Content uploaded by Yamila N. Torres Cleuren
Author content
All content in this area was uploaded by Yamila N. Torres Cleuren on Oct 22, 2019
Content may be subject to copyright.
Available via license: CC BY-NC-ND 4.0
Content may be subject to copyright.
1
Deconstructing the individual steps of vertebrate translation initiation
Adam Giess#1, Yamila N. Torres Cleuren#1, Håkon Tjeldnes1, Maximilian Krause1,2, Teshome Tilahun
Bizuayehu1, Senna Hiensch2, Aniekan Okon3, Carston R. Wagner3, and Eivind Valen*1,2
1 Computational Biology Unit, Department of Informatics, University of Bergen, Bergen 5020, Norway
2 Sars International Centre for Marine Molecular Biology, University of Bergen, Bergen 5008, Norway
3 Dept. Medicinal Chemistry, University of Minnesota, Minneapolis, MN 55455, USA
#These authors contributed equally to this work (in alphabetical order)
*To whom correspondence should be addressed: eivind.valen@gmail.com
Abstract
Translation initiation is often attributed as the rate determining step of eukaryotic protein synthesis and key
to gene expression control 1. Despite this centrality the series of steps involved in this process are poorly
understood 2,3. Here we capture the transcriptome-wide occupancy of ribosomes across all stages of
translation initiation, enabling us to characterize the transcriptome-wide dynamics of ribosome recruitment
to mRNAs, scanning across 5’ UTRs and stop codon recognition, in a higher eukaryote. We provide
mechanistic evidence for ribosomes attaching to the mRNA by threading the mRNA through the small
subunit. Moreover, we identify features regulating the recruitment and processivity of scanning ribosomes,
redefine optimal initiation contexts and demonstrate endoplasmic reticulum specific regulation of initiation.
Our approach enables deconvoluting translation initiation into separate stages and identifying the regulators
at each step.
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
2
Introduction
In eukaryotes, translation initiation is a highly orchestrated sequence of events where the ribosomal 43S
pre-initiation complex (PIC) is first recruited to the beginning of the transcript through interactions with
initiation factors and the 5’ m7G cap 4. The 43S PIC then scans the transcript in a 5’-to-3’ direction until a
suitable translation initiation site (TIS) is encountered. Upon recognition of the TIS, the large ribosomal
subunit is recruited to form an elongation-capable 80S ribosome. These initiation steps are broadly
acknowledged to be a rate limiting factor in protein synthesis 3,5,6. Despite this, our knowledge of ribosome
recruitment, scanning, and TIS recognition is limited.
Ribosome profiling (ribo-seq) has enabled global quantification and localization of translation through the
capture of footprints from elongating 80S ribosomes 7. A limitation of ribo-seq, however, is that it is blind
to ribosomes from other stages of translation. Recently, translation complex profiling (TCP-seq) was
introduced which circumvented this problem by crosslinking all stages of ribosomes to the mRNAs 8.
However, because this technique relies on first purifying 80S ribosome containing transcripts, it is limited
to studying 40S-ribosome positioning to transcripts that have at least one 80S ribosome and are thus
actively translated. Here, we have expanded this approach to capture footprints from all ribosome-
associated mRNAs, including transcripts not bound by any 80S subunit. Our approach immobilizes all
ribosomal subunits on the mRNA by paraformaldehyde crosslinking, followed by sucrose gradient
separation of the small subunits from the 80S complexes 8 (Fig. 1A). After extracting the RNA, sequencing
libraries are made of each fraction using template-switching, which enables the use of ultra-low input
material (1 ng) 9. Because our method captures different populations of ribosomes than TCP-seq, we will
refer to our modified protocol as “Ribosome Complex Profiling” (RCP-seq).
Here, we use RCP-seq to capture footprints of both 80S ribosomes and small ribosomal subunits across the
transcriptome of a developing zebrafish embryo (Fig. 1A, Methods). Mapping scanning small subunits over
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
3
5’ UTRs allows us to distinguish three distinct phases during translational initiation: 1) recruitment of small
subunits to the mRNAs, 2) progression to the start codon, and 3) conversion of scanning to elongating
ribosomes.
Results
To investigate the regulation of translation initiation in a vertebrate, we performed RCP-seq during
zebrafish embryo development (see Methods and Supplementary Notes). As expected under the scanning
model of translation, the footprints from the small subunit fraction predominantly mapped to the 5’ UTR
of the transcripts, while the elongating 80S footprints mapped to the protein coding region (CDS). A sharp
divide between the fractions occurred at the start codon consistent with the conversion of scanning 43S
PICs to elongating 80S ribosomes (Fig. 1B). Here, the distribution of footprint lengths also revealed a range
of ribosomal initiation conformations similar to those previously reported in yeast (Fig. 1D, Fig. S1) 8. As
previously reported for TCP-seq, tRNA species contained within ribosomes are also selectively protected
by RCP-seq, and consistent with capturing scanning ribosomes we found initiator MET-tRNA strongly
enriched in the small subunit fraction (Fig. 1C). Taken together, these observations provide strong support
for the selective capture of footprints from small subunits with RCP-seq.
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
4
Figure 1: RCP-seq selectively captures 80S ribosomes and small subunits in zebrafish.
A) Schematic representation of RCP-seq protocol. B) Coverage of RCP-seq reads across all transcripts. Footprints from small
subunits (blue) map predominantly to 5’ UTRs while 80S footprints (orange) map predominantly to coding regions (CDS). C)
Abundance of tRNA species (x-axis) and false discovery rate (FDR) (y-axis) between the RCP-seq small subunit (40S) and 80S
fractions. Initiator Met-tRNA is highlighted (blue). D) Over-representation of RCP-seq small subunit (upper) and 80S (lower)
fraction footprints around start codons. Counts of 5’ (left) or 3’ (right) ends of fragments are summed across highly expressed
genes (>= 10 FPKM). The barplots show the proportion of read counts per position (x-axis), while the heatmaps show the same
counts stratified by length (y-axis) and coloured by total count.
We first sought to understand how the 43S PIC is recruited to the mRNA. Previous studies have suggested
two alternative models for the 43S PIC binding to mRNA 2. In the first, mRNA is “threaded” through the
mRNA channel of the complex while in the second, mRNA “slots” directly into the channel, possibly
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
5
leading to suboptimal scanning of the first nucleotides (Fig. 2A) 2. The slotting and threading models are
predicted to lead to substantially different profiles of protected fragments over the 5’ end (Fig. 2A).
Zebrafish mRNAs have a strong enrichment of small subunit footprints coinciding with the 5’ end of
transcripts (Fig. S2). The 5’ peak is not present in non-coding RNAs arguing that it is a feature only of
translated RNA molecules and not an artifact of the method (Fig. S3). The start of these footprints all
coincide with the transcription start site, and have a wide range of read lengths from the lower detection
limit (~15 nt) up to about 80nt which is slightly longer than scanning 43S PICs (Fig. 2B and Fig. S2, S4).
The majority of footprints downstream of this peak corresponds to the range commonly reported for 43S
PICs (60-70 nt) 8,10. A similar pattern was also observed when realigning data from TCP-seq in yeast to
high-resolution mapping of transcription start sites 11 (Fig. S5, Methods). These patterns of increasing
lengths of small subunit footprints at the 5’ end of the transcript up to the size of the longest small subunit
footprints are consistent with footprints from successive threading of the transcript through the mRNA
channel of the 43S PIC complex.
Under the threading model, the cap-binding initiation factor eIF4E is placed at the leading edge of the 43S
PIC and mRNA is threaded through the mRNA-binding channel 2. To test the response of the 5’ end peaks
to eIF4E inhibition, we sequestered eIF4E using the small molecule inhibitor 4Ei-10 12 (a cell-permeable
prodrug improving upon 4Ei-1 13) thereby specifically blocking eIF4E-cap binding, leading to a small but
general inhibition of translation (Fig. S6, Supplementary Notes) followed by RCP-seq. This resulted in a
global depletion of 5’ peaks with the ratio of reads at the 5’ peak relative to reads internal to the 5’ UTR
reduced to ~63% of WT levels. (Fig. 2C, Fig. S3, S7). In transcripts with very short 5’ UTRs only threading
is expected to be able to initiate translation as slotting would deposit the small subunit too far downstream
to scan the start codon 2,14. Consistently, we observed a strong 5’ peak reduction in response to eIF4E
inhibition in transcripts initiated through the Translation Initiator of Short 5’ UTR (TISU) motif (Fig. S3E-
F), in line with previous reports that these transcripts are eIF4E-sensitive 14. Collectively, this suggests
threading is dependent on eIF4E and is a common recruitment pathway during early development.
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
6
We next asked which features could influence the recruitment of 43S PICs to the 5’ cap. To measure the
amount of 43S PICs present on 5’ UTRs, we defined the scanning efficiency (SE) as the number of small
subunit footprints over a 5’ UTR relative to its mRNA abundance (Fig. S8). This metric is conceptually
identical to the widely used translational efficiency (TE), which measures elongating ribosomes relative to
mRNA abundance 15 (Supplementary Notes). Using this metric, we observed that transcripts with a 5’ C,
and to a lesser extent 5’ T, showed reduced SE and TE compared to transcripts beginning with an A or G
(Fig. 2D, Fig. S9A). This is consistent with biochemical studies which have shown that transcripts
beginning with a pyrimidine (C/T) have a lower affinity for eIF4E binding than those starting with a purine
(A/G) 16,17. An initial C is a feature of transcripts containing a 5’ Terminal Oligo-Pyrimidine (TOP) tract, a
motif often present in mRNAs encoding the protein synthesis machinery and a target of mTOR-mediated
translation control 16,17. We found that while C has an effect, the overall effect on SE and TE reduction is
dominated by mRNAs with the TOP-motif (Fig. 2E, Fig. S9B). This demonstrates that during early
development a reduced number of 43S PIC are recruited to TOP-motif containing transcripts resulting in
reduced translation.
Figure 2: 43S PIC recruitment and impact of 5’ transcript features.
A) Schematic representation of two canonical recruitment models (upper panel): “Threading” (left) and “Slotting” (right), the
resulting protected fragments (middle panel) and the location of the mapped reads relative to the transcription start site (bottom
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
7
panel). B) Heatmap of counts from 3’ ends of small subunit reads stratified by length (y-axis) over each position (x-axis) relative
to transcription start site. Dotted line shows the beginning of the transcript with all reads ending at the enriched diagonal start. C)
Small subunit 5’ counts, relative to transcription start site. Dotted lines show the median values per condition (3 conditions shown,
control, 0.1 μM and 10 μM 4Ei-10 treatment). D-E) Empirical cumulative density of scanning efficiency for highly expressed
transcripts (>10 FPKM). D) Transcripts colored by their first nucleotide. An initial pyrimidine (C/T) results in lower scanning
efficiency than a purine (A/G) (C: p < 2.2x10-16, T: p < 2.2x10-16; number of transcripts per group, A=10388, C=1663, G=8335,
T=967). E) Transcripts starting with a TOP motif show reduced scanning efficiency (p < 2.2x10-16), while non-TOP transcripts
starting with a C show reduced SE but small effect compared to the other non-TOP transcripts (number of transcripts per group
Not TOP C starting = 1257, Not TOP other = 19690, TOP = 406).
As the 43S PIC progresses through the 5’ UTR, it can encounter obstacles that can lead to termination of
scanning. The RCP-seq data revealed this as a slight decline of scanning ribosomes throughout the 5’ UTR
(Fig. S10A). Under the assumption that averaged across all transcripts scanning proceeds at a uniform pace
throughout the 5’ UTR, we compared the density of small subunit complexes of all transcripts at the 5’ end
of the mRNA to the density proximal to the start codon (Fig. S10B). Based on this analysis, we estimate
that on average across all transcripts about 68 % of all ribosomes recruited to the 5’ end reach the start
codon. The loss of scanning ribosomes is largely contingent on whether the 5’ UTR contains one or more
upstream open reading frames (uORF)18,19 (Table S1) with only a modest correlation (Spearman’s rho: -
0.03) with 5’ UTR length if you control for number of uORFs (Fig. S11). In transcripts that lack a uORF,
scanning overall maintains high processivity, consistent with previous reports 20, with a median of 95% of
ribosomes retained. Collectively, this argues that scanning is highly stable, and globally regulated through
5’ UTR elements promoting disassociation.
In transcripts containing uORFs, the CDS is translated either from ribosomes that fail to recognise the often
sub-optimal uORF TIS 7,21,22,23, or by reinitiating ribosomes that continue scanning after translating the
uORF 24,25. Therefore, uORFs typically lead to reduced protein synthesis by consuming scanning 43S PICs
(Fig. 3A-C) 26. Consistent with our global estimates, we find a local decline of 43S PIC footprints coinciding
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
8
with an increase in 80S footprints at uORF TIS’ (Fig. 3D). The ratio of the 43S PIC density upstream vs
downstream of a uORF TIS can therefore quantify to what extent uORFs consume scanning 43S PICs (Fig.
3D). As expected, uORFs starting with an ATG start codon (Fig. 3E-F) and with a TIS context similar to
the Kozak sequence (Fig. 3E) have the highest 43S PIC consumption.
Surprisingly, we found the ability of the small subunit to resume scanning after uORF translation to be
highly dependent on the choice of stop codon. For proteins, TAA and TGA have been reported as the most
and least efficient termination codons, respectively 27. Consistently, uORFs with TGA have the greatest
reduction of downstream scanning small subunits (Fig. 3F), the highest density of downstream 80S
footprints (Fig. S12A) and the lowest ratio of small subunit to 80S complexes directly over their stop codons
(Fig. S12B). This less efficient stop codon recognition globally leads to a small, but significant effect on
the TE of the downstream CDS (Fig. S12C). This suggests that failure to recognise a stop codon results in
extended uORF translation and decreased rates of reinitiation after the translation of the extended uORF 28–
31. The choice of uORF stop codon can therefore regulate synthesis of the downstream protein.
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
9
Figure 3: uORFs reduce the number of 43S PICs scanning across 5’ UTRs.
A-C) The impact of the number of uORFs on (A) scanning subunits on 5’ UTR, (B) the translational efficiency of the 5’ UTR, and
(C) the translational efficiency of the protein (*** = p-values < 0.001). D) Coverage of small subunit (40S) footprints (upper, blue)
and ribo-seq 80S complex footprints (lower, orange) in fixed windows of 100 nt up- and downstream of the first ATG uORF. E)
Heatmaps showing the rate of scanning subunit consumption as measured by the ratio of small subunit reads upstream versus
downstream of all uORFs stratified by surrounding Kozak score and start codon. F) Same as E, but with ranking of start and stop
codon.
Whether a 43S PIC will recognize the TIS and trigger initiation of translation depends on the sequence
surrounding the start codon. For many species, studies have defined an optimal consensus sequence for
translation initiation (the Kozak sequence 32,33), often defined from indirect measures such as sequence
conservation 34 or reporter protein expression 35. Uniquely, RCP-seq enables us to directly measure the
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
10
average initiation rate (IR) on individual transcripts as the ratio of 80S ribosomes in the CDS to small
subunit complexes in the 5’ UTR (Fig. S8). By calculating the median IR of all transcripts containing a
specific nucleotide at a specific position, this model revealed that the consensus that maximized IR is
identical to the known zebrafish Kozak sequence 34 (Fig. 4A). This model, however, considers positions
independently and therefore only reflects an average over sequences with high IR and not the efficiency of
any particular sequence. To obtain this, we grouped all genes with identical sequence context and ranked
these sequences by their median IR (Fig. 4B, S13). The resulting ranking was consistent with a previous
assessment of a small number of sequences in zebrafish 34, but surprisingly revealed that the Kozak
sequence is not the optimal context. The highest scoring sequence was AGGCATG which differs by two
bases (G at -3 and -2). More surprisingly, several sequences that differ strongly to the reported Kozak
sequence rank above it. To test this ranking, we constructed GFP mRNA reporters with three different
initiation sequences, but otherwise identical: 1) AAGC: a sequence highly similar to the Kozak but with
low IR, 2) AAAC: the Kozak sequence previously defined for zebrafish, and 3) TGGA: a sequence differing
at 4 bases from the Kozak, but with greater IR. The translational efficiency (see Methods) of these reporters
when injected into zebrafish embryos confirmed our direct measurements from IR (Fig. 4C). Taken together
with previous reports of weaker than expected correlations between Kozak sequences and translation 36,37,
this demonstrates that a Kozak-similarity measure does not capture the complexity of start codon
recognition, but can be obtained by transcriptome-wide quantification through methods such as RCP-seq.
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
11
Figure 4: Direct measurements of initiation rate.
A) Median initiation rates (IR) for all transcripts containing nucleotide (y-axis) at a specific position (x-axis) relative to the protein
TIS. The zebrafish Kozak sequence is highlighted with black borders (AAACATGGC). B) Median IR for entire sequence from -4 to
-1 (upper) with corresponding Kozak strength (lower). Arrows indicate the sequences selected for reporter constructs. C) Relative
protein abundance for GFP reporter constructs for three different initiation contexts, as measured by the protein/RNA abundance
ratios in zebrafish embryos at 24 hours post fertilization. D) Gene set enrichment analysis of ranked genes based on Initiation Rate
(IR) of start contexts. Q-value (color) is based on False Discovery Rate. Gene ratio (x-axis) and size of dots represent the number
of genes in each category out of the 2489 included in the analysis. Only significant terms shown. E) Cumulative fraction of the fold
change of observed / expected IR between ER-associated genes vs other genes (background). The median fold change (arrow) is
1.68 (number of transcripts per group ER = 449, background = 8648).
Given the range of observed IRs we next asked whether the efficiency of initiation could be linked to gene
function. Using gene set enrichment analysis we identified gene ontologies associated with extreme IR
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
12
values (Fig. 4D, Table S2). Interestingly, genes with the highest IR were strongly enriched for membrane
genes and other proteins synthesized exclusively at the endoplasmic reticulum (ER). Overall, these genes
exhibited more CDS translation relative to their abundance of scanning compared to genes translated in the
cytosol. To investigate whether this increased initiation is simply a consequence of better start codon
contexts, we calculated their observed IR relative to the expected IR given the initiation sequence. This
revealed a significantly higher IR compared to other genes with identical initiation contexts (Fig. 4E,
median enriched 1.68 fold, p < 0.009) suggesting that these genes benefit from an overall increased
initiation rate unrelated to the initiation sequence.
Discussion
In this study we expanded the TCP-seq protocol in two key aspects: 1) we capture all small subunits, not
only those that co-occur on transcripts with 80S ribosomes, and 2) we use template-switching in the library
preparation to enable the use of less input material 9. Our modified RCP-seq therefore captures ribosomal
complexes globally from all stages of the translation process and can be easily applied to other systems
with limited input material, such as specific polysomal fractions or cell types. We used RCP-seq to study
the dynamics of translation initiation during early stages of development in a vertebrate system, zebrafish.
The longer 5’ UTRs of zebrafish allows for a detailed analysis of initiation by spatial separation of
recruitment, scanning and start codon recognition.
Our data supports the threading model of ribosome recruitment to mRNA. At the 5’ end of mRNAs we
observed a “ladder” of differentially sized fragments with 5’ ends coinciding with the transcription start
site (Fig 2B). Fragment sizes shorter than the length of the 40S mRNA tunnel are consistent with the mRNA
gradually entering the tunnel, but conflicts with a slotting model where single-sized fragments would be
expected (Fig 2A). However, two alternative explanations could also potentially account for these
fragments. In the first, the SSU could be slotted adjacent to the 5’ cap, but then proceed to back-slide in the
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
13
5’ direction. However, previous studies have shown that mRNA-binding by factors eIF4A, eIF4B/H and
eIF4F prevent the SSU from back-sliding 38,39, which makes it unlikely that the abundant 5’ reads
(suggesting a frequent occurrence) are due to back-sliding. The second possibility is that these reads are
simply 3’-to-5’ degradation intermediates. However, two observations argue against this possibility: first,
non-coding RNAs have very few 5’ reads arguing for a translation-dependent origin (Fig. S3G), and second,
sequencing reads from degradation intermediates (and other possible artifacts) would be expected to
increase when ribosome scanning is inhibited. Instead, upon eIF4E inhibition we observe that these short
5’ fragments disappear together with fragments derived from SSU scanning (Fig. 2C). We therefore
conclude that threading of mRNAs is the most likely explanation for the presence of these fragments.
Moreover, TCP-seq libraries from yeast realigned to CAGE-defined transcription start sites revealed a
similar distribution of short 5’ fragments (Fig. S5) supporting threading as a universal mechanism,
Our data allowed to analyze the global processivity of scanning ribosomes. We found that the majority of
scanning ribosomes reach the protein coding start codon and identified uORFs as a major cause of
detachment. Consistent with previous studies 25,40,41,42, we find that the choice of stop codon affects uORF
termination across the transcriptome, and furthermore that poor stop codons lead to an increase of read-
through 80S ribosomes, decreasing the ability of SSUs to reinitiate at the CDS. This results in a reduction
of CDS translation demonstrating that the choice of uORF stop codon can globally affect protein
expression.
Finally, by accounting for the number of small ribosomal subunit complexes available for initiation, we
confirmed previous observations that the Kozak sequence provides a strong initiation context. Nevertheless,
our data revealed that there are additional endogenous sequences that give rise to equal or better rates of
initiation. This is consistent with reports of weaker than expected correlations between Kozak sequences
and gene expression in human 36 and yeast 37. We furthermore found that genes with the most efficient
initiation rates were strongly enriched for proteins synthesized in the ER. While mRNAs at the ER have
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
14
been previously shown to be more efficiently translated than mRNAs in the cytosol 43,44, our results further
show that this increased translation is independent of the initiation sequence, but rather a result of SSU
complexes proceeding more efficiently to translation elongation at the ER. Together, this shows that the IR
metric can capture regulation of initiation in subsets of mRNAs and reveal novel features of
compartmentalized translation.
Overall, our approach enables the deconvolution of translation initiation into distinct substeps. The RCP-
seq protocol can be further applied to study samples with limited input material, which will allow addressing
heterogeneity and specialization of the translation machinery, compartmentalized translation or tissue-
specific translation. This opens for the possibility to obtain novel insights into scanning and initiating
mechanisms across organisms and disease models.
Methods
Embryo sample collection and crosslinking
D. rerio embryonic samples were collected using standard zebrafish husbandry 45. In short, AB males and
females were separated the day before mating. Shortly after first light, fish were put together and allowed
to mate for 10 min. Embryos were collected in E3 medium (5 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl2,
0.33 mM MgSO4,), cleaned, and dechorionated using pronase (1 mg/ml, Sigma) for 5 minutes. Embryos
were cleaned thoroughly after dechorionation and grown on 1% agarose plates containing E3 medium until
the desired stage. Embryos were staged-matched for sample collection (Table S3). 200 embryos were
transferred to 2 ml Eppendorf tubes per sample and washed twice in PBS with protease inhibitors (1:100
dilution, cOmplete, Mini, EDTA-free protease inhibitor cocktail) immediately prior to crosslinking, and
left in 250 μl.
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
15
Embryos were snap-chilled by addition of 750 μl ice-cold PBS with 4% paraformaldehyde (PFA, freshly
prepared) and immediately placed on ice. Samples were incubated for 15 min on ice, with gentle agitation.
PFA medium was fully removed and 1 ml lysis buffer (20 mM HEPES-KOH pH 7.4, 100 mM KCl, 2 mM
MgCl2) added. Glycine was added to a final 0.25 M concentration for PFA quenching, and samples
incubated for 5 min on ice. Embryos were then washed twice in lysis buffer and resuspended in lysis buffer
supplemented with 0.5 mM DTT, 2 μl Superase-In (RNase Inhibitor) and 1x protease inhibitor. Samples
were immediately flash frozen in liquid nitrogen and stored at -80oC until used.
Separation of ribosomal small subunit and ribosomal complexes
Samples were lysed in a cold room (4oC) by first shaking at 1,300 rpm for 10 min, followed by passing six
times through a 27G needle. Samples were clarified by centrifuging for 15 min at 14,500g, 4oC. The OD260
absorbance of the supernatant was measured by nanodrop. 1/10th of each lysate was kept for RNA
sequencing of that sample. Based on the absorbance readings, samples were digested using RNase I (0.0383
U x sample volume (μl) x OD 260 absorbance) for 45 min, at 23oC, 300 rpm shaking. Linear sucrose
gradients were made from 5% to 30% sucrose solutions (containing 50 mM Tris-HCl pH 7.0, 50 mM
NH4Cl, 4 mM MgCl2, 1 mM DTT) with Biocomp Gradient Station (long cap programme, 5% to 30%).
Gradients were cooled down for 45 min at 4oC. Samples were layered on top of each gradient and tubes
were centrifuged in a SW-41 rotor (Beckman-Coulter) at 4oC, 38,000 rpm for 4 hours. The gradients were
fractionated using Biocomp Gradient Station and the small subunit and 80S fractions were identified and
collected by monitoring the absorbance profile at 254 nm.
RNA isolation
The collected fractions and RNA controls were supplemented with 1% SDS, 10 mM EDTA, 10 mM Tris-
HCl pH 7.4, and 10 mM glycine. One volume of phenol:chloroform.isoamyl alcohol (pH 4.5) was added
to each sample, immediately placed on a shaker at 65oC, 1,300 rpm for 45 min. After a 5 min centrifugation
at 15,000g at room temperature, the aqueous phase was transferred to a new tube and precipitated by
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
16
addition of 20 μg glycogen, 0.1 volume 3M sodium acetate (pH 4.5) and 2.5 volumes absolute ethanol.
Samples were precipitated at -20oC for at least 3 hours. RNA was pelleted by centrifugation at 21,000g for
40 min, followed by two washes with 80% ethanol and 20 min centrifugation at 21,000g, 4oC. After drying
the pellet, it was resuspended in 17 μl water, and concentration assessed on nanodrop.
Construction of RCP-seq sequencing libraries
RNA samples were end-repaired by first incubating for 2 min at 80oC and on ice for 5 min, followed by the
addition of 2 μl 10x T4 PNK buffer, 1 μl SUPERase-In, and 1 μl T4 PNK (10U/μl), and incubated for 2 h
at 37oC. rRNA fragments were removed by using Ribo-Zero Magnetic Gold Kit (Illumina) following
manufacturer’s instructions and purifying RNA using clean & concentrator-5 spin columns (Zymo
Research) in a final volume of 13 μl. Libraries were constructed using the TaKaRa SMARTer smRNA-Seq
Kit for Illumina, following manufacturers’ instructions without intermediate freezing points. In general,
ATP was used for polyadenylation depending on the amount of starting material (less than 25 ng, no ATP).
The number of cycles used was optimized per sample in the PCR amplification step and samples were
eluted in a final volume of 20 μl. RCP-seq library sizes were checked on Agilent Bioanalyzer DNA High
Sensitivity chips. Depending on size distribution, small and/or large fragments were removed by using
AMPure XP beads (in order to remove adapter dimers and too large fragments not resulting from ribosomal
protection). Sequencing was performed on a NextSeq 500 (Illumina) at the Norwegian Sequencing Centre
in Oslo, high output mode with single reads of 75 bp or 150 bp.
eIF4E inhibitor assays
The eIF4E inhibitor 4Ei-10 (or 6a 12) was synthesized at the Wagner lab (University of Minnesota, USA).
The inhibitor was diluted in DMSO to a concentration of 100 μM, and kept frozen as stock. Further dilutions
were performed in water. One nl of two concentrations (10 μM and 100 nM) were injected into
dechorionated zebrafish embryos between 1-4 cell stages, in parallel with DMSO injections as controls.
Embryos were allowed to continue development and samples were collected for RCP-seq at 64-cell and
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
17
Shield stages (as described above). Flash frozen samples at Shield stages were also collected for polysome
profiling in order to quantify effects on global translation for both 4Ei-10 and DMSO injected samples.
Polysome profiling
Control (DMSO-injected) and inhibitor-injected samples were collected at Shield stage and flash frozen to
halt ribosomes. Samples were lysed in polysome lysis buffer (10 mM Tris-HCl (pH 7.4), 5 mM MgCl2, 100
mM KCl, 1% Triton X-100) by shaking for 10 min at 1,300 rpm and 4oC, followed by 6x lysing through a
27-gauge needle. Samples were centrifuged at 14,500 g, for 15 min, 4oC. The supernatant was transferred
to a new tube and its absorbance at 260 nm was measured with Nanodrop One to quantify RNA content.
Linear sucrose gradients (15-45%) were prepared in a buffer containing 20mM HEPES KOH (pH 7.4), 5
mM MgCl2, 100 mM KCl, 2 mM DTT and 10 μl Superase-In. Gradients were cooled down for 45 min at
4oC. Samples were layered atop the gradients and ultracentrifuged in a Beckman-Coulter SW-41 rotor at
36,000 rpm, 2 h, 4oC. Polysome profiles of the gradients were obtained by in-line 254 nm absorbance
measuring with Biocomp Gradient Station.
Reporter assays
Translational efficiency of initiation contexts was tested using eGFP reporters. Three eGFP reporters with
different initiation contexts were synthesized. The coding sequence of eGFP was amplified from pCS2+-
eGFP vector using High Fidelity Phusion MasterMix (ThermoFisher, #F-531L). The forward primers for
the PCR included the SP6 promoter sequence, followed by 26 bp of eIF3d leader sequence and the
respective start codon context; the reverse primer (M13-rev) was located after the common SV40
termination signal included in the pCS2+ vector backbone (see Table 1). Following synthesis by a two-step
PCR, samples were gel-purified and RNA was synthesized with SP6 mMessage mMachine (Thermo Fisher,
#AM1340), following manufacturer’s recommendations, and cleaned-up using Zymo RNA Clean &
Concentrator-25 columns (Zymo Research, #R1013).
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
18
To control for GFP expression changes, we further synthesized RFP mRNA with fixed start codon context.
The coding sequence of RFP was amplified from pT2KXIGdeltaIn-MCS-birA-tagRFP vector (AddGene
#58378) using High Fidelity Phusion MasterMix. The forward primers for the PCR included the SP6
promoter sequence and the reverse Primer was located after the common SV40 termination signal included
in the pT2KXIG vector backbone (see Table 1). Following synthesis by a two-step PCR, samples were
processed as above to obtain mRNA.
Stock RNA solutions for injections at 120 ng/ul were made based on nanodrop & Qubit RNA concentration
measurements. RNA solutions containing RFP and eGFP reporters were co-injected at a final concentration
of 50ng/ul per reporter (with phenol red added for injection visualization) and injected at 1 nl per embryo
at 2-4 cell stage (dechorionated embryos). Embryos were collected and dechorionated as described above.
Groups of 25 embryos for each of the samples were collected at 24 hpf and used for qRT-PCR and eGFP
protein quantification. For qRT-PCR, RNA was extracted using Trizol, and processed following the
protocol previously described 46. eGFP RNA expression was quantified against RFP for injection control
and endogenous controls serp1 and Elf1a.
For eGFP protein quantification, the 25 embryos collected were homogenized in 100 mM Tris (pH 7.5)
with 1% Triton X-100, and processed as described in 34. After quantifying total protein concentration with
the Pierce BCA kit (Thermo Scientific), the samples were diluted to equal total protein concentrations. A
Qubit fluorometer (Life Technologies) was used to measure eGFP fluorescence, using 9 technical replicates
for each sample, and with GFP dilutions as measurement controls.
Translational efficiency was calculated as described previously 34, by quantifying RNA with qRT-PCR and
quantifying eGFP protein by measuring eGFP fluorescence in Qubit. Relative eGFP protein (average of all
9 technical replicates for each sample) was calculated compared to AAAC (zebrafish Kozak), divided by
the relative abundance of eGFP RNA in each corresponding sample as measured by normalised qPCR
abundance. For the AAAC samples, we compared their eGFP values to the average of the group to get their
individual TE values.
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
19
Table 1: primers used for reporter synthesis and qPCR
Primer name
Sequence
SP6-AAAC-fw
CTTGATTTAGGTGACACTATAGTACGGATTCGTACACCAGTAAAGGCGAAACATG
GTGAGCAAGGGCGAG
SP6-AAGC-fw
CTTGATTTAGGTGACACTATAGTACGGATTCGTACACCAGTAAAGGCGAAGCATG
GTGAGCAAGGGCGAG
SP6-TGGA-fw
CTTGATTTAGGTGACACTATAGTACGGATTCGTACACCAGTAAAGGCGTGGAATG
GTGAGCAAGGGCGAG
M13-rev
GGAAACAGCTATGACCATG
SP6-RFP-AAAC-
fw
CTTGATTTAGGTGACACTATAGTACGGATTCGTACACCAGTAAAGGCGAAACATG
GTGTCTAAGGGCGAAGA
RFP-rev
CAGCCATACCACATTTGTAGAG
eGFP qPCR fw
TCAAGGAGGACGGCAACATC
eGFP qPCR rev
AACTCCAGCAGGACCATGTG
RFP qPCR fw
TGCAGAAGAAAACACTCGGC
RFP qPCR rev
TGTCGGCCTCCTTGATTCTT
Serp1 qPCR fw
GTGGATCAGCGATATTCCAG
Serp1 qPCR rev
AGAGAAGCGGAATGGTCGAG
Elf1a qPCR fw
CTCCTCTTGGTCGCTTTGCT
Elf1a qPCR rev
GCCTTCTGTGCAGACTTTGTGA
Transcript definitions
The analysis was performed on the most highly expressed transcript from each gene, calculated from total
RNA-seq coverage (datasets are described in Table S4). Cap analysis gene expression (CAGE) was used
to update the 5’ UTR on a per sample basis as follows. The highest CAGE peak was selected in a search
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
20
region from the 3’ most end of the transcript, to the greater of; the 5’ most end of the 5’ UTR or 1000 nt
upstream of the annotated start codon (If the highest CAGE peak was called downstream of the protein TIS
the transcript was excluded for further analysis). Transcripts were excluded from further analysis if they
overlapped with the most highly expressed transcript of another gene, or with an annotated non-coding
transcript (defined as all Ensembl transcripts with biotypes other than protein_coding).
TISU motif containing transcripts were defined as those that with 5’ UTRs of <= 30 nt in length and a PWM
scores against the consensus TISU sequence SAASATGGCGGC (where S is C or G) of >= -12. TOP motif
containing transcripts were defined as those beginning with a C followed by at least 4 T nucleotides. Kozak
sequence strength was determined through PWM scores against the zebrafish kozak matrix taken from 34.
Three genes (ENSDARG00000077330, ENSDARG00000102873, ENSDARG00000089382) contained
strong coverage peaks across repetitive regions in their 3’ UTR and were excluded from plots showing
densities over 3’ UTRs (Fig. 1B, S14).
Read trimming and alignment
RCP-seq reads were trimmed with cutadapt searching for the “AAAAAAAAAA” added to the 3’ of each
fragment during library preparation, allowing 1 mismatch and at least 5 nt of overlap. The first 3 nt of each
read were removed, reads shorter than 15 nt after trimming were discarded. The remaining reads were
aligned to, in order, the PhiX genome, rRNA from the silva database47 (version 119), organism-specific
ncRNA as defined by Ensembl (zebrafish GRCz10), organism-specific tRNA produced with tRNA-scan
SE 48 (default settings). Reads that did not match to any of the above were aligned to the D. rerio GRCz10
genome. Total RNA-seq reads were trimmed and aligned to the D. rerio GRCz10 genome. Ribo-seq reads
were trimmed and aligned to rRNA and ncRNA as above, unaligned reads where then aligned to the D.
rerio GRCz10 genome. CAGE reads were trimmed and aligned to the D. rerio GRCz10 genome.
Alignments were performed using tophat2 49 against the D. rerio GRCz10 genome and ensembl version 81
gene annotations (Table S5), reporting up to 20 hits for reads mapping to multiple locations (later filtered
with MAPQ, see below).
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
21
TCP-seq data from Saccharomyces cerevisiae 8 (SRA: SRP074093) was processed as above (with the
exception of 1st 3nt removal and using STAR as aligner with default parameters) to the R64_1_1 genome
with Ensembl version 79 gene annotations. S. cerevisiae 5’ UTR were defined with CAGE 11 (GEO:
GSE69384).
RCP-seq fractionation
The RCP-seq sedimentation fractions corresponding to small subunits and 80S complexes were determined
from sequencing all sedimentation fractions. Based on coverage profiles (Fig. S14), fractions 12-14 were
determined to contain small subunit fragments. Fractions 18-19 were determined to contain 80S complex
fragments. RCP-seq small subunit counts in this study are reported from a pooled set of all relevant fractions
(Table S3), unless otherwise stated. In order to determine the maximum length of small subunit RCP-seq
reads, one sample was re-sequenced for 150 cycles (shown in Fig. S4), as opposed to the 75 cycles used for
the rest of the samples.
Further read processing
Inappropriately truncated RCP-seq reads after polyA trimming were updated by extending alignments
where trimmed regions exactly matched transcriptomic references. Unusually high RCP-seq coverage
peaks were removed from transcripts by filtering out reads with the same 5’ and 3’ coordinates that were
present at >=200 times the average coverage of each transcript. A subset of small subunit reads (~25-35 nt
in length), were observed to show 3nt periodicy over the CDS region (Fig. 1D, lower left, CDS region).
This periodicity is indicative of translation, but it was not clear if these reads represent the leaky scanning
of 43S PICs, queued behind translating ribosomes, or footprints of translating complexes, where possibly
the 60S subunit has become detached, before sedimentation. As such, reads corresponding to the length of
typical translating fragments (length 25-35 nt) were considered ambiguous and removed from RCP-seq
small subunit libraries. Conversely the RCP-seq 80S libraries used in figure 1B, were filtered to only
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
22
include the lengths of typical translating fragments (length 25-35 nt), akin to ribo-seq libraries. RCP-seq
reads that mapped to positions overlapping the last 10 nt of each transcript were discarded, to remove 3’
peaks that likely result from polyA selection during the library preparation.
Read counting
The relative enrichment of tRNA species between small subunit and 80S complex fractions was calculated
with edgeR 50 using a binomial generalized log-linear model and likelihood ratio test. Read counts were
summed per tRNA anticodon type. CAGE reads counts were normalised using the power law method from
the cageR package 51. The FPKMs (fragments per kilobase, per million mapped reads) of RCP-seq, Ribo-
seq and Total RNA-seq reads with a MAPQ >= 10 were calculated for transcript 5’ UTR, CDS, and 3’ UTR
regions. Reads that overlapped multiple regions were preferentially assigned to CDS > 5’ UTR > 3’ UTR.
Fragment lengths heatmaps and fragment distribution
Metaplots of RCP-seq footprint distributions were plotted in windows across at the 5’ most end of
transcripts (TSS) or around protein TIS, for transcripts with at total RNA-seq FPKM>10 and 5’ UTRs at
least 100 nt long. Fragment counts are assigned to either the 5’ or the 3’ end of the fragment. The heatmaps
of counts per fragment length are coloured by sum of counts from all transcripts at a given position for a
given fragment length. The proportion of coverage window counts are summed for all transcripts at given
position. The transcripts of genes ENSDARG00000036180, ENSDARG0000001479 were observed to
have strong artefactual peaks caused by premature read trimming in polyA regions upstream of the protein
TIS and were removed from the TIS plots in Fig. 1D.
For the yeast TCP-seq data, the footprints were similarly plotted as heatmaps for the area around the TSS
in Fig. S5. Using a median based filter to remove transcripts that had extreme peaks in the +2 to +40 region
relative to TSS (total of 21 genes). Filter removed transcript if: peak at any position in +2 to +40 > median
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
23
footprints in transcript per position * 99 quantile rank of transcript’s footprints per position + 99.9 quantile
rank of the medians of all transcripts.
Scaled coverage meta plots
The coverage, or 5’ counts of RCP-seq and ribo-seq reads with a MAPQ >= 10 was calculated across
transcript 5’ UTRs, CDS, and 3’ UTR. Transcripts with 5’ UTRs, CDS, and 3’ UTR greater than a length
cutoff (typically 100 nt) were scaled to the length value. Values are displayed as the sum of all selected
transcripts, mean normalised values or z-score for all selected transcripts. Counts from each transcript
normalised by z-score across the whole transcript allow for comparisons of transcripts across wide
expression ranges.
Estimates of 43S PIC loss across 5’ UTRs
The coverage of small subunit footprints in regions proximal to the beginning of the transcript and the
protein TIS were used to infer the number of 43S PICs recruited to a transcript and those available for
initiation at the protein TIS. These regions were defined based on coverage metaplots (Fig. S10) as +70 to
+120 nt relative to the beginning of the transcript and -100 to -50 nt relative to the protein TIS. The ratio of
coverage in these regions was used to estimate the loss of 43S PICs across 5’ UTRs for all protein coding
transcripts, with RNA-seq FPKM>= 10 and 5’ UTR >= 220 nt in length.
5’ feature plots (Fig. 2)
Empirical cumulative density for scanning efficiency (Fig. 2D,E) and translational efficiency (Fig. S9) were
plotted for all protein coding transcripts with >= 10 RNA FPKM. For groups of transcripts starting with; i)
an A, C, G or T; or ii) transcripts starting with a TOP motif, transcripts starting without a TOP motif and
also starting with a C, and transcripts starting without a TOP motif and also starting with a A, G or T.
uORF plots (Fig. 3)
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
24
Upstream open reading frame coverage metaplots were produced for all protein coding transcripts with an
ATG uORF starting > 100 nt from the 5’ most end of the transcript and the TIS, centering on the first (5’
most) ATG uORF within the transcript.
The proportion of small subunit reads mapping upstream (from the beginning of the transcript to the uORF
start codon) or downstream of the uORF (uORF start codon to the protein TIS) were calculated for all ATG,
CTG and GTG uORFs, starting >= 50 nt from the beginning of the transcript and the protein TIS, in
transcripts with RNA FPKM>=1, stratified by uORF start codon, and Kozak strength quantile. Similarly
the proportion of small subunit reads mapping upstream (from the beginning of the transcript to the uORF
start codon) or downstream of the (uORF stop codon to the protein TIS) were calculated for all ATG, CTG
and GTG uORFs starting >= 50 nt from the beginning of the transcript and >=50 nt from uORF stop codon
to the protein TIS, in transcripts with >=1 RNA FPKM, stratified by uORF start codon, and uORF stop
codon.
Scanning efficiency, translational efficiency and 5’ UTR translational efficiency (Fig. S8, S15) were
calculated for all protein coding transcripts with 5’ UTRs >= 100 nt in length and RNA FPKM>=1, that
contained 0,1,2,3 or 4 ATG uORFs. Statistical significance was reported between transcripts containing 0
ATG uORFs versus those containing 1-4 ATG uORFs as: W = 112420000, p-value < 2.2x10-16 for 5’ UTR
RCP-seq 40S FPKM, W = 138210000 p-value < 2.2x10-16 for translational efficiency and W = 73953000,
p-value < 2.2x10-16 for 5’ UTR translational efficiency.
The ratio of RCP-seq small subunits to ribo-seq 80S complexes over the uORF stop codon, per uORF stop
codon; the density of 80S complexes between the uORF stop codon and protein TIS, normalised by CDS
total RNA-seq; and the translational efficiency (Fig. S12) were calculated for all ATG, CTG or GTG uORFs
starting >= 50 nt from the beginning of the transcript and >=50 nt from uORF stop codon to the protein
TIS, in transcripts with >=1 RNA FPKM. Statistical significance was reported as the median log2 ratio of
small subunits to 80S over the stop codon of transcripts containing TAA uORFs vs TGA uORFs (W =
250820000, p-value < 2.2x10-16 the median normalised log2 downstream ribo-seq 80S density of TAA
uORFs vs TGA uORFs (W = 1190600000, p-value < 2.2x10-16), and the median log2 translational
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
25
efficiency of transcripts containing TAA uORFs vs TGA uORFs (W = 1298200000, p-value = 6.128x10-
11).
Initiation plots (Fig. 4)
Initiation rates (see Fig. S8) were calculated as the ratio of RCP-seq small subunits in the 5’ UTR to ribo-
seq 80S complexes in the CDS, for all protein coding transcript sequences with RNA FPKM >= 10 and 5’
UTRs >= 100 nt in length. Initiation rates for transcripts were then grouped by each nucleotide, per position
in a -4 to +5 window surrounding the protein coding start codon, excluding the start codon. Median
initiation rate was calculated for nucleotides that were present more than 1000 times. Initiation rates were
also calculated over continuous sequence contexts, using a smaller window of -4 to +3 nucleotides
surrounding protein coding start sites, calculating the median initiation rate for all sequences present >= 20
times in the selected transcripts. This smaller window was used in order to increase the number of transcripts
present in each sequence bin.
GO enrichment analysis
GO analysis was performed with GOrilla, detecting GO terms based on ranking of genes according to
Initiation Rate (IR). P-values computed according to the mHG model, and FDR q-values for the correction
of the p-value for multiple testing. Genes were filtered by leader lengths >= 100nt, CAGE peaks on
transcript > 5 reads, and FPKM of RNA-seq > 0 and FPKM of ribo-seq and SSU libraries > 0. Giving a
total of 449 ER enriched genes and 8648 as the background set.
Statistical testing and plotting
Significance testing was performed in R using the Wilcoxon rank sum test with continuity correction. The
metrics used to investigate relationships between small subunits and 80S footprints are summarised in Fig.
S15. Boxplot upper whiskers extend from the 1st quartile to the largest value no further than 1.5 times the
distance between the first and third quartile, from the first quartile. Boxplot lower whiskers extend from the
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
26
third quartile to the smallest value no further than 1.5 times the distance between the first and third quartile,
from the third quartile.
Data and code availability
Custom scripts used to process the RCP-seq libraries are available at the following link:
https://github.com/agiess/RCP_processing. The RCP-seq libraries have been uploaded to the ENA
database privately until publication (accession number PRJEB33323). Access will be granted upon
request.
Acknowledgements
We wish to thank members of the Valen and Thompson labs (University of Bergen, Norway), Ignatova and
Duncan labs (University of Hamburg, Germany), and Preiss lab (Australian National University, Australia)
for helpful discussions of methodologies. We wish to thank Andrea Pauli (IMP, Austria), Sushma Nagaraja-
Grellscheid and Eric Thompson (University of Bergen, Norway) for providing valuable feedback on the
manuscript.
The project was funded by Bergen Research Foundation, the Norwegian Research Council (#250049) and
core funding from the Sars International Centre for Marine Molecular Biology.
Author contributions
AG, YTC and EV designed the research. YTC, MK, TBB and SS performed the experiments. AG, YTC,
HT and EV analyzed the data. AO and CRW designed and synthesized the 4Ei-10 compound. All authors
discussed the results and contributed to writing the paper.
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
27
References
1. Schwanhäusser, B. et al. Global quantification of mammalian gene expression control. Nature 473,
337–342 (2011).
2. Kumar, P., Hellen, C. U. T. & Pestova, T. V. Toward the mechanism of eIF4F-mediated ribosomal
attachment to mammalian capped mRNAs. Genes Dev. 30, 1573–1588 (2016).
3. Shirokikh, N. E. & Preiss, T. Translation initiation by cap-dependent ribosome recruitment: Recent
insights and open questions. Wiley Interdiscip. Rev. RNA e1473 (2018).
4. Sonenberg, N. & Gingras, A.-C. The mRNA 5′ cap-binding protein eIF4E and control of cell growth.
Current Opinion in Cell Biology 10, 268–275 (1998).
5. Arava, Y. et al. Genome-wide analysis of mRNA translation profiles in Saccharomyces cerevisiae.
Proc. Natl. Acad. Sci. U. S. A. 100, 3889–3894 (2003).
6. Shah, P., Ding, Y., Niemczyk, M., Kudla, G. & Plotkin, J. B. Rate-limiting steps in yeast protein
translation. Cell 153, 1589–1601 (2013).
7. Ingolia, N. T., Ghaemmaghami, S., Newman, J. R. S. & Weissman, J. S. Genome-Wide Analysis in
Vivo of Translation with Nucleotide Resolution Using Ribosome Profiling. Science 324, 218–223
(2009).
8. Archer, S. K., Shirokikh, N. E., Beilharz, T. H. & Preiss, T. Dynamics of ribosome scanning and
recycling revealed by translation complex profiling. Nature 535, 570–574 (2016).
9. Hornstein, N. et al. Ligation-free ribosome profiling of cell type-specific translation in the brain.
Genome Biol. 17, 149 (2016).
10. Kozak, M. & Shatkin, A. J. Migration of 40 S ribosomal subunits on messenger RNA in the presence
of edeine. J. Biol. Chem. 253, 6568–6577 (1978).
11. Wery, M. et al. Nonsense-Mediated Decay Restricts LncRNA Levels in Yeast Unless Blocked by
Double-Stranded RNA Structure. Mol. Cell 61, 379–392 (2016).
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
28
12. Okon, A. et al. Anchimerically Activated ProTides as Inhibitors of Cap-Dependent Translation and
Inducers of Chemosensitization in Mantle Cell Lymphoma. J. Med. Chem. 60, 8131–8144 (2017).
13. Smith, K. A. et al. Transforming Growth Factor-β1 Induced Epithelial Mesenchymal Transition is
blocked by a chemical antagonist of translation factor eIF4E. Sci. Rep. 5, 18233 (2015).
14. Elfakess, R. et al. Unique translation initiation of mRNAs-containing TISU element. Nucleic Acids
Res. 39, 7598–7609 (2011).
15. Ingolia, N. T., Lareau, L. F. & Weissman, J. S. Ribosome profiling of mouse embryonic stem cells
reveals the complexity and dynamics of mammalian proteomes. Cell 147, 789–802 (2011).
16. Tamarkin-Ben-Harush, A., Vasseur, J.-J., Debart, F., Ulitsky, I. & Dikstein, R. Cap-proximal
nucleotides via differential eIF4E binding and alternative promoter usage mediate translational
response to energy stress. Elife 6, (2017).
17. Meyuhas, O. & Kahan, T. The race to decipher the top secrets of TOP mRNAs. Biochim. Biophys.
Acta 1849, 801–811 (2015).
18. Calvo, S. E., Pagliarini, D. J. & Mootha, V. K. Upstream open reading frames cause widespread
reduction of protein expression and are polymorphic among humans. Proc. Natl. Acad. Sci. U. S. A.
106, 7507–7512 (2009).
19. Chew, G.-L. et al. Ribosome profiling reveals resemblance between long non-coding RNAs and 5′
leaders of coding RNAs. Development 140, 2828–2834 (2013).
20. Berthelot, K., Muldoon, M., Rajkowitsch, L., Hughes, J. & McCarthy, J. E. G. Dynamics and
processivity of 40S ribosome scanning on mRNA in yeast. Mol. Microbiol. 51, 987–1001 (2004).
21. Lee, S. et al. Global mapping of translation initiation sites in mammalian cells at single-nucleotide
resolution. Proc. Natl. Acad. Sci. U. S. A. 109, E2424–32 (2012).
22. Fritsch, C. et al. Genome-wide search for novel human uORFs and N-terminal protein extensions
using ribosomal footprinting. Genome Res. 22, 2208–2218 (2012).
23. Kozak, M. Pushing the limits of the scanning mechanism for initiation of translation. Gene 299, 1–
34 (2002).
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
29
24. Kozak, M. Effects of intercistronic length on the efficiency of reinitiation by eucaryotic ribosomes.
Mol. Cell. Biol. 7, 3438–3445 (1987).
25. Grant, C. M. & Hinnebusch, A. G. Effect of sequence context at stop codons on efficiency of
reinitiation in GCN4 translational control. Mol. Cell. Biol. 14, 606–618 (1994).
26. Jackson, R. J., Hellen, C. U. T. & Pestova, T. V. The mechanism of eukaryotic translation initiation
and principles of its regulation. Nat. Rev. Mol. Cell Biol. 11, 113–127 (2010).
27. Bonetti, B., Fu, L., Moon, J. & Bedwell, D. M. The efficiency of translation termination is
determined by a synergistic interplay between upstream and downstream sequences in
Saccharomyces cerevisiae. J. Mol. Biol. 251, 334–345 (1995).
28. Luukkonen, B. G., Tan, W. & Schwartz, S. Efficiency of reinitiation of translation on human
immunodeficiency virus type 1 mRNAs is determined by the length of the upstream open reading
frame and by intercistronic distance. J. Virol. 69, 4086–4094 (1995).
29. Kozak, M. Constraints on reinitiation of translation in mammals. Nucleic Acids Res. 29, 5226–5232
(2001).
30. Szamecz, B. et al. eIF3a cooperates with sequences 5’ of uORF1 to promote resumption of scanning
by post-termination ribosomes for reinitiation on GCN4 mRNA. Genes Dev. 22, 2414–2425 (2008).
31. Mohammad, M. P., Munzarová Pondelícková, V., Zeman, J., Gunišová, S. & Valášek, L. S. In vivo
evidence that eIF3 stays bound to ribosomes elongating and terminating on short upstream ORFs to
promote reinitiation. Nucleic Acids Res. 45, 2658–2674 (2017).
32. Kozak, M. Point mutations define a sequence flanking the AUG initiator codon that modulates
translation by eukaryotic ribosomes. Cell 44, 283–292 (1986).
33. Kozak, M. An analysis of 5’-noncoding sequences from 699 vertebrate messenger RNAs. Nucleic
Acids Res. 15, 8125–8148 (1987).
34. Grzegorski, S. J., Chiari, E. F., Robbins, A., Kish, P. E. & Kahana, A. Natural Variability of Kozak
Sequences Correlates with Function in a Zebrafish Model. PLoS One 9, e108475 (2014).
35. Noderer, W. L. et al. Quantitative analysis of mammalian translation initiation sites by FACS-seq.
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
30
Mol. Syst. Biol. 10, 748 (2014).
36. Vogel, C. et al. Sequence signatures and mRNA concentration can explain two-thirds of protein
abundance variation in a human cell line. Mol. Syst. Biol. 6, (2010).
37. Pop, C. et al. Causal signals between codon bias, mRNA structure, and the efficiency of translation
and elongation. Mol. Syst. Biol. 10, 770 (2014).
38. Siridechadilok, B., Fraser, C. S., Hall, R. J., Doudna, J. A. & Nogales, E. Structural roles for human
translation factor eIF3 in initiation of protein synthesis. Science 310, 1513–1515 (2005).
39. Spirin, A. S. How Does a Scanning Ribosomal Particle Move along the 5′-Untranslated Region of
Eukaryotic mRNA? Brownian Ratchet Model. Biochemistry 48, 10688–10692 (2009).
40. Beznosková, P., Gunišová, S. & Valášek, L. S. Rules of UGA-N decoding by near-cognate tRNAs
and analysis of readthrough on short uORFs in yeast. RNA 22, 456–466 (2016).
41. Cridge, A. G., Crowe-McAuliffe, C., Mathew, S. F. & Tate, W. P. Eukaryotic translational
termination efficiency is influenced by the 3′ nucleotides within the ribosomal mRNA channel.
Nucleic Acids Research 46, 1927–1944 (2018).
42. McCaughan, K. K., Brown, C. M., Dalphin, M. E., Berry, M. J. & Tate, W. P. Translational
termination efficiency in mammals is influenced by the base following the stop codon. Proc. Natl.
Acad. Sci. U. S. A. 92, 5431–5435 (1995).
43. Stephens, S. B. & Nicchitta, C. V. Divergent regulation of protein synthesis in the cytosol and
endoplasmic reticulum compartments of mammalian cells. Mol. Biol. Cell 19, 623–632 (2008).
44. Reid, D. W. & Nicchitta, C. V. Primary role for endoplasmic reticulum-bound ribosomes in cellular
translation identified by ribosome profiling. J. Biol. Chem. 287, 5518–5527 (2012).
45. Avdesh, A. et al. Regular care and maintenance of a zebrafish (Danio rerio) laboratory: an
introduction. J. Vis. Exp. e4196 (2012).
46. Peterson, S. M. & Freeman, J. L. RNA isolation from embryonic zebrafish and cDNA synthesis for
gene expression analysis. J. Vis. Exp. (2009). doi:10.3791/1470
47. Quast, C. et al. The SILVA ribosomal RNA gene database project: improved data processing and
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
31
web-based tools. Nucleic Acids Res. 41, D590–6 (2013).
48. Lowe, T. M. & Eddy, S. R. tRNAscan-SE: a program for improved detection of transfer RNA genes
in genomic sequence. Nucleic Acids Res. 25, 955–964 (1997).
49. Kim, D. et al. TopHat2: accurate alignment of transcriptomes in the presence of insertions, deletions
and gene fusions. Genome Biol. 14, R36 (2013).
50. Robinson, M. D., McCarthy, D. J. & Smyth, G. K. edgeR: a Bioconductor package for differential
expression analysis of digital gene expression data. Bioinformatics 26, 139–140 (2010).
51. Haberle, V., Forrest, A. R. R., Hayashizaki, Y., Carninci, P. & Lenhard, B. CAGEr: precise TSS data
retrieval and high-resolution promoterome mining for integrative analyses. Nucleic Acids Res. 43,
e51 (2015).
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/811810doi: bioRxiv preprint first posted online Oct. 21, 2019;
