ABSTRACT

   Acetogens synthesize acetyl-CoA via the CO[2]-fixing Wood-Ljungdahl
   pathway. Despite their ecological and biotechnological importance,
   their translational regulation of carbon and energy metabolisms remains
   unclear. Here, we report how carbon and energy metabolisms in the model
   acetogen Acetobacterium woodii are translationally controlled under
   different growth conditions. Data integration of genome-scale
   transcriptomic and translatomic analyses revealed that the acetogenesis
   genes, including those of the Wood-Ljungdahl pathway and energy
   metabolism, showed changes in translational efficiency under
   autotrophic growth conditions. In particular, genes encoding the
   Wood-Ljungdahl pathway are translated at similar levels to achieve
   efficient acetogenesis activity under autotrophic growth conditions,
   whereas genes encoding the carbonyl branch present increased
   translation levels in comparison to those for the methyl branch under
   heterotrophic growth conditions. The translation efficiency of genes in
   the pathways is differentially regulated by 5′ untranslated regions and
   ribosome-binding sequences under different growth conditions. Our
   findings provide potential strategies to optimize the metabolism of
   syngas-fermenting acetogenic bacteria for better productivity.

   IMPORTANCE Acetogens are capable of reducing CO[2] to multicarbon
   compounds (e.g., ethanol or 2,3-butanediol) via the Wood-Ljungdahl
   pathway. Given that protein synthesis in bacteria is highly energy
   consuming, acetogens living at the thermodynamic limit of life are
   inevitably under translation control. Here, we dissect the
   translational regulation of carbon and energy metabolisms in the model
   acetogen Acetobacterium woodii under heterotrophic and autotrophic
   growth conditions. The latter may be experienced when acetogen is used
   as a cell factory that synthesizes products from CO[2] during the gas
   fermentation process. We found that the methyl and carbonyl branches of
   the Wood-Ljungdahl pathway are activated at similar translation levels
   during autotrophic growth. Translation is mainly regulated by the
   5′-untranslated-region structure and ribosome-binding-site sequence.
   This work reveals novel translational regulation for coping with
   autotrophic growth conditions and provides the systematic data set,
   including the transcriptome, translatome, and
   promoter/5′-untranslated-region bioparts.

   KEYWORDS: acetogen, acetogenesis, translational regulation,
   Wood-Ljungdahl pathway

INTRODUCTION

   Acetogenesis has received immense attention because it offers an
   efficient route for microbial metabolism to convert one-carbon (C[1])
   gaseous feedstocks, such as carbon monoxide (CO) and carbon dioxide
   (CO[2]), to acetyl-CoA via the reductive acetyl-CoA pathway, known as
   the Wood-Ljungdahl (WL) pathway ([41]1, [42]2). This unique pathway is
   presumed to be the first carbon fixation pathway on earth ([43]3) and
   enables acetogens to grow autotrophically using C[1] gases as the sole
   carbon sources ([44]4, [45]5). A few system-level analyses for
   acetogens revealed that the expression of acetogenesis-related genes
   was upregulated under autotrophic growth conditions ([46]6[47]–[48]10);
   however, the corresponding protein abundances presented low
   correlations with the mRNA levels ([49]7). While the transcriptional
   regulation of acetogenesis-related genes is well described, the
   mechanism by which acetogens consistently maintain efficiently balanced
   translation levels under autotrophic growth conditions remained unclear
   ([50]11). In general, protein synthesis in bacteria is highly energy
   consuming, using up approximately half of the energy utilized for cell
   growth ([51]12). Since acetogens grow at the thermodynamic limit of
   life ([52]13), translational regulation is inevitable, and thus,
   resource allocation should be optimized ([53]14). Therefore, we
   speculated that the synthesis of essential proteins for autotrophic
   growth is tightly regulated and that the balanced expression of
   acetogenesis-related proteins should be crucial for efficient metabolic
   reactions under autotrophic growth conditions.

   Here, we provide the first systematic analysis of gene expression in
   Acetobacterium woodii, one of the model acetogens, at both the
   transcriptional and translational levels under heterotrophic (fructose)
   and autotrophic (H[2] plus CO[2] [H[2]+CO[2]]) growth conditions. Based
   on multi-omics analysis, we found that A. woodii efficiently controls
   the translation of acetogenesis-related proteins under autotrophic
   growth conditions. Importantly, the genes encoding the methyl and
   carbonyl branches are translated at similar levels during autotrophic
   growth, whereas the translation levels of the genes from the carbonyl
   branch are increased compared to those of genes from the methyl branch
   under heterotrophic growth conditions. In addition, promoter and
   5′-untranslated-region (5′-UTR) analyses suggest that the translation
   efficiency is differentially regulated by 5′-UTR and
   ribosome-binding-site (RBS) sequences under autotrophic growth
   conditions. Our findings provide a crucial basis for understanding
   acetogenesis from organic and inorganic substrates and further
   engineering acetogens to produce value-added products from C[1] gaseous
   feedstocks with better productivity.

RESULTS

Genome-scale determination of translation levels.

   Ribosome profiling (Ribo-Seq) enables the monitoring of protein
   synthesis efficiency at a genome-wide scale by using deep sequencing of
   ribosome-protected mRNA fragments (RPFs) ([54]15). Changes in the ratio
   between the RPF and mRNA transcript levels can be used to identify
   translational regulation under the conditions of interest
   ([55]15[56]–[57]17). To determine the changes in the translation levels
   of A. woodii genes under autotrophic and heterotrophic conditions, we
   generated more than 10.3 million Ribo-Seq reads with an average read
   length of 30 to 32 bp, indicating at least an 87.8× sequencing depth
   (see [58]Table S1 in the supplemental material). The number of RPF
   reads per gene was then normalized using DEseq2 ([59]18)
   ([60]Table S2), resulting in a high degree of correlation between the
   biological duplicates (Pearson’s r > 0.97) ([61]Fig. S1). We also
   obtained 1.2 million to 4.7 million RNA sequencing (RNA-Seq) reads,
   corresponding to at least 42.8-fold coverage ([62]Table S1). As
   expected, a positive correlation was observed between transcription and
   translation levels under heterotrophic (Pearson’s r = 0.86) and
   autotrophic (Pearson’s r = 0.90) growth conditions, respectively
   ([63]Fig. 1A). For example, the hydrogen-dependent carbon dioxide
   reductase (HDCR) genes (fdhF2, hydA2, fdhD, hycB2, and hycB3) were
   highly upregulated at the transcription (fold changes [FCs] = 5.2 to
   10.2; P value adjusted for multiple testing with the Benjamini-Hochberg
   procedure [P[adj]] of <8.42 × 10^−20) and translation (fold
   changes = 3.0 to 4.5; P[adj] < 5.97 × 10^−28) levels under autotrophic
   growth conditions ([64]Fig. 1B; see also [65]Text S1 in the
   supplemental material). However, the genes encoding selenium-free
   formate dehydrogenase (FDH) (FdhF1) and FdhF1-specific HycB1 were not
   expressed at both the transcription and translation levels under both
   growth conditions ([66]Text S1).

FIG 1.

   FIG 1
   [67]Open in a new tab

   Determination of the translational landscape of A. woodii. (A)
   Scatterplot of all pairwise log[10] expression levels between mRNA (x
   axis) and RPF (y axis), presenting a positive correlation. See
   [68]Tables S1 and [69]S2 in the supplemental material for sequencing
   statistics and all RPF levels, respectively. (B) Example of RNA-Seq and
   Ribo-Seq (RPF) profiles for hydrogen-dependent carbon dioxide reductase
   (HDCR) (Awo_c08190 to Awo_c08260), highlighted in dark blue. H,
   heterotrophic growth; A, autotrophic growth. (C) Comparison of
   translation levels to protein abundances. Points represent matching
   ORFs in the fold changes of translation levels and the changes in the
   amounts of protein previously reported ([70]20) during growth on
   fructose versus growth on H[2]+CO[2] (Pearson’s r = 0.66). (D)
   Proportional translation levels for the first and second genes in the
   operon were compared during growth on fructose or H[2]+CO[2]. (E)
   Translation levels (RPKM[RPF]) of the F[1]F[o] ATP synthase operon
   (Awo_c02150 to Awo_c02240) and the RplJ and RplL ribosomal proteins
   (Awo_c10870 to Awo_c10880) during autotrophic growth in correlation to
   their protein copy numbers in the enzyme complexes.
   TEXT S1

   Analysis of RNA-Seq and Ribo-Seq data. Download [71]Text S1, PDF file,
   0.2 MB^ (240.5KB, pdf) .

   Copyright © 2021 Shin et al.

   This content is distributed under the terms of the [72]Creative Commons
   Attribution 4.0 International license.
   FIG S1

   Generation of transcriptomics and translatomics data in A. woodii DSM
   1030. (A and B) Growth curves and metabolite concentrations were
   measured with fructose (A) and H[2]+CO[2] (B) as the carbon and energy
   sources. The data are presented as the means from three different
   biological replicates ± SEM. For RNA-Seq and Ribo-Seq, cell cultures of
   heterotrophically and autotrophically grown cells were harvested at
   OD[600]s of 0.8 and 0.15, respectively. (C) Correlation heat map of
   gene expression profiles from RNA-Seq libraries. (D)
   Principal-component (PC) analysis (PCA) of RNA-Seq libraries. (E)
   Concordance of RNA-Seq data with quantitative RT-PCR data. To validate
   RNA-Seq data independently, quantitative RT-PCR was performed using RNA
   extracted from the cultures grown under the same growth conditions.
   Normalized expression levels and quantification cycle (C[q]) values
   were compared for seven genes. Validated genes are represented by
   uppercase letters: A, secA (Awo_c28060), encoding a protein translocase
   subunit; B, gyrA (Awo_c00060), encoding DNA gyrase; C, hydA2
   (Awo_c08260) of the HDCR gene cluster; D, acsD (Awo_c10710) of the WL
   pathway carbonyl branch; E, metF (Awo_c09310) of the WL pathway methyl
   branch; F, etfB (Awo_c08710) of the electron-bifurcating lactate
   dehydrogenase genes; G, pyc (Awo_c26160), encoding pyruvate
   carboxylase. (F) Correlation heat map of gene expression profiles from
   Ribo-Seq libraries. (G) PCA of Ribo-Seq libraries indicating the
   reproducibility and quality of Ribo-Seq data. (H) Average Ribo-Seq read
   density aligned at the start and stop codons for heterotrophic (H1 and
   H2) and autotrophic (A1 and A2) growth. 5′ ends of sequencing reads
   started from 17 bp upstream of the start codon with 3-nt periodicity,
   indicating the translocation of the ribosome during translation
   elongation. Download [73]FIG S1, PDF file, 0.7 MB^ (670.3KB, pdf) .

   Copyright © 2021 Shin et al.

   This content is distributed under the terms of the [74]Creative Commons
   Attribution 4.0 International license.
   TABLE S1

   Statistics of RNA-Seq, dRNA-Seq, and Ribo-Seq. Download [75]Table S1,
   PDF file, 0.1 MB^ (108.3KB, pdf) .

   Copyright © 2021 Shin et al.

   This content is distributed under the terms of the [76]Creative Commons
   Attribution 4.0 International license.
   TABLE S2

   Transcription start sites, mRNA expression levels, ribosome-protected
   mRNA fragment (RPF) levels, and translation efficiencies (TE) according
   to heterotrophic and autotrophic growth conditions in A. woodii.
   Download [77]Table S2, XLSX file, 0.8 MB^ (824.8KB, xlsx) .

   Copyright © 2021 Shin et al.

   This content is distributed under the terms of the [78]Creative Commons
   Attribution 4.0 International license.

   We defined the translation level of each gene as the number of
   ribosomes engaged in translation elongation per gene length, with
   correspondence to reads per kilobase per million RPFs (RPKM[RPF])
   ([79]10, [80]15, [81]17, [82]19). The comparison between the fold
   changes of RPKM[RPF] values and the changes in the amount of protein
   previously reported in A. woodii during growth on fructose versus
   growth on H[2]+CO[2] ([83]20) showed a moderate correlation (Pearson’s
   r = 0.66) ([84]Fig. 1C). Discrepancies may be attributed to slight
   differences in culture conditions, the possibility that protein
   turnover rates or ribosome elongation rates are not constant for all
   genes, or experimental bias in the process of mass spectrometry
   analyses.

   For validation of data quality, we initially investigated whether
   subunits of multimeric protein complexes are translated according to
   their stoichiometry under autotrophic growth conditions ([85]15). For
   this analysis, we examined the known stoichiometry of the multimeric
   protein complexes in A. woodii. A total of 64 subunits of 17 protein
   complexes were examined, and the translation levels were linearly
   correlated between the first and second subunits of the multimeric
   protein complexes (Pearson’s r = 0.97) ([86]Fig. 1D). RPKM[RPF] values
   of nonequimolar subunits such as RplJ/L and AtpB/E were also
   proportional to the respective stoichiometries, ranging from 1- to
   9-fold, whereas RPKM[RNA] values were less correlated ([87]Fig. 1E).
   Thus, the translation of each subunit matched the respective
   stoichiometry of protein complexes for achieving the correct
   stoichiometry in multimeric protein complexes.

Translation levels of acetogenesis-related genes under autotrophic growth
conditions.

   The transcriptomic results suggest that the genes encoding acetogenesis
   as well as gluconeogenesis and the pentose phosphate pathway were
   activated for biomass formation and CO[2] reduction during autotrophic
   growth ([88]6[89]–[90]8, [91]10) (see [92]Fig. S2 and [93]Text S1 in
   the supplemental material), as is the case with Acetobacterium bakii
   ([94]9). Furthermore, we investigated the translation levels of the
   acetogenesis-associated genes under autotrophic growth conditions
   ([95]Table S2).
   FIG S2

   Transcriptome dynamics of the central metabolism of Acetobacterium
   woodii during heterotrophic and autotrophic growth. (A) MA plot of
   genes sequenced during heterotrophic and autotrophic growth.
   Differentially expressed genes are presented as red dots. See
   [96]Tables S1 and [97]S2 in the supplemental material for sequencing
   statistics and all transcript abundances, respectively. (B) Enriched
   KEGG pathways of the differentially expressed genes
   (Bonferroni-corrected P value of <0.05). (C) Differentially expressed
   genes involved in the pentose phosphate (PP) pathway,
   glycolysis/gluconeogenesis, the Wood-Ljungdahl (WL) pathway, the
   branched TCA cycle, energy conservation, and the redox-balancing system
   in A. woodii. The heat map indicates the RNA expression levels and
   log[2] fold changes. See [98]Table S2 for the complete enzyme names and
   all transcript abundances. Download [99]FIG S2, PDF file, 2.7 MB^
   (2.7MB, pdf) .

   Copyright © 2021 Shin et al.

   This content is distributed under the terms of the [100]Creative
   Commons Attribution 4.0 International license.

   First, the HDCR complex revealed dynamic translational regulation under
   autotrophic growth conditions within a 4-fold range ([101]Fig. 2A).
   Among the subunits of the HDCR complex, FdhF2 (RPKM[RPF] = 562.6) and
   HycB2 (RPKM[RPF] = 502.6) were translated at rates more than 2.2 times
   higher than those of HycB3 (RPKM[RPF] = 228.3) and HydA2 (RPKM[RPF] =
   259.9). The FdhF2 and HycB2 complex also catalyzes the reduction of
   CO[2] to formate with ferredoxin, not using H[2] ([102]21), reflecting
   in particular the protein synthesis rate of the HDCR gene cluster. Six
   genes in the methyl branch were similarly translated as expected from
   the subunit stoichiometry (coefficient of variation [CV] = 11%)
   ([103]Fig. 2B). For example, methylene-tetrahydrofolate (THF) reductase
   (MTHFR) is a unique heterotrimer complex comprising the MetF, MetV, and
   RnfC2 subunits ([104]22). They were equally translated according to
   their subunit stoichiometry ratio (1:1:1) under autotrophic growth
   conditions. However, genes from the methyl branch were not similarly
   transcribed (CV = 38%) ([105]Fig. 2B), resulting in significant
   differences between the CV[RNA] and CV[RPF] (P = 0.03).

FIG 2.

   FIG 2
   [106]Open in a new tab

   Translational status of the WL pathway under autotrophic growth
   conditions. (A) RNA expression levels and translation levels of
   hydrogen-dependent carbon dioxide reductase (HDCR) under autotrophic
   growth conditions. Illustrations and subunit ratios of the HDCR complex
   are shown. For CO[2] reduction, electrons are provided by either
   reduced ferredoxin or the HydA-HydB3 subunits, where H[2] oxidation
   occurs. (B) Comparison between the mRNA (RPKM[RNA]) and translation
   (RPKM[RPF]) levels of the methyl branch of the WL pathway under
   autotrophic growth conditions. Illustrations and subunit ratios of the
   MTHFR complex are shown. The coefficient of variation (CV) (SD of the
   mean × 100) was calculated based on the RPKM[RPF] values of all genes
   in the methyl branch gene cluster under each growth condition. (C) RPF
   levels for the carbonyl branch of the WL pathway under autotrophic
   growth conditions. Illustrations and subunit ratios of the ACS/CODH and
   MET complexes are shown. (D) Correspondence between the published
   protein abundance of C. ljungdahlii ([107]27) and the translation ratio
   of carbonyl branch proteins. The reported exponentially modified
   protein abundance index (emPAI) value for each protein was used for
   determining the protein content (moles percent) using the formula
   emPAI/Σ(emPAI) × 100. C. ljungdahlii values are plotted against the
   relative translation levels of A. woodii (Pearson’s r = 0.90).

   Second, the enzymes in the carbonyl branch were translated in several
   different ratios under autotrophic growth conditions ([108]Fig. 2C).
   The subunits of two large protein complexes (220 to 310 kDa), the
   methyltransferase/corrinoid iron-sulfur protein (MET/CoFeSP) and the
   acetyl-CoA synthase/carbon monoxide dehydrogenase (ACS/CODH) complexes,
   had relatively high translation levels (3- to 5-fold higher) than
   others in the carbonyl branch. The homodimeric MET protein (AcsE)
   ([109]23) interacts with two CoFeSPs, and each CoFeSP comprises small
   and large subunits (AcsC and AcsD, respectively) ([110]5, [111]24).
   Thus, presumably, a higher translation level of AcsE is required for
   efficient MET/CoFeSP complex assembly. Similar translation levels
   (CV = 6.5%) were observed for acsA (Awo_c10740) and acsB (Awo_c10760),
   which encode the heterotetrameric ACS/CODH complex (A2B2)
   ([112]Fig. 2C) ([113]25, [114]26). We further validated the translation
   levels of the genes in the carbonyl branch by comparing our data
   against the protein abundances of Clostridium ljungdahlii under syngas
   fermentation conditions ([115]27). Our results were in accordance with
   the protein abundance, with Pearson’s r value of 0.90 ([116]Fig. 2D).

   Third, the subunits of the electron-bifurcating hydrogenase complex
   (CV[RPF] = 24%; RPKM[PRF] = 107.3 to 204.9; CV[RNA] = 46%; RPKM[RNA] =
   239.8 to 917.3) encoded by hydA1, hydB, hydD, and hydC and the Rnf
   complex (CV[RPF] = 23%; RPKM[PRF] = 285.6 to 517.8; CV[RNA] = 23%;
   RPKM[RNA] = 166.1 to 317.5) encoded by rnfB, rnfA, rnfE, rnfG, rnfD,
   and rnfC1 were produced within a 2-fold range ([117]Table S2). These
   results are consistent with their subunit stoichiometric relationship.
   Collectively, the acetogenesis-related genes were effectively regulated
   at the translational level under autotrophic growth conditions.
   Moreover, the production of carbonyl branch enzymes was upregulated
   compared to that of other enzymes related to acetogenesis, reflecting
   their limited reaction fluxes ([118]16).

Differential translation levels of acetogenesis-associated genes under
heterotrophic and autotrophic growth conditions.

   Although the changes in the mRNA and RPF levels of the pentose
   phosphate pathway, glycolytic/gluconeogenesis pathway, and
   tricarboxylic acid (TCA) cycle are similar, some of the
   acetogenesis-related genes were determined to be insignificant from the
   fructose conditions to the H[2]+CO[2] conditions ([119]Fig. 3A). In
   particular, ACS/CODH, MET/CoFeSP, F[1]F[o] ATP synthase, and the Rnf
   complex were significantly upregulated at the transcription level (fold
   changes = 1.8 to 8.5; P[adj] < 1.72 × 10^−4); however, surprisingly,
   all genes retained a similar (P[adj] < 0.85) or lower RPF level under
   autotrophic growth conditions (fold changes = 0.68 to 0.89;
   P[adj] < 8.12 × 10^−3) (see [120]Text S1 in the supplemental material).

FIG 3.

   [121]FIG 3
   [122]Open in a new tab

   Differential translation levels of acetogenesis proteins under fructose
   and H[2]+CO[2] conditions. (A) Transcriptome and translatome dynamics
   of central metabolism during heterotrophic and autotrophic growth. Fold
   changes of genes involved in the pentose phosphate (PP) pathway,
   glycolytic/gluconeogenesis, the Wood-Ljungdahl (WL) pathway, the
   branched TCA cycle, the serine/glycine synthesis pathway, the glycine
   cleavage system (GCS), and redox-balancing systems in A. woodii are
   shown. The heat map shows log[2] fold changes of mRNA and RPF between
   H[2]+CO[2] and fructose conditions. See [123]Table S2 in the
   supplemental material for the complete enzyme names and RPKM[RPF]
   levels. (B) Correlation of gene expression between the fructose and
   H[2]+CO[2] conditions at the transcription (RPKM[RNA]) and translation
   (RPKM[RPF]) levels. Pearson’s correlations (R) between the two values
   are shown. (C) Comparison between the translation levels of WL
   pathway-related genes under fructose and H[2]+CO[2] conditions. The
   vertical scatterplot indicates individual translation levels (black,
   RPKM[RPF]) and the median value (red line). The red dotted line
   represents the median value of the methyl branch. The significance of
   differences was assessed by the Wilcoxon rank sum test (ns, not
   significant; **, P < 0.01; ***, P < 0.001). (D) Correlation between the
   protein copy numbers of proteins and the corresponding translation
   levels (RPKM[RPF]) of the F[1]F[o] ATP synthase operon under fructose
   and H[2]+CO[2] conditions. Pearson’s correlations (R) and P values
   between the two values are shown. (E) Comparison of acetate production
   from fructose and H[2]+CO[2] by A. woodii cell suspensions. The assay
   design is presented on the left. Resting cell assays were performed in
   imidazole buffer with 20 mM NaCl under strictly anoxic conditions (<5
   ppm O[2]) (right). Cell suspensions were prepared from cells grown on
   fructose or H[2]+CO[2] and analyzed for their ability to produce
   acetate from fructose or H[2]+CO[2] (80:20 [vol/vol]; 200 kPa),
   resulting in four types of samples (designated H→H, H→A, A→H, and A→A).
   Acetate production was measured via HPLC, and all values are means ±
   standard errors of the means (SEM) obtained from three independent
   experiments. (F, top) Design of plasmid-based expression for the
   characterization of GCS contribution to the heterotrophic growth of E.
   limosum. Plasmid pJIR-GCS harbors a transcription unit of gcvTHP genes
   under the control of the ptetO1 (pt1) promoter. Plasmid pJIR-NoP, which
   harbors gcvTHP genes without a promoter (NoP), was also constructed and
   used as a negative control. (Bottom) Cell growth (left) and metabolite
   profiles (right) of the GCS and NoP strains under heterotrophic growth
   conditions. Cells were grown using DSMZ medium 135 supplemented with 4
   g liter^−1 fructose, 2 g liter^−1 NaCl, 15 μg ml^−1 thiamphenicol, and
   30 ng ml^−1 anhydrotetracycline. The values are presented as the means
   from three different biological replicates ± SEM. (G) Quantitative
   RT-PCR data for GCS gene expression in the E. limosum GCS strain.
   Plasmid pJIR-NoP, which harbors the gcvTHP genes without a promoter,
   was also constructed and utilized as a negative control. Each value is
   the mean from three independent replicate experiments, and error bars
   indicate SEM. The ptsF gene (ELIM_c2548) (fructose-specific PTS) was
   used as a positive control. The housekeeping gene gyrA (ELIM_c0629)
   (DNA gyrase subunit A) was used as the reference.

   Next, the correlation between gene expression and growth conditions
   across the multimeric protein subunits associated with acetogenesis was
   investigated at the transcription and translation levels
   ([124]Fig. 3B). The analysis revealed that the RPF ratios of subunits
   in multimeric proteins are relatively constant (Pearson’s r = 0.64 to
   1.00) for an individual gene independent of the growth conditions,
   although the RNA ratio is not correlated (Pearson’s r = −0.40 to 0.13)
   for MTHFR, hydrogenase, and the Rnf complex between heterotrophic and
   autotrophic growth conditions, most likely reflecting tight
   translational regulation of acetogenesis-associated genes.

   Considering that protein production was regulated at both the
   transcriptional and translational levels, we hypothesized that the key
   proteins for acetogenesis are more efficiently translated under
   autotrophic growth conditions. Overall, we observed similar translation
   levels of genes encoding the HDCR, methyl branch, and carbonyl branch
   under autotrophic growth conditions (P > 0.05 by a Wilcoxon rank sum
   test), whereas enzymes of the HDCR and carbonyl branch had higher
   translation levels than the enzymes of the methyl branch under
   heterotrophic growth conditions (P < 0.005 by a Wilcoxon rank sum test)
   ([125]Fig. 3C). In particular, two large protein complexes (ACS/CODH
   and MET/CoFeSP) (median RPKM[RPF] = 3,833.2) were upregulated up to
   12-fold at the translation level compared to the HDCR complex (median
   RPKM[RPF] = 311.8) under heterotrophic growth conditions, suggesting
   unbalanced protein expression among the WL pathway proteins.
   Furthermore, we observed that RPKM[RPF] values of F[1]F[o] ATP
   synthases and their protein copy numbers were less correlated under
   heterotrophic growth conditions (Pearson’s r = 0.66; P > 0.05), whereas
   the RPKM[RPF] and respective stoichiometries were significantly similar
   under autotrophic growth conditions (Pearson’s r = 0.80; P = 0.01),
   indicating the efficient translation of each subunit under H[2]+CO[2]
   conditions ([126]Fig. 3D).

   To determine whether the unbalanced translation of the WL pathway
   observed under heterotrophic growth conditions affects acetogenesis,
   resting cell assays were performed. To test this, we used
   heterotrophically (H) and autotrophically (A) grown A. woodii cells,
   resulting in four types of samples (designated H→H, H→A, A→H, and A→A)
   ([127]Fig. 3E). Interestingly, the levels of acetate production from
   fructose were similar in autotrophically (A→H, 13.8 mM) and
   heterotrophically (H→H, 14.4 mM) grown cells. In contrast, acetate
   production from H[2]+CO[2] in heterotrophically grown cells (H→A,
   7.5 mM) was only about half of that in autotrophically grown cells
   (A→A, 14.7 mM), suggesting relatively low acetogenesis activity of the
   heterotrophically grown cells. Based on the data from translational
   regulation and metabolic activity, we propose that the WL pathway genes
   are translated at similar levels to achieve efficient acetogenesis
   activity under autotrophic growth conditions.

The glycine cleavage system may contribute to heterotrophic acetogenesis.

   Since the translation level of carbonyl branch enzymes was higher than
   that of enzymes from the methyl branch during growth on fructose, we
   looked for folate‐dependent single‐carbon reaction sequences that could
   replenish the carbon in the methyl branch. The intermediates
   methylene-THF and methyl-THF could be ideal candidates. In the A.
   woodii genome, we found genes encoding the glycine cleavage system
   (GCS), in which the carbon flow can be coupled to the WL pathway
   ([128]Fig. 3A and [129]Fig. S3A). Recently, we reported the functional
   cooperation of the WL pathway and the GCS in Clostridium drakei
   ([130]28), which are functionally interconnected to convert CO[2] into
   acetyl-CoA and acetyl-phosphate. 3-Phosphoglycerate (3-PG), an
   intermediate in the Embden-Meyerhof-Parnas (EMP) pathway, is converted
   to serine and glycine via the serine/glycine synthesis pathway. Serine
   transhydroxymethylase produces methylene-THF and glycine, and the
   latter is oxidized to CO[2] and methylene-THF. Therefore, this system
   is expected to contribute to the utilization of the carbonyl branch
   during heterotrophic growth.
   FIG S3

   Contribution of the glycine cleavage system to heterotrophic
   acetogenesis in A. woodii. (A) Schematic representation of the
   reactions involved in heterotrophic acetogenesis, including
   serine/glycine synthesis (blue), the glycine cleavage system (GCS)
   (blue), and the carbonyl branch of the WL pathway (blue). Carbon
   dioxide and methylene-THF produced by glycolysis and the GCS are
   highlighted in green. The NADH and reduced ferredoxin produced by
   glycolysis and the GCS are highlighted in red. See [131]Table S2 in the
   supplemental material for the complete enzyme names. (B) Translation
   levels (RPKM[RPF]) of the EMP pathway, serine/glycine synthesis, the
   GCS, and the WL pathway under heterotrophic and autotrophic growth
   conditions. See [132]Table S2 for the translation levels. (C) A
   comparison between proteins associated with serine/glycine synthesis
   and the GCS in A. woodii, E. coli, and eight acetogenic bacteria was
   performed using the BLASTp algorithm. Download [133]FIG S3, PDF file,
   0.5 MB^ (503.6KB, pdf) .

   Copyright © 2021 Shin et al.

   This content is distributed under the terms of the [134]Creative
   Commons Attribution 4.0 International license.

   Our data revealed that genes in the serine biosynthesis pathway,
   comprising 3-phosphoglycerate dehydrogenase (PHGDH), phosphoserine
   aminotransferase (PSAT), and phosphoserine phosphatase (PSP), were
   significantly upregulated at both the transcription (fold changes = 2.1
   to 2.3; P[adj] < 0.020) and translation (fold changes = 1.4 to 2.4;
   P[adj] < 0.017) levels under heterotrophic growth conditions
   ([135]Fig. S3B). Meanwhile, glyceraldehyde-3-phosphate dehydrogenase
   (GAP), phosphoglycerate kinase (PGK), triosephosphate isomerase (TPI),
   and the GCS gene cluster encoding glycine dehydrogenase (GLDC) (GcvH),
   aminomethyltransferase (AMT) (GcvT), and glycine cleavage system H
   protein (GcvH) were significantly upregulated at the translational
   level only (fold changes = 1.4 to 2.2; P[adj] < 1.30 × 10^−4) during
   heterotrophic growth. In general, all proteins required for
   GCS-mediated heterotrophic acetogenesis were significantly activated at
   the translational level ([136]Fig. S3B). This is in accordance with the
   hypothesis that the GCS pathway replenishes the methyl branch during
   growth on fructose.

   Since no efficient genetic tools for knockout have been reported yet
   for A. woodii, we investigated the contribution of the GCS to
   heterotrophic acetogenesis in other acetogens that do not have this
   system. Most acetogens contain GCS genes, whereas Eubacterium limosum
   lacks key genes of this pathway ([137]Fig. S3C). Considering that the
   carbonyl branch enzymes of E. limosum are at least 2-fold
   translationally upregulated compared to the enzymes of the methyl
   branch under heterotrophic growth conditions ([138]10), we assumed that
   GCS-mediated heterotrophic metabolism with the energy conservation
   system ([139]29) could be an energetically efficient process in E.
   limosum ([140]Fig. S4A to [141]C). To confirm this, we initially
   identified the transcription unit of the GCS via transcription start
   site (TSS) and termination site determinations ([142]Fig. 3F), showing
   that gcvHTP may form an operon. The genes were cloned into plasmid
   pJIR-GCS under the control of the ptetO1 (pt1) promoter, and the
   plasmid was transformed into E. limosum. The expression of the gcvHTP
   genes led to a 2.25-fold increase in the growth rate (P < 0.002) on
   fructose compared to the control strain ([143]Fig. 3F and [144]G).
   Along with the growth rate increase, relatively rapid fructose
   reduction and 140% higher acetate production (25.3 mM) were also
   observed in the GCS strain, compared with the control strain. These
   results indicated that the heterologous expression of the GCS in E.
   limosum conveys a fitness advantage during growth on fructose. However,
   an additional knockout study is required to validate the GCS function
   in A. woodii.
   FIG S4

   Fructose metabolism and heterotrophic acetogenesis of E. limosum. (A)
   Schematic representation of the reactions involved in heterotrophic
   acetogenesis in E. limosum. E. limosum is closely related to A. woodii,
   and the WL pathway and bioenergetic processes are similar to those of
   A. woodii. In contrast to A. woodii, E. limosum has formate
   dehydrogenase (FDH) for the reduction of carbon dioxide with reduced
   ferredoxin, instead of the HDCR complex. Therefore, when fructose is
   completely converted to acetate, 4 ATP mol^−1 fructose can be generated
   according to the energy conservation model. (B) Schematic
   representation of the reactions involved in heterotrophic acetogenesis,
   including serine/glycine synthesis (blue) and the carbonyl branch of
   the WL pathway (blue), in E. limosum. When the metabolic flux is
   changed to the serine/glycine synthesis pathway due to the absence of
   the GCS, unbalanced redox potential and low ATP yield can be expected
   (3.8 ATP plus 0.3 NADH mol^−1 fructose) due to the absence of the GCS.
   (C) Schematic representation of the reactions involved in heterotrophic
   acetogenesis, including serine/glycine synthesis (blue), the GCS
   (blue), and the carbonyl branch of the WL pathway (blue), in E.
   limosum. Under heterotrophic growth conditions, fructose is catabolized
   via the EMP pathway, and 3-phosphoglycerate (3-PG), an intermediate
   metabolite involved in the EMP pathway, can then be converted into
   serine and glycine through the serine/glycine synthesis pathway. Serine
   and THF are converted into glycine and methylene-THF (CH[2]-THF) by
   serine hydroxymethyltransferases (SHMT), and the enzymatic cleavage of
   glycine produces additional carbon dioxide (green) and methylene-THF
   (green) via the GCS, generating 1 mol carbon dioxide and 2 mol
   methylene-THF from 1 mol serine. PHGDH diverts glycolytic flux into the
   serine/glycine synthesis pathway. For the sake of simplicity, the
   particular glycolytic flux into the serine/glycine synthesis pathway is
   assumed to be based on the carbonyl branch/methyl branch translation
   ratio between the two growth conditions. The carbon dioxide and
   methylene-THF produced by glycolysis and the GCS are highlighted in
   green. The NADH and reduced ferredoxin produced by glycolysis and the
   GCS are highlighted in red. Download [145]FIG S4, PDF file, 0.7 MB^
   (673.3KB, pdf) .

   Copyright © 2021 Shin et al.

   This content is distributed under the terms of the [146]Creative
   Commons Attribution 4.0 International license.

Changes in translation efficiency between heterotrophic and autotrophic
growth conditions.

   Translation in bacteria is known to be regulated by controlling their
   quantity of active ribosomes ([147]30) by RNAs or binding of regulatory
   proteins ([148]31). Considering the dynamic and substrate-specific
   translational regulation of acetogenesis-associated genes, we
   hypothesized that differences in translational efficiency (TE)
   contribute substantially to the control of gene expression in the
   acetogenesis of A. woodii. To address this, we calculated the TE of
   each gene by dividing the RPF density by the corresponding mRNA level
   (RPF/mRNA) and examined the changes in TE between the two growth
   conditions (autotrophic/heterotrophic growth conditions). Most genes
   revealed a 20-fold range change in the TE (0.2 to 4 log[2]). By
   examining the fold change threshold (1 standard deviation [SD] from the
   mean at a log[2] scale), we identified 547 high-TE (16%) and 274 low-TE
   (8%) genes under the two growth conditions ([149]Fig. 4A). A negative
   correlation between the mRNA change and TE change (Pearson’s r = −0.49)
   indicated translational buffering ([150]Fig. 4B) ([151]32). This
   observation demonstrates that ribosome abundance is limited compared to
   the transcript level, which is supported by the fact that the TE of
   ribosomal proteins is downregulated under autotrophic growth
   conditions. Gene Ontology (GO)/KEGG pathway enrichment analysis of the
   translationally regulated genes revealed 27 transcriptional regulators,
   such as the LambdaCh01/AraC/XRE families, a lactate/cellobiose
   phosphotransferase system (PTS), and 69 putative membrane proteins
   upregulated under autotrophic growth conditions ([152]Fig. 4C). Genes
   involved in oxidative phosphorylation, methane metabolism, and
   riboflavin metabolism were significantly enriched (Bonferroni-corrected
   P value of <0.0009) in the low-TE group found under autotrophic growth
   conditions.

FIG 4.

   [153]FIG 4
   [154]Open in a new tab

   Global translational dynamics of A. woodii between fructose and
   H[2]+CO[2] conditions. (A) Distribution of differences in translation
   efficiency (TE) for all proteins (autotrophic/heterotrophic growth
   conditions). Most of the phage-related genes showed a high TE under the
   autotrophic growth conditions (top). The black dotted line indicates
   the log[2] fold change threshold (μ = 0.1; σ = 1.4 [log[2] scale]). The
   red dotted line indicates the Gaussian fit curve. See [155]Table S2 in
   the supplemental material for all TE values. (B) Fold changes (log[2]
   scale) in mRNA abundances and TEs (autotrophic/heterotrophic growth
   conditions), presenting a negative correlation. (C) Enriched GO/KEGG
   pathways of the translationally regulated genes (Bonferroni-corrected P
   value of <0.05). (D) mRNA fold changes and translation efficiency fold
   changes of all genes (autotrophic/heterotrophic growth conditions):
   hydrogen-dependent carbon dioxide reductase (HDCR) (Awo_c08190 to
   Awo_c08260), the methyl branch gene cluster (Awo_c09260 to Awo_c09310),
   the carbonyl branch gene cluster (Awo_c10670 to Awo_c10760), the
   bifurcating hydrogenase gene cluster (Awo_c26970 to Awo_c27010), the
   Rnf complex gene cluster (Awo_c22010 to Awo_c22060), F[1]F[o] ATP
   synthase (Awo_c02140 to Awo_c02240), phage-related genes, transmembrane
   genes, and genes encoding putative transcriptional regulators. We
   analyzed the translational dynamics for phage-related genes, where two
   large phage-related gene clusters (Awo_c30850 to Awo_c31230 and
   Awo_c34690 to Awo_c35270) were inspected. See [156]Table S2 for all TE
   values. To analyze translational regulation for obligate members of
   transmembrane proteins, we created a list of transmembrane proteins
   with predicted or known topologies based on data in the UniProt 2017_12
   database. Moreover, we obtained a list of regulators that were
   annotated as “transcriptional regulators” from the UniProt 2017_12
   database. Significance was assessed by a two-tailed Wilcoxon
   signed-rank test (ns, not significant; *, P < 0.05; **, P < 0.01; ****,
   P < 0.0001).

   Notably, transcripts from the WL pathway and energy conservation genes
   were increased during growth on H[2]+CO[2], while the TE was decreased
   ([157]Fig. 4D). The enriched low-TE genes with the GO term oxidative
   phosphorylation were associated with the F[1]F[o] ATP synthase operon
   and the carbonyl branch of the WL pathway. However, although the WL
   pathway, Rnf complex, and F[1]F[o] ATP synthase all had a decreased TE
   under autotrophic growth conditions (median difference of less than
   −4.11 log[2]; P < 0.002 by a Wilcoxon signed-rank test), the
   electron-bifurcating hydrogenase revealed upregulated expression
   patterns at the translational (fold change of >1.45; P[adj] < 1.96^−6)
   as well as transcriptional levels, resulting in insignificant TE
   changes. In contrast, genes encoding the transmembrane proteins and the
   transcriptional regulators increased in TE under autotrophic
   conditions, while the mRNA level decreased. Therefore, these results
   provide direct evidence that A. woodii regulates the gene expression of
   specific cellular functions at the translational level, including
   acetogenesis-related genes under autotrophic conditions.

Translation initiation plays a crucial role in protein synthesis under
autotrophic growth conditions.

   The wide range (20-fold) of TEs across the genes raises the question of
   how translation is regulated under autotrophic growth conditions.
   Although the understanding of translational regulation is limited in A.
   woodii, we hypothesized that the secondary structure of the 5′ UTR
   could control translation efficiency. In general, the 5′ UTR is a
   regulatory component that allows the ribosomal machinery to bind to
   mRNA and initiate translation. This translation initiation step could
   be a major determinant in TE, as it is the rate-limiting step for the
   translation process ([158]33). To examine the effect of the 5′ UTR on
   TE, we first determined the transcription architecture of the A. woodii
   genome using the differential RNA sequencing (dRNA-Seq) method
   ([159]34) ([160]Fig. 5A, [161]Fig. S5A to F, and [162]Table S2). A
   total of 581 TSS positions were analyzed based on the ratio of 5′-end
   read densities (>2-fold) between the RNA 5′-pyrophosphatase-treated
   (RPP^+) and -nontreated (RPP^−) libraries. Motif searches upstream of
   sequences of the TSSs resulted in the TSSs containing purine-rich TSSs
   along with an extended −10 box motif (TATAAT) and the −35 box motif
   (TTGACA). Based on these, we assigned 483 5′ UTRs for 455 genes,
   including all of the acetogenesis-associated genes ([163]Fig. S5G and
   H). For example, the HDCR operon can be transcribed starting at fdhF2,
   as concluded from the TSS determination ([164]Fig. 5A and [165]B).
   Several TSS positions were independently validated by using 5′ rapid
   amplification of cDNA ends (RACE) and Sanger sequencing ([166]Fig. 5B
   and [167]Fig. S5C and D). In A. woodii, the median 5′-UTR length was 49
   nucleotides (nt) ([168]Fig. 5C), and this length distribution is
   similar to the 5′-UTR length distribution of the psychrotolerant
   acetogen Acetobacterium bakii (median length of 46 nt; P > 0.05 by a
   Wilcoxon rank sum test) ([169]9). However, the 5′ UTR of genes related
   to the WL pathway of A. woodii (median length of 38 nt) is
   significantly shorter than that of A. bakii (median length of 97 nt;
   P < 0.03by a Wilcoxon rank sum test), showing that the expression of
   genes related to the acetogenesis of A. woodii is not regulated by the
   5′ UTR in response to low temperatures, unlike A. bakii ([170]9).

FIG 5.

   FIG 5
   [171]Open in a new tab

   Analysis of translational regulation with the 5′ UTR. (A) Example of a
   determined TSS of the HDCR operon and RNA-Seq profiles and dRNA-Seq
   profiles mapped onto the A. woodii genome. RNA
   5′-pyrophosphatase-treated (RPP^+) and -untreated (RPP^−) libraries are
   represented, and the red dotted line indicates the TSS position. See
   [172]Table S1 in the supplemental material for sequencing statistics of
   dRNA-Seq. The total TSSs are listed in [173]Table S2. (B) TSS
   validation of the HDCR operon using 5′ rapid amplification of cDNA ends
   (RACE) experiments. The red triangle indicates the expected product
   length. See [174]Fig. S5 for TSS validation of the WL pathway, Rnf
   complex operon, the bifurcating hydrogenase operon, and F[1]F[o] ATP
   synthase operon. (C) Comparison of the 5′-UTR length distributions
   between A. bakii ([175]9) and A. woodii. Violin plots show medians and
   quartile ranges of 5′-UTR lengths. The black line represents the
   statistical difference in the 5′-UTR lengths of total genes, whereas
   the blue dotted line indicates the statistical difference in the 5′-UTR
   lengths of genes related to the WL pathway according to the Wilcoxon
   rank sum test (ns, not significant; *, P < 0.05). (D) Ribosome
   occupancy in 5′ UTRs. (Left) Compared to the coding sequence (CDS), the
   RPF/RNA ratio was calculated in 5′ UTRs under heterotrophic and
   autotrophic growth conditions. (Right) Autotrophic growth conditions
   may cause accumulation of ribosomes at the 5′ UTR of mRNA. (E)
   Autotrophic growth leads to ribosome stalling at the initiation codon
   of the cooC1 gene. The position of the A-site of the stalled ribosome
   obtained from the Ribo-Seq data is also indicated. Our RPF data also
   represent cooC1 gene translation that commences upstream in frame with
   the canonical AUG. (F) TE change distributions across minimum folding
   free energies (MFEs) of the 5′ UTR. Structures of a representative set
   of the 5′ UTRs were predicted using ViennaRNA RNAfold ([176]48).
   Significance was assessed by a two-tailed Wilcoxon signed-rank test
   (ns, not significant; *, P < 0.05). (G) Conserved ribosome-binding
   sequence for the total CDS, genes with high TE changes, phage-related
   genes (most of the phage-related genes showed high TEs), and genes with
   low TE changes at the upstream region of the initiation codon. The high
   TE change and phage-related gene sets revealed more obvious canonical
   AG-rich Shine-Dalgarno motifs than the low TE change gene set.
   FIG S5

   Transcription architecture of the A. woodii genome. (A) Categorization
   of 581 TSSs determined in this study. Primary transcripts were
   sequenced and mapped to the A. woodii genome, yielding at least
   104-fold genomic coverage. A total of 581 TSS positions were analyzed
   based on the ratio of 5′-end read densities (>2-fold) between the RNA
   5′-pyrophosphatase-treated (RPP^+) and -untreated (RPP^−) libraries.
   The total TSSs are listed in [177]Table S2 in the supplemental
   material. The criteria for the categorization of TSSs are detailed in
   Materials and Methods. (B) Examples of TSS determination in A. woodii
   under heterotrophic (H) and autotrophic (A) conditions. (C) TSS
   validation of acetogenesis-related genes using a 5′ rapid amplification
   of cDNA ends (RACE) experiment. The red triangle indicates the expected
   length of the PCR product. (D) Sequences of 5′-RACE products were
   confirmed using Sanger sequencing. Red arrows indicate the primers used
   for amplification. (E) Motif searches upstream of the A. woodii TSSs
   for all TSSs. The TSSs contain purine-rich transcription initiation
   sites along with an extended −10 box motif (TATAAT) and the −35 box
   motif (TTGACA). (F) Proportions of each nucleotide around the TSS (+1).
   (G) Distribution of 5′-UTR lengths in the primary transcripts and
   conserved ribosome-binding-site (RBS) sequences for 455
   protein-encoding transcripts. (H) Promoter sequences of
   acetogenesis-related gene clusters. Download [178]FIG S5, PDF file, 1.5
   MB^ (1.5MB, pdf) .

   Copyright © 2021 Shin et al.

   This content is distributed under the terms of the [179]Creative
   Commons Attribution 4.0 International license.

   Next, we observed ribosome occupancy in 5′ UTRs. Compared to the open
   reading frame (ORF) region, the median RPF/RNA ratio was 0.48 for the
   5′ UTRs under heterotrophic growth conditions ([180]Fig. 5D),
   indicating that the ribosome density in the ORF is twice that of the 5′
   UTRs. Interestingly, we found a significantly increased RPF/RNA ratio
   (median, 1.38) in the 5′ UTR in comparison to the ORF region under
   autotrophic growth conditions (>2.8-fold; P < 0.0001 by a Wilcoxon
   matched-pairs signed-rank test). This unique pattern is closely related
   to ribosome pausing, which in certain circumstances triggers
   translational abandonment via the process of trans-translation
   ([181]35). For example, the 5′ UTR of the cooC1 gene (Awo_c10670)
   presented increased ribosome occupancy around the initiation codon
   under autotrophic growth conditions ([182]Fig. 5E). Thus, autotrophic
   growth conditions may cause the accumulation of ribosomes at the 5′ UTR
   of the gene, suggesting an impediment of translation initiation. The
   accumulation of ribosomes at the 5′ UTR is also observed in several
   Escherichia coli genes under carbon limitation ([183]36) and in
   Saccharomyces cerevisiae under oxidative stress ([184]37) and amino
   acid starvation ([185]15).

   Next, we investigated the correlation between the TE and the 5′-UTR
   sequences, indicating that highly structured 5′ UTRs (less than the
   lower median ΔG value of −30 kcal mol^−1) reduce the TE ([186]15,
   [187]16) of the downstream genes (P = 0.029 by a Wilcoxon rank sum
   test) under autotrophic growth conditions ([188]Fig. 5F), as is the
   case with E. limosum ([189]10). The Shine-Dalgarno (SD) sequence is of
   prime importance in identifying the translation initiation site on
   bacterial mRNA ([190]38), and a canonical SD motif was observed in the
   phage and high-TE genes ([191]Fig. 5G). These results show that a
   highly conserved SD sequence is more important to enhance TE under
   autotrophic growth conditions than under heterotrophic growth
   conditions, as seen previously with C. ljungdahlii ([192]14).
   Collectively, translational initiation is more important for protein
   synthesis under autotrophic growth conditions than under heterotrophic
   growth conditions. The divergent regulations of translation initiation
   are due in part to the mRNA 5′-UTR structure and SD sequence.

DISCUSSION

   Here, we provide a systematic analysis of gene expression in A. woodii
   at both the transcriptional and translational levels under
   heterotrophic and autotrophic conditions. The data from the
   transcriptomic analysis are in line with previous results in A. bakii
   ([193]9) showing that acetogenesis-related genes, as well as
   gluconeogenesis/pentose phosphate pathway genes, were activated for
   biomass formation and CO[2] reduction during autotrophic growth. In
   addition, the gene expression of glycolysis and an incomplete TCA cycle
   were not activated under autotrophic conditions. Furthermore, our
   translatome data enabled us to understand how translation is
   efficiently regulated under energy-limited conditions. At a genome
   scale, the translation level of subunits of enzyme complexes matched
   the determined stoichiometries under autotrophic growth conditions. The
   translation levels were primarily dependent on the mRNA level under
   both heterotrophic and autotrophic conditions; however, most of the
   genes changed their TEs up to 20-fold between the two growth
   conditions, indicating efficient translational regulation during growth
   on H[2]+CO[2] ([194]Fig. 4A). Given that the translational capacity of
   a cell is reallocated in the face of limited resources, protein
   synthesis rates of the essential enzymes for autotrophic growth should
   be optimized to achieve a maximal growth rate ([195]16). Under
   autotrophic conditions, the TE of ribosome biogenesis decreased,
   resulting in a greater restriction of the available ribosome protein
   than of the mRNA. Thus, A. woodii was assumed to be carrying out
   protein synthesis economically, thus focusing on the balanced
   translation of the acetogenesis-related proteins for the most efficient
   metabolic reactions. Specifically, these data revealed that the
   translation levels of the methyl branch and energy conservation enzymes
   were effectively controlled in similar amounts under autotrophic growth
   conditions. Interestingly, the carbonyl branch enzymes revealed
   relatively variable translation levels (from 1- to 5-fold), which is
   consistent with data from the C. ljungdahlii proteome study under
   syngas conditions ([196]27). In particular, MET/CoFeSP and ACS/CODH
   show higher translation levels than other WL pathway-related enzymes.
   This probably compensates for the rate-limiting catalytic step
   ([197]39) by the large conformational changes during the enzymatic
   reactions ([198]24). Although the TE of the carbonyl branch was
   significantly downregulated, it still presented high protein fractions
   (∼7.8%) of the total protein content ([199]27, [200]40) under
   autotrophic growth conditions. Presumably, this represents a strategy
   for allocating limited protein synthesis capacity, as is the case with
   C. ljungdahlii ([201]14, [202]41, [203]42). However, further proteomic
   analyses will be needed to quantify the abundance of these carbonyl
   branch enzymes at the protein level in A. woodii.

   Acetogenesis is a modular metabolism in which the oxidative part is
   combined with the reductive part by appropriate electron donors such as
   H[2], NADH, and reduced ferredoxin ([204]43). Based on the translatome
   data, we hypothesized that carbon can also flow from the oxidative part
   (glycolysis) to the reductive part (WL pathway) by the GCS in A.
   woodii. The GCS can be regarded as an anaplerotic metabolic pathway
   that fills up the methyl branch of the WL pathway. Under these
   conditions, the Rnf complex is less important for redox homeostasis,
   and thus, less ATP is formed. This is compensated for by higher ATP
   formation in the central pathways. A beneficial effect of the GCS is
   also evident from the observation that E. limosum containing the GCS
   grew faster and produced more acetate than the parental strain. Thus,
   our results suggest that the GCS contributes to CO[2] fixation and
   redox balancing with energy conservation under heterotrophic growth
   conditions. Under autotrophic growth conditions, it was also reported
   that CO[2] can be converted into acetyl-CoA and acetyl-phosphate via
   the functional cooperation of the WL pathway, GCS, and glycine
   reductase in C. drakei and E. limosum ([205]28). However, further
   analyses are required to validate the function of the GCS in A. woodii
   proposed from our data set.

   The RPF read density observed in the 5′ UTRs indicates the presence of
   the attached 70S ribosome and that ribosome assembly may occur upstream
   of the SD sequence. The 70S ribosome requires additional sliding to
   place the ribosome above the start codon. In particular, the high RPF
   read density at −10 to ∼15 nt from the start codon indicates that the
   ribosomes are impediments due to the binding of the ribosome to the SD
   sequence (see [206]Fig. S1H in the supplemental material). Although the
   regularity is low, the average RPF read density at the 5′ UTR shows
   3-nt periodicity, indicating the stepwise movement of the ribosome. As
   described in E. coli during nutrient limitation ([207]30), the number
   of active ribosomes may be reduced through inhibition of IF2-mediated
   translation initiation by ppGpp, which increases during nutrient
   limitation ([208]44). The increased ribosome density in the 5′ UTR
   during autotrophic growth indicates that the ribosome is inactivated
   within the noncoding region of the mRNA. Such ribosome inactivation can
   be beneficial to cells in an environment of insufficient energy, as for
   the regulation of specific metabolic pathways by reducing the overall
   translation capacity. Our 5′-UTR analysis suggests that translation
   initiation is more important for protein synthesis during autotrophic
   growth than during growth on fructose, and it is regulated by the
   5′-UTR structure ([209]10) and SD sequence ([210]14). Importantly,
   autotroph-responsive TSS selection of the acsD gene was identified with
   validated TSSs, resulting in alternative 5′-UTR mRNAs between the
   fructose and the H[2]+CO[2] conditions. Although the H[2]+CO[2]
   conditions form short 5′ UTRs (31 nt) in acsD, the fructose conditions
   form long 5′ UTRs (121 nt), resulting in a complex secondary structure
   for the 5′ UTR ([211]Fig. S5B).

   Collectively, the systemic approach to determine both the transcriptome
   and translatome enabled us to unravel the TE, which calculates the
   actual rate of mRNA translated into proteins, and to identify
   GCS-coupled heterotrophic acetogenesis as well as translational control
   of genes encoding the GCS, the WL pathway, and energy conservation. In
   particular, understanding the regulation of acetogenesis and collecting
   diverse bioparts derived from multi-omics data are prerequisites for
   the biotechnological application of acetogens. It may be feasible to
   improve the translation levels of key enzymes for strain engineering in
   order to optimize the metabolic process of syngas fermentation or to
   produce native or nonnative products with better productivity.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

   Acetobacterium woodii DSM 1030 and Eubacterium limosum ATCC 8486 were
   obtained from the Leibniz Institute DSMZ (German Collection of
   Microorganisms and Cell Cultures). A. woodii and E. limosum were
   cultivated strictly anaerobically at 30°C, in DSMZ medium 135 ([212]45)
   supplemented with 2 g liter^−1 NaCl. Fructose (5 g liter^−1) or a gas
   mixture of H[2]+CO[2] (80:20; 200 kPa) was used for the heterotrophic
   and autotrophic growth conditions, respectively. Escherichia coli DH5α
   was obtained from Enzynomics, Inc. (Daejeon, South Korea), and used for
   plasmid cloning and maintenance.

Analytical methods.

   Growth was monitored by measuring the optical density at 600 nm
   (OD[600]). To remove cell debris, 500-μl samples were centrifuged at
   14,000 × g for 10 min at 4°C. All metabolite samples were prepared by
   filtration through 0.2-μm Minisart RC15 syringe filters (Sartorius,
   Goettingen, Germany). Metabolic products were analyzed using a Waters
   1525 high-performance liquid chromatography (HPLC) system (Waters,
   Milford, MA, USA) equipped with a MetaCarb 87H organic acid column
   (Agilent Technologies, Wilmington, DE, USA) and a refractive index (RI)
   detector (Waters), operating at 37°C. H[2]SO[4] (0.01 N) was used as
   the mobile phase. The flow rate of the mobile phase was 0.6 ml min^−1,
   and the volume of each injection was 20 μl.

RNA extraction and RNA sequencing library preparation.

   For RNA extraction, cell cultures of heterotrophically and
   autotrophically grown cells were harvested at OD[600]s of 0.8 and 0.15,
   respectively. Cell cultures were harvested by centrifugation at
   4,000 × g for 10 min at 4°C and resuspended in 500 μl of lysis buffer
   (20 mM Tris-HCl [pH 7.4], 140 mM NaCl, 5 mM MgCl[2], and 1% Triton
   X-100). Resuspended cells were flash-frozen in liquid nitrogen and then
   ground using a mortar and pestle. Unbroken cells and cell debris were
   removed by centrifugation at 4,000 × g for 10 min at 4°C. Purified
   total RNA was then obtained using TRIzol reagent (Thermo Fisher
   Scientific, Waltham, MA, USA) according to the manufacturer’s
   instructions. To remove the remaining genomic DNA, rDNase I (Ambion,
   Thermo Fisher Scientific) was added to the purified RNA according to
   the manufacturer’s instructions. Total RNA was measured using the Qubit
   RNA HS assay kit. RNA quality was confirmed by the A[260]/A[280] ratio
   (>1.8) and by visualization of the two distinct bands of rRNAs by 2%
   agarose gel electrophoresis. rRNAs were specifically removed by the
   Ribo-Zero rRNA removal kit (bacteria) (Illumina, Inc., San Diego, CA,
   USA) according to the manufacturer’s instructions. Strand-specific RNA
   sequencing (RNA-Seq) libraries were constructed using the rRNA-depleted
   RNAs and a TruSeq strand mRNA LT sample prep kit (Illumina, Inc.) and
   then sequenced in an Illumina MiSeq instrument with a 151-bp single-end
   sequencing recipe.

Differential RNA-Seq library preparation.

   Differential RNA-Seq (dRNA-Seq) libraries were prepared as described
   previously ([213]34), with some modifications ([214]9). Briefly, 400 ng
   of the rRNA-depleted RNA was divided into two samples for constructing
   two different libraries (RPP^+ and RPP^−). For preparing the RPP^+
   libraries, the sample containing primary (5′-PPP) and processed
   (5′-P/5′-OH) transcripts was treated with RNA 5′-polyphosphatase (RPP;
   Epicentre Biotechnologies, Madison, WI, USA). For the preparation of
   RPP^− libraries, RPP was omitted from the enrichment step. 5′-RNA
   adaptors (see [215]Table S3 in the supplemental material) were ligated
   with T4 RNA ligase (Thermo Fisher Scientific) to the RPP^+ and RPP^−
   samples with a 1:3 molar ratio of RNA to 5′-RNA adaptor. The ligated
   products were then purified with Agencourt AMPure XP beads (Beckman
   Coulter, Brea, CA, USA). For cDNA synthesis, random nonamer adaptors
   ([216]Table S3) and a SuperScript III reverse transcriptase kit
   (Invitrogen, Carlsbad, CA, USA) were used, according to the
   manufacturer’s directions. The cDNA libraries were purified with 0.8×
   volumes of AMPure XP beads and then amplified with the indexed primer
   ([217]Table S3) and Phusion high-fidelity DNA polymerase (Thermo Fisher
   Scientific). Amplification was monitored with SYBR green gel stain
   solution (Invitrogen) on a CFX96 real-time PCR detection system
   (Bio-Rad Laboratories, Hercules, CA, USA). The PCR conditions were as
   follows: 98°C for 30 s and several cycles (until the reaction reached
   the plateau phase) of 98°C for 10 s, 56°C for 20 s, and 72°C for 20 s,
   followed by 72°C for 5 min. The dRNA-Seq libraries were purified with
   0.8× volumes of AMPure XP beads and then sequenced using a 100-bp read
   recipe on an Illumina HiSeq2500 instrument (Illumina, Inc.). Two
   biological replicates were sequenced, and the required oligonucleotide
   sequences are listed in [218]Table S3.
   TABLE S3

   Oligonucleotides used in this study. Download [219]Table S3, PDF file,
   0.1 MB^ (118KB, pdf) .

   Copyright © 2021 Shin et al.

   This content is distributed under the terms of the [220]Creative
   Commons Attribution 4.0 International license.

Ribosome profiling library preparation.

   Bacterial cells were pretreated at 30°C for 10 min with 100 μM
   chloramphenicol (CM) prior to harvesting by centrifugation at 4,000 × g
   for 10 min at 4°C. The collected cells were washed with 500 μl polysome
   buffer (20 mM Tris-HCl [pH 7.4], 140 mM NaCl, 5 mM MgCl[2], and 100 μM
   CM) and resuspended in ice-cold lysis buffer (20 mM Tris-HCl [pH 7.4],
   140 mM NaCl, 5 mM MgCl[2], 100 μM CM, and 1% Triton X-100). The
   resuspended cells were flash-frozen in liquid nitrogen. Subsequently,
   the frozen samples were ground by a mortar and pestle. The cell lysates
   were recovered by centrifugation at 16,000 × g for 10 min at 4°C, and
   the supernatant was then transferred to a new tube. To degrade RNA that
   was not protected by the ribosome, the supernatant was subjected to
   micrococcal nuclease (MNase; New England BioLabs, Ipswich, MA, USA)
   digestion using 400 U of MNase. Digestion proceeded for 2 h at 37°C
   with gentle rotation, and the reaction was stopped by the addition of
   10 μl of EGTA (0.5 M). The enriched monosome fraction was confirmed by
   visualization of the ribosome-protected mRNA footprints, with sizes
   ranging from 26 to 34 nt, in a 15% polyacrylamide Tris-borate-EDTA
   (TBE)-urea gel (Invitrogen). The monosome fraction was specifically
   recovered using a Microspin S-400 HR instrument (GE Healthcare Life
   Sciences, Marlborough, MA, USA) according to the manufacturer’s
   instructions. The recovered ribosome-bound RNA was isolated using the
   TRIzol reagent (Thermo Fisher Scientific), and the remaining rRNAs were
   removed by the Ribo-Zero rRNA removal kit (bacteria) (Illumina, Inc.)
   according to the manufacturer’s instructions. The footprints were
   denatured at 80°C for 90 s and then equilibrated to 37°C. Subsequently,
   footprint samples were incubated at 37°C for 1 h with a solution
   containing 10 U of T4 polynucleotide kinase (PNK; New England BioLabs),
   5 μl of 10× T4 PNK buffer (New England Biolabs), 0.5 mM ATP, and 20 U
   of SUPERase In RNase inhibitor (Thermo Fisher Scientific).
   Phosphorylated footprint samples were purified using the RNeasy
   MinElute column (Qiagen, Valencia, CA, USA), and the RNA concentration
   was evaluated using the Qubit RNA HS assay kit (Invitrogen). Ribosomal
   profiling (Ribo-Seq) libraries were constructed using the NEBNext small
   RNA library prep kit for Illumina (New England BioLabs) according to
   the manufacturer’s instructions, and the constructed libraries were
   sequenced using the 50-bp read recipe on the Illumina HiSeq2500
   instrument.

Resting cell assay.

   For the resting cell assay, heterotrophically and autotrophically grown
   A. woodii cells were harvested at OD[600]s of 0.8 and 0.15,
   respectively. Cell suspensions were prepared as described previously
   ([221]21), with some modifications. All the following steps were
   carried out under strictly anoxic conditions (<5 ppm O[2]) in an
   anaerobic chamber (Coy Laboratory Products, Grass Lake, MI, USA) filled
   with 96% N[2] and 4% H[2]. The cell samples were washed twice using
   imidazole buffer (50 mM imidazole, 20 mM KCl, 20 mM MgSO[4], 4.4 μM
   resazurin, 4 mM 1,4-dithioerythritol [pH 7.0]) and resuspended in the
   same buffer. Some of the suspension fractions were aliquoted by
   centrifugation at 6,000 × g for 10 min at 4°C, and total protein
   contents were quantified as described above. For the cell suspension
   experiments, whole cells were resuspended to a protein concentration of
   1 mg ml^−1 in a 100-ml anoxic glass bottle. In the assay, 20 mM NaCl
   was added to the sample, adjusting the final volume to 20 ml with
   buffer, and the bottles were incubated at 30°C in a shaking incubator.
   The headspace pressure was adjusted to 200 kPa with a gas mixture of
   H[2]+CO[2] (80:20 [vol/vol]; 200 kPa) for autotrophic conditions, or 5
   g liter^−1 fructose was added under 200 kPa N[2] for heterotrophic
   conditions. Acetate was measured using a Waters 1525 HPLC system
   equipped with a MetaCarb 87H organic acid column (Agilent Technologies)
   as mentioned above. All values are means obtained from three
   independent experiments.

5′ rapid amplification of cDNA ends.

   The 5′ rapid amplification of cDNA ends (RACE) analysis was conducted
   as described previously ([222]9). The validated TSSs and their expected
   amplicons are shown in [223]Fig. S5. The amplicons were visualized
   using a 2% agarose gel. The oligonucleotides used for 5′-tag-cDNA
   library construction are available in [224]Table S3.

Quantitative RT-PCR assay.

   To valid RNA-Seq data independently, quantitative reverse
   transcription-PCR (RT-PCR) was performed using the RNA extracted from
   cultures grown under equal growth conditions. The purified RNA samples
   were converted to cDNA using the SuperScript III first-strand synthesis
   system (Invitrogen) according to the manufacturer’s directions. In
   total, seven genes related to the WL pathway for constitutive
   expression were selected to validate the RNA-Seq results. Relative
   expression was measured using a Kapa SYBR Fast universal quantitative
   PCR (qPCR) kit (Kapa Biosystems, Wilmington, MA, USA) according to the
   manufacturer’s instructions. All primer sequences were determined using
   NCBI Primer-BLAST and are listed in [225]Table S3.

Plasmid construction.

   The pJIR750ai plasmid was purchased from Sigma-Aldrich. The TargeTron
   gene knockout system was removed from pJIR750ai by digestion with PvuI,
   and the plasmid was then religated using T4 DNA ligase, resulting in a
   new pJIR plasmid. To construct the pJIR-GCS plasmid, a DNA fragment of
   the ptetO1 (pt1) promoter was amplified using primer pair tetO1_F and
   tetO1_R and then inserted into the BamHI and PvuI sites of the pJIR
   vector, resulting in pJIR-Pt1. The DNA fragment containing the tetR
   gene from the pdCas9 plasmid (Addgene) was amplified using primer pair
   ptet_tetR_F and ptet_tetR_R and then inserted into the SacI and BamHI
   sites in pJIR-Pt1, resulting in pJIR-R-Pt1. Following construction,
   gcvHTP genes from A. woodii genomic DNA were amplified using two primer
   pairs (GCS_IF_F and GCS_IF_R, and GCS_IF_NoP_F and GCS_IF_NoP_R) and
   then cloned into pJIR-R-Pt1 using the NcoI and PvuI sites and into pJIR
   using the PvuI site, resulting in the pJIR-GCS and pJIR-NoP vectors,
   respectively. These cloning steps proceeded using the In-Fusion HD
   cloning kit, and all high-fidelity PCR amplifications were carried out
   using Phusion high-fidelity DNA polymerase (Thermo Fisher Scientific).
   Purification of PCR products and plasmids from E. coli DH5α was
   performed with the QIAquick PCR purification kit (Qiagen) and the
   DNA-spin plasmid DNA purification kit (iNtRON Biotechnology, Daejeon,
   South Korea), respectively. All recombinant plasmids were transformed
   into E. coli DH5α competent cells (Enzynomics, Daejeon, South Korea)
   for cloning and maintenance. All oligonucleotides used for the
   construction of plasmids are listed in [226]Table S3.

Deep-sequencing data analysis.

   The sequence reads generated from RNA-Seq (150-bp single-end reads),
   dRNA-Seq (100-bp single-end reads), and Ribo-Seq (50-bp single-end
   reads) were demultiplexed and trimmed using bcl2fastq v1.8.4. Adaptor
   sequences and low-quality bases were additionally trimmed using CLC
   Genomics Workbench 6.5.1 (Qiagen). The trimmed reads were mapped to the
   reference genome (the genome sequence and annotation were obtained from
   GenBank accession number [227]CP002987 for A. woodii DSM 1030) using
   CLC Genomics Workbench with the following parameters: mismatch cost of
   2, insertion cost of 3, deletion cost of 3, length fraction of 0.9, and
   similarity fraction of 0.9. For Ribo-Seq data, rRNA and tRNA reads were
   removed by alignment to rRNA and tRNA regions. rRNA- and tRNA-free
   reads were aligned to the reference genome with the same parameters as
   the ones mentioned above.

   For RNA-Seq and Ribo-Seq, a gene-wise read count was generated from the
   trimmed reads using the “bedtools -multicov –bam input.bam –bed
   input.bed” command. Considering Ribo-Seq data, raw reads were counted
   for reads mapping to the ORFs, excluding the first and last 5 codons to
   eliminate the effects of initiation and termination ([228]15, [229]19).
   The expression values of the genes were normalized using Bioconductor
   package DEseq2 (18) and the gene-wise read count table as the input. To
   identify the differentially expressed genes, the significance levels
   were set at an adjusted P value (P[adj]) of <0.01 and a log[2] fold
   change (log[2] FC) of greater than |1|. The translation level (protein
   synthesis rate) was defined by the RPKM ([230]10, [231]15, [232]17,
   [233]19). All RPKM values were calculated as RPKM = (10^9 × number of
   raw reads mapped to the gene)/(total mapped reads × length of the
   gene). Translational efficiencies were calculated by dividing the RPF
   level by the estimated mRNA level in arbitrary units. Additionally, the
   5′ end of the RPF was used to estimate the information on ribosomal
   positions. To define the position of the ribosome A-site, Ribo-Seq was
   additionally performed as described in a previous report ([234]46). For
   this, each RPF was estimated by adjusting 17 nt from the 5′ end of the
   RPF reads ([235]15).

   dRNA-Seq data were analyzed as described in a previous report ([236]9).
   Briefly, we performed normalization to the 5′ ends of the aligned
   dRNA-Seq read profiles with a scaling factor, and the ratio between
   RPP-treated (RPP^+) and -untreated (RPP^−) reads was calculated for
   each genomic position. The TSSs were selected from RPP^+/RPP^− ratios
   of peak densities of >2. These TSSs were manually curated and
   categorized based on their genomic position. TSSs having the highest
   peak density within a distance <300 bp upstream and 100 bp downstream
   of annotated genes were categorized as primary TSSs (P), and the others
   were categorized as secondary TSSs (S). TSSs that were located on the
   same strand as an annotation or on its opposite strand were classified
   as internal TSSs (I) and antisense TSSs (A), respectively. If they were
   not classified into any of the categories mentioned above, they were
   classified as intergenic TSSs (N). The TSSs were compared with two
   conditional libraries and were summed to give the total TSSs within the
   range of ±4 bp. All TSSs are summarized in [237]Table S2. Multilayered
   RNA-Seq, dRNA-Seq, and Ribo-Seq mapped reads were used to construct
   genome-wide coverage maps, and all maps were visualized using SignalMap
   (Roche NimbleGen, Inc., Madison, WI, USA).

Motif discovery in the promoter.

   For motif analysis in promoters, we extracted 50 nt of DNA sequence
   upstream of each detected TSS position. The conserved −10 and −35 boxes
   were analyzed by MEME v4.11.1 ([238]47) (a MEME P value cutoff of 0.05
   was used).

Statistical testing.

   Data statistical testing (Pearson’s correlation coefficient, two-tailed
   Student’s t test, two-tailed Wilcoxon signed-rank test, and two-tailed
   Wilcoxon-Mann-Whitney test) was performed using GraphPad Prism v8
   software (GraphPad, San Diego, CA, USA). P values of <0.05 were
   considered statistically significant.

Data availability.

   All transcriptome and translatome raw data in FASTQ format are
   available in the European Nucleotide Archive under study accession
   number [239]PRJEB33460.

Supplementary Material

   Reviewer comments
   [240]reviewer-comments.pdf^ (986.7KB, pdf)

ACKNOWLEDGMENTS