High-throughput interrogation of programmed ribosomal frameshifting in human cells

Fig. 1. A massively parallel reporter assay for programmed ribosomal frameshifting. a Schematics of the library design; for forward primer, rev reverse primer, BC barcode. b Outline of the experimental pipeline; the plots in the bottom constitute representative bin profiles for variants with no (left), low (middle), or intermediate (right) GFP fluorescence. c, d Mean ± 95% CI of −1 (c) and +1 (d) PRF signal (% GFP fluorescence) for barcode control groups corresponding to the indicated wild-type PRF sequences (blue), carrying single nucleotide deletions (green) or having a stop codon inserted upstream of the frameshift sequence (red); n = 22, 2, 13, 32, 10, 25, 21, 8, 13, 24, 9 (c) and 2, 2, 1, 10 (d) wild-type sequences (blue), and n = 32, 7, 12, 33, 28, 23, 23, 21, 14, 23, 40 (c) and 11, 7, 19, 20 (d) sequences with deletion or stop codon (green and red) tested, in the order they appear on the graph; the shaded area denotes the range of background fluorescence; asterisks denote significant differences between the wild type and the other groups (single nucleotide deletion and stop control) combined (Mann–Whitney U test). e Comparison of −1 and +1 PRF reporter readout (% GFP fluorescence)…

Fig. 2. Slippery sites are optimized for the respective frameshifting event. a Clustered heat map of median percent wild-type frameshifting of variants, in which the indicated region (upstream, slippery site (including the preceding codon) and downstream) differs from the native one (n = 37–325). b The median percent wild-type PRF signal of variants, in which the indicated position is changed is plotted for the entire length of the variable region; gray box: slippery site. c Percent wild-type −1 PRF signal for variants with a point mutation only in the indicated position of the slippery site; the box shows the quartiles of the dataset, while the whiskers show the rest of the distribution except for outliers (n = 22, 11, 7, 6, 7, 8, 10, and 17 sequences tested, in the order they appear on the graph). d Heat map showing median percent GFP fluorescence conferred by replacing the native slippery site in −1 PRF sites tested positive in the assay (Fig. 1c), with the indicated combination of slippery site elements. e Distribution of percent wild-type GFP fluorescence of variants, in which the slippery site was replaced by any possible combination of bases following the pattern XXXYYYZ; the box…

Fig. 3. PRF sites exhibit different sensitivities to changes in the downstream sequence. a Schematic of downstream sequence manipulations. b Percent wild-type frameshifting rates of variants, in which the native downstream region has been replaced by constant sequences, leaving the indicated number of nucleotides after the slippery site unchanged; n = 1–3 per data point and PRF event. c Fraction of point mutations in the 40 nt downstream of the slippery site resulting in <50% of wild-type GFP fluorescence (green) and background fluorescence (<1.3% GFP fluorescence, blue), for −1 (top) and +1 (bottom) PRF events; n = 51, 47, 54, 23, 47, 51, 49, 29 (upper) and 9, 10, 52 (lower) sequences tested in total. d The difference in mean % wild-type frameshifting between variants, in which the indicated position is predicted to be paired vs. unpaired along the variable region; gray box: slippery site.

Fig. 4. Downstream RNA structure affects PRF efficiency in a context-dependent manner. a Schematic of downstream structure variations. b Percent −1 GFP fluorescence of variants, in which the downstream region is replaced by a synthetic sequence predicted to fold into the respective native secondary structure (using Vienna RNA (only stem-loop; light blue) or pKiss (including pseudoknots; dark blue; Methods section)); the shaded area denotes the range of background fluorescence; n = 15, 17, 13, 19, 12, 8, 11, 11, 5, 17, 4, 8, and 8 sequences tested, in the order they appear on the graph. c % −1 GFP fluorescence of variants, in which the downstream region is replaced by a synthetic sequence predicted to fold into the HIV (blue) or SRV1 (green) secondary structure (or variations thereof; Methods section); the box shows the quartiles of the dataset, while the whiskers show the rest of the distribution except for outliers; the shaded area denotes the range of background fluorescence; n = 38/37, 32/29, 30/34, 32/29, 32/27, 16/21, 41/36, 34/31, 14/7, 23/20, 37/35, and 25/25 sequences tested for HIV and SRV1 structure variants, respectively, in the order they appear on the graph. d % GFP…

Fig. 5. Additional regulatory elements contribute to PRF. a Schematic of sequence manipulations upstream and immediately downstream of the slippery site. b Boxplots of % wild-type frameshifting rates for variants of −1 PRF sites, in which the indicated number of nucleotides were deleted (negative numbers) or inserted (positive numbers, taken from two different constant sequences) after the slippery site; the box shows the quartiles of the dataset, while the whiskers show the rest of the distribution except for outliers; n = 13, 7, 4, 17, 7, 3, 16, 54, 4, 27, 10, 12, 23, 8, 6, and 23 sequences tested, in the order they appear on the graph. c % GFP fluorescence for variants of −1 PRF sites with the sequence upstream of the slippery site being randomly recoded or replaced with sequences predicted to have the indicated secondary structure; the shaded area denotes the range of background fluorescence. d % wild-type frameshifting rates (+1 for OAZ, blue, and −1 for HIV, green) of variants, in which the native upstream region has been replaced by constant sequences up to the indicated position relative to the PRF site. e Boxplot of percent of wild-type frameshifting rates for variants, in which…

Fig. 6. Prediction of frameshifting potential and efficiency. a Outline of the process of generating features and building and testing predictive models. b, c Prediction scores (area under the receiver operating characteristic curve (ROC AUC) in the case of classification (b) and Pearson correlation coefficients in the case of regression (c)) on held-out test data for the indicated PRF event(s) or the entire library using different feature sets (slip: nucleotide identities in the canonical XXXYYYZ slippery site; tai: tAI scores around the slippery site, aa: amino acid class (unpolar, polar, and charged) around the slippery site, dg: MFE of downstream regions, sec: predicted pairedness of downstream positions). d ROC curve showing performance of a classifier trained on 80% of all designed library variants passing filtering and testing on the remaining 20%. e For the test set (20% of all designed library variants of all −1 PRF events (left) or HIV-1 (right) passing filtering), the measured % GFP fluorescence is plotted against the model prediction (trained on 80% of all library variants of all −1 PRF events (left) or HIV-1 (right) passing filtering); Pearson correlation coefficient and the…

Fig. 7. Testing of PRF sites from HIV clinical isolates reveals subtype differences and associations with viral load. a Distribution of percent GFP fluorescence of measured HIV variants, n = 368. b Percent GFP fluorescence as determined by FACSseq is plotted against the value determined by measuring the corresponding isolated clone by FACS; Pearson correlation coefficient and the associated two-tailed p-value are reported. c Boxplot showing percent GFP fluorescence of HIV gag-pol PRF variants coming from the indicated subtypes; the box shows the quartiles of the dataset, while the whiskers show the rest of the distribution except for outliers; n = 35, 46, 10, 103, 112, and 8 sequences tested, in the order they appear on the graph. d Viral load (HIV titer) is plotted against percent GFP fluorescence determined for the natural gag-pol PRF site variant from the corresponding clinical isolate; vertical line: HIV HXB2 wild-type percent GFP fluorescence, n = 77. e Predicted vs. measured percent GFP fluorescence for 20% of the HIV variants (held-out test set) with the full feature set except for the identity of slippery site positions (left) or after removing MFE features (middle) or all…