Our team aimed to increase the uptake of xylose from lignocellulosic biomass in E. coli. Despite xylose being the second most abundant sugar in lignocellulose, due to carbon catabolite repression (CCR), E. coli preferentially metabolises glucose and inhibits xylose metabolism. Through implementing a yeast-derived XR-XDH pathway, overexpressing E. coli’s xylulose kinase, and adding phosphoketolase, we worked to improve the growth of our engineered E. coli on biomass.
Project Selection & Enzymatic Design
Our project was advised and guided by researchers who were from the 2017 Macquarie_Australia iGEM team and included Louise Brown, Robert Willows and Thi Hunyh. Their project, H2ydroGEM, engineered E. coli bacteria to produce hydrogen through photosynthesis and thus demonstrated the potential of synthetic biology in creating clean energy.
Then, we discovered that a significant portion of lignocellulose was not being utilised efficiently by the diauxic growth pattern of E. coli, limiting its rate of growth and hence reducing the efficiency of bio-energy production. Hence, our team pursued research on potential methods of pursuing an improved E. coli strain which would be capable of concurrent glucose-xylose co-utilisation. We came across a figure of all natural metabolic pathways of xylose in various micro-organisms within a scientific paper called Biochemical routes for uptake and conversion of xylose by microorganisms (Zhao, Z., Xian, M., Liu, M. et al.. Biotechnol Biofuels 13, 21 (2020). https://doi.org/10.1186/s13068-020-1662-x) which sparked our design process for improving the efficiency of E. coli’s xylose metabolism.
Firstly, we analysed the current processes that E. coli utilises to metabolise xylose. E. coli makes use of the XI pathway, in which xylose isomerase (XI) converts xylose to xylulose, then xylose kinase (XK) converts xylulose to xylulose-5-phosphate (X5P), which is then further broken down in the pentose phosphate pathway.
Then, we compared the XI pathway with other major xylose metabolic pathways present in different bacteria. We noticed that a similar pattern of metabolism occurred in the XR-XDH pathway of S. cerevisiae cells, which included an extra step of converting xylose to xylitol through xylose reductase (XR), then xylitol into xylulose through xylitol dehydrogenase (XDH). Otherwise, the conversion from xylulose to X5P and onwards was similar between both the XI and XR-XDH pathways.
The introduction of the XR-XDH pathway would provide multiple ways for xylose to be catabolised into X5P, increasing its xylose uptake. Hence, we identified the induction of the S. cerevisiae XR-XDH pathway in E. coli as a viable method to promote increased xylose uptake.
Furthermore, we hypothesised that, with the addition of another xylose pathway, a flux of X5P would be created. To alleviate this flux, we would also introduce the PK pathway by inducing the expression of phosphoketolase (PK) enzymes that convert X5P into glyceraldehyde-3-phosphate (G3P), which then leads to glycolysis.
We were encouraged by the 2021 Jiangnan_China team’s documentation of their improved part (BBa_K3803016) which combined XR, XDH, and XK gene expressions to induce a higher cell growth and production yield. We were further encouraged by various scientific literature articles which sought to tackle the diauxic growth issue through different approaches, ranging from increasing transmembrane transporter rates of XylE and XylFGH, modifying AraE as a xylose transporter, rewiring transporter preferences, and so on. We had not come across any instances where both the XR-XDH pathway and PK pathway were induced in E. coli, and wanted to pursue this combination.
Experimental Design
We planned to build our parts individually through synthesis, then transform them into pSB1C3 (Chloramphenicol-resistant) or pSB1K3 (Kanamycin-resistant) plasmids, which would allow us to test and measure the function of each part separately, then mix-and-match as we assemble parts, and troubleshoot issues at a smaller scope. Furthermore, synthesising multiple parts through a single synthesis is both difficult and expensive due to its long base pairs. We then planned to combine all parts into a single plasmid through assembly.
We planned to implement the design in competent E. coli DH5-alpha and E. coli NEB-alpha substrains. We had also initially planned to test the growth of unengineered S. cerevisiae and S. stipitis cells to observe the efficiency of the XR-XDH pathway to (a) compare it with the E. coli’s native XI pathway; and (b) determine whether our engineered E. coli outperforms existing yeast cells.
To test the individual functionality of our enzymes, we planned to observe and compare the growth curves (through OD600) of E. coli with and without IPTG induction on various types of M9 media - xylose, xylitol, and glucose - with either chloramphenicol or kanamycin as antibiotics. We could then test other combinations of parts, such as XR and XDH, and eventually test E. coli with all four enzymes induced. Finally, we plan to observe the growth rates of our engineered E. coli when growing on sugar extracted from lignocellulosic biomass (which has a variety of sugars, e.g. glucose, xylose, arabinose), to test a real-world viability of our project.
The below table outlines our expected experimental results for various enzymes on different types of media:
| Enzyme | Medium | Expected growth rate when compared against non-induction |
| Xylose reductase | Xylose | No significant change |
| Xylose reductase | Xylitol | Increase |
| Xylose reductase | Glucose | No significant change |
| Xylitol dehydrogenase | Xylose | No significant |
| Xylitol dehydrogenase | Xylitol | Increase |
| Xylitol dehydrogenase | Glucose | No significant change |
| Xylulose kinase | Xylose | Slight increase |
| Xylulose kinase | Xylitol | No significant change |
| Xylulose kinase | Glucose | No significant change |
| Phosphoketolase | Xylose | Slight increase |
| Phosphoketolase | Xylitol | No significant change |
| Phosphoketolase | Glucose | No significant change |
| XR-XDH | Xylose | Increase |
| XR-XDH | Xylitol | Increase |
| XR-XDH | Glucose | No significant change |
| XR-XDH-XK-PK | Xylose | Increase |
| XR-XDH-XK-PK | Xylitol | Increase |
| XR-XDH-XK-PK | Glucose | No significant change |
And overall, if our design works as intended, our engineered E. coli should have a faster growth rate when on any lignocellulosic biomass with sufficient xylose or xylitol content.
Sequence Design
In pursuit of appropriate sequences for each of the four enzymes, we searched through the Uniprot database, which returned favourable protein sequences for each of the four enzymes.
| Enzyme | Accession number | Source organism | Source gene |
| Xylose reductase | P31867 | S. stipitis | XYL1 |
| Xylitol dehydrogenase | Q07993 | S. cerevisiae | XYL2 |
| Xylulose kinase | P09099 | E. coli | xylB |
| Phosphoketolase | Q9AEM9 | B. lactis | XFP |
We designed each enzyme to be expressed as a translational unit activated upon IPTG induction through prepending a lac promoter/operator (BBa_K4324201), RBS (BBa_K4324200), and a T1 terminator from E. coli’s rrnB gene (BBa_B0010).
We then found the corresponding nucleotide sequences through the tblastn tool
Lastly, we codon-optimised the sequence for expression in E. coli K12 through IDT’s codon optimisation tool, ensuring that EcoRI, XbaI, SpeI and PstI restriction sites were not accidentally present within the sequence.
We sent these four sequences off to IDT for synthesis, which took a few weeks to arrive. We were notified that we received “best effort” attempts for XR and XDH by IDT due to issues during synthesis.
Upon delivery, these lyophilised gene fragments were dissolved, and ligated into two plasmid backbones each (pSB1C3 [CAM] or pSB1K3 [KAN]) through standard iGEM assembly protocol. The plasmids were transformed into liquid culture, grown overnight on LB media, and picked for plasmid isolation.
We utilised the following number system for the duration of our project:
| Number | Enzyme | Antibiotic Resistance |
| 1x | XR | CAM |
| 2x | XDH | CAM |
| 3x | Phosphoketolase | CAM |
| 4x | XK | CAM |
| 5x | XR | KAN |
| 6x | XDH | KAN |
| 7x | Phosphoketolase | KAN |
| 8x | XK | KAN |
We ran an agarose gel electrophoresis for these uncut plasmids, as shown below:
Taking into consideration that uncut plasmids tend to run faster on the gel, we could identify which plasmids were likely to have inserts in them, as they would be located above the line of empty plasmids as annotated above.
| Likely Insert | 21, 31, 32, 33, 41, 42, 43, 44, 62, 81, 83, 84 |
| Likely Empty | 11, 12, 13, 14, 22, 23, 24, 34, 51, 52, 61, 71, 74 |
| Unloaded | 53, 54, 63, 64, 82 |
Knowing that we received “best effort” attempts for XR and XDH, we were not surprised to see that all of 1 and 5 (XR) plasmids showed to be empty, and most of 2 and 6 (XDH) plasmids showed to be empty, except 21 and potentially 62. For the XK and phosphoketolase plasmids, we saw significantly more plasmids with inserts.
Then, we prepared plasmids from our culture again, then digested them with EcoRI and PstI and ran an agarose gel electrophoresis.
All parts except XR were around the correct length, and it appeared that XR was not being digested properly by EcoRI and PstI. We then further tried digestion with XbaI and SpeI, which returned similar unsuccessful results. Although we knew that both enzyme restriction sites were present and our restriction enzymes were functional, the XR part we were trying to cut was either not present or had an incorrect sequence.
As we encountered consistent issues with XR, and as we could not depend on the XDH ‘best effort’ parts to be expressed correctly, we sent selected parts for NextGen sequencing. After performing BLAST queries, all sequences were as expected for the XDH (only KAN), XK and phosphoketolase parts. This hence also confirmed that part 62, which appeared to potentially contain the XDH part in the initial uncut plasmid gel, indeed had the correct insert. However, the XR part seemed to contain random DNA.
| Number | Enzyme | Antibiotic Resistance | Sequence Confirmed? |
| 1 all | XR | CAM | No |
| 2 all | XDH | CAM | No |
| 32 | Phosphoketolase | Cam | Yes |
| 43 | XK | CAM | Yes |
| 5 all | XR | KAN | No |
| 62 | XDH | KAN | Yes |
| 72 | Phosphoketolase | KAN | Yes |
| 83 | XK | KAN | Yes |
Following successful screening and sequence confirmation of XK and phosphoketolase, we wanted to measure the growth curves of these strains to study the impact of the new constructs on the E. coli’s growth. E. coli containing XK and phosphoketolase were grown in the M9 media with appropriate antibiotics, containing different carbon sources (glucose, xylose and Xylitol) over a period of 26 hours, with and without IPTG induction. OD600 were taken every 3h.
For the XFP part in the Chloramphenicol backbone, cells grew slightly better without IPTG induction. As expected, Glucose is the most preferred carbon source, with the cell growth rate more than 3 times faster than in xylose. XFP cells could not grow in xylitol.
For E.coli with XK gene, IPTG helped double the growth rate in glucose and xylose. They grew well in these two carbon sources, with glucose still the most preferable. Interestingly, this engineered strain of E.coli grew well in Xylitol, with the growth rate as fast as xylose. More tests and analysis needs to be done to understand this unusual characteristics.
Through this first DBTL cycle, we were able to establish the initial project design, build and screen for the relevant parts (with mixed success), and test its functionality through measuring growth rates.
From the design process, we learnt so much about the enzymatic pathways and processes within E. coli that drove our design choices in selecting the yeast-derived XR-XDH pathway and phosphoketolase.
To improve our testing, we further planned to assemble all parts into one and measure its growth on lignocellulosic biomass containing a variety of sugars at different concentrations.
After reflecting upon our first DBTL cycle, we had encountered a unique issue where the major problem seemed to rest with the synthesis itself. Hence, instead of redesigning the genetic sequence, we decided to send XR and XDH genes off to Twist Bioscience for another synthesis.
Screening
As with the parts from IDT, the new parts were ligated into plasmid backbones, transformed into competent E. coli K12 cells, grown overnight in LB media, and best-looking ones picked. After a plasmid preparation, each plasmid was digested with EcoRI single digest and EcoRI & SpeI double digest.
However, none of the digests were found to be at their expected sizes on the agarose gel, hence the first round of screening was not successful. We sequenced ten plasmids from the above gel with inserts using a Nanopore, which revealed that all had random E. coli DNA inserted.
Once again, we picked another ten colonies from each plate, ran plasmid preparations, digested them, and ran them on gel electrophoresis, only to discover the recurring issue of unexpected sizes. Hence, the second round of screening was also unsuccessful.
Troubleshooting
Trying to determine the cause of this issue, we noticed that the cultures we took our plasmid backbones from were not fresh. This led us to believe that the pSB1C3 linearised vector may have undergone genomic DNA contamination.
To resolve this issue, we produced new plasmid backbones from fresh overnight cultures, and the parts were ligated again using standard BioBrick assembly protocols.
To save time, we picked 4 colonies of each XR and XDH and ran a colony PCR for our third round of screening.
This time, XR1 and XR2 seemed to have PCR products at approximately the expected sizes. The other products were too small, and therefore are incorrect.
In preparation for testing our E. coli’s growth on actual lignocellulosic biomass, we extracted and quantified (through DNS assay) the sugar content within common biomass items.
Refer to the sugar extraction analysis page for a detailed analysis of the results and implications of the experiment.
In summary, we determined that wattle, banana, bread, barley grain and sifton bush were promising candidates of carbon sources for our E. coli cells to be tested on.
Plasmids for XDH, XK, and phosphoketolase were transformed into E. coli, which were grown on M9 media with their respective parts induced for a few days to check growth on xylose (CAM & KAN), xylitol (CAM), glucose (CAM) and glycerol (KAN) as carbon sources.
On the M9 media with xylose and KAN, we observed that none of the E. coli grew - not even the control cells with empty plasmids. Comparing this with the growth of every spot on the M9 media with xylose and CAM, this led us to believe that the antibiotic resistance marker on the plasmid backbone itself was not functioning properly, and likely not an issue with the part itself.
On the M9 media with glycerol and KAN, we observed minimal growth of the controls, no growth of XDH and XK, but a large growth of XK. Through literature research, we discovered that xylulose kinase could actually phosphorylate glucose to some extent, which we believed was causing it to display prominent growth on the glycerol medium (Luccio et al., Structural and Kinetic Studies of Induced Fit in Xylulose Kinase from Escherichia coli, Journal of Molecular Biology, Volume 365, Issue 3, 2007, Pages 783-798, https://doi.org/10.1016/j.jmb.2006.10.068.).
On the M9 media with glucose and CAM, we observed almost equal growth for all colonies, which is expected as our new parts are involved with improving xylose metabolism. This also verified that the addition of our parts did not hinder the cells’ growth on a glucose medium.
On the M9 media with xylose and CAM, we also observed almost equal growth for all colonies. However, as we tested these parts individually, the addition of only XR or only XDH would not complete the newly induced XR-XDH pathway, and we would not necessarily see any improvement of growth on xylose media. Also, as the main intention of adding phosphoketolase was to alleviate the flux of X5P, there may not have been enough flux generated by the XI pathway alone for there to be a visible growth increase as a result. We would, however, expect to see slight improvements when over-expressing XK, and from the two spots on the plates, it is difficult to draw conclusions about its functionality or non-functionality.
On the M9 media with xylitol and CAM (left), we saw no growth from any of the colonies. This would be expected for the control, XK and phosphoketolase, as they lack a xylitol-converting pathway (XR or XDH). We would, however, expect to see at least some growth on XR as it converts xylitol to xylose, before metabolising it through the native XI pathway. However, we knew that xylitol was not a great sole carbon source for E. coli, and thus hypothesised that the xylitol was not enough to trigger E. coli’s xylose metabolism. Hence, we reperformed the spot cultures on M9 media with CAM with a mixture of xylose and xylitol (right). Unfortunately, we were once again unable to see any growth of the XR colonies. Literature suggests that the reverse reaction from xylitol to xylose of XR has 4-5% the reaction rate of the forward reaction, which may be the cause of this poor growth (Verduyn et al. Properties of the NAD(P)H-dependent xylose reductase from the xylose-fermenting yeast Pichia stipitis. Biochem J. 1985 Mar 15;226(3):669-77. doi: 10.1042/bj2260669.).
Through troubleshooting our failed screening attempts, we have learnt a most valuable lesson - do not use plasmid backbones from old cultures!
Through the sugar extraction and quantification, we discovered which lignocellulosic biomass contained the most reducing sugars, and thus viable candidates to serve as media for growth testing.
Furthermore, through the spot growth and the above literature article, we learnt how XK had a significant growth on the M9 media with glycerol, due to its ability to phosphorylate glycerol to an extent. We also discovered a potential issue with the antibiotic resistance marker of the pSB1K3 backbone due to no samples growing on KAN M9 media. Hence, we would need to re-ligate the parts onto new pSB1K3 backbones and perform the spot cultures again to further validate our conclusions.
Overall, after having completed two DBTL cycles, we have visions for how future cycles will look like, if we were not bound by time and laboratory constraints. Firstly, we would like to perform sequencing for XR1 & XR2 to see if they are correct. Also, we would like to assemble all correct parts into one plasmid and perform relevant screening and growth measurement. Furthermore, we would like to integrate the sugar extraction and DNS assay data into the growth curve measurement of our engineered E. coli by growing them in the extracted sugars.
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |