To construct the biosensor, a deficient strain is required, where 7-MX needs to be essential for its survival. By investigating the metabolic pathways related to xanthine, a downstream metabolite of 7-MX in Escherichia coli (E. coli )(Figure 1), it was determined that the cell’s DNA and RNA precursors are synthesized through two pathways. The primary pathway involves the guaB gene, while the secondary pathway utilizes xanthine as a substrate and is mediated by the gpt gene. Therefore, knockout of the guaB gene to disrupt the primary pathway will activate the secondary pathway and create the desired deficient strain.

Figure 1 Xanthine metabolism chart

To achieve this, the RED homologous recombination method was used to knock out the guaB gene and replace it with a kanamycin (Kana) resistance gene for subsequent experiments.Detailed experimental procedures can be found on our experiment page.

First, we constructed the BW/pKD46 electrocompetent strain and utilized PCR to amplify a homologous recombination fragment containing a resistance marker (Kana) to delete the guaB gene, resulting in the BW-ΔguaB strain. Meanwhile, a targeting fragment was prepared by amplifying approximate 500 bp sequences flanking the guaB gene through PCR.

Next, the assembled fragment was electroporated into the induced BW-ΔguaB electrocompetent cells to achieve gene knockout.The resulting knockout strains were then subjected to colony PCR screening , with original strains used as negative controls. The results are as follows (Figure 2).

Figure 2 Colony PCR to screen for the knockout clones using the upper and lower homology arms as primers

Here, the length of the Kana resistance gene used for replacement is 1600 bp, while the length of the guaB gene to be knocked out is 2000 bp. Preliminary evidence indicated successful knockout, followed by sequence analysis of the fragment and testing under selective culture conditions, with results consistent with expectations.

Summary：These results showed that the knockout process of guaB gene was smooth and the strain performed well under selective culture conditions, which laid a foundation for subsequent experiments.

Successful knockout strains were isolated and subjected to single-colony expansion to produce competent BW25113-ΔguaB cells, which are now xanthine-dependent strains.Subsequently,based on previous literature[1] the ndm gene family responsible for methylxanthine degradation was utilized to design the pYB1s-ndmDCE plasmid (pDCE) (Figure 3).

This plasmid carrys a system that degrades 7-MX into xanthine.

Figure 3 Plasmid pYB1s-ndmDCE

Using the Golden Gate method, the PCR-amplified ndmC, ndmD, and ndmE coding sequences were ligated into the pYB1s vector. Subsequently, colony PCR was performed using ndmC-F and ndmE-R primers to screen for the successful insertion of these genes. The results are shown in Figure 4.

Figure 4 Results of colony PCR screening validation

The lengths of the PCR products were as expected, indicating the pDCE was successfully constructed.

The plasmid was then transformed into competent BW25113-ΔguaB cells, which can allow these cells to utilize 7-MX for DNA and RNA synthesis, and thus make 7-MX essential for their survival. Consequently, this process converted the strain into a sensor to detect 7-MX levels.

Thereby preliminarily constructing a biosensor. We then proceeded to test the functionality of the sensor, focusing on two key aspects: substrate specificity and substrate concentration sensitivity.

We first examined the substrate specificity by dissolving seven types of methylxanthines and xanthines in M9 medium. During the induction of the bacterial cells, to enhance bacterial proliferation, we used ZY medium supplemented with double the amount of yeast extract and LB medium. In the subsequent steps of transferring the cells, the induced bacterial cells were washed with water to remove the yeast extract from the ZY medium, which contains xanthine, and were then washed and resuspended in M9 medium. Finally, various methylxanthine compounds were individually added to a 96-well plate, followed by the addition of 200 μL of the bacterial suspension to each well. The growth curves were monitored using a microplate reader. The results are shown in Figure 5.

Figure 5 OD value of the strain at 36 h

The growth curve results obtained from the microplate reader indicate that the BW25113-ΔguaB strain transformed with the pDCE exhibited the expected dependence on 7-MX and xanthine. Specifically, when the strain was cultured in medium containing 7-MX, its growth rate increased significantly, suggesting that 7-MX serves as an essential compound for this strain. Furthermore, by testing the effects of other methylxanthine compounds, we found that the biosensor exhibited a high specificity for 7-MX, as other methylxanthines did not show the same growth-promoting effect.

Thus, the substrate specificity of the sensor has been successfully confirmed.

Next, we tested the concentration gradient of 7-MX. Using the same method as before, we dissolved 7-MX in the M9 medium at concentrations ranging from 0.1 to 0.5 mM/L. Following the same procedure, we used a microplate reader to monitor the growth curves. The results are shown below (Figure 7).

Figure 7 Growth curves of experimental bacteria in different concentrations of 7-MX in an enzyme labeling apparatus over a period of 36 h

The experimental results demonstrate that the biosensor exhibits good sensitivity to changes in 7-methylxanthine (7-MX) concentration. As the 7-MX concentration increases, the growth rate of the cells shows a clear dependence. Under low 7-MX concentration (0.1 mM/L), the cell growth rate is relatively slow, but as the concentration rises (to 0.3 mM/L and 0.5 mM/L), cell growth accelerates significantly, indicating that the strain's demand for and response to 7-MX strengthens with increasing concentration.

Summary: These experiments confirm that the biosensor exhibits the expected substrate specificity and concentration sensitivity. 7-MX can indeed be accurately detected by the sensor and serves as an essential factor for cell growth, providing a solid foundation for further experimental applications.

We conducted site-directed mutagenesis on five key predicted sites for both NdmA and NdmB, as reported in the literature. Corresponding degenerate primers were designed to perform site-directed random mutagenesis, thereby constructing mutant libraries for these two genes.

For the first round of screening and validation (Figure 7), the products of the directed mutations underwent DNA purification and DpnI digestion to minimize the background caused by the original template plasmid. The mutant fragments were then ligated to the vector and transformed into DH5α competent cells. After overnight cultivation, the monoclonal colonies on the plates were washed off and pooled using sterile water. Plasmids were then extracted from the mixture and transformed into BW25113 competent cells. The transformation products were amplified by shaking incubation, and single colonies were isolated using a dilution plating method.After incubating the plates for a period of time, single colonies were picked and inoculated into deep-well plates for further cultivation. Caffeine substrate was added to the deep-well plates to induce expression and perform whole-cell catalysis. Subsequently, 200 μL of the transformation solution and 10 μL of the induced biosensor strain were transferred into a 96-well plate. A microplate reader was then used to detect whether the mutated strain could produce 7-MX, allowing the biosensor strain to grow normally, thus validating the function of the mutations.

Figure 8 The overall process of mutation screening

Subsequent screening of the mutated strains revealed that the NdmA mutants did not meet expectations, although a few NdmB mutants showed promising results (Figure 9). Consequently, we decided to continue further mutational screening of the ndmB gene to obtain strains more aligned with the desired outcomes (Figure 8).

Figure 9 Growth of mutant bacteria detected by microplate reader

In the second round of mutations, to clarify the key active sites, the five critical pocket residues in NdmB were divided into two groups based on their proximity. One group consisted of two closely located sites (Gln289 and Leu293), while the other group included three more dispersed sites (Trp256, Cys267, and Met271). Site-directed mutagenesis was performed on both groups to obtain mutants from random combinations of mutations at these sites (Figure 10).

Figure 10 The five key pocket residues in NdmB

The screened mutant strains underwent a second round of validation and screening.

After recombining the mutant plasmids, we first transformed them into DH5α. Single colonies were grown, and the plasmids were extracted and subsequently transformed into BW25113 and plated. Single colonies were then picked and cultured in deep-well plates with 800 μL of ZY medium containing Ara, IPTG and S50, and induced at 25°C for 18 h. The deep-well plates were centrifuged at 4200 rpm for 10 min, and the supernatant was discarded. Then, 400 μL of 1× M9 medium supplemented with 0.5 mM caffeine (representing a 2x concentration of the cell mass) was added. Whole-cell catalysis was performed at 25°C for 18 h, followed by centrifugation. 200 μL of the supernatant was transferred into a 96-well plate, and 10 μL of a supplement solution (containing ZYM5052, calcium chloride, magnesium chloride, trace elements, inducers, and antibiotics) was added. Another 10 μL of pre-induced pDCE (using double the amount of yeast extract) was added (as pDCE may grow slowly, the 96-well plate was sealed in a plastic bag and stored at 4°C until pDCE induction was complete).

The OD values of each strain were measured using a microplate reader (Figure 11).

Figure 11 OD value for 24 h of growth

Mutant strains with good growth (OD ≥ 1.5) were selected for HPLC analysis to determine whether any products were produced (Figure 12).

Figure 12 The well-growing mutant strains were detected by HPLC

The strains producing only 7-MX with the highest yield were identified as target mutant strains. A selected highest 7-MX-producing strain was then sequenced, and the sequencing results revealed that the NdmB protein had only one single mutation at the position 289, where glutamine was substituted by alanine, while other minor mutations were synonymous. This suggests that the 289 position is a critical for the catalytic activity of NdmB. At this stage, we cannot exclude the possibility that mutating this position to other amino acids could yield strains with higher 7-MX production yields.

To explore this hypothesis, a site-directed saturation mutagenesis was performed at position 289, followed by the same cultivation and screening process as described above. Ultimately, the results showed that the strain with the highest yield remained the one with the Q289A (glutamine to alanine) mutation, confirming its optimal performance.

Initially, a gradient experiment to test substrate concentration was performed. It was anticipated that higher substrate concentrations would lead to better 7-MX yield. Using a biomass of 50 OD, caffeine at a gradient of concentrations was used as the substrate for the whole-cell catalysis, and the product yields were measured by HPLC. The results are shown in Figure 14.

Figure 14 Bar chart of the effects of different substrate concentrations on 7-MX production with experimental strains under 50 OD conditions

Based on the results, we concluded that, under the 50 OD biomass condition, a substrate concentration of 4 mM allowed complete conversion of caffeine, with a 7-MX yield of approximate 3.5 mM. However, there was an accumulation of the intermediate product PX. When the substrate concentration exceeded 4 mM, the results deviated from the expected outcome. Although caffeine was almost entirely converted, the yield of 7-MX did not increase significantly; instead, the accumulation of PX continued to rise.

Based on these data, the maximum yield of 7-MX under 50 OD biomass conditions is limited to 4 mM, suggesting that biomass might be the primary limiting factor to improve 7-MX production.

Considering that a 10 mM substrate concentration can still be significantly converted and that the accumulation of intermediate products is high, we conducted the subsequent experiment using a 10 mM substrate to explore whether biomass limitations affect 7-MX production. The results are shown in Figure 15.

Figure 15 The bar chart of the effects of different OD value on 7-MX production with experimental strains under 10mM caffeine

The HPLC results fully met the experimental expectations. Increasing the biomass directly enhanced the molar ratio of 7-MX in the final system. Ultimately, under conditions of 100 OD biomass and 10 mM substrate concentration, a conversion rate of 90% from caffeine to 7-MX was achieved. As biomass increased, the substrate conversion rate was noticeably improved. However, if the biomass is increased before the complete conversion of caffeine, there is a significant accumulation of the intermediate product PX. It is only after the complete consumption of caffeine that increasing biomass helps reduce the accumulation of PX.

Given the substantial accumulation of the intermediate product PX, we hypothesized that the conversion of PX to 7-MX is the rate-limiting step. To validate this hypothesis, timed sampling during the whole-cell catalysis process was conducted to analyze the changes in the concentrations of different substances over time, and corresponding catalysis time curves were plotted (Figure 16).

Figure 16 Time curve of the production of 7-MX with experimental strains at 10mM caffeine, 50OD

Under conditions of 100 OD biomass and 10 mM caffeine substrate, the whole-cell catalysis process for 18 h achieved complete conversion of caffeine within 4 h. Subsequently, the intermediate product PX was slowly converted into the product 7-MX. However, the conversion rate significantly decreased over time, possibly due to the rapid inactivation of the enzyme. Due to time constraints, further experiments to verify changes in enzyme activity during this process were not conducted. Investigating this in more detail could potentially optimize the procedure to further increase the yield of 7-MX.

In the sensor screening phase, three strains were identified with mutations at the position 289, where the amino acids were mutated to Thr, Ser, and Ala, respectively. After optimizing the whole cell catalysis conditions, the yields of these three strains were re-evaluated (Figure 17).

Figure 17 Different mutated strains were detected by HPLC

Using the same conditions of 100 OD biomass and 10 mM caffeine substrate, the results still indicate that the NdmB Q289A strain achieves the highest yield of 7-MX. Interestingly, an additional observation was made: the original NdmB enzyme only removes the methyl group at the N3 position, converting caffeine into PX (the intermediate product observed in the experiment). However, with mutations leading to smaller R groups, while the enzyme's ability to remove the N3 methyl group remains effective, there is an increasing capability to remove the N1 methyl group. This suggests that the smaller R group at position 289 allows the substrate to enter the active site from different orientations, resulting in different outcomes. This finding raises a hypothesis that NdmB and NdmA may have originally been the same enzyme, which through a long evolutionary process, gradually diverged in their functions.

Initially, coffee grounds were collected from various coffee shops, such as Starbucks, Luckin Coffee, Kudi, and 818(Figure 16). During the collection process, inquiries were made regarding the disposal methods for coffee grounds. As anticipated, the majority of coffee shops opted to discard the coffee grounds directly.

Figure 18 Team members go to a coffee shop to get coffee grounds

Figure 19 Different brands of coffee grounds obtained

Upon observing the collected coffee grounds, it was noted that the grounds from Starbucks and Luckin Coffee were initially processed into clumps, while those from other sources remained unprocessed. The coffee grounds were then spread out in foam trays and placed in an oven set at 60°C to dry for 48 h, resulting in dehydrated coffee grounds (Figure 19).

Base on previous literature, we attempted to extract caffeine from the coffee grounds. Gradient dilution extraction was first performed on the coffee grounds from Luckin Coffee and Starbucks to extract caffeine(Figure 20).

Figure 20 Extract caffeine from Luckin and Starbucks coffee grounds

However, according to the liquid chromatography data, it was concluded that the caffeine content in each gram of coffee grounds is between 90-100%. This significantly differs from the theoretical values initially reviewed, which should be in the range of 0.5-1.5% [2](Figure 21).

Figure 21 Bar chart of the concentration of caffeine extracted from Luckin and Starbucks coffee grounds

Since the results did not meet expectations, an analysis of the process was conducted. We predicted that during the ultrasonic extraction of caffeine, the ultrasonic waves might have caused partial decomposition of substances in the coffee grounds (such as lipids and proteins), leading to inaccurate measurement. Additionally, insufficient dilution might have resulted in incomplete extraction. Therefore, we decided to increase the dilution factor while simultaneously use heating method to improve caffeine extraction.

For each brand of coffee grounds, 5 grams were weighed and dissolved in 100 mL of water. The mixture was then heated in a microwave until boiling, which was maintained for five times. After samples cooling down, the solution for each sample was filtered using filter paper. The filtered liquid was then analyzed by HPLC. The filtered solution was passed through the 0.22μm membrane one more time to prevent the presence of any solid residues to clog the HPLC chromatographic column (Figure 22).

Figure 22 The overall process of extracting caffeine from coffee grounds

The final results indicate that the coffee grounds content of the four brands falls within the range of 0.5–1.5%. Furthermore, using the extraction method referenced in the literature, our results, after accounting for the loss of caffeine, are largely consistent with the reported value of 2 milligrams of caffeine per gram of coffee grounds (Figure 23).

Figure 23 Bar chart of the concentration of caffeine extracted from 4 different brands

After measuring the caffeine content in the coffee grounds, we decided to directly use the coffee grounds to prepare a whole-cell catalysis reaction mixture. Subsequently, the optimized strains were employed to utilize this reaction mixture derived from the coffee grounds for the production of 7-MX.

Figure 24 The whole cell catalyzed conversion solution was made from 4 different brands of coffee grounds

The extraction was carried out using 10 mL centrifuge tubes. Each brand of coffee grounds was weighed at 2 grams and dissolved in 5 mL of Tris buffer (Figure 24). Due to the small size of the container and the risk of boiling over, a water bath heating method was used(Figure 25a and b) The centrifuge tubes were placed in boiling water and heated for 10 min. After cooling, the mixture was centrifuged at 4000 rpm for 10 min. The supernatant was then transferred to a 2 mL Eppendorf tube and subjected to a second round of centrifugation at 15,000 g for 10 min (Figure 25c).

Figure 25.a Water bath heating fixture device .b Integrated device for water bath heating method.c Procedures before 18 h fermentation.d Mix the NdmB Q289A strain with transformation liquid containing caffeine.

Then, whole-cell catalysis was performed using the induced NdmB Q289A strain. One milliliter of the transformation liquid was added to 100 mL of the culture medium with 100 OD of the bacteria in a flask.(Figure 25d)The whole-cell catalysis was conducted under the optimized conditions for 18 h. Starting from 0 h, a 50 μL aliquot of samples was collected every 2 h, diluted by a factor of two, and prepared for HPLC analysis to generate the whole-cell catalysis time curve (Figure 26).

Figure 26 The fermentation curves with 100 OD NdmB Q289A over 18 h

According to the HPLC results, the outcomes fully align with the experimental expectations, showing that, from 0 h with no 7-MX present, caffeine was almost entirely converted into 7-MX and PX by the end. However, the conversion rate of caffeine to 7-MX was lower than that using the pure caffeine. This suggests that some components in the coffee grounds may interfere with the enzyme's activity.

[1] Kim, J. H., Kim, B. H., Brooks, S., Kang, S. Y., Summers, R. M., & Song, H. K. (2019). Structural and Mechanistic Insights into Caffeine Degradation by the Bacterial N-Demethylase Complex. Journal of molecular biology, 431(19), 3647–3661. https://doi.org/10.1016/j.jmb.2019.08.004

[2] Vandeponseele, Alexandre, Micheline Draye, Christine Piot, and Gregory Chatel (2021). Study of Influential Parameters of the Caffeine Extraction from Spent Coffee Grounds: From Brewing Coffee Method to the Waste Treatment Conditions. Clean Technologies 3, no. 2: 335-350.

https://doi.org/10.3390/cleantechnol3020019