Skip to main content
NCBI home page
ACS AuthorChoice logoLink to ACS AuthorChoice
. 2023 Jun 22;18(10):2211–2223. doi: 10.1021/acschembio.3c00166

Direct Nanopore Sequencing for the 17 RNA Modification Types in 36 Locations in the E. coli Ribosome Enables Monitoring of Stress-Dependent Changes

Aaron M Fleming 1,*, Praneeth Bommisetti 1, Songjun Xiao 1, Vahe Bandarian 1, Cynthia J Burrows 1,*
PMCID: PMC10594579  PMID: 37345867

Abstract

graphic file with name cb3c00166_0006.jpg

The bacterium Escherichia coli possesses 16S and 23S rRNA strands that have 36 chemical modification sites with 17 different structures. Nanopore direct RNA sequencing using a protein nanopore sensor and helicase brake, which is also a sensor, was applied to the rRNAs. Nanopore current levels, base calling profile, and helicase dwell times for the modifications relative to unmodified synthetic rRNA controls found signatures for nearly all modifications. Signatures for clustered modifications were determined by selective sequencing of writer knockout E. coli and sequencing of synthetic RNAs utilizing some custom-synthesized nucleotide triphosphates for their preparation. The knowledge of each modification’s signature, apart from 5-methylcytidine, was used to determine how metabolic and cold-shock stress impact rRNA modifications. Metabolic stress resulted in either no change or a decrease, and one site increased in modification occupancy, while cold-shock stress led to either no change or a decrease. The double modification m4Cm1402 resides in 16S rRNA, and it decreased with both stressors. Using the helicase dwell time, it was determined that the N4 methyl group is lost during both stressors, and the 2′-OMe group remained. In the ribosome, this modification stabilizes binding to the mRNA codon at the P-site resulting in increased translational fidelity that is lost during stress. The E. coli genome has seven rRNA operons (rrn), and the earlier studies aligned the nanopore reads to a single operon (rrnA). Here, the reads were aligned to all seven operons to identify operon-specific changes in the 11 pseudouridines. This study demonstrates that direct sequencing for >16 different RNA modifications in a strand is achievable.

Chemical modification of RNA occurs naturally in all phyla of life that includes over 140 different structures.1 These modifications are essential for RNA to adopt specific secondary and tertiary structures.2 They are also used for epitranscriptomic regulation,3 and cells employ them to mark self vs nonself RNA and to send foreign RNA to decay pathways.4 The application of mass spectrometry has enabled many of these understandings, but to perform these experiments, the RNA polymers are enzymatically degraded to their nucleoside monomers for analysis.5 Thus, all of the sequencing information is lost. Alternatively, limited digestion to short oligomers can provide the local sequence to map a modification after the full-length sequence is reconstructed.6 The sequence information for some RNA modifications is often found one modification at a time using custom approaches to introduce specific signatures that can be found in cDNAs after high-throughput sequencing.7,8 Expanding this approach to inspect more than one modification at a time is challenging because it is difficult to find chemical tools that are applicable to more than one modification under similar reaction conditions.9,10 The payoff of sequencing multiple modifications in a single strand of RNA at one time is to allow researchers to understand how these chemical decorations are used together for cellular control over phenotype and in response to stress.

Direct sequencing of RNA with nanopore technology has the potential to locate and quantify more than one modification in the strand.11 Nanopore sequencing of RNA involves electrophoretically passing the strand through a small aperture protein nanopore. As the nucleotides pass through the narrow zone of the protein, the ionic current is modulated based on the sequence. The ionic current levels are then deconvoluted with a recurrent neural network to yield the base calls. Roughly five nucleotides of RNA contribute to the current levels analyzed, termed a k-mer. Success in this approach requires an ATP-dependent 3′,5′-helicase to slow the movement of the RNA through the nanopore and allow enough time to record the small current level differences. Lastly, available base-calling algorithms for RNA have been trained on canonical RNA sequences and are not modification aware. Nevertheless, chemical modifications to RNA can be identified by their unique signatures compared to the canonical forms in the base call, ionic current, and/or dwell time data as they pass the helicase active site and then through the constriction zone of the pore.12−19

Modifications to RNA resulting in base calling differences have enabled sequencing for pseudouridine (Ψ), N6-methyladenine (m6A), inosine (I), the 2′-O-methylation of A, C, G, and U, inosine, and N7-methylguanosine (m7G), as examples.12−19 Some RNA modifications, such as 5-methylcytidine (m5C), do not strongly impact the base calling and are not easily found by this method.15 Alternatively, the ionic current level can differ for RNA modifications, such as Ψ, as it passes through the nanopore providing data to determine the frequency of a modification.12−14 This approach has a challenge: RNAs do not necessarily impact the current level when centered in the k-mer, and prior knowledge of this impact is needed for quantitative analysis of the data.14 Another domain in which RNA modification can be revealed in nanopore sequencing is in the active site of the helicase motor protein where the translocation rate differs for the modification relative to the canonical form.14,20 Pseudouridine provides a recognized example of an RNA modification that impacts the helicase dwell time data.14,20 While many studies have focused on analyzing a single modification type in RNA by nanopore sequencing, there have been a few examples of multimodification analysis. These include sequencing E. coli tRNA with a nanopore to demonstrate these hypermodified RNAs produce base call data differences at known modification sites,21 and rRNA has been sequenced to find the 16S Ψ516 and m7G527 concurrently.13 Lastly, the four 2′-O-methyl ribose modifications and Ψ in rRNA were profiled in E. coli and yeast.12,20

In this work, we focused our sequencing efforts on E. coli 16S and 23S RNA because these strands have 36 well-established chemical modifications at known sites with known levels under standard growth conditions (aerobic LB media, 37 °C, grown to stationary phase).2,6 There exist 17 different modified RNA structures occurring on the four canonical nucleotides (A family: m6A, N6,N6-dimethyladenosine (m62A), and 2-methyladenosine (m2A); C family: m5C, 2′-O-methylcytidine (Cm), N4,2′-O-dimethylcytidine (m4Cm), and 5-hydroxycytidine (ho5C); G family: 2′-O-methylguanosine (Gm), N1-methylguanosine (m1G), N2-methylguanosine (m2G), and m7G; U family: Ψ, N3-methylpseudouridine (m3Ψ), 5-methyluridine (m5U), N3-methyluridine (m3U), dihydrouridine (D or hU), and 2′-O-methyluridine (Um); Figure 1A). The positions of these modifications are known (Figures 1B and S1; numbering is based on the rrnA operon). The goal of the present work was not to find new modifications with nanopore sequencing. Instead, it was to discover how the sequencer can be used for the inspection and quantification of a diverse set of RNA modifications on the same RNA strand. These studies are the first to catalog the base call, ionic current, and helicase dwell signatures for the 36 modification sites at once. The findings are then used to address whether and how the modifications change in E. coli rRNA after exposure to metabolic or cold-shock stress. Lastly, the E. coli genome possesses 7 rRNA operons (rrn),22 allowing us to profile the operon expression levels and determine the operon-specific Ψ and m3Ψ modification levels before and after stress.

Figure 1.

Figure 1

Open in a new tab

(A) Structures and (B) locations of the established RNA modifications found in the 16S and 23S rRNA strands from E. coli.

Results

Defining the Sequencing Signatures for the 36 RNA Modifications in E. coli rRNA

The E. coli K12 DH5 alpha strain with a plasmid expressing the ampicillin resistance gene was grown aerobically in ampicillin-containing LB media at 37 °C to stationary phase after which the total RNA was harvested (detailed methods provided in the Supporting Information). The RNA integrity was verified by gel analysis (Figure S2) and then heat denatured for 10 min at 65 °C, flash cooled, and finally 3′-poly-A tailed using a commercial kit. The 3′-poly-A tailed RNA was the input for library preparation using the Oxford Nanopore Technologies (ONT) direct RNA sequencing kit that ligates a poly-T adaptor on the 3′ ends of the poly-A tailed RNA followed by reverse transcription to yield a DNA:RNA heteroduplex for sequencing. A final sequencing adaptor is then ligated onto the duplex. The library-prepared RNA was sequenced on either a MinION flow cell or Flongle flow cell, both with the R9.4 chemistry, and reads with Q > 7 were analyzed. Synthetic 16S and 23S rRNAs with sequences from the rrnA operon were prepared by in vitro transcription (IVT) and sequenced to provide a control for the RNA nanopore sequencing profile without modifications. The raw sequencing data in fast5 format were base called using Guppy (6.0.7) to generate the base called data in fastq format. The reads were aligned to the 16S and 23S sequences found in the rrnA operon using minimap2,23 sorted with SAMtools,24 and visualized with Integrative Genomics Viewer (IGV; Figure S3).25 The base call data were quantified with Nanopore-Psu16 and ELIGOS2.15 Lastly, the ionic current and dwell time data were analyzed and quantified with Nanopolish,26 Nanocompore,17 and Tombo.27 All computational tools were used following the instructions in their online manuals.

Sequencing RNA with the commercial nanopore system occurs from the 3′ to 5′ ends impacting the data in a few ways (Figure 2A).11 First, it is well established that there is greater read depth for the 3′-end of the sequence than for the 5′-end. Fortuitously, most of the rRNA modifications are found in the 3′ halves of both the 16S and 23S RNA strands. Second, the modifications first pass through the helicase, and if they alter the translocation kinetics of the enzyme, the impact will be observed ∼11 nucleotides 3′ to the modification because this is the sequence in the nanopore sensor that will be base called and aligned to the reference sequence (Figure 2B).14 In the current level data, the RNA modification may or may not impact the current level when located in the center of the k-mer resulting in challenges in identifying the current level for more than one modification in proximity to one another (Figure 2C).14 In the base call data, the impact is usually observed at the location of the modification, but not always (Figure 2C).15

Figure 2.

Figure 2

Open in a new tab

Direct RNA nanopore sequencing for chemical modifications to E. coli rRNA. (A) Schematic of the nanopore sequencer illustrating the protein nanopore and helicase sensor positions. (B) Helicase dwell time is altered by RNA modifications, as illustrated for 23S m2G1835 in E. coli rRNA and 23S G1835 in synthetic rRNA. (C) The nanopore ionic current levels can be impacted by RNA modifications, as shown for 23S Ψ955. This example demonstrates that the current impact may not occur when centered in the 5-nt k-mer, and the signal is lost when the Ψ955 writer protein rluC is knocked out as verification that the signal comes from the modification. (D) Base calling can be impacted by an RNA modification that is illustrated in the IGV plot for the U955 site in wild-type E. coli with Ψ, and the rluC gene knockout strain and synthetic rRNA without Ψ at this site. In the plot, blue represents a miscall of C and gray is a correct call relative to the reference sequence. (E) Color-coded plot of base calling error, ionic current level, and dwell time nanopore sequencing signatures for all 36 E. coli rRNA modifications. (F) Color-coded plot of E. coli knockout strains and synthetic RNAs with and without modifications that were sequenced to understand the modification signatures. *The writer for the N4 methyl group (rsmH) was studied for m4Cm. ‡An RNA generated by IVT containing m4C was studied, and an attempt to synthesize an RNA strand by IVT with m4CmTP failed.

The base call data, ionic current, and dwell time data impacts observed at 36 E. coli RNA modification sites allowed most of them to be identified (Figure 2E). Synthetic E. coli rRNA was used for comparison and served as a control to eliminate any nonrandom errors in the data, which is an inherent feature of nanopore direct RNA sequencing data. Seventeen of the modifications are in the primary sequence at sites 5 nucleotides or more distant from other modifications, and therefore, the changes between the biological and synthetic data sets could easily be attributed to the RNA modification (16S: Ψ516, m7G527, m2G1207, m4Cm1402, m5C1407, and m3U1498; 23S: Ψ955, m6A1618, m2G1835, m5U1939, m5C1962, m6A2030, m7G2069, Gm2251, Ψ2457, Um2552, and Ψ2580; Figures 1, 2E gray, and S4–S28). In a few cases, confirmation that the signals were a result of the modification came from sequencing rRNA derived from writer knockout E. coli (Figures 2C and S1 and S4–S28). Figure 2C illustrates loss of the signal from the 23S Ψ955 when its writer protein rluC is knocked out and the rRNA is sequenced. This example also shows the current level change occurs at the k-mer with the 5′ C centered in the nanopore sensor.

Nineteen of the modifications were clustered with another modification within 5 or fewer nucleotides in the sequence, and their individual signatures, particularly in the ionic current and helicase dwell time data, could not easily be determined (16S: m2G966 and m5C967; m2G1516, m62A1518, and m62A1519; 23S: m1G745, Ψ746, and m5U747; Ψ1911, m3Ψ1915, and Ψ1917; m2G2445 and D2449; Cm2498, ho5C2501, m2A2503, and Ψ2504; and Ψ2604 and Ψ2605; Figures 1, 2E, and S4–S28). To understand the signatures for each of these clustered modifications, RNA was sequenced from an E. coli strain in which an established writer for a modification was knocked out (Figure S1);28−30 the data were compared to the RNA sequencing data from the wild-type strain to allow identification of the unique signature for the members of these clustered modifications (Figure 2F).

The approach outlined enabled us to systematically determine the signatures for all modifications with a few exceptions (Figure 2E gray). In the 16S rRNA, there are two m62A residues at positions 1518 and 1519 written into the RNA by the same methyltransferase;29 base call data allowed quantification, even though they gave current signatures that could not be easily deconvoluted (Figure 2E red). Second, in the 23S rRNA, there exist Ψ1911, m3Ψ1915, and Ψ1917 in which the isomerization of U to Ψ at the three sites is catalyzed by the same synthase, after which the methyl group on N3 of Ψ1915 is installed.5 It was not possible to definitively determine nanopore ionic current signatures for these modifications (Figure 2E red); nonetheless, the base call data could be used for their identification and quantification (Figure 2E gray). Finally, synthetic RNAs prepared by IVT containing Ψ,14,31 m5C, m4C, Cm, and ho5C were installed with their commercially available NTPs, and additionally, NTPs were synthesized for m5U and m4Cm to install them in long RNAs (Figures S29–S33). Two noteworthy points are, first, that CmTP was installed during T7 RNA polymerase catalyzed IVT with low yields, as previously reported,32 which is not an issue in this case. Second, the success in the Cm-containing RNA synthesis encouraged us to attempt to synthesize the m4Cm-containing RNA by IVT but the reaction failed to produce full-length RNA after repeated attempts; therefore, this was not pursued. For all other modifications in the list above, this study allowed independent verification of their base call, ionic current, and dwell time signatures in 6 or more different k-mer contexts (Figure 2F). Inspection of the data identified base call signatures for 27 of the modifications, nanopore ionic current level signatures for 28 of the modifications, and helicase dwell time signatures for 26 of the modifications (Figure 2E). These signatures were used in the studies that follow.

Monitoring Stress-Dependent Changes to rRNA Modifications by Nanopore Sequencing

The knowledge of nanopore signatures for the rRNA modifications was used to detect and measure their occupancy in the strands under normal growth conditions and after exposure to metabolic or cold-shock stress. Prior reports have used mass spectrometry or gel analysis to quantify each of the E. coli rRNA modifications under stationary phase growth in LB media at 37 °C, similar to the present conditions, to find they exist at >85% occupancy and many at near-quantitative levels.6,33−36 In the first analysis, the data from the nanopore sensor were interrogated (Figures 3A and S34). The current levels were inspected and quantified using Tombo;27 recall, RNA modifications do not necessarily alter the current most when centered in the k-mer, and using the prior analysis allowed picking the correct position for each modification to conduct the quantification (Figure 2C). Tombo provides quantification of RNA modifications by comparing current levels between the E. coli rRNA with those found in the synthetic RNA without modifications.27 The base call data derived from the current levels can also be analyzed for the quantification of RNA modification. The computational tools ELIGOS2 and Nanopore-Psu were employed. The tool ELIGOS2 can be used for modification quantification by comparing the error at specific bases (ESB) for the E. coli rRNA strands against the synthetic rRNA (Figure S34);15 Nanopore-Psu was trained on mostly cellular rRNA to predict Ψ occupancy specifically in nanopore-sequenced RNA.16

Figure 3.

Figure 3

Open in a new tab

Stress-dependent changes in E. coli rRNA modifications found by direct nanopore RNA sequencing from the nanopore protein sensor. (A) The nanopore sequencing setup has two sensors, the helicase and protein nanopore, in which the nanopore provides current levels used in base calling analysis. (B) Current-level quantification via the Tombo tool for changes in rRNA modifications. Inset illustrates biological replicate reproducibility. (C) Base-call quantification for Ψ sites using Nanopore-Psu. Inset shows reproducibility in biological replicates. Changes in the E. coli rRNA modifications identified after (D) metabolic or (E) cold-shock stress. The changes were color-coded based on a Student’s t test of the RNA modification level changes between normal growth and stress conditions (yellow = decrease with P < 0.05; red = decrease with P < 0.01; and green = increase with P < 0.05). The ELIGOS2 data used for quantification are provided in Figure S34.

Nanopore derived signatures were initially inspected (Figure 3A). The E. coli before and after stress were inspected by a colony-unit-forming assay for verification that they were alive at the time of harvesting the RNA (Figure S35). Biological replicates were used for sequencing E. coli rRNA obtained from cells grown in LB media at 37 °C found that the modification quantification from current-level and base-call data was generally reproducible (Figure 3B,C and S36). Next, scatter plot analysis of the RNA modification levels under normal conditions vs the stress conditions identified many sites that deviated between them (Figures 3B,C and S36). Duplicate data sets were analyzed for each condition and the Student’s t test was used to find those with significant changes; the changes were color-coded to be yellow for a lower level of significance in the decrease (*P < 0.05), red for a larger significance in the decrease (**P < 0.01), and green for a lower level of significance with an increase in the modification level (*P < 0.05; Figures 3D,E and S37).

Metabolic stress to the E. coli was induced by growing them on minimal media (M9 with 0.5% glucose) under aerobic conditions at 37 °C for 20 h following literature reports.22 Cold-shock stress to the E. coli was induced by first growing the cells aerobically in LB media at 37 °C for 20 h, followed by placing the cells at 20 °C and then allowing them to grow for 1 h at the changed temperature. In the Supporting Information, heat-shock stress was studied and data are provided with the knowledge that the RNA sequenced was fragmented from the high temperature, a known phenomenon (Figure S38).37 The cells exposed to metabolic stress showed significant decreases in the rRNA modifications based on current-level analysis (8 sites), ELIGOS2 base-call analysis (10 sites), and Nanopore-Psu base-call analysis (3 sites; Figure 3D). In all three data analysis approaches, 23S Ψ2604 was found to increase in occupancy (Figure 3D). The analysis provided contradictory findings for 23S Ψ746, in which a decrease in occupancy was found with Tombo and ELIGOS2 but an increase in occupancy was found for Nanopore-Psu (Figure 3D). Cold-shock stress to E. coli resulted in rRNA modification level decreases based on current-level analysis (17 sites), ELIGOS2 analysis (13 sites), and Nanopore-Psu analysis (6 sites; Figure 3E).

The rRNA modification levels found to change with metabolic or cold-shock stress can be grouped. The first group is m2G, for which the rRNAs have 5 sites of this modification (Figure 1B), four of which decreased with metabolic and cold-shock stressors (Figure 3D,E). The second group is Ψ, in which E. coli rRNAs have 11 sites (10 Ψ sites and 1 m3Ψ site; Figure 1B), six changed with metabolic stress, and nine changed with cold-shock stress (Figure 3D,E). The third group consists of a few sites of single modification structures that changed and include cold shock, resulting in a decrease of 16S m4Cm1402, 16S m62A at positions 1518 and 1519, 23S m6A2030, 23S ho5C2501, and 23S m2A2503 (Figure 3E). Metabolic stress resulted in a decrease at 16S m4Cm1402, the two 16S m62A sites at positions 1518 and 1519, and 23S m2A2503 (Figure 3D). These findings identify that rRNA modifications can change with metabolic and thermal stress in E. coli. This is a finding in line with a hypothesis that rRNAs play a central role in sensing stress.38

The Helicase Sensor Identifies the Methyl Group Lost from m4Cm during Stress

The hypermodified RNA m4Cm stands out as one for further investigation because its levels decrease with both forms of stress studied. Does stress cause the loss of both the sugar and base methyl groups or only one methyl group? And if it is just one, which one is it (Figure 4A)? The 16S m4Cm1402 is the only modification of this type in E. coli, and studying its change by mass spectrometry is very challenging; further, if only one methyl group is lost, then analyzing rRNA oligomer fragments obtained by limited nuclease digestion by mass spectrometry cannot easily differentiate whether it was the base or sugar methyl group. The helicase sensor was found to provide a solution to this challenge using the same approach we previously reported for finding Ψ in the SARS-CoV-2 RNA genome (Figure 4B).14

Figure 4.

Figure 4

Open in a new tab

Helicase sensor allows identification of the methyl group lost from the E. coli 16S m4Cm1402 site during metabolic or cold-shock stress. (A) The data provided in Figure 3D and E found that the bismethylated base 16S m4Cm1402 had changed. Were one or two methyl groups lost, and if one, which one? (B) The helicase sensor identifies RNA modifications by a change in dwell time; however, as a result of the base calling coming from ionic current levels in the nanopore sensor, the modifications are observed in the reference transcriptome alignment as a change in dwell time that is 3′ distal by 10–12 nucleotides from the modification position in the RNA. (C) Nanocompore analysis was used to compare two biological replicates of the E. coli 16S rRNA vs the synthetic rRNA without modifications to find dwell times that differ significantly between the samples. The P-values reported from Nanocompore were negative-log transformed for visualization, in which an increase in value represents a greater degree of difference between the samples. (D) Histograms of dwell time values obtained from Nanopolish analysis of 16S rRNA at positions 1412 and 1414 from nonstressed E. coli. (E) Average dwell times were found for the rRNAs being unwound by the helicase sensor by fitting the dwell time histograms to a Gaussian distribution (panel D). The errors reported are for the fitting errors of the average value plotted. The original plots with fitting statistics for the data are provided in Figure S39.

The Nanocompore tool was used to identify statistically significant differences (Kolmogorov–Smirnov (KS) test) in the helicase sensor dwell times between the E. coli 16S rRNAs and the synthetic 16S rRNA without any modifications.17 The P-values from the KS test were negative-log transformed for visualization purposes such that an increase in value represents a greater degree of statistical significance between the data sets. The statistical analysis of E. coli grown at 37 °C in LB media identified two positions (1412 and 1414) for which 16S m4Cm1402 impacted the helicase dwell time (Figure 4C bottom panel). We hypothesize that one resulted from the base methyl group, and the other is from the sugar methylation. Surprisingly, the metabolic and cold-shock stressed E. coli rRNAs, when statistically analyzed against the control, were found to have an altered dwell time only at position 1414 supporting the conclusion that one methyl group was lost from the same site under both stressors (Figure 4C interior panels). Confirmation that it was the loss of the base methyl group at N4 was derived from sequencing the 16S rRNA from E. coli in which the methyltransferase for writing the methyl group on N4 of C 1402 was knocked out (rsmH). Analysis of rsmH knockout E. coli 16S rRNA found that only the 1414 site had a significant difference in dwell time compared to the synthetic rRNA (Figure 4C top panel). The dwell time measurements showed that in E. coli metabolic and cold-shock stress results in the 16S rRNA having a Cm residue at position 1402 instead of a hypermodified m4Cm. Finally, we set out to validate this nanopore finding by an independent approach, namely, LC-MS analysis of the 16S rRNA that possesses the m4Cm residue (Figure S40). The LC-MS analysis found m4Cm decreased in the rsmH writer KO, as well as under the cold-shock and metabolic stress conditions. The m4C was not uniquely identifiable because it coelutes with natural m5C in this strand, while the data show no increase in this nucleotide under stress conditions. Unfortunately, an eluting peak for Cm did not emerge for any of the conditions studied, which are fully described in Figure S40. Nonetheless, the present work characterized the helicase signature for Cm in synthetic RNA that is similar to a prior study20 and matched the results from the stressed cells. This demonstration provides an example of the power of the helicase as a sensor to learn details about RNA modifications. This would not be easily achieved by other methods.

To gain a better biophysical understanding of how the m4Cm modification was changing the helicase activity (i.e., dwell time), hundreds of events were analyzed with Nanopolish to measure the dwell time values at positions 1412 and 1414 in the samples.26 The dwell time values were log-transformed and then binned, and histograms were made, followed by fitting with a Gaussian distribution function (Figure 4D). At position 1412 for the N4 methyl group on C, the distribution was fit to a single Gaussian function, and at position 1414 for the 2′-O-methyl group on C, the distribution was fit to two Gaussian functions. The base methyl group impacting position 16S 1412 was found in the modified 37 °C grown E. coli to have a dwell time average of ∼3 ms; loss of the methyl group in the stressed E. coli, rsmH KO cells, and synthetic 16S rRNA gave an average dwell time of 2-fold or longer (6–8 ms). This shows that the presence of the base methyl group results in faster helicase activity compared to the canonical nucleotide, likely because the N4 methyl group is slightly destabilizing to the dG:m4Cm base pair. Inspection of the data at position 1414 found two populations in all of the E. coli rRNAs (no stress, stress, and rsmH gene knockout) with an average dwell time of less than 10 ms and a population with an average dwell time increased by about 10-fold (80–110 ms; Figure 4E). The most revealing observation is that in the synthetic 16S rRNA at position 1414 only one population of helicase dwell times was observed with an average value of ∼9 ms (Figure 4E). This final analysis identifies the 2′-O methyl group on C and impacts helicase processivity with two different time values, one that is 10-fold longer than observed for the unmodified C nucleotide and the other with a similar time value as measured for the C nucleotide.

Identification of E. coli rRNA Operons and Their Deposition of the 11 Ψ Residues

In the final inspection of the data, the rRNA sequencing reads were aligned against the E. coli reference genome with all seven rRNA operons (rrnA, B, C, D, E, G, and H; Figure 5A) instead of the rrnA operon reference in the previous studies. The operons each possess their own sequence variations that allow alignment to each operon (Figures 5B and S41).22 One drawback to this is that the data at each site have lower coverage, and therefore, the operon-specific analysis used Nanopore-Psu focused on 10 Ψ sites and 1 m3Ψ in the E. coli rRNA sequences. As a consequence of Ψ being the most abundant modification in the rRNA and these modifications being subject to change, the limited analysis provides a picture of operon-specific modification changes before and after stress. In Figure 5, the data are for metabolic stress. The operon-specific analysis of cold-shock stress is provided in Figures S42–S44 with the limitation that not all sites could be analyzed as a result of the decreased sequencing depth when aligning with all seven operons.

Figure 5.

Figure 5

Open in a new tab

Operon-specific rRNA changes in Ψ levels in E. coli rRNAs. (A) Genome map illustrating the positions of the rRNA operons in the E. coli genome. (B) Example IGV plot of E. coli 16S rRNA from positions 78–93 showing five different natural operon-specific sequence variations. (C) Relative expression levels for the seven different E. coli rRNA operons when the cells were grown aerobically at 37 °C in either LB media or under metabolic stress. (D) Heat plots for Nanopore-Psu predicted levels for the 11 Ψ sites in the 16S and 23S rRNAs. The values plotted are averages from a set of biological replicates.

First, the expression levels of rRNA from each operon were determined under normal and metabolic stress conditions. A prior report normalized the data to the expression of rrnC because it is the closest to the origin of replication (Ori; Figure 5A),22 and we maintain that convention herein. When E. coli were grown in LB media at 37 °C, the most highly expressed operon was rrnC followed by rrnB, rrnE, rrnH, rrnA, rrnD, and rrnG (Figure 5C red bars). In general, the highly expressed operons were those in the direction of replication (rrnC, B, E, H, and A), and the two operons oriented in the opposite direction were expressed at lower levels (Figure 5A and C). Next, we found that under metabolic stress, the operon expression levels changed (Figure 5C). Those more distal to the Ori (rrnH, D, and G) showed the greatest increase in expression, while those more proximal to the Ori (rrnA, rrnB, and rrnE) decreased in expression (Figure 5A and C). A previous report conducted a similar analysis to that described here,22 by sequencing the E. coli rRNAs via conversion to cDNAs, and they obtained nearly identical results as those found from nanopore direct RNA sequencing conducted herein. In the Supporting Information are operon expression levels for E. coli under cold-shock stress (Figure S43).

The Nanopore-Psu analysis for Ψ in each of the rRNAs from the seven operons is the first experiment of this type and provides deeper insight into the changes found in Figure 3. Pseudouridine levels in the rRNAs under normal growth conditions, in general, were similar across all seven rRNA operons, with the exceptions of 23S Ψ746 and 23S Ψ2604 (Figure 5D left panel). The installation of 23S Ψ746 was predicted by Nanopore-Psu to exist with the highest frequency of 80% in the rRNA from the rrnA operon and at the lowest frequency of 67% in the rRNA from the rrnC operon. In contrast, 23S Ψ2604 was installed with the highest frequency in the rRNA from the rrnC operon at 93% and the lowest frequency in the rRNA from the rrnD operon at 65%.

Under metabolic stress, operon-specific Ψ levels were similar at eight sites, with three noteworthy differences. The first is 23S Ψ 2604, in which the installation increased the most on all operons except rrnC, which was already at a high level as described above. When the Ψ levels were determined by alignment to a single reference (Figure 3E), 23S Ψ2604 was found to increase under metabolic stress; however, the operon-specific analysis finds this increase occurs on all operons except rrnC (Figure 5D). The second is 23S Ψ2457, which decreased in installation in the rRNA when aligning the data against a single reference (Figure 3E), and this claim holds when aligning the reads against all seven operons (Figure 5E right panel). An interesting point is that the decrease in writing 23S Ψ2457 occurred in the rrnD operon. The second is 23S Ψ746, which was found in the single operon alignment data to not change upon metabolic stress; in contrast, alignment of the data to all seven operons finds this Ψ increases in most operons and decreases in rrnH leading to the null result in the single operon alignment data (Figure 5E right panel and Figure3E). This final observation identifies that stress can result in underlying rRNA modification changes with operon specificity that are not observable when aligning the data to a single reference. Differences were also found in cells experiencing cold-shock stress (Figure S44).

Discussion

Nanopore Direct RNA Sequencing Can Monitor 16 Different E. coli rRNA Modifications

In the present work, E. coli rRNA was sequenced directly with the commercial nanopore platform from ONT. This was done to evaluate the ability to sequence and quantify up to 17 different RNA modification structures that occur 36 times in these RNAs. Prior structural and in cellulo studies have determined the sites, quantified the installation, and identified the writers for all but one of these 36 modifications in E. coli rRNA (Figure S1).2,6,29 This provides an excellent case to study nanopore direct RNA sequencing for more than one modification type at a time during sequencing. Finally, a prior study reported two modifications in 16S E. coli rRNA (m7G527 and Ψ516), and the present work that expands the analysis to all rRNAs is fully consistent with the earlier report on these two sites.13

Seventeen of the 36 modifications are located in the primary sequence distant from other rRNA modifications, and they were easily evaluated to determine their ionic current, dwell time, and base calling signatures (Figure 2E). All of these could be revealed in one or more of the nanopore sequencing data types with the exception of m5C. This observation is consistent with a prior report on mRNA sequencing.15 Another observation is with respect to the two m6A sites in the 23S rRNA (positions 1618 and 2030), in which 1618 was only observable by a helicase dwell signature, while 2030 yielded a base call and dwell signature. Computational tools such as Epinano39 and ELIGOS215 that inspect base call data can locate and quantify m6A in mRNA, in which this modification is deposited at high frequency in DRACH motifs (D = A or G; R = A or G; H = A, C, or U; bold = m6A). In contrast, 1618 and 2030 in E. coli rRNA are written in the sequence contexts 5′-ACAGG and 5′-UGAAG, respectively. This observation nicely demonstrates the strong sequence context dependency in nanopore direct RNA sequencing signatures, a feature we previously described for Ψ.31

Nineteen of the 36 RNA modifications in the E. coli rRNA are present as clusters in the primary sequence, representing a challenge to deconvolute the nanopore and helicase signatures (Figure 2E). We approached this challenge by sequencing rRNA strands from E. coli that had known writer enzymes knocked out to identify the signal changes in comparison to the rRNAs from native E. coli cells (Figure 2F). This approach worked with two exceptions that include the two adjacent m62A sites at 1518 and 1519 in the 16S rRNA, and the three closely spaced 23S Ψ1911, m3Ψ1915, and Ψ1917 sites; monitoring these was achievable in the base call data (Figure 2E). A key success in this approach for determining the modification signatures was the four closely spaced modifications in the 23S rRNA (Cm2498, ho5C2501, m2A2503, and Ψ2504). As a final point, synthetic RNAs made by IVT provided independent verification of the signatures for Ψ,14,31 m5C, Cm, m4C, ho5C, and m5U.

Acknowledgment of Limitations to the Data Analysis

Nanopore direct RNA sequencing occurs from the 3′ to 5′ ends, resulting in higher depth reads on the 3′ end of the sequences (Figure S3). Compared to DNA sequencing by this approach, the data for RNA are more error prone, which decreases accuracy and confidence in the results. Nonetheless, use of controls and many standards, as conducted in these studies, allows RNA modifications to be inspected by this method. The significant prior studies on E. coli rRNA allow comparison of the nanopore data reported in the present work.2,6,29 With respect to RNA modification quantification, a convenient aspect of nanopore data is the ability to predict the modification occupancy. Quantification of the data is successful, as reported here and in other studies,12 but these values when compared to high-accuracy mass spectrometry data are quite variable. For example, m7G (16S 527 and 23S 2069) is quantifiable by the current level to give values similar to MS measured values (16S m7G527: nanopore = 98% and MS > 95%;6Figure S37). In a less ideal case, 23S Gm 2251 is present at >95% occupancy by MS analysis,6 while the current-level analysis predicts 28% occupancy, base call analysis fails to make a prediction (Figure S37), and helicase analysis is challenging to quantify, as described below. Similar large differences between MS and nanopore values exist for m6A, m5C, m5U, D, and ho5C, as well as a single m2G site (23S 2445; Figures S37). Nanopore direct RNA sequencing can provide data not obtainable by MS analysis, as demonstrated by the two Ψ sites at positions 2604 and 2605 in the 23S rRNA. In this final example, MS analysis is challenged to quantify the modifications,33 while base call data and ionic current level data can be quantified, albeit they give different results (base call: 2604 = 72% and 2605 = 98%; current level: 2604 = 50% and 2605 = 55%; Figure S37). These comparisons indicate that nanopore data can be quantified, but the values may differ from more accurate MS data. For this reason, we looked for relative changes in values instead of absolute quantification of occupancy.

Another noteworthy limitation is in regards to using base calling as a means of RNA modification detection and quantification. The seven E. coli rRNA operons serve as an excellent example of caution in this analysis. For instance, in the 16S rRNA at position 93 there exists a natural C/U sequence variation between the operons (Figure 5B). Interestingly, a C miscall is the most common Ψ signature in nanopore base call data;12,14,18,31 thus, this natural sequence variation masquerades as a Ψ if only base call data is considered, when in actual fact one is sequencing a mixture of C and U variants at that site. This example nicely illustrates why in our prior studies the use of helicase dwell time, for which Ψ yields a signature, serves as a check for verification when using nanopore direct RNA sequencing for the discovery of new Ψ sites.14

Stress-Dependent rRNA Changes in E. coli Were Monitored by Nanopore Sequencing

E. coli cells were subjected to either micronutrient or cold-shock stress, followed by reanalysis of the rRNA modifications to inspect for changes in occupancy. Keeping the quantification challenges in mind, we analyzed the data for changes in the rRNA modifications relative to the cells grown under normal conditions (Figure 3). When the E. coli cells were allowed to grow and divide under either stress condition, the modifications either did not change or were found to decrease in occupancy with the exception of one that increased in occupancy. Modifications that consistently decreased with the stress based on a change in one of the data types were m2G (e.g., 16S 1516 and 23S 2445) and Ψ (23S 746, 955, 2457, and 2504).

There are two exceptions to this observation; under metabolic stress, 23S Ψ2604 was found to increase in its occupancy, and mixed results were found for 23S Ψ746 (current and ELIGOS2 predict a decrease, and Nanopore-Psu predicts an increase; Figure 3D). Clarification of the inconsistent result for 23S Ψ746 comes from the operon-specific Ψ analysis (Figure 5D). This position was found to have mixed operon-specific Ψ deposition in which the rrnA operon had less Ψ installed with metabolic stress, and the others showed an increase in Ψ installation. The data in Figure 3 were obtained by aligning to only the rrnA operon as a reference; hence, the alignment is likely the culprit. For the current-level analysis with Tombo and base call analysis with ELIGOS2, minimap2 was used with one set of mapping parameters for alignment, while Nanopore-Psu conducts the alignment with minimap2 using a different set of mapping parameters; this difference likely impacted the reads mapped and the Ψ quantified. This demonstrates how small differences in computational tools can affect the results of the analysis.

Helicase Sensor Dwell-Time Analysis Provides Clarity on the m4Cm Change with Stress

In the 16S rRNA sequence at position 1402, there is a hypermodified C nucleotide m4Cm that decreased with metabolic and cold-shock stress (Figures 2D and 2E). First, there exists an m5C at position 1407 in this sequence (Figure 1B), but as described, this modification does not impact nanopore data, and therefore, the signature results from m4Cm. What was the change, loss of one or both methyl groups? If one, which methyl group was lost? In E. coli, this is the only m4Cm, and studying this by complete digestion to nucleosides followed by quantitative MS analysis would be challenging, especially because E. coli have many Cm residues in other RNAs; moreover, using MS to analyze rRNA fragments would be challenging with the loss of a single methyl group.

The helicase sensor was found to provide the data to answer this question. The hypermodification m4Cm impacts the helicase activity in two different positions that are found when comparing helicase dwell time differences between the wild-type E. coli rRNA with synthetic rRNA without modifications. A prior study noted that this residue could impact the helicase dwell time,20 consistent with our findings. Under the stressors imposed, only one helicase dwell time difference was observed in the statistical comparison (Figure 4C). The deduction of which methyl group was lost came from studying the rRNA from E. coli with the base methyl writer gene rsmH knocked out. The comparative analysis found that when only a Cm resides at position 1402, a single helicase dwell time difference is observed that tracks with those observed in the stressed E. coli (Figure 4C). This confirms that under stress in the 16S rRNA at position 1402, the cells only methylate the sugar to yield Cm. Furthermore, we validated this conclusion by an independent analysis performed by LC-MS in which a 50-mer oligonucleotide spanning the m4Cm1402 site was isolated, digested, and analyzed (see Figure S40).

In principle, a histogram for a population of dwell times that differ between the modified and unmodified states can be integrated to quantify the presence of modification. The histogram shown in Figure 4D, as well as in Figure S29, illustrates a challenge with the quantification modifications via helicase dwell time. The nanopore is a single-molecule sensor, and the dwell times represent the monitoring of helicase activity one molecule at a time. At this molecular level, the data identify how stochastic enzymes are in processing their substrates, resulting in the broad time distributions measured.40 Therefore, any unmodified RNA may be masked by the presence of the modification in the helicase dwell time, resulting in less reliable quantification. Nonetheless, the presence of the dwell time change provides a check for the presence of an RNA modification when one exists. This feature may also exist when sequencing DNA modifications with the nanopore.41

Two remaining questions regarding the helicase activity on m4Cm include the following: why are there two different points at which this double modification changes the helicase activity? Second, why does Cm yield two different helicase dwell times? The patent literature suggests the helicase used by ONT is a mutated version of a Hel308 helicase,42 but the actual mutations are not publically known. With this limitation and using the native structure for archaeal Hel308 solved bound to DNA, we can hypothesize why a base and sugar methyl group would impact the helicase at different positions.43 In this structure, Lys 289 interacts with the ribose backbone, impacting the 2′-O-methyl group, while the base interacts with Phe 350 (Figure S45). Interestingly, these two interactions occur two nucleotides apart, consistent with the nanopore helicase dwell time analysis for m4Cm. A reason for two helicase dwell time populations (Figure 4D) may result from Cm favorably adopting 3′-endo over the 2′-endo sugar pucker, in which the 3′-endo conformation likely stabilizes RNA duplexes by preforming the A helix and the 2′-endo does not;44 however, studies to understand the kinetics of this equilibrium in a helicase active site have yet to be conducted to support whether they are sufficiently long-lived to impact the helicase activity differently.

Biological Implications for rRNA Modification Changes during Stress

The implications of rRNA modification changes in E. coli under metabolic or cold-shock stress are described for a few of the modifications described in this study. The first is the decrease of 16S m4Cm1402 yielding Cm at this position. This residue is conserved in bacteria and is located in the P-site of the ribosome where it makes contact via the N4 nitrogen with the second and third positions of the P-site codon.45 The function of this contact is proposed to fine-tune the P-site structure, resulting in correct recognition of the initiation codon. Loss of the N4 methyl group from 16S m4Cm1402 results in increased efficiency of non-AUG start codons and greater UGA read-through, resulting in decreased ribosome fidelity.45 This may be a feature used by the cell to assist in survival during stress.

A second noteworthy rRNA modification that decreased under cold-shock stress is 23S ho5C2501 (Figure 2E). This site is of interest to us because it is the only modification resulting from a two-electron oxidation of the base. This modification site is located in proximity to the peptidyl transferase center and is hydroxylated in most bacteria.46 The presence of this modification is hypothesized to provide tolerance to oxidative stress;46 however, why it is lost during cold-shock stress remains unknown. Two possibilities include (1) transferring E. coli at stationary phase growth at 37 to 20 °C leads to new rRNA synthesis, and because this is a late-stage modification in the 23S rRNA,46,47 it was not yet installed, resulting in the lower yields. (2) The ho5C is easily further oxidized, and cold-shock stress could have resulted in conditions that led to its further oxidation yielding a highly distorted base compared to the parent.48 Signals were not observed that would suggest the presence of the distorted structure, but it is possible this structure could not easily pass the helicase and therefore was not detected in the longer reads. Future work is needed to understand why 23S ho5C2501 changed.

The changes in Ψ observed with stress are not easily understood from a structural perspective. Loss of any Ψ modification in E. coli rRNA has no or a small phenotype; actually, they can all be removed and the cells only show modest defects in ribosome biogenesis, function, and cell growth.49 As for metabolic stress, prior proteome analysis provides details about changes in protein writers during stress that can be tied to the changes observed in the sequencing experiments in the present work.50 An increase in 23S Ψ2604 occurs with metabolic stress, and the expression level of the writer for this Ψ (rluF) increases by 150% (Figure S45). Furthermore, rluF installs Ψ in tRNATyr at position 35, and this could also be impacted by metabolic stress. Stress-dependent changes in tRNA modifications have been noted.51

In contrast, lower levels of 23S m6A2030, m2G2445, and Ψ2504 were observed that track with the reduction in expression of the corresponding writers (rlmJ, rlmL, and rluC, respectively)50 for these rRNA modifications during stress (Figure S46). The most significant was 23S m2G1835, which gave a large decrease in occupancy in the rRNA, and the proteome analysis failed to identify the presence of the writer protein rlmG in E. coli experiencing the same form of metabolic stress as imposed in these studies.50 This argument for the RNA modification changes is not the complete picture, and future studies likely assisted with nanopore direct RNA sequencing will be needed to further address the biology of the changes. The changes in E. coli rRNA modifications under stress differ from those reported for S. cerevisiae found by direct RNA nanopore sequencing.12 The eukaryotic vs prokaryotic cell types studied and the difference in the time of the imposed stress relative to the cell replication rate best address the differing findings. These comparisons to prior studies support a conclusion in E. coli that rRNA and its modification status are sensitive to environmental conditions. These rRNA modifications may play a larger role in cell signaling beyond supporting ribosome structure for protein translation. The present work now provides nanopore current levels, base call information, and helicase dwell time analyses for 10 additional RNA modifications (m4Cm, m2G, m5U, m3U, m1G, m3Ψ, D, ho5C, m2A, and m4C). These details will assist future nanopore direct RNA sequencing experiments for the discovery of how many more RNA modifications, beyond those already known, are impacted when cells are exposed to environmental challenges.

Methods

Complete experimental details and an approach for data analysis are provided in the Supporting Information.

Preparation of Biological RNA

The E. coli K12 DH5 alpha strain with an ampicillin-resistant gene supplied from a plasmid (psiCheck2) was grown aerobically in ampicillin-containing LB media at 37 °C to stationary phase conditions (∼20 h). The knockout E. coli K12 strains obtained from the Keio collection are kanamycin resistant and, therefore, were grown aerobically in kanamycin-containing LB media at 37 °C to stationary phase. Induction of metabolic stress was achieved following the literature,22 in which the wild-type E. coli were grown in M9 medium with 0.5% glucose to the stationary phase. The thermal stress was induced following the literature52 by growing the E. coli grown for 15 h and then transferring them to either 20 or 42 °C and then growing for 1 h under the new conditions. A colony-forming unit assay was conducted on all E. coli samples to verify that the cells were still viable (Figure S36). After the growth, the total RNA was extracted from pelleted cells using a commercial kit following the protocol (Zymo Direct-zol RNA Miniprep kit). The RNA was verified by agarose gel electrophoresis and then stored at −80 °C until sequencing was commenced.

Preparation of Synthetic RNA

The duplex DNA utilized for the synthesis of the 16S and 23S sequences found in the rrnA E. coli operon and the judiciously designed RNAs to study individual modifications were obtained from a commercial source with the T7 RNA polymerase promoter before the sequence was made. The RNAs were synthesized using the T7MegaScript Kit (Thermo Fisher Scientific) following the manufacturer’s protocol. In the cases to install a modified nucleotide in the RNA, the canonical NTP was replaced with the modified NTP at the same concentration. Success in the synthesis of the RNA was achieved by agarose gel electrophoresis.

RNA Library Preparation and Data Analysis

The purified RNAs were first 3′-poly-A tailed with a commercial kit (Life Sciences) to allow the first step of library preparation to occur, which ligates on a poly-T adaptor. The SQK-RNA002 kit provided by ONT was used without modification to the protocol. The prepared RNA libraries were loaded on either a MinION flow cell (∼10 ng) or a Flongle flow cell (∼5 ng) and sequenced using the standard settings in the MinKnow software. The ionic current vs time trace data were stored in fast5 file formats that used the standard pass cutoff (Q > 7). The passed data were base called using Guppy (6.0.7). A newer version of Guppy exists and was not used in this work; however, we previously reported Guppy 6.3.2 gave similar miscalls for Ψ as 6.0.7.31 The reads were aligned to the reference using minimap2,23 sorted with SAMtools,24 and visualized with IGV.25 The base-called data were quantified using either Nanopore-Psu16 or ELIGOS215 following the procedures provided with these tools. The passed fast5 files were analyzed using Nanopolish,26 Nanocompore,17 and Tombo27 as described in the literature citations for these tools. All data were visualized using Origin or Excel.

Acknowledgments

The National Institutes of Health provided financial support for this project (R01 GM093099 and R35 GM145237 to CJB, and R35 GM126956 to VB).

Data Availability Statement

The base called data for all sequencing experiments has been publicly deposited at Zenodo a public repository for data at doi: 10.5281/zenodo.7746124 that is searchable in the OpenAIRE explorer.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acschembio.3c00166.

  • Complete materials and methods; IGV plots; gel analysis of RNA; base call, ionic current level, and dwell time analysis for all 36 RNA modifications; analysis of synthetic RNA containing m5C, Cm, m4C, ho5C, and m5U; RNA modification writer knockout studies; calibration curves for modification quantification; raw data used in the quantitative plots; all bar graphs; LC-MS analysis of 16S rRNA; rRNA operon specific sequence variations; helicase dwell time analyses; operon sequence variations; colony forming assay; heat-shock data; cold-shock operon data; Hel308 structural details; and proteome analysis (PDF)

The authors declare the following competing financial interest(s): AMF and CJB have a patent licensed to Electronic BioSciences, and AMF occasionally consults for this company.

Special Issue

Published as part of ACS Chemical Biologyvirtual special issue “Nucleic Acid Regulation”.

Supplementary Material

cb3c00166_si_001.pdf (9MB, pdf)

References

  1. Jones J. D.; Monroe J.; Koutmou K. S. A molecular-level perspective on the frequency, distribution, and consequences of messenger RNA modifications. WIREs RNA 2020, 11, e1586 10.1002/wrna.1586. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Watson Z. L.; Ward F. R.; Méheust R.; Ad O.; Schepartz A.; Banfield J. F.; Cate J. H. Structure of the bacterial ribosome at 2 Å resolution. eLife 2020, 9, e60482 10.7554/eLife.60482. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Shi H.; Wei J.; He C. Where, when, and how: Context-dependent functions of RNA methylation writers, readers, and erasers. Mol. Cell 2019, 74, 640–650. 10.1016/j.molcel.2019.04.025. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Thompson M. G.; Sacco M. T.; Horner S. M. How RNA modifications regulate the antiviral response. Immunol. Rev. 2021, 304, 169–180. 10.1111/imr.13020. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Su D.; Chan C. T.; Gu C.; Lim K. S.; Chionh Y. H.; McBee M. E.; Russell B. S.; Babu I. R.; Begley T. J.; Dedon P. C. Quantitative analysis of ribonucleoside modifications in tRNA by HPLC-coupled mass spectrometry. Nat. Protoc. 2014, 9, 828–841. 10.1038/nprot.2014.047. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Popova A. M.; Williamson J. R. Quantitative analysis of rRNA modifications using stable isotope labeling and mass spectrometry. J. Am. Chem. Soc. 2014, 136, 2058–2069. 10.1021/ja412084b. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Dai Q.; Zhang L.-S.; Sun H.-L.; Pajdzik K.; Yang L.; Ye C.; Ju C.-W.; Liu S.; Wang Y.; Zheng Z.; Zhang L.; Harada B. T.; Dou X.; Irkliyenko I.; Feng X.; Zhang W.; Pan T.; He C. Quantitative sequencing using BID-seq uncovers abundant pseudouridines in mammalian mRNA at base resolution. Nat. Biotechnol. 2023, 41, 344–354. 10.1038/s41587-022-01505-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Hawley B. R.; Jaffrey S. R. Transcriptome-wide mapping of m(6) A and m(6) Am at single-nucleotide resolution using miCLIP. Curr. Protoc. Mol. Biol. 2019, 126, e88 10.1002/cpmb.88. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Khoddami V.; Yerra A.; Mosbruger T. L.; Fleming A. M.; Burrows C. J.; Cairns B. R. Transcriptome-wide profiling of multiple RNA modifications simultaneously at single-base resolution. Proc. Nat. Acad. Sci. U.S.A. 2019, 116, 6784–6789. 10.1073/pnas.1817334116. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Behm-Ansmant I.; Helm M.; Motorin Y. Use of specific chemical reagents for detection of modified nucleotides in RNA. J. Nucleic Acids. 2011, 2011, 408053 10.4061/2011/408053. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Branton D.; Deamer D.. Nanopore Sequencing An Introduction; World Scientific Publishing Co. Pte. Ltd., 2019. [Google Scholar]
  12. Begik O.; Lucas M. C.; Pryszcz L. P.; Ramirez J. M.; Medina R.; Milenkovic I.; Cruciani S.; Liu H.; Vieira H. G. S.; Sas-Chen A.; Mattick J. S.; Schwartz S.; Novoa E. M. Quantitative profiling of pseudouridylation dynamics in native RNAs with nanopore sequencing. Nat. Biotechnol. 2021, 39, 1278–1291. 10.1038/s41587-021-00915-6. [DOI] [PubMed] [Google Scholar]
  13. Smith A. M.; Jain M.; Mulroney L.; Garalde D. R.; Akeson M. Reading canonical and modified nucleobases in 16S ribosomal RNA using nanopore native RNA sequencing. PloS One 2019, 14, e0216709 10.1371/journal.pone.0216709. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Fleming A. M.; Mathewson N. J.; Howpay Manage S. A.; Burrows C. J. Nanopore dwell time analysis permits sequencing and conformational assignment of pseudouridine in SARS-CoV-2. ACS Cent. Sci. 2021, 7, 1707–1717. 10.1021/acscentsci.1c00788. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Jenjaroenpun P.; Wongsurawat T.; Wadley T. D.; Wassenaar; Trudy M.; Liu J.; Dai Q.; Wanchai V.; Akel N. S.; Jamshidi-Parsian A.; Franco A. T.; Boysen G.; Jennings M. L.; Ussery D. W.; He C.; Nookaew I. Decoding the epitranscriptional landscape from native RNA sequences. Nucleic Acids Res. 2021, 49, e7 10.1093/nar/gkaa620. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Huang S.; Zhang W.; Katanski C. D.; Dersh D.; Dai Q.; Lolans K.; Yewdell J.; Eren A. M.; Pan T. Interferon inducible pseudouridine modification in human mRNA by quantitative nanopore profiling. Genome Biol. 2021, 22, 330. 10.1186/s13059-021-02557-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Leger A.; Amaral P. P.; Pandolfini L.; Capitanchik C.; Capraro F.; Miano V.; Migliori V.; Toolan-Kerr P.; Sideri T.; Enright A. J.; Tzelepis K.; van Werven F. J.; Luscombe N. M.; Barbieri I.; Ule J.; Fitzgerald T.; Birney E.; Leonardi T.; Kouzarides T. RNA modifications detection by comparative nanopore direct RNA sequencing. Nat. Commun. 2021, 12, 7198. 10.1038/s41467-021-27393-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Tavakoli S.; Nabizadeh M.; Makhamreh A.; Gamper H.; McCormick C. A.; Rezapour N. K.; Hou Y.-M.; Wanunu M.; Rouhanifard S. H. Semi-quantitative detection of pseudouridine modifications and type I/II hypermodifications in human mRNAs using direct long-read sequencing. Nat. Commun. 2023, 14, 334. 10.1038/s41467-023-35858-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Ramasamy S.; Sahayasheela V. J.; Sharma S.; Yu Z.; Hidaka T.; Cai L.; Thangavel V.; Sugiyama H.; Pandian G. N. Chemical probe-based nanopore sequencing to selectively assess the RNA modifications. ACS Chem. Biol. 2022, 17, 2704–2709. 10.1021/acschembio.2c00221. [DOI] [PubMed] [Google Scholar]
  20. Stephenson W.; Razaghi R.; Busan S.; Weeks K. M.; Timp W.; Smibert P. Direct detection of RNA modifications and structure using single-molecule nanopore sequencing. Cell Genom. 2022, 2, 100097 10.1016/j.xgen.2022.100097. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Thomas N. K.; Poodari V. C.; Jain M.; Olsen H. E.; Akeson M.; Abu-Shumays R. L. Direct nanopore sequencing of individual full length tRNA strands. ACS Nano 2021, 15, 16642–16653. 10.1021/acsnano.1c06488. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Kurylo C. M.; Parks M. M.; Juette M. F.; Zinshteyn B.; Altman R. B.; Thibado J. K.; Vincent C. T.; Blanchard S. C. Endogenous rRNA sequence variation can regulate stress response gene expression and phenotype. Cell Rep. 2018, 25, 236–248. 10.1016/j.celrep.2018.08.093. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Li H. Minimap2: pairwise alignment for nucleotide sequences. Bioinformatics 2018, 34, 3094–3100. 10.1093/bioinformatics/bty191. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Li H.; Handsaker B.; Wysoker A.; Fennell T.; Ruan J.; Homer N.; Marth G.; Abecasis G.; Durbin R. The sequence alignment/map format and SAMtools. Bioinformatics 2009, 25, 2078–2079. 10.1093/bioinformatics/btp352. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Robinson J. T.; Thorvaldsdóttir H.; Winckler W.; Guttman M.; Lander E. S.; Getz G.; Mesirov J. P. Integrative genomics viewer. Nat. Biotechnol. 2011, 29, 24–26. 10.1038/nbt.1754. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Loman N. J.; Quick J.; Simpson J. T. A complete bacterial genome assembled de novo using only nanopore sequencing data. Nat. Meth. 2015, 12, 733–735. 10.1038/nmeth.3444. [DOI] [PubMed] [Google Scholar]
  27. Stoiber M.; Quick J.; Egan R.; Eun Lee J.; Celniker S.; Neely R. K.; Loman N.; Pennacchio L. A.; Brown J. De novo identification of DNA modifications enabled by genome-guided nanopore signal processing. bioRxiv 2017, 094672 10.1101/094672. [DOI] [Google Scholar]
  28. Baba T.; Ara T.; Hasegawa M.; Takai Y.; Okumura Y.; Baba M.; Datsenko K.; Tomita M.; Wanner B.; Mori H. Construction of Escherichia coli K-12 in-frame, single-gene knockout mutants: the Keio collection. Mol. Syst. Biol. 2006, 2, 2006.0008. 10.1038/msb4100050. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Sergeeva O. V.; Bogdanov A. A.; Sergiev P. V. What do we know about ribosomal RNA methylation in Escherichia coli?. Biochimie 2015, 117, 110–118. 10.1016/j.biochi.2014.11.019. [DOI] [PubMed] [Google Scholar]
  30. Kimura S.; Sakai Y.; Ishiguro K.; Suzuki T. Biogenesis and iron-dependency of ribosomal RNA hydroxylation. Nucleic Acids Res. 2017, 45, 12974–12986. 10.1093/nar/gkx969. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Fleming A. M.; Burrows C. J. Nanopore sequencing for N1-methylpseudouridine in RNA reveals sequence-dependent discrimination of the modified nucleotide triphosphate during transcription. Nucleic Acids Res. 2023, 51, 1914–1926. 10.1093/nar/gkad044. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Chelliserrykattil J.; Ellington A. D. Evolution of a T7 RNA polymerase variant that transcribes 2′-O-methyl RNA. Nat. Biotechnol. 2004, 22, 1155–1160. 10.1038/nbt1001. [DOI] [PubMed] [Google Scholar]
  33. Bakin A.; Kowalak J. A.; McCloskey J. A.; Ofengand J. The single pseudouridine residue in Escherichia coli 16S RNA is located at position 516. Nucleic Acids Res. 1994, 22, 3681–3684. 10.1093/nar/22.18.3681. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Conrad J.; Sun D.; Englund N.; Ofengand J. The rluC gene of Escherichia coli codes for a pseudouridine synthase that is solely responsible for synthesis of pseudouridine at positions 955, 2504, and 2580 in 23 S ribosomal RNA. J. Biol. Chem. 1998, 273, 18562–18566. 10.1074/jbc.273.29.18562. [DOI] [PubMed] [Google Scholar]
  35. Del Campo M.; Kaya Y.; Ofengand J. Identification and site of action of the remaining four putative pseudouridine synthases in Escherichia coli. RNA 2001, 7, 1603–1615. [PMC free article] [PubMed] [Google Scholar]
  36. Lesnyak D. V.; Sergiev P. V.; Bogdanov A. A.; Dontsova O. A. Identification of Escherichia coli m2G methyltransferases: I. the ycbY gene encodes a methyltransferase specific for G2445 of the 23 S rRNA. J. Mol. Biol. 2006, 364, 20–25. 10.1016/j.jmb.2006.09.009. [DOI] [PubMed] [Google Scholar]
  37. René O.; Alix J. H. Late steps of ribosome assembly in E. coli are sensitive to a severe heat stress but are assisted by the HSP70 chaperone machine. Nucleic Acids Res. 2011, 39, 1855–1867. 10.1093/nar/gkq1049. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Cheng-Guang H.; Gualerzi C. O. The ribosome as a switchboard for bacterial stress response. Front. Microbiol. 2021, 11, 619038 10.3389/fmicb.2020.619038. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Liu H.; Begik O.; Lucas M. C.; Ramirez J. M.; Mason C. E.; Wiener D.; Schwartz S.; Mattick J. S.; Smith M. A.; Novoa E. M. Accurate detection of m6A RNA modifications in native RNA sequences. Nat. Commun. 2019, 10, 4079. 10.1038/s41467-019-11713-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Craig J. M.; Laszlo A. H.; Brinkerhoff H.; Derrington I. M.; Noakes M. T.; Nova I. C.; Tickman B. I.; Doering K.; de Leeuw N. F.; Gundlach J. H. Revealing dynamics of helicase translocation on single-stranded DNA using high-resolution nanopore tweezers. Proc. Natl. Acad. Sci. U. S. A. 2017, 114, 11932–11937. 10.1073/pnas.1711282114. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Nookaew I.; Jenjaroenpun P.; Du H.; Wang P.; Wu J.; Wongsurawat T.; Moon S. H.; Huang E.; Wang Y.; Boysen G. Detection and discrimination of DNA adducts differing in size, regiochemistry, and functional group by nanopore sequencing. Chem. Res. Toxicol. 2020, 33, 2944–2952. 10.1021/acs.chemrestox.0c00202. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Heron A.; Clarke J.; Moysey R.; Wallace J.; Bruce M.; Jayasinghe L.; Caprotti D.; Soeroes S.; McNeill L.; Alves D.; Bowen R.; Milton J.. Modified Helicases US2015/0191709A1, July 9, 2015.
  43. Büttner K.; Nehring S.; Hopfner K. P. Structural basis for DNA duplex separation by a superfamily-2 helicase. Nat. Struct. Mol. Biol. 2007, 14, 647–652. 10.1038/nsmb1246. [DOI] [PubMed] [Google Scholar]
  44. Abou Assi H.; Rangadurai A. K; Shi H.; Liu B.; Clay M. C; Erharter K.; Kreutz C.; Holley C. L; Al-Hashimi H. M 2′-O-Methylation can increase the abundance and lifetime of alternative RNA conformational states. Nucleic Acids Res. 2020, 48, 12365–12379. 10.1093/nar/gkaa928. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Kimura S.; Suzuki T. Fine-tuning of the ribosomal decoding center by conserved methyl-modifications in the Escherichia coli 16S rRNA. Nucleic Acids Res. 2010, 38, 1341–1352. 10.1093/nar/gkp1073. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Fasnacht M.; Gallo S.; Sharma P.; Himmelstoß M.; Limbach P. A; Willi J.; Polacek N. Dynamic 23S rRNA modification ho5C2501 benefits Escherichia coli under oxidative stress. Nucleic Acids Res. 2022, 50, 473–489. 10.1093/nar/gkab1224. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Koculi E.; Cho S. S. RNA post-transcriptional modifications in two large subunit intermediates populated in E. coli cells expressing helicase inactive R331A DbpA. Biochemistry 2022, 61, 833–842. 10.1021/acs.biochem.2c00096. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Rivière J.; Klarskov K.; Wagner J. R. Oxidation of 5-hydroxypyrimidine nucleosides to 5-hydroxyhydantoin and its alpha-hydroxy-ketone isomer. Chem. Res. Toxicol. 2005, 18, 1332–1338. 10.1021/tx050121i. [DOI] [PubMed] [Google Scholar]
  49. O’Connor M.; Leppik M.; Remme J. Pseudouridine-free Escherichia coli ribosomes. J. Bacteriol. 2018, 200, e00540 10.1128/JB.00540-17. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Schmidt A.; Kochanowski K.; Vedelaar S.; Ahrné E.; Volkmer B.; Callipo L.; Knoops K.; Bauer M.; Aebersold R.; Heinemann M. The quantitative and condition-dependent Escherichia coli proteome. Nat. Biotechnol. 2016, 34, 104–110. 10.1038/nbt.3418. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Huber S. M.; Leonardi A.; Dedon P. C.; Begley T. J. The versatile roles of the tRNA epitranscriptome during cellular responses to toxic exposures and environmental stress. Toxics 2019, 7, 17. 10.3390/toxics7010017. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Henderson B. R.; Pfister G.; Boeck G.; Kind M.; Wick G. Expression levels of heat shock protein 60 in human endothelial cells in vitro are unaffected by exposure to 50 Hz magnetic fields. Cell Stress Chap. 2003, 8, 172–182. . [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

cb3c00166_si_001.pdf (9MB, pdf)

Data Availability Statement

The base called data for all sequencing experiments has been publicly deposited at Zenodo a public repository for data at doi: 10.5281/zenodo.7746124 that is searchable in the OpenAIRE explorer.

Articles from ACS Chemical Biology are provided here courtesy of American Chemical Society

RESOURCES