Heterologous expression of a glycosyl hydrolase and cellular reprogramming enable Zymomonas mobilis growth on cellobiose

Plant-derived fuels and chemicals from renewable biomass have significant potential to replace reliance on petroleum and improve global carbon balance. However, plant biomass contains significant fractions of oligosaccharides that are not usable natively by many industrial microorganisms, including Escherichia coli, Saccharomyces cerevisiae, and Zymomonas mobilis. Even after chemical or enzymatic hydrolysis, some carbohydrate remains as non-metabolizable oligosaccharides (e.g., cellobiose or longer cellulose-derived oligomers), thus reducing the efficiency of conversion to useful products. To begin to address this problem for Z. mobilis, we engineered a strain (Z. mobilis GH3) that expresses a glycosyl hydrolase (GH) with β-glucosidase activity from a related α-proteobacterial species, Caulobacter crescentus, and subjected it to an adaptation in cellobiose medium. Growth on cellobiose was achieved after a prolonged lag phase in cellobiose medium that induced changes in gene expression and cell composition, including increased expression and extracellular release of GH. These changes were reversible upon growth in glucose-containing medium, meaning they did not result from genetic mutation but could be retained upon transfer of cells to fresh cellobiose medium. After adaptation to cellobiose, our GH-expressing strain was able to convert about 50% of cellobiose to glucose within 24 h and use it for growth and ethanol production. Alternatively, pre-growth of Z. mobilis GH3 in sucrose medium enabled immediate growth on cellobiose. Proteomic analysis of cellobiose- and sucrose-adapted strains revealed upregulation of secretion-, transport-, and outer membrane-related proteins, which may aid release or surface display of GHs, entry of cellobiose into the periplasm, or both. Our two key findings are that Z. mobilis can be reprogrammed to grow on cellobiose as a sole carbon source and that this reprogramming is related to a natural response of Z. mobilis to sucrose that promotes sucrase production.

Introduction Advances in synthetic biology and lignocellulosic hydrolysate production have encouraged development of the α-proteobacterium Zymomonas mobilis as a platform microbe for production of renewable biofuels and chemicals (e.g., ethanol, C 4 and C 5 alcohols, or C 5 -C 15 terpenoids) from lignocellulosic biomass [1][2][3]. Efficient conversion of lignocellulose to biofuels and bioproducts is essential for the development of sustainable sources of fuels and chemicals that minimize competition with food production [4]. Z. mobilis is promising for lignocellulosic conversions because it rapidly and efficiently converts glucose to ethanol, tolerates high ethanol concentrations, tolerates other inhibitors present in the lignocellulosic hydrolysates, and has a small (2.1 Mb) and increasingly well-defined genome amenable to synthetic biology approaches [5]. Z. mobilis uses the Entner-Doudoroff pathway for glycolysis, which reduces the amount of protein synthesis required to convert glucose to pyruvate relative to Embden-Meyerhof-Parnas glycolysis used by yeast and many other microbes [6]. Additionally, fermentation to ethanol is not tightly linked to cell growth, enabling exceptionally high flux that continues even when cell growth stops.
Despite its potential as a platform microbe for lignocellulosic biofuel production, wild-type Z. mobilis has a limited substrate range consisting of glucose, fructose, and sucrose [7], and thus requires engineering to convert the diverse sugar monomers and oligomers present in lignocellulosic hydrolysates. Engineered strains that convert the 5-carbon sugars xylose [8] and arabinose [9] or cellobiose [10] to ethanol have been developed, but use of sugar oligomers other than its native substrate sucrose at levels that support Z. mobilis cell growth has not been reported. Sugar oligomers including cello-and xylo-oligomers can be significant components of lignocellulosic hydrolysates, representing about 18-25% of total soluble sugars in corn stover hydrolysates, especially when hydrolysates are generated under milder conditions that minimize inhibitor production or enable recovery of intact lignin [11]. Residual glycosyl hydrolases (GHs) in enzymatically prepared hydrolysates may enable conversion of these oligomers to monomers as conversion reduces end-product inhibition of GHs, but there is growing interest in enzyme-free chemical deconstruction methods like those enabled by the renewable solvent γ-valerolactone [12]. These enzyme-free hydrolysates will lack the residual GHs necessary for continued oligomer hydrolysis during fermentation. Thus, equipping Z. mobilis with an ability to convert sugar oligomers efficiently by production and extracellular release of GHs would improve prospects for its use as a platform microbe for lignocellulosic conversions.
As cellobiose is the most fundamental unit of these unusable sugar oligomers, we targeted production and release of GHs that would enable efficient cellobiose conversion by Z. mobilis by heterologous expression of Cellvibrio japonicus Cel3A and Caulobacter crescentus CC_0968, both belonging to glycosyl hydrolase family-3 (GH3). We selected Cel3A because it has already been shown to enable E. coli to grow on cellobiose in minimal medium [13]. CC_0968 was selected because C. crescentus is an α-proteobacterium like Z. mobilis. Past studies have successfully expressed eglX endoglucanase from Pseudomonas fluorescens var. cellulosa (later reclassified as Cellvibrio japonicus sp. nov) [14], celZ endoglucanase from Erwinia chrysanthemi [15], and endoglucanase from Cellulomonas uda CB4 [16] in Z. mobilis. Two cellulolytic enzymes from Acidothermus cellulolyticus, E1 and GH12, were expressed in Z. mobilis and their activities verified by a zymogram test with carboxymethylcellulose [17]. Similarly, β-glucosidase from Ruminococcus albus was tagged N-terminally with a signal peptide from the Z. mobilis periplasmic enzyme glucose-fructose oxidoreductase and overexpressed in Z. mobilis, which enabled fermentation of cellobiose to ethanol [10]. However, these previous studies only demonstrated cellobiose conversion in resting cells, and heterologous gene expression in Z. mobilis was not reported to produce significant growth in cellular biomass using cellobiose as a carbon source. Growth on oligosaccharides is crucial to enable genetic dissection of the Z. mobilis systems required for GH secretion and the use of selective pressure to improve GH production and secretion by Z. mobilis.
Here we report successful expression and extracellular release of a GH3 β-glucosidase encoded by CC_0968 from Caulobacter crescentus at levels that enabled growth of Z. mobilis on cellobiose as a sole carbon source. Growth on cellobiose correlated with increased expression and release of GH3, which was induced by prolonged incubation (adaptation) in cellobiose or by exposure to sucrose medium. Proteomic analysis revealed that both cellobiose and sucrose adaptation included numerous changes to protein levels relative to growth on glucose that suggest a cellular remodeling program that enables extracellular display or release of GHs.

Heterologous expression of GHs in Escherichia coli and Zymomonas mobilis
Two GH-encoding genes, cel3A from Cellvibrio japonicus and CC_0968 (encoding GH3) from Caulobacter crescentus, were cloned and expressed separately from an expression vector derived from pIND4 [18], which we call pVector and which has an isopropyl β-D-1-thiogalactopyranoside (IPTG)-inducible T7A1 promoter driving GH expression, a spectinomycin-resistance gene, a Rhodobacter-derived broad host range origin of replication for Z. mobilis, and a pUC ori for E. coli (Fig 1A). E. coli DH5α and Z. mobilis ZM4 were transformed with pCel3A, pGH3, or control plasmid pVector and tested for growth in minimal medium containing either cellobiose or glucose as the sole carbon source. As expected, growth on cellobiose was not observed in either E. coli or Z. mobilis transformed with pVector (Fig 1B and 1E). However, for E. coli both pCel3A and pGH3 facilitated growth on cellobiose (Fig 1C and 1D). A corresponding effect was not observed for Z. mobilis, for which neither pCel3A nor pGH3 resulted in growth on cellobiose (Fig 1F and 1G). Both E. coli and Z. mobilis were able to grow on glucose when transformed with pVector, pCel3A, or pGH3, indicating that these plasmids do not negatively impact Z. mobilis viability (Fig 1B-1G). Furthermore, Z. mobilis pVector and Z. mobilis pGH3 exhibited consistent growth in rich medium plus glucose (RMG) supplemented with spectinomycin and IPTG indicating that GH expression does not induce an inhibitory metabolic or protein synthesis burden in Z. mobilis pGH3 (S1 Fig).

Non-native glycosyl hydrolases are readily secreted from E. coli but not Z. mobilis
Bacteria express and localize proteins to various subcellular locations. Depending on the presence or absence of an N-terminal signal sequence and other features, proteins are either retained in the cytoplasm or localized to different cellular compartments. Secreted proteins may be targeted to the inner membrane, outer membrane (OM), periplasm, or extracellular region depending on the accessibility of secretory apparatuses, the nature of signal sequences, or both [21]. To determine why heterologous expression of GHs in Z. mobilis was unable to support growth on cellobiose, we tested for differences in expression and localization of the GHs between E. coli and Z. mobilis.
Extracellular (culture broth), whole cell, spheroplast, periplasmic, and cytoplasmic protein fractions representing equal numbers of cells were prepared as described previously [22] (see Methods). Using 4-methylumbelliferyl β-D-glucopyranoside (MUG), a fluorogenic analogue of cellobiose, GH activity was measured for each protein fraction and normalized to the total protein concentration in the fraction ( Table 1; Materials and methods).
We compared GH activity in Z. mobilis transformed with pGH3 encoding the C. crescentus CC_0968 (called Z. mobilis GH3 hereafter), or with pVector, grown in rich medium glucose (RMG) to E. coli transformed with pGH3, or pVector, grown in Luria-Bertani (LB) medium. Among the isolated protein fractions of Z. mobilis GH3, GH activity was highest (~68%) in the periplasmic fraction, with~99% of total GH activity localized to intracellular fractions (periplasm, spheroplast and cytoplasm). GH activity localization closely followed the localization predictions for CC_0968 by LipoP 1.0 [23] and PSORTb [24] (S3 Table). Only a small fraction of activity (~1% of total activity) was observed from the extracellular fraction (i.e., culture medium; Table 1). E. coli pGH3 also exhibited about half of total GH activity in the periplasm but, in contrast to Z. mobilis, significant activity was also present in the culture medium and accessible in isolated whole cells (i.e., washed whole cells could hydrolyze assay substrate added to the resuspended cells; Table 1). These results suggest that extracellular or surfaceaccessible GH activity may be important for growth on cellobiose. Strains containing the pVector control plasmid did not exhibit GH activity in any fractions, indicating that Z. mobilis ZM4 and E. coli DH5α express little or no endogenous GH activity.
Although we detected expression and activity from both C. crescentus (pGH3) and C. japonicus (pCel3A) enzymes, subsequent experiments demonstrated that only pGH3 enabled adaptation of Z. mobilis for growth on cellobiose. Hence, we report here the subcellular GH activity for only GH3 from C. crescentus.

Z. mobilis growth on cellobiose requires a long adaptation but not genetic mutation
After confirming GH activity in Z. mobilis GH3, we tested whether adaptation or selection for mutations could enable growth on cellobiose. Z. mobilis GH3 and a control strain (pVector) were inoculated into rich medium containing 2% cellobiose and 0.05% glucose (RMCG) with 0.4 mM IPTG at apparent OD600 of~0.05. The cultures were incubated at 30˚C and monitored for change in turbidity. The presence of 0.05% glucose enabled the cells to grow to an apparent OD 600 of~0.13, at which point glucose was consumed from the medium. The cells  [19] and Z. mobilis ZM4 in Zymomonas minimal medium [20] with glucose or cellobiose as a carbon source. (A) Expression of GH genes in a pVector backbone and a summary of the constructs and their effects on growth in minimal medium supplemented with cellobiose. (B-D) Growth of E. coli DH5α containing plasmids pVector or pCel3A or pGH3 in MOPS minimal medium supplied with 0.4% glucose or cellobiose. (E-G) Growth of Z. mobilis ZM4 containing plasmids pVector or pCel3A or pGH3 in a Zymomonas minimal medium containing 2% glucose or 2% cellobiose. The growth curves are averages of three replicates. � Gene from Cellvibrio japonicus, # Gene from Caulobacter crescentus. then entered a long (~48 h) lag phase before again showing growth (Fig 2A). After this lag period, only the GH3-expressing strain and not the control strain resumed growth.
To determine if growth on cellobiose was enabled by a permanent genetic alteration we grew RMC-adapted and unadapted Z. mobilis GH3 in RMCG+0.4 mM IPTG. As a control, we also grew RMC-adapted Z. mobilis GH3 in RMG before returning cells to RMCG+0.4 mM IPTG (reRMG). Washed, RMC-adapted cells inoculated into RMCG no longer exhibited a long lag phase after consumption of glucose but instead continue growing on cellobiose ( Fig  2B). RMC-unadapted cells exhibited a long lag phase after glucose consumption consistent with previous observations. Interestingly, the RMC-adapted cells grown in RMG before returning to RMCG lagged similarly to RMC-unadapted cells (Fig 2B) suggesting that growth on RMG reverted the cells back to a state that required a long lag phase before growth on cellobiose could resume. Based on these observations, we concluded that Z. mobilis GH3 adaptation to RMC cannot be attributed to a permanent genetic change. Instead we hypothesized that a slow remodeling of gene expression and cellular state occurs during adaptation (i.e., the long lag phase) that allows for growth on cellobiose.
We also tested the growth of Z. mobilis GH3 as colonies on solid agar medium containing RMC. Cells were struck out on RMC agar plates with or without 0.4 mM IPTG alongside control cells containing pVector and incubated for 3-6 d at 30˚C. After 3 d, adapted Z. mobilis pGH3 cells grew only in the presence of IPTG (S2 Fig), verifying that growth of Z. mobilis GH3 on cellobiose requires induction of GH3 synthesis. We also verified that heterologous GH3 protein was produced in Z. mobilis GH3 by analyzing proteins produced with and without IPTG by SDS-PAGE of cell lysates. The GH3 protein band was evident in all IPTGinduced cells expressing GH3 as a distinct, detectable band of~80 kDa (predicted MW of GH3 after removal of its signal sequence is 79652 Da) (S3 Fig). Estimation of the extent of induction by densitometry revealed several-fold increases in GH3 in most strains grown in the presence of IPTG.
To understand the changes that allow Z. mobilis GH3 to grow on cellobiose, we used MUG to assay and compare extracellular (supernatant) and whole-cell (pellet) GH activity between adapted and unadapted cells. RMCG-adapted and unadapted cells were grown in RMC and samples were collected at zero and 24 h. At zero h both adapted and unadapted cells exhibited low extracellular GH activity. After 24 h, adapted cells showed a marked increase in extracellular GH activity, whereas extracellular GH activity remained unchanged for unadapted cells (Fig 2C). Adapted whole cells exhibited higher GH activity than unadapted whole cells at both zero and 24 h, with GH activity increasing several fold after 24 h for both adapted and unadapted cells (Fig 2D). These results further indicate that GH release to the extracellular medium is important for growth on cellobiose. Adapted whole cells also displayed considerable levels of GH activity, suggesting that adaptation either increased permeability of the OM to the substrate (i.e., MUG or cellobiose) or increased display of GH3 on the cell surface. Taken together, either greater transfer of GH to the extracellular space, alteration of OMs, or both appear to contribute to growth on cellobiose.
Although a comparable non-genetic remodeling of Z. mobilis has not been described previously, the cell surface proteins of some bacteria, including E. coli, are known to undergo reversible, non-genetic changes in composition [25]. Subpopulations of bacteria are also known to change growth state in reversibly without underlying genetic change [26]. Thus, a reversible remodeling of the Z. mobilis cell envelope is not without precedent.
We also measured extracellular and whole cell GH activity of the RMCG-adapted cells that were regrown in RMG before returning cells to RMC (reRMG). These reRMG cells showed low levels of extracellular GH activity and only moderate whole cell GH activity on par with unadapted cells (Fig 2E). Based on these findings, we concluded that adaption to RMC likely Growth and physiological changes induced by adaptation to cellobiose medium. (A) Growth of Z. mobilis ZM4 transformed with control pVector or pGH3 in 2% cellobiose medium with 0.05% glucose (RMCG). Z. mobilis GH3 growth in RMCG can be described in three stages: initial growth on glucose, a long lag phase, and growth on cellobiose. (B) Growth comparison of Z. mobilis GH3 in RMCG. Cells were either adapted to cellobiose, unadapted to cellobiose, or adapted to cellobiose and then regrown in RMG (reRMG). Extracellular (C) and whole-cell (D) GH activity for RMCG-adapted or -unadapted Z. mobilis GH3. (E) Extracellular and whole cell GH activity of RMCG-adapted Z. mobilis GH3 grown in RMG after adaptation before returning to RMCG. GH activity reported as relative fluorescence signal produced per min normalized by input cell number (apparent OD 600 ). Error bars are standard deviations of biological triplicates. https://doi.org/10.1371/journal.pone.0226235.g002 consists of a reversible remodeling of Z. mobilis rather than a genetic change that permanently altered the properties of the cells. When wild-type Z. mobilis ZM4 was transformed with plasmids recovered from RMC-adapted Z. mobilis GH3, the newly transformed strain behaved like unadapted cells, further indicating that the adaptation occurred in the host cell itself and not by mutation of the plasmid.
Having demonstrated that Z. mobilis GH3 can grow solely on cellobiose, we next measured cellobiose conversion to ethanol at successive times during adaptation and growth on cellobiose (0, 3, 6, 12, 24, 48, 96 and 168 h after inoculation in RMCG medium). No cellobiose conversion occurred during the initial growth on glucose or during the approximately 48-h adaptation period in Z. mobilis GH3. After 48 h, Z. mobilis GH3 began consuming cellobiose as indicated by depletion of cellobiose from the medium (S4A

Serial transfer of culture also enhanced cellobiose conversion
We also tested cellobiose adaptation of Z. mobilis GH3 by serial passage using Z. mobilis pVector as a control (Fig 3). The first passage was performed from RMG to RMC with 0.4 mM IPTG and all subsequent passages performed in RMC+IPTG (Fig 3A). To prevent carryover of extracellular GH between passages, cells were pelleted, washed, and resuspended in fresh medium. Growth of cells was monitored by measuring apparent OD 600 and the supernatant collected at the beginning and end of each passage to measure cellobiose depletion and ethanol accumulation.
After the first passage to RMC, little to no growth was observed over the course of 3 d for both Z. mobilis pVector and Z. mobilis GH3 (Fig 3B). After the second passage, Z. mobilis GH3 cell density more than doubled after 2 d whereas control cells did not grow (Fig 3C). Z. mobilis GH3 continued to grow robustly after the third passage with a doubling time of <24 h (Fig 3D). During this serial passage growth of Z. mobilis GH3 on cellobiose was concomitant with cellobiose depletion from the medium (Fig 3E-3G) and accumulation of ethanol ( Fig  3H-3J). From the serial passage experiment, we conclude that after the second passage Z. mobilis GH3 was able to utilize cellobiose for growth and ethanol production.

Growth in sucrose medium enabled growth on cellobiose
Sucrose is a natural disaccharide substrate for Z. mobilis, catabolism of which depends on secreted sucrase(s) [27] and uncharacterized changes in cellular state. Thus, we hypothesized that exposure to sucrose might induce changes in Z. mobilis that enable growth on cellobiose. To test this hypothesis, we grew Z. mobilis GH3 in rich medium containing 2% sucrose (RMS) or RMG for 48 h. The cells were then washed and inoculated into the fresh RMC containing 0.1% sucrose (RMCS) and growth was monitored (Fig 4A). We found that sucrose-grown cells resumed growth efficiently on cellobiose without a long lag phase, similar to RMC-adapted cells (Fig 2B), but neither RMG-grown Z. mobilis GH3 nor pVector control cells grew significantly. However, RMG-grown Z. mobilis GH3 did eventually resume growth after a lag, as seen previously (Fig 2). This finding suggests that sucrose can induce changes in Z. mobilis that enable GH3-mediated growth on cellobiose. To investigate the effects of sucrose on GH activity, we assayed and compared GH activity of cellobiose-, sucrose-, and glucose-grown Z. mobilis GH3. We found that both cellobiose-and sucrose-grown cells exhibited higher extracellular and whole cell GH activity than glucose-grown cells (S5 Fig) consistent with our previous observations of RMC-adapted cells. These results suggest that growth on sucrose induces a cellular response in Z. mobilis that is similar to the remodeling that occurs during RMC adaptation.
To determine the minimum amount of sucrose needed to remodel cells for growth on disaccharides, we adapted Z. mobilis GH3 in RMC supplemented with 0.2-0.8% sucrose and compared growth in RMC+sucrose to unadapted cells and control pVector (Fig 4B). pVector control cells were only able to grow in RMC+0.8% sucrose, suggesting that 0.8% sucrose is the minimum amount of sucrose that will support growth of Z. mobilis in RMC medium (Fig 4B). After 72 h, unadapted Z. mobilis GH3 grown in RMG before inoculating in RMC+sucrose showed little-to-no growth on RMC+0.2% sucrose and only modest growth on RMC+0.4% sucrose, the latter possibly supported by some cellobiose consumption (S6   mobilis GH3, here defined as cells grown in RMS for 48 h, were also inoculated into fresh RMC with increasing amounts of sucrose (i.e., RMC+0.2-0.8% sucrose). In each culture, the sucrose-adapted cells consumed almost all sucrose in the medium by 24 h and after which cells continued to grow on cellobiose (Fig 4B and S5C Fig). These results suggest that 0.2% sucrose is sufficient to remodel Z. mobilis GH3 for growth on cellobiose, but that adaptation in higher sucrose concentrations will support more cell growth and greater rates of cellobiose consumption.

Adaptation in cellobiose or sucrose medium remodeled Z. mobilis similarly
Adaptation to both cellobiose and sucrose similarly promote growth on cellobiose and induce increases in extracellular and whole cell GH activity. However, it is unclear what specific cellular changes occur in response to sucrose and cellobiose adaptation and what changes are needed for growth on cellobiose. To address this question, we compared protein levels of Z. mobilis GH3 adapted in cellobiose and sucrose and compared to unadapted cells grown in glucose (see Methods) using unlabeled mass spectrometry proteomics. Both extracellular and intracellular protein fractions were collected and analyzed. A total of 1539 proteins were identified from intracellular samples, representing >80% of annotated protein-coding genes in Z. mobilis ZM4 ATCC 31821 [5]. A total of 1231 proteins were identified from extracellular samples, but most extracellular proteins (1215 out of 1231) overlapped with the intracellular fraction of proteins. This result suggests that many proteins detected in the growth medium (i.e., the extracellular fraction) are likely derived from the cytoplasm either by cell breakage or by incomplete separation of cells from the extracellular medium.
We observed greater similarity in intracellular protein levels (R 2 = 0.37) between cellobioseand sucrose-adapted cells when normalized to glucose-grown cells whereas extracellular protein levels were less similar across conditions (R 2 = 0.17) (S7 Fig). Given our observations that extracellular GH activity increases during adaptation to cellobiose and that sucrase and levansucrase are known to be secreted from Z. mobilis [27] in response to sucrose, we looked at levels of secretion-related proteins in both sucrose-and cellobiose-adapted cells. Notably, the levels of a majority of annotated transport and secretion-related proteins increased in both sucrose-and cellobiose-adapted cells (Fig 5, S7 Fig). Interestingly, GH3 CC_0968 was also upregulated in both extracellular and intracellular fractions in both cellobiose and sucrose media (Fig 5, S8 Fig) despite a consistent amount of IPTG in the glucose, sucrose, and cellobiose media. This increase in GH3 expression may be explained by an increase in pGH3 plasmid copy number, reduced turnover of GH3 after release to the extracellular medium, a change in promoter activity on pGH3, or a combination of these effects.
To understand the nature of protein remodeling during sucrose and cellobiose adaptation more completely, we performed K-means clustering on normalized log2-fold change values (normalized to glucose-grown cells) for proteins that were measured in both the intracellular and extracellular fractions (1199 proteins in total). In total, 50 clusters were produced revealing similar remodeling patterns between sucrose-and cellobiose-adapted cells (Fig 6). Two prominent clusters are present in which protein levels are similarly downregulated or upregulated in sucrose-and cellobiose-adapted cells. We performed Gene Ontology (GO) enrichment analysis of KEGG pathways [28] for each cluster and identified several pathways that were statistically enriched across six clusters. Of the clusters primarily comprised of upregulated proteins, several transport-related and cell membrane pathways were enriched such as ABC transporters, integral membrane components, receptor activities, transport activities, and cell outer membrane pathways (Figs 5 and 6). These results are consistent with GO-term enrichment analysis of differential protein levels in each sample compared to glucose-grown cells where GO terms related to stress (oxidoreductase), secretion, and transport (periplasmic space, OM, transport, and protein secretion) were enriched in both intra-and extracellular fractions of cellobiose-and sucrose-adapted strains (S4 Table). Within the large cluster of primarily downregulated proteins several growth-related pathways were enriched including the ATP synthase, ribosome, translation, and purine and pyrimidine biosynthesis pathways (Fig 6). A decrease in growth-related proteins can be attributed to the reduced growth rate of cells metabolizing cellobiose or sucrose relative to cells growing on glucose.
Despite the similarities in protein expression between cellobiose-and sucrose-adapted cells, we observed sets of proteins uniquely differentially expressed between the two conditions (S7 Fig). Of note, we observed a disparity in the expression patterns of SacB (extracellular levansucrase) and SacC (extracellular sucrase) [29][30][31]. Although SacB was upregulated in both cellobiose and sucrose adapted cells (S1 Data), SacC expression was only upregulated in sucrose-adapted cells. Previous work has shown that SacB and SacC can be expressed as both a bicistronic transcript and individually as monocistronic transcripts with SacC under control of a strong promoter [32]. Our results indicate that SacC upregulation is dependent on a sucrose-specific signal whereas SacB monocistronic expression can be activated in a sucroseindependent manner.

Discussion
Our study of xenogeneic GH-enabled growth of Z. mobilis on cellobiose and its conversion to ethanol revealed two key discoveries. First, even though high expression of a β-glucosidase from a closely related α-proteobacterium C. crescentus occurs relatively quickly (within 12 h) in rich medium glucose, Z. mobilis GH3 depends on a transient adaptation period of several days before growth on cellobiose and high-flux ethanol production becomes possible. Second, growing Z. mobilis on sucrose induces cellular changes similar to cellobiose adaptation that also allow Z. mobilis GH3 to readily grow on cellobiose without a long lag phase.

Cellobiose adaptation coincides with extensive changes to the Z. mobilis membrane proteome
In this study, we showed that adaptation of Z. mobilis GH3 to cellobiose medium correlated with extensive changes to its membrane proteome that coincided with an increased presence of GH3 in the extracellular medium. When grown in RMG, GH3 was located mostly in the periplasm and cytoplasm with only a small fraction of GH3 secreted to the culture medium ( Table 1). These initial low levels of extracellular or periplasmic GH were insufficient to generate enough glucose for growth and the strain required an approximately 72-h lag phase before cells became adapted and slow growth could commence (Fig 2A). Growth on cellobiose consistently coincided with increased GH activity in the extracellular or outer membrane space ( Table 1, Fig 2, S5 Fig) underscoring the importance of GH localization for cellobiose conversion. Once cells expressing GH3 were adapted to cellobiose, they could continue to grow without a long lag when placed in fresh cellobiose medium. However, this adaptation to cellobiose was reversible, and when cellobiose-adapted cells were regrown in RMG they would again require a long lag phase (Fig 2B) in cellobiose medium before growth could occur. Thus, cellobiose adaptation consisted of a reversible change to the cellular composition that was not conferred by a stable genetic adaptation.
Our proteomics results revealed that adaptation to growth on cellobiose changed the protein composition of Z. mobilis GH3 compared to a glucose-grown control. The protein expression patterns clustered into groups of proteins of related functions in both the intracellular and extracellular fractions (Fig 6). Notably, proteins related to secretion and transport were upregulated (Fig 5), which we hypothesize leads to increased extracellular release of GH3, increased display of GH3 on the outer membrane, entry of substrates (cellobiose) into the periplasm, or a combination of these effects (Fig 2D).
Our results merit comparison to those of Yanase et al. [10], who expressed a clostridial ßglucosidase fused to a twin-arginine-translocation signal in Z. mobilis and observed conversion of cellobiose to ethanol by cells grown to saturation in glucose-containing broth. Whereas

PLOS ONE
Yanese et al. found the ß-glucosidase activity primarily in the periplasm and to a lesser extent on the cell surface and did not report growth of Z. mobilis on cellobiose, we found that Z. mobilis GH3 adapted for growth on cellobiose released at least half of the ß-glucosidase activity into the culture broth. It is likely that the adaptation documented here that facilitates growth of Z. mobilis GH3 on cellobiose represents a different mode of β-glucoside production, but a direct comparison of the Yanase et al. cellobiose-converting strain to Z. mobilis GH3 would be needed to draw definitive conclusions.

Cellobiose and sucrose adaptation may be comprised of a native scavenging response to glucose depletion
Similar cellular changes were observed during adaptation on cellobiose and adaptation to sucrose, a native disaccharide substrate in Z. mobilis. This similarity is apparent in the protein level changes in cellobiose-and sucrose-adapted cells, including increased extracellular GH3, and by the observation that adaptation to sucrose enables growth on cellobiose. Although sucrose is a natural substrate for Z. mobilis, we also observed a long lag phase before cells began growing on sucrose. Given these findings, we hypothesize that the changes in the membrane proteome and protein secretion that accompany both sucrose and cellobiose adaptation are part of a generalized scavenging response to glucose-depleted conditions. The changes in protein secretion and envelope composition during this scavenging response could improve acquisition of nutrients by Z. mobilis from the extracellular medium. Should a suitable substrate be found while scavenging, then substrate-specific and possibly more energy intensive responses would be activated as we observed with sucrose-specific upregulation of SacC. In the case of cellobiose adaptation this cellular remodeling increased GH3 release or surface display of cells incubated with cellobiose and ultimately enabled cells to grow on cellobiose.
We note that our results do not distinguish whether GH3 is secreted to the cell surface and released via secretion mechanisms related to Z. mobilis sucrase secretion or if the adaptation simply changes the outer membrane of Z. mobilis in ways that enable nonspecific GH3 release or release via cell lysis. Possibly, membrane stress from protein overexpression could contribute to cell lysis or membrane leakage. The presence of some intracellular proteins in the culture broth of adapted Z. mobilis GH3 indicates that at least some lysis or leakage occurred. However, the presence of peptides in the yeast-extract containing Z. mobilis growth medium precluded absolute quantitation of protein concentrations to assess the amounts of cell proteins vs. GH3. Further work will be required to understand the mechanism of Z. mobilis adaptation for GH3-mediated growth on cellobiose.

Prospects for engineering Z. mobilis for broader utilization of oligosaccharides
Z. mobilis is being developed as a platform for the conversion of lignocellulosic hydrolysates to biofuels and bioproducts. Complete utilization of oligosaccharides will improve product yields and economic feasibility of lignocellulosic biomass conversion. Our findings highlight important challenges to engineering broader oligosaccharide utilization in Z. mobilis. First and foremost, we lack a comprehensive understanding of native secretion and transport pathways in Z. mobilis. Broadening the substrate specificity to oligosaccharides beyond cellobiose will require a more complete understanding of the rules governing both GH secretion and oligosaccharide entry to the periplasm in Z. mobilis. It is clear from our work and the work of others [33,34] that GH localization and substrate accessibility is crucially important. Likewise, the impact of species origin on heterologous gene expression in Z. mobilis may not be fully appreciated. That expression of celA from C. japonicus did not enable Z. mobilis to grow on cellobiose, even when allowing time for adaptation, suggests that important differences exist in how xenogeneic glycosyl hydrolases are recognized by and interact with endogenous Z. mobilis pathways.
Although GH3 expression allows Z. mobilis to metabolize cellobiose, the long adaptation time required to remodel Z. mobilis for growth on cellobiose is not practical for industrial applications. Continuous fermentations may be a suitable option given that cellobiose adaptation is retained upon transfer to fresh RMC. However, this option does not eliminate the initial long adaptation time. Further mechanistic dissection of cellobiose and sucrose adaptation is needed to identify key regulators governing adaptation (e.g., the proposed scavenging response). With greater understanding of the basic mechanisms, it is plausible that GHexpressing strains could be engineered by rewiring the natural response to sucrose to induce the necessary cellular changes to support oligosaccharide metabolism and eliminate the need for a long adaptation to achieve cellular remodeling.

Strains, plasmids and culture conditions
Zymomonas mobilis ZM4 (ATCC #31821), Escherichia coli DH10B (Invitrogen, Carlsbad, CA, USA) and E. coli DH5α (New England BioLabs, Ipswich, MA, USA) were used in this study. The E. coli DH10B strain was used for cloning and E. coli DH5α was used for expressing the recombinant proteins. Unless otherwise specified, all the E. coli strains were grown in LB or MOPS minimal medium [19] at 37˚C with shaking. Z. mobilis strains were grown in rich medium containing 1% yeast extract, 15 mM KH 2 PO 4 plus 2% glucose (RMG), 2% cellobiose (RMC), varying concentrations of sucrose (RMS), or in Zymomonas minimal medium (ZMM) [20] at 30˚C without shaking. Plasmid pIND4-spec, a derivative of the Rhodobacter-derived, broad-host-range shuttle vector pIND4 [18] in which the kanamycin-resistance gene was replaced with a spectinomycin resistance gene, was used to clone and express glycoside hydrolases. Spectinomycin was used at 50 μg mL -1 for E. coli and 100 μg mL -1 for Z. mobilis.

Plasmid construction
All oligonucleotide primers used for cloning are listed in S1 Table and were obtained from  Integrated DNA Technologies (Coralville, IA, USA). For plasmid construction, the vector backbone and gene fragments were amplified by PCR using Q5 DNA polymerase (New England Biolabs, Ipswich, MA, USA) following the manufacturer's protocol. DNA fragments were purified by agarose gel electrophoresis and assembled using Gibson Assembly mix (New England Biolabs, Ipswich, MA, USA) following the manufacturer's protocol. All plasmids used in this study were verified by DNA sequencing and restriction enzyme digestion analysis and are listed in S2 Table. The plasmids pCel3A and pGH3 were constructed by cloning the respective glycosyl hydrolases encoded by cel3A from Cellvibrio japonicus and CC_0968 from Caulobacter crescentus in pIND4-spec (pVector). The predicted localizations of Cel3A and CC_0968 are shown in S3 Table. Transformations of E. coli and Z. mobilis cells Electro-competent cells were prepared following a standard protocol [35]. Plasmid DNA or Gibson assembly reactions (~100 ng) was used to transform of E. coli competent cells (~10 9 cells in 50 μl 10% v/v glycerol in water). Electroporation of E. coli was performed with a Bio-Rad gene pulser with a setting of 200 O, 25 μF, and 1.75 kV in a 0.1 cm cuvette. Immediately after electroporation, SOC medium (1 mL) was added and cells were incubated (37˚C for 1 h); the recovered cells (100 μL)were then spread on LB agar containing appropriate antibiotic for selection and the plates were incubated (37˚C overnight).
For Z. mobilis transformation, plasmid DNA (~1 μg) was used to transform cells (~10 9 cells in 50 μl 10% v/v glycerol in water). Type 1 restriction inhibitor (1 μL; Epicentre) was added to the plasmid DNA prior to mixing with competent cells. Electroporation of Z. mobilis was performed with a Bio-Rad gene pulser with a setting of 200 O, 25 μF, and 1.6 kV in a 0.1 cm cuvette. Immediately after electroporation, recovery broth (1 mL; 5 g glucose/L, 10 g yeast extract/L, 5 g tryptone/L, 2.5 g (NH 4 ) 2 SO 4 /L, 0.2 g KH 2 PO 4 /L, and 0.25 g MgSO 4 •7H 2 O/L) was added and the cells were incubated (2-3 h at 30˚C). The recovered cells were spread on RMG-agar containing the appropriate antibiotic for selection and the plates were incubated (30˚C for 2-4 d) to obtain transformed colonies.

Extraction of cellular and subcellular fractions, protein quantification, and activity assay
For activity measurements, seed cultures were prepared by overnight cultivation in the desired conditions. LB or RMG (5 mL) supplemented with required concentration of spectinomycin were inoculated with seed cultures of E. coli and Z. mobilis, respectively, and incubated until the apparent OD 600 reached~0.4. For protein induction, IPTG (0.2 mM) was added to both cultures and incubation was continued overnight. Extracellular and intracellular fractions (supernatant, periplasm, cytoplasm, spheroplast and whole cells) were prepared from the same number of cells quantified by measuring apparent OD 600 and adjusting the volume to obtain an apparent OD 600 of 1.5.
Protein concentration was measured by using Bicinchoninic Acid (BCA) assay (Thermo Scientific) following the manufacturer's protocol. Cell samples (20 μL) and standards of diluted bovine serum albumin (BSA) were transferred to clean 1.5 mL microtubes. A no-protein control was also included. Cold (-20˚C) acetone (80 μL) was added and the sample was vortexed vigorously and incubated (-20˚C, 1 h). The proteins were pelleted by centrifugation (15, 000 × g, 15 min, 4˚C). The supernatant was then carefully removed and discarded. Protein pellets were washed with cold acetone (100 μL) by adding it around the walls and centrifugation (15,000 × g, 2 min, 4˚C). After removal of supernatant, the pellets were dried in vacuo at room temperature (10 min). The dried pellets were resuspended in 2% SDS, 9 mM Tris-Cl pH 8.0 (10 μL) and incubated (70˚C,10 min). Protein concentration was measured following the manufacturer's protocol. Samples were then incubated (37˚C, 30 min to 2 h). Absorbance was measured at 562 nm in a Tecan M1000 plate reader (Tecan Group Ltd., Männedorf, Switzerland) and protein concentration was determined by comparison to BSA standards.
The glycosyl hydrolase activity assay was performed by adding protein sample (10−25 μL) to 50 μL of 2 mM 4-methylumbelliferyl β-D-glucopyranoside (MUG; Sigma-Aldrich, St. Louis, MO, USA) in a 96-well plate. The reactions were monitored in a Tecan M1000 plate reader with fluorescence excitation at 365 nm and emission at 455 nm for 180 min with readings every 5 min. The fluorescence produced was plotted as a function of time and the enzyme activity was determined from the slope of this plot. The activity value was normalized by the amount of protein present in the reaction.

Growth adaptation
Triplicate samples of Z. mobilis GH3 (expressing CC_0968) and the pVector control were grown overnight in RMG medium containing spectinomycin (5 mL, 30˚C). A sample (1 mL) was taken at apparent OD 600~1 .0, centrifuged, and washed twice with sterile RMC medium. Washed cell pellets were then resuspended in a culture tube RMC medium (10 mL) containing spectinomycin and 0.4 mM IPTG. Culture tubes were then incubated (30˚C) without shaking. Growth was monitored initially after 3, 6, 12 and 24 h, and then every 24 h. After a significant growth was seen for Z. mobilis GH3, cells were collected, washed with RMC medium, and resuspended again in RMC with spectinomycin and IPTG. This process was repeated with RMC medium supplemented with 0.05% glucose (RMCG). To evaluate whether the adaptation was due to permanent genetic change, cellobiose-adapted Z. mobilis GH3 was regrown in RMG medium (reRMG) and transferred back to fresh RMCG medium after washing. Growth of three types of Z. mobilis GH3 strains, unadapted, adapted and adapted regrown in RMG (reRMG), were then compared in identical conditions.
A serial passage experiment also was performed using Z. mobilis GH3 and pVector. The cells were grown in RMG overnight, collected by centrifugation and washed with RMC to remove residual glucose. The cells were then resuspended in RMC containing spectinomycin and 0.4 mM IPTG to an apparent OD 600 of~0.1 and incubated for the first passage (30˚C, 3 d). The apparent OD 600 was then measured and a second passage was performed after centrifugation and resuspension of the cell pellets in fresh RMC-spectinomycin-IPTG medium at a similar starting apparent OD 600 . Incubation was continued for 2 d and then the third passage was performed after similar recovery by centrifugation and resuspension of the cell pellets in fresh RMC-spectinomycin-IPTG medium. For the third passage, incubation was continued for one d. Cultures were then quantified for apparent OD 600 and stored for metabolite analysis by HPLC.
For adaptation in sucrose medium, Z. mobilis GH3 was grown in rich medium containing 2% sucrose (RMS) without shaking (30 ºC, 48 h). Cells were then collected by centrifugation (5000 × g, 5 min, room temperature). The supernatant was then discarded and the pellets were washed with sterile deionized water. Finally, the pellets were resuspended in fresh growth medium containing glucose, sucrose, cellobiose or a combination of these sugars depending on experiment to be performed.

Growth and activity measurement from adapted vs unadapted culture
Adapted cultures were derived from cells grown in RMCG for 24 h. Unadapted cultures were derived from cells grown in RMG for 12 h. After centrifugation, the pellets were washed with RMC and resuspended in RMCG medium containing spectinomycin and IPTG. For growth measurement, samples were removed at time intervals, apparent OD 600 was recorded and cells were stored at -20˚C prior to metabolite analysis. For GH activity measurements, the required volume of cells was withdrawn from the culture tube, centrifuged (20,000 × g, 5 min, 4˚C). The supernatant was transferred to a fresh tube to assay for extracellular GH activity and the pellets were washed with water and resuspended in water at a calculated volume to give an equivalent number of cells in all samples. For GH activity measurement, extracellular or pellet fractions (20-25 μL) were transferred to a 96-well plate in triplicate, 2 mM MUG (50 μL) was added, and the readings in a Tecan 1000 plate reader were immediately started (fluorescence excitation at 365 nm and emission of 455 nm and, for cell samples, cell density at 600 nm). Normalization was performed by dividing the fluorescence value with corresponding apparent OD 600 .

SDS PAGE and GH3 protein signal measurement
For analysis of GH3 protein induction, both adapted and unadapted Z. mobilis strains with plasmids pGH3 or pGH3T were grown in RMG supplemented with appropriate concentration of antibiotics (S3 Fig). The plasmid pGH3T encodes, in addition to GH3, a C. crescentus TonB receptor for cellobiose (CC_0970) that was found to have little or no effect on cellobiose utilization in Z. mobilis. When the cultures reached apparent OD 600 of~0.4, they were divided into equal portions and IPTG was added to one portion to final concentration of 0.5 mM after which growth of both portions was continued for 24 h. Final apparent OD 600 was measured for all samples. Approximately equal numbers of cells equivalent to one mL of apparent OD 600 3.0 were centrifuged (10,000 × g, 5 min, 4˚C). The pellets were resuspended in SDS loading solution (20 μL of 62 mM Tris�Cl, pH6.8, 2% w/v SDS, 10% v/v glycerol, 5% v/v β-mercaptoethanol, 0.05% w/v bromphenol blue), incubated (98˚C, 10 min), and then immediately cooled on ice for 5 min. The samples were centrifuged again (10,000 × g, 2 min, 4˚C) and portions of the supernatants (10 μL) were loaded on a 4-12% Tris-Glycine slab gel connected to Tris-Glycine SDS running buffer (Invitrogen). The gel was electrophoresed at 200 volts for 1 h, stained with Coomassie Brilliant Blue R-250 solution (BioRad, #1610436), imaged using a white light transilluminator and a CCD camera equipped with an 595 ± 55 nm bandpass filter (Fluorichem 8000; Protein Simple, Inc.), and then quantified using Imagequant software (GE Healthcare).

Quantification of cellobiose conversion and ethanol production
Cellobiose and ethanol were quantified by adaptation of a method described by Ha et al., 2011 [36]. Samples were withdrawn after apparent OD 600 measurement and stored at −20˚C until all required time points were collected. The frozen samples were thawed at room temperature and vortexed and centrifuged (10000 × g, 5 min, 4˚C). Portions (100 μL) of the supernatants were transferred to labeled 1.5 mL autosampler vials and 900 μL of pure water was added to each vial and mixed properly. The vials were capped and placed in a 4˚C cooled autosampler tray. Fractions (50 μL) of the samples were then injected to an Agilent 1260 Infinity HPLC system with a quaternary pump, vacuum degasser, and refractive index detector (Agilent Technologies, Inc., Palo Alto, CA) and separated on an Aminex HPX-87H with Cation-H guard column (BioRad, Inc., Hercules, CA, USA; 300 × 7.8 mm, cat #125-0140). The mobile phase was 0.02 N H 2 SO 4 and was used at a flow rate of 0.5 mL min -1 ; both column and detector temperatures were maintained at 50˚C. Data were analyzed using ChemStation C.01.06 software (Agilent Technologies, Inc., Palo Alto, CA, USA). The metabolites of interest (cellobiose, glucose, and ethanol) were analyzed and quantified using standard calibration curve prepared from the respective pure compounds obtained from Sigma-Aldrich (St. Louis, MO, USA).

Proteomics analysis of cellobiose-and sucrose-adapted vs unadapted Z. mobilis GH3
Z. mobilis GH3 was grown in RMG, RMC, or rich medium + 2% sucrose (RMS) to late log phase (apparent OD 600 of~0.8). Cells were harvested and extracellular and intracellular fractions were collected by centrifugation (20,000 × g, 5 min, 4˚C). Proteins were digested, analyzed by LC-MS/MS, and peptide identity was verified with Z. mobilis genome peptide library as described below.
To prepare samples, cells were lysed by suspension in 6 M guanidine hydrochloride (GnHCl), followed by addition of methanol to 90% as described previously [37]. Samples were centrifuged (15,000 × g, 5 min, 4˚C). Supernatants were discarded and pellets were allowed to air dry for~5 min. Pellets were resuspended in 200 μL 8 M urea, 100 mM Tris pH 8.0, 10 mM (tris(2-carboxyethyl)phosphine) (TCEP), and 40 mM chloroacetamide, then diluted to 2 M urea in 50 mM Tris�HCl, pH 8. Trypsin was added at an estimated 50:1 ratio, and samples were incubated overnight at room temperature. Each sample was desalted over a PS-DVB solid phase extraction cartridge and dried in vacuo. Peptide mass was assayed with a peptide colorimetric assay.
For each analysis, peptides (2 μg) were loaded onto a 75 μm inner diameter, 30 cm long capillary with an imbedded electrospray emitter and packed with 1.7 μm C18 BEH stationary phase. The mobile phases used were A: 0.2% formic acid and B: 0.2% formic acid in 70% acetonitrile. Peptides were eluted with an increasing gradient of acetonitrile from 0% to 53% B over 75 min followed by a 5 min 100% B wash and a 10 min equilibration in 0% B.
Eluting peptides were analyzed with an Orbitrap Fusion Lumos (Thermo Fisher Scientific, Waltham, MA, USA). Survey scans were performed at R = 240,000 with wide isolation analysis of 300-1,350 mz. Data-dependent top speed (1 s) MS/MS sampling of peptide precursors was enabled with dynamic exclusion set to 20 s on precursors with charge states 2 to 4. MS/MS sampling was performed with 1.6 Da quadrupole isolation, fragmentation by HCD with NCE of 25, analysis in the ion trap with maximum injection time of 10 ms, and AGC target set to 3 x 10 4 .
Raw files were analyzed using MaxQuant 1.6.0.1 [38,39]. Spectra were searched using the Andromeda search engine against a Z. mobilis subsp. mobilis ZM4 (Genbank accessions CP023715, CP023716, CP023717, CP023718, CP023719) protein database and a target decoy database generated in house. Label-free quantitation (LFQ) [40] and match between runs were toggled on, and ion trap tolerance was set to 0.4 Da. All other parameters were set by default. Peptides were grouped into subsumable protein groups and filtered to 1% FDR, based on target decoy approach. Downstream analysis of protein group LFQ values were performed using the Perseus software platform [41]. First, all LFQ values were log2 transformed and any protein groups missing a value from � 6 samples were removed followed by missing value imputation. Fold changes for each protein were calculated for the cellobiose and sucrose conditions by comparison of the LFQ against that of the control sample from the appropriate condition (intra-or extracellular). A Student's two sample t-test was performed for each fold change measurement and p-values were corrected for multiple hypothesis testing by the Benjamini-Hochberg method to generate quantitative FDR values.
K-means clustering was performed on glucose normalized log2-fold change values for proteins that were measured in both the intracellular and extracellular fraction experiments (1199 proteins). The desired number of clusters was set to 50 using Euclidian distance and average linkage. Gene ontology enrichment in each cluster was performed using Fishers exact test with Benjamini-Hochberg correction for multiple hypotheses (p<0.05). Gene ontology annotations were downloaded from UniProt [42].
Supporting information S1  Table. Enriched Gene Ontology (GO) terms in both upregulated and downregulated intracellular and extracellular fractions of cellobiose-and sucrose-adapted Z. mobilis GH3 strain relative to glucose-grown cells. Landick.