Using previously published datasets (Cho et al., 2014), transcripts were screened for their expression in the 5′UTR regions of the adjacent coding regions to discover putative 5′UTR regions that could contribute to the gene regulation in Z. mobilis. Initially, 392 potential candidates were identified, leading to the need of filtering these 5′UTR candidates from the large number and diversity of other types of transcripts. For this purpose, a bioinformatics pipeline was developed to select candidates from the large number of initial candidates (Figure 1). Only candidates that showed comparable levels of expression with the adjacent genes were retained as potential UTRs. All transcripts <35 base pairs were filtered out, as this is the shortest known length of a UTR regulatory element (Roth et al., 2007). As well-characterized metabolic enzymes are highly regulated by their 5′ UTRs, these types of candidates were prioritized in the list. Collectively, this analysis resulted in a total of 101 potential candidates that were selected for experimental analysis (Table S1).

To further discern the properties of potential 5′UTR candidates, bioinformatics analyses were performed using Rfam (a database of multiple sequence alignments), consensus secondary structures, and covariance models representing RNA families (Griffiths-Jones et al., 2003). There are three RNA categories in Rfam: non-coding RNAs, structured cis-regulatory elements, and self-splicing RNAs (Nawrocki et al., 2015). Four candidates were matched with predicted riboswitches in Z. mobilis: TPP, crcB, and two cobalamin switches (Table 1). TPP and cobalamin riboswitches are widely conserved and have been demonstrated in E. coli and other bacteria, such as Bacillus subtilis, to control the regulation of downstream genes by direct binding to thiamine pyrophosphate and cobalamine, respectively (Nahvi et al., 2002; Winkler et al., 2002). The crcB RNA motif identified by Rfam is known as a fluoride riboswitch (Baker et al., 2012) adjacent to the 5′ end of the chloride channel protein gene (ZMO0547). This fluoride switch regulates gene expression based on its structural change in response to fluoride ions. Genes encoding this fluoride-specific type of chloride channel protein have been shown to be regulated by fluoride riboswitches in a variety of organisms (Stockbridge et al., 2012). Importantly, these results supported the presence of 5′UTRs before experimental confirmation and also validated the UTR bioinformatics prediction methods.

Given that structural conservation is closely associated with the regulatory roles of RNA (Yang et al., 2010), LocARNA was used to determine conservation based on both sequence and structure (Will et al., 2012). NCBI BLAST (Camacho et al., 2009) was used to identify sequence homology for each 5′UTR candidate, as required for LocARNA input. Of the initial 101 5′UTR candidates, 28 contained structurally conserved motifs, many of which showed complexity. Figure S1 shows representative data of this analysis.

Lastly, differential expression data under ethanol, acetate, and xylose stresses were compiled for the genes associated with the 5′UTR candidates (He et al., 2012a; Yang et al., 2013, 2014a,b). Genes with differential expression under one or more stresses may be regulated by a 5′UTR or other mechanism. As shown in Table 1 and Table S3, 17 mRNAs and 7 proteins corresponding to the 5′UTR candidates were up- or down- regulated under stress. Due to their association with stress response(s), these candidates were further investigated in this work.

The cellular expression of all 101 final candidates (Table S1) was further confirmed by RT-PCR analysis, for which two primer sets were designed (Figure 2A). Primer set A was designed for the amplification of a long transcript from hypothetical 5′UTR regions to the middle of the corresponding mRNA coding region. Primer set B was designed to amplify a relatively short transcript inside the adjacent mRNA coding region as a positive control representing the adjacent mRNA expression level (Figure 2A). As a negative control, reverse transcriptase was excluded from the reaction to confirm the absence of genomic DNA contamination. Representative data are illustrated in Figure 2B and all RT-PCR results are shown in Figure S2. PCR bands from both primer sets A and B proved the expression of transcript containing potential UTRs upstream. However, the presence of a band from primer set B and not from set A, indicated that the UTR was not detected upstream of the mRNA. From the experimental analysis, fifty 5′UTR candidates that showed contiguous expression in the transcript along with the mRNA, which are marked as confirmed candidates in Table S1. Lack of detectable expression of other candidates could be explained by unsuccessful PCRs, false positives in our candidate selection from the transcriptomic data, or different growth conditions between RT-PCR and previous transcriptomic studies.

Precise transcription start sites (TSS) for all candidates of interest were determined by 5′ RACE (Figure 2C). A summary of the 36 final candidates that were further investigated and their gene function is included in Table 1 and additional information in Table S2. It should be noted that 14 candidates were excluded from further analysis out of the 50 experimentally confirmed candidates due to their overlapping transcription with the adjacent gene (3′ end of the adjacent gene was connected with potential 5′ UTRs), length <30 bp, or lack of experimental validation by RT-PCR and/or 5′RACE (Figure S3).

To test each 5′UTR candidate's ability to regulate downstream gene expression, an in vivo fluorescence-based reporter gene screening system was developed, taking advantage of previous successful efficient expression of GFP in Z. mobilis (Douka et al., 2001). As shown in Figure 3A, inducible expression of GFP was confirmed under the P_tet promoter, and then the 60-nucleotide sequence of the theophylline riboswitch element was cloned in front of the GFP gene to establish the functionality of the GFP screen in Z. mobilis (Figure 3B). The fluorescence shifted when it is induced with 10 μg/mL tetracycline compared to that of the un-induced control sample (Figure 3C). After confirming the functionality of the fluorescence system in Z. mobilis, the well-characterized theophylline synthetic riboswitch was used as a test case to establish the screening system that was able to elicit a fluorescence change specific to the activation of a 5′UTR. The theophylline riboswitch has been engineered as a synthetic riboswitch system to control gene expression in various bacterial species and natively controls gene expression at the translational level by binding to the small molecule, theophylline (Suess et al., 2004; Lynch et al., 2007; Topp et al., 2010).

To establish the functionality of the GFP screen, the 60-nucleotide sequence of the theophylline riboswitch element was cloned in front of the gfp gene in Z. mobilis (Figure 3B) and the level of GFP expression with 2 mM theophylline was compared with a DMSO control in Z. mobilis. Control-GFP exhibited about 10-fold GFP expression when it was induced (Figures 3C,D). In contrast to control-GFP, a 2-fold increase in GFP expression from theophylline RNA stability element (RSE)-GFP construct was observed. Although the fold change in GFP was not high (limited to about 2-fold), this was enough to screen for positively activating candidates. It is also worth noting that this is not an unusual fold change of expression when using the same theophylline riboswitch in different bacteria species (Topp et al., 2010) without engineering pairing strength of the region between the aptamer and RBS (Ribosome Binding Site) in the theophylline switch within a particular species. Importantly, the activation of the theophylline switch was successfully demonstrated, as well as the use of this 5′UTR in vivo fluorescence-based screening system in Z. mobilis for the first time. This could be a useful tool for screening the control of gene expression in metabolic pathways related with ethanol tolerance or other stress responses in Z. mobilis.

Utilizing the Z. mobilis pEZ minimal shuttle vector system (Yang et al., 2016b), a high-throughput cloning strategy was developed for construction of each 5′UTR-containing GFP plasmid by the combination of Golden Gate assembly and PheS counter selection marker (Figure 3E). Given that nucleotides in the coding region may affect the structure of 5′UTRs for the regulation of the gene, each 5′UTR sequence (verified by 5′ RACE) was cloned with an additional 90 base pairs of the downstream coding region of the corresponding gene for the generation of 5′UTR-GFP libraries. Primers used for the generation of 5′UTR-GFP libraries are listed in Table S3. Using this plasmid construct, a library of 36 5′UTR-GFP candidates was generated (Table 1) for further testing. The sequences and features of the final 36 candidates were shown in Table S2.

Since high tolerance to ethanol is one of the desirable features of Z. mobilis, the initial screen of 5′UTR-GFP libraries was conducted under ethanol stress (5% (v/v) ethanol-supplemented media) to evaluate potential 5′UTR activation relative to a standard medium (RM) control. A strain with only GFP and no 5′UTR (Control-GFP) was used as a negative control. All experiments were done in biological triplicates. After the signal difference was observed to be highest at 10 h post-induction, the fluorescence measurements were taken at this time point for all other experiments. This is consistent with previous data that showed maximum GFP fluorescence in late exponential phase (Douka et al., 2001). Upon screening all candidates, two were identified that exhibited significant fluorescence changes under 5% ethanol supplementation compared to the control without added ethanol: the 5′UTR of ZMO0347 (RNA binding protein Hfq, UTR_ZMO0347) and the 5′UTR of ZMO1142 (thioredoxin reductase, UTR_ZMO1142) (Figure S4).

To explore responsiveness of these 5′UTRs under different levels of ethanol, 1, 3, or 5% (v/v) of ethanol were added to the cultures before fluorescence and western blotting analysis (Figure 4). In this plasmid system, GFP fluorescence uniformly decreased under any level of ethanol stress when regulated by the UTR_ZMO0347 (RNA binding protein Hfq), whereas the fluorescence under the control of UTR_ZMO1142 (thioredoxin reductase) showed an ethanol concentration-dependent decrease (Figure 4). The changes in GFP expression in cells were also confirmed by western blotting analysis, which corresponded well with fluorescence data (Figure 4). All 5′UTR candidates were also tested for responses to xylose and acetate stresses in this fluorescence screening system (Figure S5). Interestingly, UTR_ZMO1142 did not appear to regulate gene expression under xylose and acetate stresses, but only under ethanol stress (Figure S5). Consistent with these results, it has been reported that thioredoxin reductase (ZMO1142) is less abundant under 6% ethanol stress at the protein level, although not at the transcript level, indicating a layer of translational regulation (Yang et al., 2013). Additionally, for Hfq (ZMO0347), a protein that is highly associated with stress response in various organisms, a decrease of transcripts has been previously detected under 6% ethanol stress (Guisbert et al., 2007; Yang et al., 2013; Torres-Quesada et al., 2014). However, inconclusive data about changes in Hfq protein levels are found throughout the literature (and our own studies) given the difficulty in detecting this protein by mass spectrometry. Importantly, the responses of UTR_ZMO0347 and UTR_ZMO1142 to ethanol indicate that UTR regulation could be part of a mechanism of ethanol stress protection in Z. mobilis.

In addition to ethanol stress responsive 5′UTRs, a few acetate and xylose responsive elements were identified (Figure S5). A previous study confirmed that regulatory mechanisms responding to acetate stress and xylose utilization mainly included differential expression of carbon and energy metabolism genes to reduce the impact of stress on the cell (Yang et al., 2014a). Acetate stress repressed genes associated with flagellar system and glycolysis, but induced genes related to stress responses and energy metabolism (Yang et al., 2014b). 6-phsphogluconolactonase catalyzes hydrolysis of the ester linkage of lactone, resulting in production of 6-phophogluconate in the pentose phosphate pathway. Up-regulation of ZMO1478 (6-phosphogluconolactonase) transcript in acetate stress has been previously reported (Yang et al., 2014a), and in this study, UTR_ZMO1478 (6-phosphogluconolactonase) was shown to activate GFP expression under acetate stress (10 g/L sodium acetate). Taken together, these results suggest that 5′UTR could be responsible for the natural increase of ZMO1478 expression under acetate stress. Additionally, UTR_ZMO0172 (thiamine biosynthesis protein) showed down-regulation of GFP expression under both acetate and xylose stresses, indicating that this gene could be involved in general stress response mechanisms. Further studies on protein expression levels of ZMO0172 under stressful conditions could help reveal the underlying regulatory mechanism of UTR_ZMO0172.

To evaluate the physiological effects of ethanol stress-responsive regulatory 5′UTRs in Z. mobilis, a mutant strain was constructed with a genomic deletion of UTR_ZMO0347. Unfortunately, UTR_ZMO1142 could not be deleted due to the limitation of the technique used in this study for short length deletion (e.g., 57 bp in this study for UTR_ZMO1142). The UTR_ZMO0347 deletion was constructed utilizing a homologous recombination deletion technique to disrupt the 5′UTR region (except for promoter and RBS regions) of the hfq gene in Z. mobilis 8b strain. The spectinomycin resistance gene was used as a selection marker between 1 kb up- and down-stream homology arms. The deletion of this region was confirmed by PCR and Sanger sequencing (data not shown). Primers used for the construction and confirmation of deletion constructs are shown in Table S3. Growth data for the Z. mobilis parental strain 8b and its hfq 5′UTR deletion mutant strain were collected (Figure S6). A growth defect in the ΔUTR_ZMO0347 strain was observed relative to the wild type strain under no ethanol stress.

Transcript levels of the ZMO0347 (Hfq) mRNA in the deletion strain were measured via qRT-PCR with and without 5% (v/v) ethanol supplementation (Figure 5A). Although the deletion strain shows an overall decrease in transcription, a consistent up-regulation of the hfq transcript is observed in both the wild type and deletion mutant strains in the presence of ethanol. This suggests that the ethanol-responsive regulatory effect of the 5′UTR is not at the transcriptional level (e.g., due to promoter effects).

At the protein level, native Hfq levels were also evaluated in the ΔUTR_ZMO0347 mutant strain relative to its wild-type ZM4 strain via western blotting analysis (in chromosomally FLAG-tagged Hfq strains for both the wild type ZM4 and the ΔUTR_ZMO0347 strain). As shown in Figures 5B,C for the wild type and deletion strains respectively, the Hfq protein showed a small (less than a 2-fold increase) change in response to ethanol. In the wild type strain, increases in Hfq protein expression were observed starting with 5% (v/v) ethanol supplementation (but not with 1% ethanol supplementation). In contrast, in the ΔUTR_ZMO0347 deletion strain, a more sensitive response was observed upon supplementation with ethanol, in which the increase in Hfq protein was observed with only 1% (v/v) ethanol supplementation (Figure 5). This suggests that the native 5′UTR region serves as a post-transcriptional regulator, akin to a resistor, that keeps Hfq protein levels down-regulated in response to lower ethanol stress. Given the fact that the isolation of 5′UTRs in our GFP assays (in the absence of their native promoters and genetic context) indicated down-regulation of protein expression by the 5′UTR, the overall increase of hfq transcript and protein levels in the wild type strain with ethanol suggests that additional regulatory effects play a role in overall Hfq regulation. As discussed below, this is not surprising given that, in E. coli and other organisms, Hfq is globally regulated by highly complex schemes that involve its native promoter and global protein-sRNA regulators (Vecerek et al., 2005; Sobrero and Valverde, 2012; Ramos et al., 2014; Sowa et al., 2017).

Z. mobilis 8b was provided by NREL (Zhang et al., 1995). Cells were cultured in 5 mL RMG (glucose, 20.0 g/L; yeast extract, 10.0 g/L; KH₂PO₄, 2.0 g/L; pH 6.0) (Yang et al., 2009) overnight at 33°C and then inoculated into 100 mL fresh medium. Initial OD_600nm was ~0.05 and cells were induced with 10 μg/mL of tetracycline when OD_600nm reached ~0.4. 1, 3, 5, or 8% (v/v) ethanol was supplemented into RMG for ethanol stress experiments from the beginning of growth. 10 g/L of sodium acetate (NaAc) (Yang et al., 2014b) was supplemented into RMG for acetate stress. In the case of xylose stress, 10 g/L glucose + 10 g/L xylose was used instead of 20 g/L glucose in the media. Cells were collected 4 or 12 h post-induction for measuring GFP expression.

Upstream and downstream fragments (each about 1 kb) homologous to the target deletion gene were assembled with the spectinomycin gene aadA in the middle. The upstream homologous arm was PCR amplified using primers Fup and Rup, and the downstream homologous arm amplified by Fhfq-1, Fhfq-2, and Rhfq. Primers Fsp, Rsp-1, and Rsp-2 were used to introduce a terminator at the end of the spectinomycin resistance gene and primers Fhfq-1 and Fhfq-2 were used to introduce the 3xFLAG. The fragments were then assembled similarly to Gibson assembly. Briefly, primers were designed with 15-nt overhangs of desired upstream or downstream sequences and used to amplify the pieces for the assembly and purified from the primer dimers by gel electrophoresis followed by gel purification. Fragments and the pUC57 vector were mixed in a molar ratio of 3:1 (about 120 ng total DNA) and added to 0.5 U T5 exonuclease (NEB, USA), 0.5 μL Buffer 4 (NEB, USA), and nuclease-free water (up to 5 μL). The reaction mix was kept on ice for 5 min and then transformed into E. coli competent cells by chemical transformation. Cells were plated on LB agar plates with 200 μg/mL of spectinomycin. Transformants were screened by colony PCR and confirmed by Sanger sequencing (Sangon, China). The purified plasmid was electroporated into the Z. mobilis. Transformants appearing on RM agar plates with 200 μg/mL of spectinomycin were cultured and screened by colony PCR. Colonies with correct PCR product sizes were selected as deletion candidates after Sanger sequencing confirmation.

RACE experiments were performed on total RNA samples using FirstChoice® RLM-RACE kit (Ambion, CA) according to the manufacturer′s protocol and to previously published work (Cho et al., 2014). Briefly, 8 μg RNA was treated with Tobacco Acid Pyrophospatase (TAP) at 37°C for 1 h, followed by ligation of the 5′ RACE kit adapter at 37°C for 1 h. The resulting RNA was then reverse transcribed according to the manufacturer′s protocol and PCR was performed on the resulting cDNA. All primer sequences used for RACE are listed in Table S3. Resulting PCR products were purified using QIAquick PCR purification kit (Qiagen, MD) and RNase-free water (Ambion, CA) for final elution. Final PCR products were sequenced and results were compared with the genome. 5′ RACE adapter sequences (5′-GCUGAUGGCGAUGAAUGAACACUGCGUUUGCUGGCUUUGAUGAAA-3′) and adjacent mRNA sequences were used for the detection of TSS in the 5′UTRs.

The tetracycline-inducible promoter with GFP construct (pEZ-tet-GFP) was derived from the minimal shuttle vector pEZ15Asp of Z. mobilis and contains the replication origin of E. coli and Z. mobilis (Yang et al., 2016b). A parental plasmid containing pheS counter-selection marker (Kast, 1994; Miyazaki, 2015) was constructed with pheS counter-selection marker incorporated in front of the gfp gene flanked by BsmBI sites (Type IIS enzyme), which is one of the Type IIS restriction endonuclease to enable Golden Gate cloning for efficient cloning of each 5′UTR-containing GFP (Engler et al., 2008; Figure 3E). With this design, 5′UTRs along with the first 90-bps of the downstream mRNA were cloned for each candidate right in front of the gfp gene in frame. This design was implemented to preserve the native structure of the UTR by better mimicking its native context while still allowing GFP expression. Primers used for the amplification of UTR+90-bps are listed in Table S3. Each primer contains a BsmBI enzyme site on the 5′ end.

Cells were analyzed by flow cytometry using the FACSCalibur™ (BD Biosciences, CA) according to standard methods as described in a previous study (Miyazaki, 2015). Briefly, cells were pelleted and resuspended in phosphate buffered saline (PBS: 137 mM sodium chloride, 2.7 mM potassium chloride, and 10 mM phosphate buffer, pH 7.5) to a concentration on the order of 10⁷ cells/mL. The cells were excited with the 488 nm argon laser and the cell population was determined from the forward scatter and side scatter distributions reported by the cytometer. Data were collected for at least 50,000 active cells, ensuring enough events to assume that the population distribution would be unaffected by rare events. Sample data were analyzed using CellQuest Pro (BD Biosciences) with a user-defined gate. Averages of median values for each sample were calculated from biological triplicates and error bars were calculated as standard error of the mean.

RT-PCR was used to experimentally confirm the presence of 5′UTRs from the candidate list (Table S1). Total RNA was prepared using Trizol reagent (Invitrogen, CA) and resulting RNA was treated with DNase I (RNase free, ThermoScientific, MA) to prevent of genomic DNA contamination as described in the manufacturer′s protocol. Subsequently, RNA was precipitated with isopropanol and then washed with ethanol. Total RNA rehydrated with nuclease-free water (Ambion, CA) was used as the template for reverse transcription. Two hundred nanograms of RNA was incubated for the first strand synthesis with 100 ng of random hexamer and 10 mM dNTPs at 65°C for 5 min. According to the manufacturer′s protocol, SuperScript™ III Reverse Transcriptase (Invitrogen, CA) was added to RNA-primer mix with RNaseOUT™ RNase Inhibitor, 0.1 M DTT, 5x First-strand buffer and then incubated for 5 min at room temperature. The final reaction mixture was incubated at 55°C for 1 h and then heat inactivated at 70°C for 15 min.

The cDNA product from first-strand synthesis was used as a template in a 20 μL PCR reaction containing Power SYBR® Green PCR master mix (Invitrogen, CA). Specific primers were used for each target. For example, primers F (5′- GCGTCTTGTTGACCCGTAAT−3′) and R (5′- AATCCTCGTCTCGCCTTTCT−3′) were used for the amplification of the hfq gene. Three primers were designed for 5′UTR RT-PCR: the first forward primer (with reverse primer, set A) was located in the middle of the 5′UTRs and the second forward primer (with reverse primer, set B) was designed to bind in the front part of following mRNA regions. The reverse primer was designed to bind in the middle of the mRNA regions. All primers used in this study are listed in Table S3. Phusion® High-Fidelity DNA Polymerase (NEB, MA) was used for PCR amplification. No reverse transcriptase was used for the negative control to exclude potential genomic DNA contamination. Primer set B was used for the positive control as it represents amplification of the mRNA coding region. The temperature cycle used for the PCR reactions is as follows: 95°C for 10 min, 40 cycles of 95°C for 15 s and 60°C for 60 s. Relative quantification of hfq transcript between each condition was performed using Viia 7 Software (Life Technologies, CA) following the comparative delta-delta threshold cycle (ΔΔC_T) method. Samples were collected in biological triplicates.

Western blotting analysis was performed to detect GFP expression using anti-GFP antibody (Roche 11814460001). Standard western blotting protocols were used as previously described (Cho et al., 2016). Briefly, total cellular lysates were loaded onto a 12% denaturing SDS-PAGE gel. Gels were transferred to methanol activated PVDF membranes using the Trans-Blot® Semi-Dry Electrophoretic Transfer Cell (Bio-Rad, CA) and run for 40 min at 15 V or 20 min at 25 V. Membranes were blocked with 5% dry non-fat milk in Tris-buffered saline (TBS, pH 7.5) for 1 h at room temperature. The proteins were detected with anti-GFP antibody at 1:1000 dilutions or anti-FLAG at 1:5000 (Proteintech, China). As a secondary antibody, goat anti-mouse IgG (H + L) HRP Conjugate (Promega #W4021) was used at a dilution of 1:2500 (or 1:5000 with Proteintech, China). All images were developed using Clarity™ ECL Western Blotting Substrate (BioRad, #170-5060) and the ChemiDoc^TM MP Imaging System (BioRad, CA) or West Dure Extended Duration Substrate Kit (AntGene, China) with AI600 Imaging System (GE, USA). Bradford assay measurements were used to normalize the loading of all protein analyses by total protein mass. Specific proteins were detected on the membrane by western blot analysis and quantified using ImageQuant TL 8.1 (GE Healthcare, CA) or Gel-Pro analyzer (LiuYi, China). Each protein was detected using anti-GFP. The level of GFP expression was measured and then normalized using the expression of RecA as an internal control.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.