fusions to the full-length proteins was inhibited by PF-263 06446846 (Fig. 5K). Four control sequences predicted not to be PF-06446846-sensitive were 264 inhibited at comparable levels as luciferase alone (Fig. 5L). We next tested the translation inhibition of PF-06446846 towards four example "stall 267 sequences" identified only in the 10-minute dataset. PF-06446846 inhibited translation of all of 268 these transcripts only slightly more than luciferase alone (Fig. 5M). These results indicate that 269 the most sensitive PF-06446846 targets are those identified using the 1-hour treatment time. The 270 additional effects seen in the 10-minute treatment could be due to a partial adaptation of the cells were also identified using the D max approach (Fig. S14T). For the hits identified using only 281 center-of-density analysis, VPS25 and TM2D3 had stalls but no decrease in downstream reads, 282 indicating that the stall was unlikely to result in a decrease in protein production ( Fig. S14P-Q).

283
MAPRE1 did not have a clear stall site (Fig. S14R). COX10 (Fig. S14S) had a series of PF-284 06446846-induced stalls near the stop codon which would be difficult to detect using the D max 285 approach because of a lack of downstream reads to quantify. We also used center of density 286 analysis to confirm the bias for stalling near the 5' ends of the CDS by repeating the center of 287 density analyses omitting the first 50, 100 and 150 codons respectively (Fig. 6B-D). In all cases, the cluster of outliers disappeared, indicating that, while there are exceptions, PF-06446846-289 induced stalls most often occur in the first 50 codons.

291
To confirm that the decrease in reads downstream from PF-06446846 induced stalls 292 occur as a result in changes in translation as opposed to mRNA levels, we plotted the change Z-293 score transformed(22) (see materials and methods) in mRNA-seq reads versus the changes in 294 ribo-seq reads (Fig. 6E). For all identified PF-06446846-sensitive sequences, the changes in 295 ribosome footprints, were due solely to changes at the level of translation (Fig. 6E). We   stalling near the N-terminus could impede initiation, thus lowering the overall initiation rate.

355
Evidence for "queued" ribosomes ( Fig. 1H-J) suggest impeding of initiation could occur for 356 stalls near the N-terminus. However, an impact on translation initiation could not explain the effects of more C-terminal stalling on CDH1 levels ( Fig. S9G-H), and would predict that PF-358 06446846-sensitive transcripts would display a bias for higher TE, which they do not (Fig   359   S15C). Alternatively, the stalled ribosomes may be removed from the transcript by a "rescue" 360 mechanism and not complete translation. Future experiments will be required to determine the 361 underlying mechanisms of protein reduction due to ribosome stalling by PF-06446846.  percentage of reads aligning at or 5' to that codon. In all plots data from 1.5 µM PF-06446846