Development of fluorescent Escherichia coli for a whole-cell sensor of 2ʹ-fucosyllactose

2′-Fucosyllactose (2′-FL), a major component of fucosylated human milk oligosaccharides, is beneficial to human health in various ways like prebiotic effect, protection from pathogens, anti-inflammatory activity and reduction of the risk of neurodegeneration. Here, a whole-cell fluorescence biosensor for 2′-FL was developed. Escherichia coli (E. coli) was engineered to catalyse the cleavage of 2′-FL into l-fucose and lactose by constitutively expressing α-l-fucosidase. Escherichia coli ∆L YA, in which lacZ is deleted and lacY is retained, was employed to disable lactose consumption. E. coli ∆L YA constitutively co-expressing α-l-fucosidase and a red fluorescence protein (RFP) exhibited increased fluorescence intensity in media containing 2′-FL. However, the presence of 50 g/L lactose reduced the RFP intensity due to lactose-induced cytotoxicity. Preadaptation of bacterial strains to fucose alleviated growth hindrance by lactose and partially recovered the fluorescence intensity. The fluorescence intensity of the cell was linearly proportional to 1–5 g/L 2′-FL. The whole-cell sensor will be versatile in developing a 2′-FL detection system.


Results
Evaluating 2′-FL-dependent cell growth and fluorescence using a whole-cell biosensor. A whole-cell biosensor, which grows and emits fluorescence only in the presence of 2′-FL, was designed (Fig. 1A). The plasmid pConFUC was constructed to constitutively express α-l-fucosidase under the control of the J23100 promoter (Fig. 1A). The 2′-FL can be hydrolysed by α-l-fucosidase in E. coli so that the amount of released l-fucose and lactose are proportional to 2′-FL concentration. The cells were engineered not to metabolize lactose by deleting their endogenous β-galactosidase gene (lacZ) while retaining lacY to import 2′-FL. Two E. coli BL21 (DE3) variants, ∆L YA and ∆L M15, with different levels of residual β-galactosidase activity, were cultured in R medium containing 2 g/L lactose 13,16 . Both variants did not grow with lactose as a sole carbon source in a minimal medium for 24 h, whereas wild-type E. coli BL21 (DE3) grew well using lactose (Fig. 1B). The difference of growth rate of ∆L M15 and ∆L YA in lactose was marginal. For the strict restriction of lactose consumption, the E. coli ∆L YA strain which has much lower β-galactosidase activity was used in the following studies 13,16 .
The active expression of α-l-fucosidase in E. coli ∆L YA was examined by analysing the 2′-FL cleavage in the soluble fraction of E. coli ∆L YA cell extracts. The cell extracts of wild type E. coli BL21 (DE3) and E. coli ∆L YA ( Fig. S1A and Fig. 1C) did not digest 2′-FL. In contrast, 2′-FL was rapidly digested to release lactose and L-fucose by the cell extracts of E. coli BL21 (DE3) pConFUC (Fig. S1B) and E. coli ∆L YA pConFUC (Fig. 1D). These results suggest that α-l-fucosidase was actively expressed in E. coli regardless of lacZ deletion. 2′-FL detection by the increase in fluorescence intensity. E. coli ∆L YA or ∆L YA pConFUC were inoculated with R medium containing no carbon source, 2 g/L 2′-FL, 2 g/L lactose, or a mixture of 2 g/L 2′-FL coli. 2′-FL is cleaved to L-fucose and lactose by α-l-fucosidase. When lacZ is deleted, L-fucose released from 2′-FL is the sole carbon source for cell growth and fluorescence protein expression. Schematic was designed using Adobe Illustrator CS6 (v16.0.0, Adobe Systems Inc., USA). (B) Growth of E. coli strains using lactose as a carbon source. E. coli strains were cultured in R medium, which included 2 g/L lactose. Symbols denote the cell densities of E. coli BL21 (DE3) (filled circle), ∆L YA (filled square), and ∆L M15 (filled upward traingle). Cleavage of 2′-FL by soluble lysates of (C) E. coli ∆L YA and (D) E. coli ∆L YA pConFUC. Symbols denote the concentrations of 2′-FL (filled circle), Lactose (filled square), and L-fucose (filled upward traingle).  Fig. 2A,B, Table S2) and did not grow using lactose as a sole carbon source. However, lactose did not affect the cell growth of E. coli ∆L YA pConFUC using 2′-FL as the carbon source (Fig. 2B). E. coli ∆L YA pConFUC was co-transformed with pConRFP expressing a red fluorescence protein (RFP) under the control of the J23100 promoter. E. coli ∆L YA pConFUC/pConRFP (FLS1) cells had a long lag period of 4 days after their inoculation with R medium containing 2 g/L 2′-FL or a mixture of 2′-FL and lactose. Cells did not grow in R medium containing only lactose. RFP signal began to increase on the fifth day of bacterial inoculation in R medium containing 2′-FL (Fig. 2C). Although FLS1 showed a 2′-FL-dependent cell growth and a fluorescence increase, there was a few-days of long lag period. We thought that the cell was not ready to metabolize fucose even after 2′-FL cleavage to fucose and lactose.
Adaptation of the whole-cell biosensor to L-fucose for faster response toward 2′-FL. As the l-fucose metabolic pathway should be activated to use l-fucose as a carbon source 30 , we tested whether the delayed response can be resolved by pre-adapting cells to l-fucose to shorten the lag period. Cells precultured in LB broth showed a long lag period lasting a day in R medium containing L-fucose (Fig. 3A). When E. coli ∆L YA was preadapted in R medium containing 10 g/L of L-fucose before inoculation for several rounds, the lag period was dramatically shortened, and the cells immediately entered the exponential growth phase upon inoculation (Fig. 3B). When adapted cells were precultured in the LB broth a second time, cell growth on the L-fucose was retarded similarly to the non-adapted cells (Fig. 3C). These results suggest that this fast-growing phenotype arose because of simple adaptation to the substrate.
E. coli ∆L YA pConFUC/pConRFP preadapted to L-fucose (denoted as FLS2) was cultured in R medium containing 2 g/L L-fucose. While E. coli ∆L YA lacking the plasmid for fucosidase did not grow in any media for 4 days, FLS2 showed exponential growth after a short lag period (< 8 h; Fig. 3D,E). Notably, the RFP fluorescence reached its maximal fluorescence within 10-20 h when FLS2 was grown in R medium containing 2′-FL (Fig. 3F). The fluorescence emission lasted for 4 days at a stable emission. Cells did not emit fluorescence using lactose as a sole carbon source and 2 g/L lactose present together with 2′-FL did not alter fluorescence intensity either. Although RFP intensity increased by only ~ 50%, these results suggest that RFP fluorescence was a relatively fast and reliable reporter of 2′-FL.

Enhanced fluorescence intensity.
To quicken the 2′-FL detection and increase fluorescence intensity, the expression cassettes of RFP and fucosidase were recombined using different vectors. The small copy number (10-12) replicon p15A was replaced by the high copy number  replicon ColA for RFP expression (Fig. 4a) 31 . The plasmid pET-ConFUC was also constructed by replacing the kanamycin resistance cassette of pConFUC with the ampicillin resistance cassette of pET ( Fig. 4B and Table 1). The E. coli ∆L YA strain was then transformed with the resulting plasmids pET-ConFUC and pColA-ConRFP. When E. coli ∆L YA pET-ConFUC/pColA-ConRFP (FLS3) was inoculated with R medium containing 2′-FL, fluorescence increased after a short lag phase (< 2 h; Fig. 4C). Furthermore, the maximum fluorescence intensity change was ~ 7 times higher than previous transformants containing pConRFP. The new whole-cell sensor not only exhibited much stronger fluorescence than the previous one the fluorescence intensity change was proportional to a 2′-FL concentration of 1-5 g/L ( Fig. 4D and Fig. S2). The signal was strong enough to enable the visualization of the fluorescence (Fig. 4E).
Reduced lactose toxicity by fucose adaptation. We observed that the enhancement of fluorescence intensity by the new combination of expression cassettes was hindered by the presence of only 2 g/L lactose E. coli symports lactose and protons through lacY, the lactose permease structural gene, and acidifies the cytoplasm via lactose transport. This acidification causes cytotoxicity followed by the induction of cellular acid shock, resulting in the reduction of proton motive force, intracellular ATP levels, and cell viability 32 . As mentioned above, to accelerate fucose use, preadaptation to fucose was applied a second time to relieve the negative effects of cytoplasmic acidification by lactose. E. coli symports L-fucose and H + by a L-fucose/H + symporter, and the internalized cytoplasmic proton would induce E. coli to activate its acid resistance system and relieve cytoplasmic acidification 33 . Therefore, preadaptation to fucose can be used not only to activate L-fucose metabolic pathway but also to enhance acid resistance by bacterial cells (Fig. 5A). Indeed, when ∆L YA pET-ConFUC/ pColA-ConRFP was preadapted to fucose (denoted as FLS4), 2 g/L lactose did not hinder the 2′-FL detection (Fig. 5C). Furthermore, pre-adaption to fucose enabled much larger fluorescence intensity change than the nonadapted cells, even in the presence of 50 g/L lactose (Fig. 5D,E). Pre-adaption no longer improved the fluorescence intensity when 2′-FL was added to bovine milk. However, the fluorescence intensity was still proportional to 2′-FL concentration (Fig. 5F). Both biosensors were still able to detect the biologically relevant concentration of 2′-FL mixed in bovine milk (Fig. S3).

Discussion
We designed and developed a functional E. coli whole-cell 2′-FL biosensor based on the growth-coupled red fluorescence emission produced after the cleavage of 2′-FL by recombinant α-l-fucosidase, and we demonstrated that the biosensor could quantify 2′-FL. As the working principle of the biosensor, 2′-FL enters E. coli ∆L YA perhaps via lacY and then is hydrolysed into L-fucose and lactose by the recombinant α-L-fucosidase in the cytosolic space. Because lactose is an abundant disaccharide in milk, it can be used after hydrolysing into glucose and galactose by β-galactosidase in the wild-type E. coli. Therefore, endogenous lacZ was deleted completely (∆L YA) or partially disrupted (∆L M15). We confirmed that both mutants did not grow in lactose medium at all, and there was no red fluorescence emitted in ∆L YA in the absence of a carbon source (Fig. 1B). Among the E. coli mutants,  (Fig. 1D). L-fucose could be consumed, even though small amount, by E. coli cells extracts suggesting that L-fucose can be used as a sole carbon source. The expression of recombinant α-l-fucosidase and complete deletion of lacZ allowed cell growth in 2′-FL-containing media (Fig. 2B). After 2′-FL hydrolysis inside the E. coli cell, L-fucose, a hexose, is metabolized into dihydroxyacetone phosphate and L-lactaldehyde by the sequential actions of a permease, an isomerase, a kinase, and an aldolase 30 . Aerobically, L-lactaldehyde is oxidized in two steps to pyruvate using NAD-dependent lactaldehyde dehydrogenase and flavin-linked lactate dehydrogenase, and thus channelling all carbons from L-fucose into the central metabolic pathways involved in ATP and amino acid synthesis 30 . Preadaptation to fucose is likely to accelerate cell growth and RFP production because this pathway is activated.
Lactose causes cytotoxicity in several ways. When lactose is transported into a cell that cannot metabolize lactose, the cellular membrane is damaged, and the membrane potential is disrupted by the so-called lactosekilling effect 34 . Lactose can induce cytoplasmic acidification as it is transported into the cytoplasm with a proton through the lactose permease. The resulting acidification of the cytoplasm induces cellular acid shock and reduces  www.nature.com/scientificreports/ proton motive force, intracellular ATP levels, and cell viability 32 . Our results show that cell preadaptation to fucose might allow to overcome lactose acidification. Because fucose is also transported with a proton, it is likely that the acid response system of E. coli can be activated during bacteria exposure to fucose 33 .
To the best of our knowledge, this is the first demonstration of a simple and easy quantification method of 2′-FL using a whole-cell biosensor (Table S3). Our biosensor might be applicable for high-throughput screening applications using 96-or 384-well microplates. It has the potential to improve process development, colony selection, quality management, 2′-FL kit development and other steps involved in the production 2′-FL.

Material and methods
Chemicals and materials. 2′-FL was purchased from AP Technology (Suwon, Korea). L-Fucose was purchased from Carbosynth (Compton, Berkshire, UK). Lactose, trace elements for Riesenberg medium (R medium) and antibiotics were purchased from Sigma-Aldrich (St. Louis, MO, USA).
Strains and plasmids. The list of strains and plasmids used in this study are listed in Table 1. The gene encoding α-l-fucosidase from Xanthomonas manihotis was synthesized from IDT (Coralville, IA, USA) 29,35 . The synthesized gene was cloned into the pET28b expression vector (Invitrogen, Carlsbad, CA, USA). Next, T7 promoter of pET28b was substituted with the J23100 promoter (BBa_J23100, https ://parts .igem.org/Promo ters/ Catal og/Ander son) to construct pConFUC. pBbA5aRFP was purchased from Addgene (Addgene_35280) 36 . The lac UV5 promoter of pBbA5aRFP plasmid was also substituted with the J23100 promoter to construct pConRFP that constitutively expressed RFP. To construct pET-ConFUC, the kanamycin resistance gene of pConFUC was substituted for the ampicillin resistance gene amplified from pETduet (Novagen). pColA-ConRFP was constructed by inserting expression regions from pConRFP (Promoter-RBS-CDS) into pColAduet (Novagen).  www.nature.com/scientificreports/ tryptone, 0.5% yeast extract, and 1% sodium chloride). When the optical density (OD) at 600 nm reached 0.5, culture was proceeded at 16 °C and 120 rpm for 16 h. Cells were harvested and the pellet was disrupted using an ultrasonic processor (20% amplitude, cycles of 1 s ON and 1 s OFF for 1 m 30 s, on ice). Cell debris and insoluble proteins were removed by centrifugation at 15,000 × g for 1 h. The supernatant was collected and concentrated using Centricon 10 (Amicon Co., Beverly, MA, USA). A 5 g/L of concentrated supernatant was mixed with 2′-FL in R medium. The reaction was performed at 37 °C in the shaking incubator. The reaction mixture was centrifuged at 15,000 × g for 10 min, and the supernatant was filtered using 0.45 μm syringe filter. Analytical method. Cell concentration was measured by observing the OD at 600 nm. The fluorescence intensity of RFP was measured using the excitation at 584 nm and emission at 615 nm. The OD and fluorescence intensity were measured using the spectrophotometer (Spectramax M2, Union City, CA, USA). Obtained values in the presence of 2′-FL were divided by the values measured in the absence of 2′-FL to calculate relative optical density (ROD) and relative fluorescence unit (RFU). Concentrations of lactose, L-fucose, and 2′-FL after 2′-FL cleavage were measured by using HPLC (Waters Corporation, Milford, MA, USA) equipped with the Rezex ROA-Organic Acid H + column (Phenomenex, Torrance, CA, USA) and a refractive index (RI) detector. A quantity of 0.01 N H 2 SO 4 was used as the mobile phase at a flow rate of 0.6 ml/min and 50 °C 16 . All experiments were performed in triplicate.

Statistics.
No statistical method was used to determine the sample size in advance. No random method or blind test was used in the experiment and interpretation of results. Numerical analyses were performed using Graph Pad Prism (Graph Pad Software, San Diego, CA, USA).

Data availability
The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.