Overview

The application of spent coffee grounds (SCG) as organic fertilizer is limited due to the phytotoxicity of residual caffeine. To address this challenge, we engineered E.coli BL21(DE3) to express the caffeine-degrading ndm gene cluster (ndmA–E) from P.putida and tested the caffeine degradation in Cycle 1. Due to its suboptimal degradation efficiency, in Cycle 2, we also evaluated E.coli DH5α whose degradation activity was significantly enhanced despite the absence of visible protein bands on SDS-PAGE. In Cycle 3, further optimization of fermentation conditions, including temperature (28 °C), duration (48 h), and medium composition (1 g/L corn starch, 2 g/L NaNO₃, and optimized inorganic salts) resulted in approximately 90% caffeine removal from sterile SCG. Taken together, this project establishes an efficient and scalable biological method for detoxifying SCG, demonstrating great potential for the sustainable production of high-value organic fertilizers.

Cycle 1: Testing the experimental design

Design 1.0

Expression plasmids for NdmA, NdmB, NdmC, NdmD and NdmE

Our project aimed to address the challenges of caffeine phytotoxicity in SCG through designing a biological decaffeination system. The enzymes, N-demethylases (Ndm; including NdmA, NdmB, NdmC, NdmD and NdmE), encoded by P.putida can degrade phytotoxic caffeine into neutral xanthine via sequential demethylations (Summers et al., 2012; Summers et al., 2011; Summers et al., 2013) (See Figure 1). Due to the unsuitability of P.putida for fermentation, we intended to express these ndm genes in E.coli, our chassis organism, through a synthetic biology method (Kallio et al., 2014).

Figure 1. N-demethylases-dependent sequential demethylation. Phytotoxic caffeine is degraded through the stepwise removal of methyl groups by the specific enzymes NdmA, NdmB, NdmC, NdmD, and NdmE, originally identified in P.putida. This enzymatic pathway converts caffeine into theobromine, then to paraxanthine, and finally to non-toxic xanthine, effectively eliminating its plant-growth inhibitory effects. Our project involves expressing these ndm genes in E.coli to enable biological decaffeination of SCG. The caffeine-xanthine metabolic pathway is adapted from a previous review (Summers et al., 2015).

The coding sequences of ndmA, ndmB, ndmC, ndmD and ndmE genes were commercially synthesized and used as PCR templates. Their specific primers were designed to incorporate a ribosome binding site (RBS) and/or a restriction enzyme recognition site (EcoRI or SalI) at the 5' end of each gene. Following PCR amplification, the products were cloned into a modified pET28a expression vector, which was selected for its advantageous features, including: (i) low-copy origin of replication, (ii) kanamycin resistance for selection, and (iii) a strong T7 promoter system for inducible expression. To eliminate potential interference with Ndm protein structure or function, all affinity tag (His-tag, thrombin cleavage site, and T7 epitope) sequences were removed from the vector backbone. To place ndm genes in the multiple cloning site (MCS; between EcoRI and SalI) of the expression vector, the amplified gene fragments and linearized vector were subjected to double restriction digestion (EcoRI/SalI), followed by ligation to yield expression constructs for ndm genes (See Figure 2).

Figure 2. Schematic of plasmid construction and vector map. (A) Cloning strategy for the insertion of ndm genes into the modified pET28a vector. Coding sequences were amplified with primers introducing EcoRI and SalI sites, followed by restriction digestion and ligation. (B) Map of the modified pET28a expression vector.

Given that NdmD, a reductase with a Rieske [2Fe-2S] cluster, a plant-type [2Fe-2S] cluster, and a flavin mononucleotide (FMN) domain, is involved in the first step for assimilating the carbon and nitrogen in caffeine, its activity is essential for transferring electrons from NADH to their substrates. NdmA and NdmB, containing their independent [2Fe-2S] clusters and Mo cofactors, are monooxygenases with N1- and N3-specific demethylation activity, respectively(Summers et al., 2012; Summers et al., 2011), whereas NdmE containing the FMN domain and NdmC (with N7-specific demethylation activity) containing Mo cofactor, forms a complex to cooperatively transfer electrons(Summers et al., 2013).

Initially, we proposed to place all 5 ndm genes within a single pET28a expression vector, with each gene separated by a ribosome binding site (RBS) and dual 6-bp spacer sequences. However, preliminary assessment suggested a low probability of achieving functional expression of all gene products in a single E.coli host through this polycistronic design, due to the attenuated translation along the extended transcript and metabolic burden from concurrent expression of multiple heterologous proteins. Finally, we revised our gene arrangement strategy as follows (See Figure 3).

Figure 3. Design of gene circuits of Ndm expression. Three distinct polycistronic constructs were assembled: (A) ndmD-ndmA for N1-demethylation, (B) ndmD-ndmB for N3-demethylation, and (C) ndmD-ndmC-ndmE for N7-demethylation. Each gene is preceded by a ribosome binding site (RBS: TTTGTTTAACTTTAAGAAGGAGA) and flanked by 6-bp spacer sequences (AATAAT).

Build 1.0

Inducible expression of ndm genes

Plasmids pET28a-NdmDA, pET28a-NdmDB, and pET28a-NdmDCE were transformed into E.coli BL21(DE3). After their inducible expressions, we performed SDS-polyacrylamid gel electrophoresis (SDS-PAGE) and Coomassie staining. Our result revealed five robust bands whose sizes were around ~40-kDa, ~41-kDa, ~32-kDa, ~65-kDa and ~26-kDa, were detected upon IPTG induction, corresponding to NdmA, NdmB, NdmC, NdmD and NdmE, respectively (See Figure 4), verifying that our gene circuit designs were successfully achieved.

Figure 4. Recombinant protein expression in BL21(DE3). SDS-PAGE analysis of total protein from BL21(DE3) strains harboring the indicated Ndm constructs, with or without IPTG induction. Arrows indicate the positions of the overexpressed Ndm proteins with their predicted molecular masses. M: Protein molecular weight marker.

Test 1.0

1. Preliminary fermentation

To verify whether caffeine affects the E.coli growth and the expression of the ndm genes in BL21(DE3), we added standard caffeine (400 μg/mL) to IPTG-induced samples for a 12 h incubation and then monitored the E.coli growth by measuring optical density at 600 nm (OD₆₀₀) over time and performed SDS-PAGE. Here, a caffeine concentration exceeding 500 μg/mL has been reported to significantly inhibit cellular growth and metabolic activity and this concentration falls within the linear detection range of ultraviolet (UV) spectrophotometry at OD₂₇₄, the way we subsequently quantified the residual caffeine content, ensuring accurate and reproducible quantification(Newton, 1979). The growth curve of E.coli BL21(DE3) revealed that although BL21(DE3) exhibited a delayed stationary phase, its final concentration was not significantly altered (See Figure 5). Besides, the SDS-PAGE results showed no significant difference in the Ndm protein bands between the caffeine-treated group and the untreated control (See Figure 6). Therefore, we assumed that the presence of caffeine did not affect the E.coli growth and the expression of the ndm genes in BL21(DE3).

Figure 5. Effect of caffeine on E.coli BL21(DE3) growth. Growth curves of E.coli BL21(DE3) strain cultured in the presence or absence of 400 μg/mL caffeine, monitored by measuring OD₆₀₀ over time.

Figure 6. Effect of caffeine on the expression of Ndm proteins in BL21(DE3). SDS-PAGE analysis of total protein from BL21(DE3) strains harboring the indicated Ndm constructs, cultured in the presence or absence of 400 μg/mL caffeine for 12 h.

2. Caffeine detection

To determine whether exogenously expressed ndm genes enhance caffeine degradation, we supplemented IPTG-induced fermentation broth with standard caffeine for a 24 h fermentation, followed by quantification of caffeine content using ultraviolet (UV) spectrophotometry. Notably, metal ions, proteins and some other interfering substances in the broth were removed by treatment with zinc acetate and potassium ferrocyanide solutions, and their supernatant was stabilized with sodium sulfite-potassium thiocyanate solution, prior to the spectrophotometry detection. Compared to the original E.coli bacteria, BL21(DE3) expressing NdmDA, NdmDB, and NdmDCE showed a moderate reduction in caffeine content. A further reduction was observed in the BL21(DE3) expressing the full set of NdmA-E genes (See Figure 7).

Figure 7. Caffeine degradation by engineered BL21(DE3) strains. Relative concentration (measured by OD₂₇₄) of residual caffeine after 24 h of fermentation in strains harboring the indicated Ndm constructs. The control strain harbors an empty vector (EV). Data are presented as mean ± SD (standard deviation) of three biological replicates (n=3).

These data are consistent with the UV absorption characteristics of xanthine derivatives. Their maximum absorption wavelengths (λmax, in nm) are as follows: 274 nm for caffeine, 272 nm for theobromine (product of NdmDA), 272 nm for paraxanthine (product of NdmDB), 278 nm for 7-Methylxanthine (product of NdmDCE), and further 267 nm for Xanthine (product of NdmA-E). Particularly, xanthine may exhibit an additional shoulder peak or strong absorption around 245~250 nm under acidic culture conditions. Therefore, the measured values from the NdmA-E, which demonstrated the lowest caffeine levels, likely reflected a more bona fide quantification. Furthermore, NdmA-E may facilitate the demethylation reaction step by step, thereby enhancing overall caffeine degradation and resulting in a lower residual caffeine content. Taken together, the expression of ndm genes promote the caffeine degradation to some extent.

Learning 1.0

Although the Ndm proteins were over expressed in E.coli BL21(DE3), their caffeine degradation efficiency did not meet our expectation. This discrepancy may result from the theory that the Ndm proteins may lack correct folding or form insoluble aggregates (e.g., inclusion bodies) during their rapid expressions, leading to a reduced enzymatic activity. Nevertheless, the underlying mechanism remains to be better clarified.

Cycle 2: Evaluating bacterial strains

Design 2.0

Considering that the suboptimal caffeine degradation efficiency observed in E.coli BL21(DE3) despite successful protein expression, we hypothesized that the high-level expression of Ndm proteins might lead to misfolding or inclusion body formation, thereby reducing enzymatic activity. To address this, we evaluated an alternative host strain, E.coli DH5α, which had been previously transformed with the Ndm expression constructs during plasmid construction.

Build 2.0

The expression plasmids and the fermentation protocol remained identical to those used in Cycle 1 to ensure comparability.

Test 2.0

According to the established protocol, we determined whether caffeine affects the DH5α growth. Like BL21(DE3) (See Figure 5), DH5α exhibited a delayed stationary phase, and their final concentrations were not significantly altered (See Figure 8).

Figure 8. Effect of caffeine on E.coli DH5α growth. Growth curves of E.coli DH5α strain cultured in the presence or absence of 400 μg/mL caffeine, monitored by measuring OD₆₀₀ over time.

Caffeine degradation was quantified using UV spectrophotometry (OD₂₇₄) following the established protocol. Control strains without ndm genes showed negligible caffeine degradation (mean OD₂₇₄ ≈ 0.97 and 0.96, respectively), indicating that the original DH5α and BL21(DE3) were not capable of degrading caffeine. Also, BL21(DE3) expressing NdmA-E reduced caffeine content by approximately 35%, consistent with Test 1.0 (See Figure 7). Notably, DH5α expressing NdmA-E achieved a 60% reduction in caffeine content compared to the control (See Figure 9), indicative of a significantly higher degradation efficiency despite the absence of visible protein bands on SDS-PAGE (See Figure 10).

Figure 9. Comparative analysis of caffeine degradation between BL21(DE3) and DH5α. Relative concentration of residual caffeine after fermentation with BL21(DE3) or DH5α strains harboring the indicated Ndm constructs.

This discrepancy suggests that low-level expression in DH5α may facilitate proper protein folding and functionality, while high-level induction in BL21(DE3) could lead to a reduced enzymatic activity. This discrepancy was also observed in some research (Al-Janabi et al., 2022) and the former iGEM project (2024 Hangzhou-SDG Team ID: #5990).

Figure 10. Total proteins from engineered DH5α and BL21(DE3). Protein profiles of DH5α or BL21(DE3) strains harboring the indicated Ndm constructs.

Learning 2.0

The pET28a vector, though facilitating high-level expression in BL21(DE3), may impose substantial metabolic stress. In contrast, “leaky expression” of DH5α is sufficient for efficient caffeine degradation, though not optimized for protein expression. This aligns with prior reports that moderate expression can enhance enzymatic activity by avoiding inclusion body formation and metabolic overload. Further validation, e.g., fine-tuning protein purification or high-performance liquid chromatography (HPLC), is recommended to confirm functional enzyme expression in DH5α.

Cycle 3: Optimizing the fermentation conditions

Design 3.0

To maximize caffeine degradation efficiency in E.coli DH5α expressing NdmA-E, we systematically optimized these key fermentation parameters, temperature, duration, and medium composition (carbon source, nitrogen source, and inorganic salts).

Build 3.0

The expression plasmids remained identical to those used in Cycle 1. The fermentation was conducted in M9 medium (minimal medium for E.coli) supplemented with various components. Bacterial growth (OD₆₀₀) and residual caffeine (OD₂₇₄) were monitored to assess the optimization effects.

Test 3.0

1. Temperature

The caffeine degradation was evaluated at temperatures ranging from 16°C to 36°C. Maximum degradation was observed at 28°C (See Figure 11), which was consequently identified as the optimal temperature for this process. Notably, a pronounced reduction in caffeine degradation in engineered DH5α occurred when the temperature exceeded 32°C, consistent with a previously reported pattern where the optimal temperature for caffeine degradation in native P.putida is observed at 30°C (Fan FangYuan et al., 2011) . Given that E.coli exhibits maximal metabolic activity at 37°C, this shift (37 to 28°C) likely represents a compromise between the metabolic efficiency of E.coli (37°C) and the functional constraints of the P.putida-derived Ndm enzymes (30°C). This temperature (28°C) may preserve enzyme stability

Figure 11. Optimization of fermentation temperature. Relative concentration of residual caffeine in cultures of DH5α harboring the NdmA-E construct after 24 h of fermentation at the indicated temperatures.

2. Duration

Time-course analysis demonstrated that caffeine degradation reached a plateau after 48 h of fermentation at 28°C (See Figure 12), indicating no significant further reduction in caffeine concentration with extended incubation beyond this point. Notably, this duration is shorter than that typically required for maximal degradation in the native host P.putida, suggesting that the expression system in E.coli may facilitate enzymatic processing of caffeine under the applied conditions. This accelerated degradation kinetics could be attributed to higher intracellular enzyme concentrations, optimized cofactor availability, or reduced metabolic competition in the recombinant strain compared to the native organism.

Figure 12. Optimization of fermentation duration. Relative concentration of residual caffeine in cultures of DH5α harboring the NdmA-E construct after indicated time of fermentation at 28°C.

3. Medium Composition

Caffeine exhibits a growth-inhibitory effect on E.coli, particularly when employed as the sole carbon and nitrogen source (Lin et al., 2023; Summers et al., 2012). Under such conditions, caffeine not only fails to support robust microbial growth but also perturbs the normal physiological cycle of the engineered strain, thereby diminishing overall fermentation efficiency. To mitigate this limitation and achieve an optimal balance between bacterial growth and caffeine degradation, we systematically evaluated the effects of supplementary nutrients. M9 minimal medium was augmented with graded concentrations of alternative carbon sources, nitrogen sources, and essential inorganic salts. Bacterial growth was quantified via optical density at 600 nm (OD₆₀₀), and residual caffeine was monitored spectrophotometrically at 274 nm (OD₂₇₄) to assess degradation efficiency.

Carbon Source: Corn starch (1 g/L) showed the best degradation efficiency (See Figure 13).

Nitrogen Source: NaNO₃ (2 g/L) proved most effective (See Figure 14).

Inorganic Salts: Optimal concentrations were determined as ZnSO₄ (0.1 g/L), FeSO₄ (0.1 g/L), and MgSO₄ (0.5 g/L) (See Figure 15).

Carbon Sources

Figure 13. Effect of carbon source supplement on bacterial growth and caffeine degradation. Relative concentration of residual caffeine (OD₂₇₄) and final optical density (OD₆₀₀) after fermentation of DH5α harboring the NdmA-E construct in M9 minimal medium supplemented with different carbon sources (left) and corn starch of different concentrations (right).

Nitrogen sources

Figure 14. Effect of nitrogen source supplement on bacterial growth and caffeine degradation. Relative concentration of residual caffeine and final optical density after fermentation of DH5α harboring the NdmA-E construct in M9 minimal medium supplemented with different nitrogen sources (left) and NaNO3 of different concentrations (right).

Inorganic salts

Figure 15. Effect of inorganic salt supplement on bacterial growth and caffeine degradation.

Relative concentration of residual caffeine and final optical density after fermentation of DH5α harboring the NdmA-E construct in M9 minimal medium supplemented with (up left) ZnSO₄, (up right) FeSO₄, and (down) MgSO₄ of different concentrations.

4. Validation with Spent Coffee Ground

Under optimized conditions, NdmA-E-expressing DH5α achieved approximately 90% caffeine degradation in sterile SCG supplemented with LB medium and nutrients (See Figure 16), demonstrating practical applicability.

Figure 16. Validation of caffeine degradation in spent coffee grounds. Relative concentration of residual caffeine in sterile SCG after treatment with engineered DH5α (NdmA-E) under optimized fermentation conditions for 48 h. The control was treated with sterile medium only.

Learning 3.0

Optimized fermentation conditions significantly enhance caffeine degradation efficiency in E.coli DH5α. The use of corn starch and NaNO₃ as carbon and nitrogen sources, respectively, along with balanced inorganic salts, supports robust bacterial growth and enzymatic activity. The 90% degradation in SCG highlights the potential for industrial-scale application. Future work should focus on scaling up the process and evaluating the ecological safety of the resulting fertilizer.

Future work

Our future efforts will focus on four key areas to advance the technology toward practical application.

1. Enzyme Engineering and Expression Optimization

We will test alternative expression systems (e.g., pCold, pBAD) or different microbial hosts to better balance expression levels and enzymatic activity, potentially improving the overall efficiency and yield of the caffeine degradation pathway.

2. Fermentation Process Scale-up and Pilot Testing

The optimized conditions will be validated in a bioreactor, with careful control and monitoring of parameters such as pH and dissolved oxygen (DO), and the implementation of fed-batch strategies. We will also assess the system's performance in non-sterile spent coffee grounds to simulate real-world application environments.

3. Product Analysis and Ecological Safety Assessment

High-Performance Liquid Chromatography-Mass Spectrometry (HPLC-MS) will be employed to definitively identify and quantify degradation products, such as xanthine. Furthermore, seed germination assays will be conducted to verify the elimination of phytotoxicity in the treated SCG, ensuring its safety for use as a fertilizer.

4. Biosafety and Compliance

A thorough risk assessment will be conducted to evaluate the survival and potential spread of the engineered bacteria in open environments. To ensure biological safety, we will design and validate robust biocontainment systems, such as kill switches, to prevent the unintended persistence of genetically modified organisms.

References

[1] Al-Janabi, S. S., Shawky, H., El-Waseif, A. A., Farrag, A. A., Abdelghany, T. M., & El-Ghwas, D. E. (2022). Stable, efficient, and cost-effective system for the biosynthesis of recombinant bacterial cellulose in Escherichia coli DH5α platform. J Genet Eng Biotechnol, 20(1), 90. https://doi.org/10.1186/s43141-022-00384-7

[2] Fan FangYuan, F. F., Xu Yan, X. Y., Liang YueRong, L. Y., Zheng XinQiang, Z. X., Borthakur, D., & Lu JianLiang, L. J. (2011). Isolation and characterization of high caffeine-tolerant bacterium strains from the soil of tea garden.

[3] Kallio, P., Pásztor, A., Thiel, K., Akhtar, M. K., & Jones, P. R. (2014). An engineered pathway for the biosynthesis of renewable propane. Nature Communications, 5(1), 4731. https://doi.org/10.1038/ncomms5731

[4] Lin, Z., Wei, J., Hu, Y., Pi, D., Jiang, M., & Lang, T. (2023). Caffeine Synthesis and Its Mechanism and Application by Microbial Degradation, A Review. Foods, 12(14). https://doi.org/10.3390/foods12142721

[5] Newton, J. M. (1979). Spectrophotometric determination of caffeine in coffee products: collaborative study. J Assoc Off Anal Chem, 62(4), 705-708.

[6] Summers, R. M., Louie, T. M., Yu, C. L., Gakhar, L., Louie, K. C., & Subramanian, M. (2012). Novel, highly specific N-demethylases enable bacteria to live on caffeine and related purine alkaloids. J Bacteriol, 194(8), 2041-2049. https://doi.org/10.1128/jb.06637-11

[7] Summers, R. M., Louie, T. M., Yu, C. L., & Subramanian, M. (2011). Characterization of a broad-specificity non-haem iron N-demethylase from Pseudomonas putida CBB5 capable of utilizing several purine alkaloids as sole carbon and nitrogen source. Microbiology (Reading), 157(Pt 2), 583-592. https://doi.org/10.1099/mic.0.043612-0

[8] Summers, R. M., Mohanty, S. K., Gopishetty, S., & Subramanian, M. (2015). Genetic characterization of caffeine degradation by bacteria and its potential applications. Microb Biotechnol, 8(3), 369-378. https://doi.org/10.1111/1751-7915.12262

[9] Summers, R. M., Seffernick, J. L., Quandt, E. M., Yu, C. L., Barrick, J. E., & Subramanian, M. V. (2013). Caffeine junkie: an unprecedented glutathione S-transferase-dependent oxygenase required for caffeine degradation by Pseudomonas putida CBB5. J Bacteriol, 195(17), 3933-3939. https://doi.org/10.1128/jb.00585-13