Volume 4, Issue 5, Pages 938-944 (September 2013) Differential Translation Tunes Uneven Production of Operon-Encoded Proteins  Tessa E.F. Quax, Yuri I. Wolf, Jasper J. Koehorst, Omri Wurtzel, Richard van der Oost, Wenqi Ran, Fabian Blombach, Kira S. Makarova, Stan J.J. Brouns, Anthony C. Forster, E. Gerhart H. Wagner, Rotem Sorek, Eugene V. Koonin, John van der Oost  Cell Reports  Volume 4, Issue 5, Pages 938-944 (September 2013) DOI: 10.1016/j.celrep.2013.07.049 Copyright © 2013 The Authors Terms and Conditions

Cell Reports 2013 4, 938-944DOI: (10.1016/j.celrep.2013.07.049) Copyright © 2013 The Authors Terms and Conditions

Figure 1 Cistron Properties Correlated with Stoichiometry of Operon-Encoded Protein Complexes Analysis of the ribosomal protein operon L7/L12 (high expression), the F-type ATPase (moderate expression), and the type I-E Cascade complex (low expression). (A) Selected operons (block arrows) and the stoichiometry in the corresponding protein complexes. (B) Predicted mRNA folding energy (ΔE) of the RBS region of each cistron (−20 to +20 bp relative to the start codon). (C) Codon bias; ΔF (optimal codon usage) is shown by dark gray and ΔCAI is shown by light gray. (D) Codon co-occurrence (ΔCo). (E) Ribosome density profiles per gene. The green arrows represent genes in each operon that encode the most abundant subunit(s), and green rectangles denote the corresponding positive deviations in codon bias (ΔF > 0.02), codon co-occurrence (highest value), low RNA folding potential (highest energy value), and/or ribosome density (highest value). Error bars represent one unit of standard deviation. See also Figures S1, S2, S3, S4, S5, and Table S2. Cell Reports 2013 4, 938-944DOI: (10.1016/j.celrep.2013.07.049) Copyright © 2013 The Authors Terms and Conditions

Figure 2 Translation Efficiency Influences Protein Expression from Individual Genes within Operons (A) Expression constructs under control of the T7 promoter (gray arrow) encoding different combinations of two identical GFP polypeptides, with synonymous mutations, resulting in low (yellow block arrow, gfpL) or high (green block arrow, gfpH) translation efficiency. The single genes, as well as the downstream gene of the operons, are translational fusions to a Strep-tag (white block arrow; in construct name indicated by an asterisk). SD, Shine-Dalgarno sequence (Figure S6). (B) Quantification of western blot with GFP antibody on whole-cell lysates of the variant GFP-expressing constructs after expression in E. coli. Equal total amounts of cellular proteins were loaded in order to allow comparison between different samples (Figure S6). Cell Reports 2013 4, 938-944DOI: (10.1016/j.celrep.2013.07.049) Copyright © 2013 The Authors Terms and Conditions

Figure 3 Models Describing Tuning of the Translation Process Five scenarios are depicted with different rates of translation initiation and/or translation elongation. The corresponding ribosome density profiles and the expected protein yield are based on the assumption that differential transcription is insignificant (as has been demonstrated experimentally for the majority of the operons analyzed in this study). Scenario 1: relatively little protein is produced (e.g., single subunit per operon-encoded complex). Scenario 2: relatively little expression is required (no significantly different codon bias). Scenario 3: a hypothetical case in which high translation initiation results in ribosome jamming because the elongation rate is not optimized. Scenario 4: relatively much protein is produced (e.g., multiple subunits per operon-encoded complex); high translation initiation (low RNA fold), elevated but not maximal elongation rate (codon adaptation), elevated but not maximal ribosome densities (experimental profiles), and high protein yield (experimental protein complex stoichiometry values). Scenario 5 will also lead to high protein yields, but the experimentally detected elevated ribosome densities indicate that (at least under the tested conditions) elongation rates are not maximal. In conclusion, scenario 4 (red box) appears to most closely approach the in vivo situation of translation-controlled overexpression. Cell Reports 2013 4, 938-944DOI: (10.1016/j.celrep.2013.07.049) Copyright © 2013 The Authors Terms and Conditions

Figure 4 Model for Translation of Polycistronic Messengers in Prokaryotes (A) Transcript consisting of two cistrons, each preceded by an SD sequence. (B) Translational coupling where the 30S ribosomal subunit remains associated (red arrow) after termination, and 50S joins for reinitiation, resulting in stoichiometric output from both cistrons (Movie S1). (C) De novo internal recruitment (red arrows) of both 50S and 30S subunits allows for differential translation initiation rates between cistrons. Depending on the translation elongation rate (codon bias) of each cistron, this may result in different ribosome density profiles (Movie S2); different types of broken arrows reflect different elongation rates (yellow cistron, low; green cistron, high); only in case of concommitant elevated initiation rate (panel-c, not in panel-b) this will result in increased protein production. Yellow and green cistrons have low and high translation efficiency, respectively (Figure 2). Cell Reports 2013 4, 938-944DOI: (10.1016/j.celrep.2013.07.049) Copyright © 2013 The Authors Terms and Conditions

Figure S1 Transcription of Operons, Related to Figure 1 Analyses of mRNA levels of the following operons are depicted. (A) Ribosomal protein operon L7/L12 (Dunkle et al., 2011), the F-type ATPase (Abrahams et al., 1994) and the Type I-E Cascade complex (Jore et al., 2011). (B) The Type I-F Cascade complex (P. aerugionosa) (Wiedenheft et al., 2011), the V-type ATPase (Beyenbach and Wieczorek, 2006), the archaeal flagellum (S.solfataricus). (C) The FtsZ operon (Errington, 2003), the Type II secretion system (Johnson et al., 2006), TAT secretion system (Jakob et al., 2009). (D) The ribosomal protein operon L5 (Dunkle et al., 2011) and the NADH dehydrogenase (Efremov et al., 2010). I: the selected operons (block arrows), and the stoichiometry in the corresponding protein complexes are shown. II: RNaseq data shown in Reads Per Kb per Million (RPKM) from three representative prokaryotes. All RNaseq data shown in this figure is originating from E.coli unless specified otherwise. The green arrows are genes in each operon that encode the most abundant subunit(s), and green rectangles highlight corresponding positive transcription values (RKPM). Cell Reports 2013 4, 938-944DOI: (10.1016/j.celrep.2013.07.049) Copyright © 2013 The Authors Terms and Conditions

Figure S2 Cistron Properties Correlated with Stoichiometry of Operon-Encoded Protein Complexes, Related to Figure 1 (A and B) Translational factors that contribute to fine tuning of uneven stoichiometry potentially occur. As representative examples are depicted the analyses of the (A) Type I-A, I-B, I-C Cascade complex (Lintner et al., 2011), Type I-F Cascade complex (Wiedenheft et al., 2011), the archaeal flagellum (Ghosh and Albers, 2011) (B) the V-type ATPase (Beyenbach and Wieczorek, 2006), the FtsZ operon (Errington, 2003). I, selected operons (block arrows), and the stoichiometry in the corresponding protein complexes. II, mRNA folding energy in the region around the RBS (−20 to +20 bp relative to the start codon) (ΔE) as measured by UNAFold program. III codon bias (ΔF optimal codons (dark gray), and ΔCAI (light gray)), IV codon co-occurrence (Δ Co) Error bars indicate standard deviations. The green arrows represent genes in each operon that encode the most abundant subunit(s), and green rectangles highlight corresponding positive deviations in codon bias (ΔF > 0.02), codon co-occurrence (highest value) and/or low RNA folding potential (highest energy value). Cell Reports 2013 4, 938-944DOI: (10.1016/j.celrep.2013.07.049) Copyright © 2013 The Authors Terms and Conditions

Figure S3 Cistron Properties Correlated with Stoichiometry of Operon-Encoded Protein Complexes, Related to Figure 1 (A and B) Translational factors that contribute to fine tuning of uneven stoichiometry potentially occur. As representative examples are depicted the analyses of the (A) the ribosomal protein operon L5 (Dunkle et al., 2011) and the TAT secretion system (Jakob et al., 2009) (B) the NADH dehydrogenase (Efremov et al., 2010) and Type II secretion complex (Johnson et al., 2006). I: selected operons (block arrows), and the stoichiometry in the corresponding protein complexes. II: mRNA folding energy in the region around the RBS (−20 to +20 bp relative to the start codon) (ΔE) as measured by UNAFold program. III codon bias (ΔF optimal codons (dark gray), and ΔCAI (light gray)), IV codon co-occurrence (Δ Co) Error bars indicate standard deviations. The green arrows represent genes in each operon that encode the most abundant subunit(s), and green rectangles highlight corresponding positive deviations in codon bias (ΔF > 0.02), codon co-occurrence (highest value) and/or low RNA folding potential (highest energy value); false positives (ΔF > 0.02, stoichiometry < 2; see Figure S4) are indicated with red rectangles. Cell Reports 2013 4, 938-944DOI: (10.1016/j.celrep.2013.07.049) Copyright © 2013 The Authors Terms and Conditions

Figure S4 Codon Bias and Subunit Stoichiometry in Bacterial and Archaeal Operons, Related to Figure 1 The horizontal axis shows the family mean codon usage shift (ΔF), i.e., the fraction of optimal codons; Figure 1C); the vertical axis shows the (approximate) number of subunits in the complex. Diamonds show genes belonging to the complexes with uneven stoichiometry; crosses show genes belonging to the complexes with even stoichiometry. The red lines show the ΔF and stoichiometry thresholds providing the best separation between 13 true positives (upper right quadrants), 41 true negatives (lower left quadrants, respectively) and 2 false positives (lower right quadrant; discussed in text). Details on the selected complexes are provided in Table S2. Cell Reports 2013 4, 938-944DOI: (10.1016/j.celrep.2013.07.049) Copyright © 2013 The Authors Terms and Conditions

Figure S5 Ribosome Density Profiles of Polycistronic Transcripts Coding for Protein Complexes with Uneven Stoichiometry, Related to Figure 1 (A and B) Analysis of selected E.coli operons: (A) the ribosomal protein operon L5 (Dunkle et al., 2011) and the TAT secretion system (Jakob et al., 2009), and (B) The NADH dehydrogenase (Efremov et al., 2010) the FtsZ operon (Errington, 2003). I, selected operons (block arrows), and the stoichiometry in the corresponding protein complexes; II, Density of ribosomes per cistron; The green arrows represent genes in each operon that encode the most abundant subunit(s), and green rectangles denote the corresponding positive deviations of ribosome density. Cell Reports 2013 4, 938-944DOI: (10.1016/j.celrep.2013.07.049) Copyright © 2013 The Authors Terms and Conditions

Figure S6 Expression of Synthetic Operons, Related to Figure 2 (A) Sequence of synthetic operon as depicted in Figure 2A, with two GFP genes with synonymous mutations which was cloned in pET52b (LH∗). Relevant features are indicated: gfpL (low efficiency of translation; yellow), gfpH (high efficiency of translation; green), cloning sites (purple; GCATGC, SphI; Compatible GAATTC/ CAATTG, EcoRI/ MfeI; Compatible ATGCAT/ CTGCAG, NsiI/ PstI; CCTAGG, AvrII; GCTAGC, NheI), T7-promoter (blue) SD- motif (bold, red), start/stop codons (bold, underlined), His-10-tag (light gray), strep-tag (dark gray). Theoretical molecular weights are: 28.6 kDa (GFP-L) and 29,8 kDa (GFP-H+strep-tag). (B) Coomassie stained SDS-PAGE gels of whole cell lysate of E. coli containing pET52b-based expression constructs (Figure 2A) encoding different combinations of two identical GFP polypeptides, with synonymous mutations, resulting in low (gfpL) or high (gfpH) efficiency of translation. The SDS-PAGE gel is a replicate of the one used for Western blot (C). Based on a marker (#7709, NEB®) molecular masses (kDa) are indicated. (C) Western blot with GFP antibody on whole cell lysates of the variant GFP-expressing constructs after expression in E.coli; the upper band corresponds to the slightly larger Strep-tagged GFP-variant (∗). Cell Reports 2013 4, 938-944DOI: (10.1016/j.celrep.2013.07.049) Copyright © 2013 The Authors Terms and Conditions