Engineered xylose-utilizing Z. mobilis 2032 (ATCC31821-5C Pgap_taltkt/Pgap xylAB, PTA-6977) was obtained from American Type Culture Collection (ATCC). The rich media with glucose (ZRMG, containing 10 g yeast extract/L, 2 g KH₂PO₄/L, 20 g glucose/L) or rich media with both glucose and xylose (RMGX, containing 10 g yeast extract/L, 2 g KH₂PO₄/L, 100 g glucose/L, 20 g xylose/L) were used for initial cultivation of 2032. To increase buffer capacity and maintain pH, extra phosphate (0.8 g KH₂PO₄/L and 2 g K₂HPO₄/L) was added into RMGX, which was named as RMGXP.

Two synthetic hydrolyzate media used in these studies (SynH3^– and SynH3) were based on a previously described synthetic hydrolyzate medium (SynHv2.1) that mimics 6% glucan-loading AFEX-corn stover hydrolyzate (6% ACSH) (Serate et al., 2015). To mimic 9% ACSH, SynH3^– (Table 1) (aka SynHv3.8) was made by increasing the media composition of SynHv2.1 by 1.5-fold with the following additional modifications. To prevent Fe loss due to filtration of FeCl₃ precipitate, FeCl₃ was replaced with Fe-citrate, 1.5 mM sodium citrate was added, and 90 g/L of cellobiose was added to match the osmolarity of 9% ACSH (∼1920 mmol/kg). Z. mobilis 2032 is unable to metabolize cellobiose meaning that the osmolarity of SynH3^– will remain high throughout our fermentation experiments. We hypothesize that the lower osmolarity of SynH3^–, prior to the addition of 90 g cellobiose/L, may be due to 9% ACSH components that are missing in our SynH recipe, likely stemming from an inability to identify or measure these compounds. This was an important modification as our previous work with E. coli highlighted osmotic stress as a key aspect of the stress response of E. coli to SynH and ACSH (Schwalbach et al., 2012; Keating et al., 2014). SynH3 (aka SynHv3.8) was generated by adding 25 LDIs (Table 2) to SynH3^– at LDI concentrations largely based on the composition analysis of 9% ACSH. All chemicals used for SynH, except two LDIs, were purchased from Fisher Scientific (Hampton, NH, United States) and Sigma-Aldrich (St. Louis, MO, United States). Feruloyl amide and coumaroyl amide were synthesized by in-house as described previously (Keating et al., 2014).

Multiomic fermentations were conducted in 3 L bioreactors (Applikon Biotechnology) containing 1.2 L of SynH3^– or SynH3 media. Starter culture were first grown in ZRMG media with 12.5 μg/L each of Tetracycline (Tc) and Chloramphenicol (Cm) under aerobic conditions at 30°C. After ∼7 h, the cultures were diluted into RMGXP medium containing 12.5 μg/ml each of Tc and Cm at an initial OD₆₀₀ (optical density at 600 nm) of 0.2, and then grown in an anaerobic chamber on a stir plate at 30°C for about 14 h. Cells from this culture were then centrifuged at 18000 × g for 3 min, cell pellets were resuspended in appropriate SynH media and inoculated into the bioreactors to give an initial OD₆₀₀ of 0.5. Fermentations were conducted at 30°C with continuous stirring (300 rpm), headspace sparging with 100% N₂ (20 mL/min) and pH controlled at 5.8. Samples were periodically removed from the bioreactors for an OD₆₀₀ measurement using a Beckman Coulter DU720 (Beckman Coulter, Inc., Brea, CA, United State) to monitor cell growth and for YSI 2700 Biochemistry Analyzer (YSI, Inc., Yellow Springs, OH, United States) to monitor the concentration of glucose and xylose.

During the fermentation, samples were collected at four growth stages: mid-glucose (about 50% glucose utilized), late-glucose (about 5–10 g glucose/L remaining), early-xylose (about 1 h after glucose was completely depleted) and mid-xylose (about 24 h after glucose was completely depleted) for transcriptomic, metabolomic, and proteomic analysis.

For RNA isolation and transcriptomic analysis, 10 ml of culture was collected in 15 ml conical tubes containing 1.25 ml ice-cold unbuffered phenol and ethanol mixture (5:95, vol/vol) and processed as described previously (Schwalbach et al., 2012).

For metabolomic sampling, 3–5 ml of culture was taken out from the bioreactor using 5 ml syringes. After quick capped the syringe, samples were transfer into an anaerobic chamber for subsequent processing to avoid the exposure of oxygen and the disturbance of metabolites. Samples were then loaded on a pre-wet 0.45-micron Nylon membrane filter (Whatman, Cat# 7404-002, 25 mm) under vacuum filtration (Hoefer, Holliston, MA, United States Cat# FH225V) to quickly separate the cells from the liquid medium. The membrane with cells was submerged into 3 ml ice-cold Extracting Solution containing acetonitrile:methanol:water (50:25:25) mixture and quickly vortexed. Samples were then removed from the anaerobic chamber and centrifuged at 18000 × g for 3 min at 4°C. One ml of supernatant was transferred in 1.5 ml Eppendorf tube, and quickly frozen in dry ice/ethanol bath and stored at −80°C.

For proteomic sampling, 2 ml of culture was collected and transferred into chilled 2 ml Microcentrifuge tubes. Cells were pelleted by centrifugation at 18000 × g at 4°C for 3 min, supernatant was removed, and cell pellets were quickly frozen in dry ice/ethanol bath for 5 min, and then stored at −80°C.

For dry cell weight (DCW) measurement, 25 ml of cells were collected, and after an OD₆₀₀ measurement, 10–20 ml cells (total OD₆₀₀ is about 30–40) were then loaded on a pre-wet 0.22 μM S-PaK membrane filter (Millipore, Cat# GSWG047S6, 47 mm) on the Nalgene filtration assembly (Nalgene, now Thermo Fisher Scientific, for 47 mm filter). The filter and cells were then washed with 5 ml Nanopure water and cells were dried in a desiccant chamber for 3–5 days and then weighed. The S-PaK membrane filters before the filtration were also pre-dried in desiccant chamber and weighed. The DCW values in milligrams per mL per OD₆₀₀ of culture was calculated by the subtraction of the filter weight after filtration to that before filtration, and then normalized by volume of cells collected and its OD₆₀₀.

RNA was isolated from cells that were captured in phenol and ethanol solution as described previously (Rhodius and Gross, 2011). RNA samples were DNase treated and submitted to the University of Wisconsin Biotechnology Center (UWBC) Gene Expression Center for rRNA subtraction by Illumina RiboZero Bacteria Kit and paired-end library generation using the Illumina TruSeq Stranded Total RNA Library kit. Libraries were sequenced at 2 bp × 125 bp on an Illumina HiSeq2500 using v4 chemistry at the UWBC DNA Sequencing Facility. Reads were filtered for low quality and adapter readthrough using trimmomatic version 0.30 (Bolger et al., 2014) using the following parameters: ILLUMINACLIP:TruSeq3-PE.fa:2:22:10 SLIDINGWINDOW:4:28 MINLEN:75. Reads were aligned to Z. mobilis 2032 GenBank chromosome and plasmid sequences CP023677, CP023678, CP023679, CP023680, and CP023681 (Yang et al., 2018) using RSEM version 1.2.3 (Li and Dewey, 2011) and Bowtie version 1.0.0 (Langmead et al., 2009) using the following parameters: –paired-end –calc-ci –estimate-rspd –forward-prob 0 –phred33-quals.

RSEM expected counts were used for downstream differential expression analysis. Pearson correlation between vectors of counts that belong to biological replicates was employed to detect outlier libraries. All the retained libraries had inter-replicate Pearson correlation of at least 0.95. Features representing rRNA and tRNA, and genes with count sums less 5000 across all remaining samples were removed from the matrix. Gene count normalization and differential expression testing was performed using DESeq2 version 1.14.1 (Love et al., 2014) run on R version 3.3.0. Biclustering analysis of gene counts was performed using backSPIN run on python 2.7 (Zeisel et al., 2015).

Several different methods were used for measuring intracellular metabolites: AEX-LC-MS and IP-LC-MS were used for most intracellular metabolites, HILIC-MS was used for extracellular amino acids and GC-MS was used for intracellular sugar compounds.

RNA-Seq data are available through the National Biotechnology Center Gene Expression Omnibus under accession number GSE135718. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (Perez-Riverol et al., 2019) partner repository with the dataset identifier PXD015007.

As described in Section “Materials and Methods,” we developed SynH3^– to mimic of highly concentrated 9% ACSH. Twenty-five LDIs were added to SynH3^– to create SynH3 (Tables 1, 2). This created a system in which we could directly compare the impacts of LDIs on the physiological responses and fermentative capabilities of Z. mobilis 2032. To this end we performed batch fermentations of Z. mobilis 2032 in both SynH3^– (i.e., without LDIs) and SynH3 (i.e., with LDIs) in biological triplicates. Cell density (by light scattering), glucose, xylose, and ethanol production were monitored for 85 h (Figure 1). Because Z. mobilis 2032 is unable to co-metabolize both glucose and xylose, our fermentations resulted in a characteristic biphasic growth pattern in which glucose was metabolized first followed by xylose once glucose was nearly depleted. Extracellular ethanol reached a maximal concentration of approximately 50 g/L (5%) in both SynH3 and SynH3^– once glucose was depleted. During glucose fermentation Z. mobilis 2032 showed similar growth, glucose uptake, and ethanol production in both SynH3 and SynH3^– and was comparable to fermentation of Z. mobilis 2032 in 9% ACSH (Yang et al., 2018). One significant difference was that a growth lag was observed in 9% ACSH, whereas no lag was seen in either SynH media (Yang et al., 2018). In contrast, although more than 80% of xylose was consumed in SynH3^– after 85 h, only about 50% of xylose was utilized in SynH3. This result is consistent with previous observations in which xylose utilization was notably reduced when Z. mobilis 2032 was grown in 9% ACSH (Yang et al., 2018). These results indicated that growth in SynH3 was able to recapitulate the LDI-induced inhibition of xylose utilization that was observed in 9% ACSH.

These results indicated that LDIs do not appear to cause cell death, since glucose consumption and cell growth were not significantly different between SynH3^– and SynH3 fermentations. It is also possible that additive or synergistic effects of ethanol with LDIs on xylose utilization occur in our experiments. Previous studies investigating the effects of ethanol toxicity on Z. mobilis have observed negative impacts on cell growth and glucose consumption in Z. mobilis ZM4 in the presence of 5% ethanol in rich medium (He et al., 2012a). Our SynH3^– and SynH3 fermentations reached approximately 5% ethanol after glucose depletion; however, in the absence of LDIs, 5% ethanol does not inhibit xylose metabolism.

To characterize the cellular responses of Z. mobilis 2032 to SynH3^– and the LDIs present in SynH3, we collected samples for metabolomic, proteomic, and transcriptomic analyses at four different timepoints during our batch fermentations: mid-glucose (T1, about 50% glucose utilized), late glucose (T2, about 5–10 g glucose/L remaining), early xylose (T3, ∼1 h after glucose was completely depleted), and late xylose (T4, ∼24 h after glucose was completely depleted). Using multiomics sampling, we were able to profile the relative expression levels of 1699 genes by RNA-Seq (Supplementary Table S1), 1587 proteins by LC-MS/MS proteomics (Supplementary Table S2), and using absolute levels, 63 intracellular and extracellular metabolites, including 18 intracellular and extracellular amino acids (Figure 2 and Supplementary Table S3).

RNA-Seq transcriptomics identified numerous differentially expressed genes between SynH3 and SynH3^– during the course of the fermentations. Overall, the number differentially expressed genes increased with subsequent timepoints and almost all genes in Z. mobilis 2032 were found to be differentially expressed at atleast one timepoint (Figure 3A). In contrast, our proteomics data identified far fewer proteins as differentially expressed, with T2 (about 5–10 g glucose/L remaining) having the highest number of differential protein levels (Figure 3B). Principal component analysis (PCA) of transcriptomic and proteomic data revealed that sample timepoint accounted for the greatest variance across samples, with biological replicates forming tight and distinct clusters in the transcriptomic data (Figure 4). Bi-clustering of transcriptomic data was consistent with PCA and produced distinct clustering by timepoint, with biological replicates clustering together into distinct clades at each time point (Figure 5). A total of seven gene clusters were identified, with each cluster corresponding to the timepoint and condition at which the highest expression of those genes was observed across all timepoints. Gene ontologies (GO) and KEGG pathways were used for gene-set-enrichment analysis (GSEA) of each cluster via hypergeometric tests and Holm multiple hypothesis testing correction (Supplementary Table S4). GSEA showed that transcription in both SynH3^– and SynH3 at T1 was enriched for growth genes such as ribosomal proteins and aminoacyl-tRNA biosynthesis, which is consistent with effects on glucose metabolism and logarithmic growth at this timepoint. At T1, SynH3 samples (gene cluster 2) also showed enrichment for sulfur, purine, seleno-compound metabolism, and aminoacyl-tRNA biosynthesis KEGG pathways. At T4, SynH3^– samples (gene cluster 7) were highly enriched for glycolysis–gluconeogenesis KEGG pathway and to lesser extent cell redox homeostasis and proteolysis GO terms. Our GSEA broadly highlights various impacts of LDIs at early glucose stages of fermentation and potential key differences in cellular activity between SynH3 and SynH3^– later when xylose was the primary carbon source.

At T1 approximately 100 genes were identified by RNA-Seq as differentially expressed in SynH3 relative to SynH3^– (at an FDR < 0.02; Supplementary Table S1). T1 differentially expressed genes are highly enriched for upregulation of KEGG sulfur metabolism pathway (KO00920). Most of the top 30 upregulated genes at T1 fall into one of four gene clusters: ZMO0003 to ZMO0009 (cysCNDG1HIJ), ZMO0282-0285 (efflux transporter), ZMO1260-1265 (ssuACB and others), and ZMO1459-1463 (sseA and others) (Supplementary Table S5).

ZMO0003 to ZMO0009 contains genes encoding cysteine biosynthetic proteins (sulfate assimilation). In E. coli, upregulation of cys genes is induced by sulfur limitation (van der Ploeg et al., 2001). Several other upregulated genes also indicate a disruption of sulfur homeostasis by LDIs in SynH3 including ZMO1261-1263 (ssuACB), ZMO1460 (sulfurtransferase), ZMO0055 (sulfite exporter) and ZMO0748 (cysteine synthase) (Supplementary Table S5). Upregulation of cys genes was also observed in E. coli grown in SynH containing LDIs (Keating et al., 2014). Miller et al. (2009) also found that the expression of genes and regulators associated with the biosynthesis of cysteine and methionine was increased by furfural treatment in E. coli. Most of the T1 upregulated genes involved in the biosynthesis of cysteine (sulfate assimilation) were only upregulated at this timepoint and returned to levels equivalent to non-LDI challenged cells at subsequent timepoints. These results suggest that disruption of sulfur homeostasis is an early and substantial effect of LDIs.

There are at least four predicted efflux pump systems annotated in Z. mobilis 2032. The ZMO0282-0285 RND efflux transporter was upregulated at T1 and, together with ZMO0281, was also upregulated at T2, T3, and T4 (Figure 6A). ZMO0282, ZMO0283, and ZMO0285 protein levels were likewise elevated in SynH3 relative to SynH3^– at all four timepoints (Figure 6B). The RND efflux system at ZMO1429-1432 was also upregulated at T3 and T4 when xylose was the primary carbon source (Figure 6B). Both ZMO0282-0285 and ZMO1429-1432 encode RND family efflux transporter systems. In E. coli the AcrAB-TolC RND efflux pump is responsible for efflux of multiple antibiotics, dyes, bile salts, and detergents (Pradel and Pagès, 2002; Seeger et al., 2008; Li and Nikaido, 2009; Nikaido and Takatsuka, 2009). ZMO0281 encodes a TetR family transcriptional repressor of RND efflux transport system, which we hypothesize regulates ZMO0282-0285. As shown in Figure 6A, the expression of ZMO0281 was elevated from stage T2 to T4, and the high expression of ZMO0281 is correlated to the decreased expression of ZMO0282-0285 genes at T4. The transcriptional regulator of ZMO1429-1432 is unknown.

In addition to upregulation of cellular efflux, LDIs also appear to impact expression of transporter and secretion systems in Z. mobilis. RNA-Seq identified the putative multidrug and toxic compound extrusion (MATE) efflux family protein (ZMO0214) as highly upregulated at T4 (Figure 6A). Additionally, five major facilitator superfamily (MFS) transporters (ZMO0099, ZMO0566, ZMO1111, ZMO1457, and ZMO2018) are either up or downregulated at T4. MFS transporters are single-polypeptide secondary carriers capable only of transporting small solutes in response to chemiosmotic ion gradients (Pao et al., 1998). Interestingly, transcriptomic data showed that all genes of the predicted Type I secretion system at ZMO0252-0255 were downregulated at T1 (Figure 7A). Proteomic data likewise showed the corresponding gene products to be highly downregulated at all four timepoints (Figure 7B). ZMO0253, ZMO0254, and ZMO0255 are proposed to encode a HylD family Type I secretion system and ZMO0252 encodes a major intrinsic protein.

In our metabolomics data we observed an accumulation of intracellular mannose-6-phoshpate and mannose-1-phosphate over the course of our fermentations in both SynH3 and SynH3^–. We note that our LC-MS assay is unable to distinguish between mannose-1P and glucose-1P, but given the concomitant accumulation of mannose-P, we reason that the change occurred primarily in mannose-1P, not glucose-1P. By T4, mannose-6P and mannose-1P reached 10.1 and 0.8 μmol per mg dry cell weight, respectively, in SynH3, which was > 4-fold higher than in SynH3^– at T4 (Figure 2, Supplementary Figure S1 and Supplementary Table S3). As there is no mannose-6P isomerase predicted in the Z. mobilis 2032 genome, mannose sugar phosphate accumulation likely occurred via uptake of D-mannose from the media since mannose is present in both SynH3 and SynH3^– at 1.8 mM. Phosphorylation of D-mannose may be catalyzed by two predicted nagE genes, ZMO0037 and ZMO1336 and further to D-mannose-1P by phosphomannomutase (ZMO0339). Interestingly, we observed a 2.7-fold upregulation of mannose-1-phosphate guanylyltransferase (ZMO1233) in SynH3 at T4 which may be contributing to higher accumulation of mannose-1P and mannose-6P relative to SynH3^–.

The role of glutathione in detoxification of drugs, xenobiotics, and oxidants is well-documented (Forman et al., 2009). Metabolomics data revealed a striking depletion of reduced glutathione (GSH) specifically in SynH3 between T1 and later timepoints. In SynH3^– reduced glutathione was only depleted approximately 1.6-fold from T1 to T4, whereas in SynH3 reduced glutathione was depleted almost 7-fold (Figure 8A and Supplementary Table S3). At all four timepoints reduced glutathione was lower in SynH3 compared to SynH3^–. Oxidized glutathione levels may have also been impacted by LDIs, however we could not accurately measure oxidized glutathione in SynH3 as levels were below the range of our standard curve. Glutathione is synthesized from glutamate, cysteine, and glycine in two steps catalyzed by GshA (ZMO1556) and GshB (ZMO1913). Both gshA and gshB expression were downregulated at T3 in SynH3 moderately, but statistically significantly (Figure 8B and Supplementary Table S1). Intracellular levels of glutamic acid were reduced > 4-fold at T4 relative to T1 in both SynH3^– and SynH3 while intracellular glycine was relatively unchanged during the course of our fermentations with a moderate increase in SynH3 at T3. We were unable to measure cysteine with our metabolomics assays.

Many genes were differentially expressed between SynH3 and SynH3^– at T3 and T4 (431 and 817, respectively; Figure 3A and Supplementary Table S1). The sheer number of differentially expressed genes, many of which are annotated as hypothetical proteins, makes it difficult to fully delineate the impact of LDIs during these xylose-metabolizing timepoints. However, several stress response genes were upregulated including groES (ZMO1928), groEL (ZMO1929), ZMO0199 (SOS-response transcriptional repressor, LexA), organic solvent tolerance protein (ZMO1311), and ZMO1996 (transcription termination factor Rho, which is known to respond to oxidative stress) (Supplementary Table S1). Downregulated genes at T4 included the ZMO0263-0270 gene cluster of hypothetical proteins and 44 plasmid-encoded genes (∼30% of total plasmid-encoded genes), many of which are predicted to encode transcriptional regulators.

The LDIs present in SynH3 have a consistent and profound negative effect on xylose metabolism by Z. mobilis 2032. We used our multiomics data to investigate how LDIs impact xylose metabolism in cells grown in SynH3. We compared the expression of all genes involved in the conversion of glucose and xylose to ethanol (Figure 9), including xylose-specific genes that are heterologously expressed in Z. mobilis 2032 (Zhang et al., 2007; Yang et al., 2018). Expression of glucose facilitated diffusion protein glf, which is involved in the uptake of both glucose and xylose, was also compared. With the exception of alcohol dehydrogenase adhA, there was no statistically significant difference in the expression of these genes or proteins in the presence of LDI at any timepoint (Supplementary Table S6). The expression of adhA was slightly down-regulated at T3 and T4. However, expression of two other alcohol dehydrogenases in Z. mobilis 2032, adhB and adhC, did not change, which could have compensated for reduced adhA expression. Thus, it is unlikely that slight down-regulation of adhA would cause the poor xylose utilization. These results indicated that the poor xylose utilization in SynH3 was not due to changes in the expression of sugar assimilation pathway genes.

Although the expression of glucose and xylose metabolism genes was not affected by LDIs, the enzymes in these pathways may themselves be inhibited. We used our metabolomics to look for the accumulation of intermediate metabolites, which could indicate an enzymatic block in metabolism pathways. All metabolites in glucose and xylose assimilation pathways, except 6-phosphogluconate, 1,3-phosphoglycerate, phosphoenolpyruvate, acetaldehyde, and ethanol, were quantitated (Figure 9, Supplementary Figures S1, S2, and Supplementary Table S3). Many metabolites, including of 2-keto-3-deoxy-6-phosphogluconate, 2-phosphoglycerate, 3-phosphoglycerate, 6-phosphogluconate, fructose-6phophase, fructose, pyruvate, ATP, GTP, and glucose decreased from T1 to T4 in both SynH3 and SynH3^–. However, there were no significant differences in the levels of these metabolites in SynH3 compared to SynH3^–. Lower levels of these metabolites during xylose metabolism (T3 and T4) can be attributed to glucose depletion, slow xylose uptake, and reduced carbon flux. Levels of pentose phosphate pathway metabolites xylulose-5-phosphate, ribulose-5-phosphate, ribose-5-phosphate, sedoheptulose-7-phophate, and shikimate were elevated at T2, T3, or both, indicating flux from xylose metabolism, but ultimately these metabolites decreased by T4. Importantly, the levels of these metabolites were similar or slightly reduced in SynH3 compared to SynH3^–. The levels of xylulose, xylonic acid, cis-aconitate, xylitol, and GMP increased in both SynH3^– and SynH3 from T1 to T4, but their levels were similar or higher in SynH3^– than SynH3 at T4 (Supplementary Figure S1). Greater consumption of xylose in SynH3^– likely contributes to the higher levels of these metabolites compared to SynH3. The production of xylonic acid and xylitol was an important contributor to reduced efficiency of xylose conversion to ethanol in the presence of LDIs.