Enhancing Biofilm Production

---

1. Phosphodiesterases

Phosphodiesterases (PDEs) degrade c-di-GMP and our goal was to therefore knock down one of the native PDEs in P. sp. IsoF. We chose to knock down PisoF_02645, a strong PDE according to Hailing Nie et al.1. Since we want the gene to only be downregulated when there is xylose, we decided to use a CRISPR interference system (CRISPRi), instead of a knockout.

After cloning our sgRNA into the plasmid and transforming it into P. sp. IsoF we performed a c-di-GMP assay, a biofilm staining assay and a swimming motility assay.

1.3 Swimming Motility assay:

To assess the swimming motility of the different bacterial strains we measured the growth diameter after incubating them overnight on 0.3% semisolid agar plates. We performed two swimming motility assays, although we were not able to record two datasets for all strains as some of the plates melted. As PDEs degrade c-di-GMP, we expect to see less motility in strains with an overexpressed PDE and vice versa more in strains where the PDE is knocked down.

Figure 5 shows that the knocked down PDE shows a decreased motility when rhamnose is added, as expected. With the PDE knocked down, c-di-GMP is not degraded, increasing the potential for biofilm formation, which aligns with our expectations.

2. Diguanylate cyclases

The second enzyme family we want to target are the diguanylate cyclases (DGCs). DGCs produce cyclic di-GMP. We tested 5 strains with different DGCs: two wildtype sequences, one native to P. sp. IsoF (PisoF_00565) and one native to P. aeruginosa (WspR). Additionally a mutant version of WspR (WspR R242A) and two mutated sequences of PisoF_00565, that we designed and mutated by ourselves (DGC PisoF R196A and DGC PisoF R240A) were tested. The aim of these mutations is to enhance the enzyme’s activity by inhibiting their I-site.

2.1 c-di-GMP assay

To test our DGC constructs, we applied the same c-di-GMP assay as described for the PDE, with the findings illustrated in Figure 1.

First, we wanted to assess whether the c-di-GMP level increases when the strains are induced with rhamnose, as our constructs are only activated when induced with rhamnose. If a change in c-di-GMP levels is observed, it would suggest that this change results from the introduced DGC.

Our mutations, DGC PisoF_00565 R196A and DGC PisoF_00565 R240A, labeled as DGC PisoF R196A and DGC PisoF R240A in the graph, showed a significant increase in activity when induced with rhamnose compared to the condition without rhamnose. This suggests that our constructs were functioning as expected and that the introduction of a DGC increases c-di-GMP levels. (R196A: t = 6.4794, df = 7.7563, p-value = 0.0002199, R240A t = 6.4794, df = 7.7563, p-value = 0.0002199).

When comparing the difference in average c-di-GMP levels between the rhamnose and no rhamnose conditions, we observed a larger difference between our DGC mutations and our control, confirming that this variation is not merely due to background noise:

The other P. sp. IsoF strains containing a DGC (DGC PisoF, DGC WspR and DGC WspR R242A) also showed a visible increase in c-di-GMP levels, however the difference was not significant. This indicates that different DGCs lead to different increases in intracellular c-di-GMP levels but that in general the intracellular c-di-GMP concentration is elevated by introducing a DGC.

Further we aimed to assess whether the mutations of the DGC PisoF we created indeed leads to increased activity compared to the wild-type DGC. The mutation of the DGC PisoF R196A, in which we exchanged an arginine at position 196 with alanine, showed a significantly higher c-di-GMP level than the wildtype DGC PisoF sequence (t = 2.5872, df = 8.9978, p-value = 0.02936).

Our mutated enzyme therefore shows higher activity than the wildtype enzyme. This indicates that we successfully mutated DGC PisoF_00565, specifically the enzyme’s I-site, by removing its negative regulation through c-di-GMP. So by altering the sequence of the DGC we prevented c-di-GMP from binding to the negative allosteric site, which enhanced the enzyme’s activity.

This has never been done or shown before for any DGC native to P. sp. IsoF. Our second mutation, DGC PisoF R240A on the other hand did not show a significant increase compared to the wildtype enzyme, suggesting that the I-site was not successfully modified.

Next, we compared our constructs to a control P. sp. IsoF dCas9 without plasmid to quantify the change in c-di-GMP production. Several of our tested strains show a notable increase in c-di-GMP production compared to the control P. sp. IsoF dCas9 no plasmid, however only the mutated DGC WspR (DGC WspR R242A) showed statistical significance for it ( t = 2.4238, df = 7.3234, p-value = 0.04433). The high variability in the DGC WspR might suggest that while the construct has a promising potential to achieve high c-di-GMP levels, further optimization might be required to achieve more consistent results. Although our mutations of the DGC native to P. sp. IsoF (DGC PisoF R196A and DGC PisoF R240) did not show a significant increase in c-di-GMP levels compared to the empty plasmid, we believe that obtaining more data points could provide a clearer picture, as there appears to be a noticeable difference in c-di-GMP levels visually. Further, when compared to P. sp. IsoF dCas9 empty plasmid, DGC PisoF R196A shows a significant increase (t = -2.9931, df = 5.3119, p-value = 0.02814) indicating inconsistencies in the data, since no difference between the two controls is expected.

A third experiment was excluded from the previous analysis because the values observed in the no-rhamnose condition were significantly different from those in the first two experiments. This discrepancy suggested that the experiments might not be directly comparable, possibly due to unintended variations in the experimental setup. The results of the third experiment can be seen in Figure 6.

Since there was only one data point per strain and condition, no statistical test was conducted. Differences to the empty plasmid are only visible for the two mutated DGCs, DGC PisoF R196A and DGC PisoF R240A. The difference in fluorescence signal of PisoF R196A compared to the empty plasmid is 239.00 and 329.00 for DGC PisoF R24 compared to the empty plasmid respectively. This supports the findings from the data shown in figure 1. Introducing a DGC elevates c-di-GMP concentrations and again, it seems like the mutation of PisoF_00565 at position 196 worked and the modified enzyme seems to have increased activity.

Based on the data of the different assays, we conclude that introducing a DGC increases the intracellular c-di-GMP concentration. Additionally, the mutated enzyme DGC PisoF_00565 R196A seems to have an increased activity compared to the wildtype DGC PisoF_00565, indicating that its I-site was successfully mutated and c-di-GMP stopped from binding to it.

However, our findings also underscore the need for additional experiments to validate and better characterize these effects.

2.2 Biofilm staining

Figure 4 represents the fluorescence data of the strains containing the five different DGCs (alongside the strain containing the sgRNA to knockdown PDE). The figure illustrates the mean fluorescence for each strain. P. sp. IsoF dCas9 and P. sp. IsoF with an empty pBBR1MCS5 plasmid served as negative controls, while P. sp. IsoF YedQ once again served as a positive control. Each strain and condition (with and without rhamnose) was measured in duplicate.

As observed in the first biofilm staining assay, the control strains showed the lowest fluorescence, suggesting minimal polysaccharide production. In contrast, the bacterial strain expressing the constitutively active DGC YedQ showed the highest fluorescence, as expected. Among the other tested strains, three showed a significant increase in fluorescence, indicating enhanced polysaccharide production: P. sp. IsoF DGC PisoF_00565, P. sp. IsoF DGC WspR and P. sp. IsoF DGC WspR R242A.

The strain expressing the mutated WspR DGC (P. sp. IsoF WspR R242A) showed the highest fluorescence level, even exceeding the positive control (P. sp. IsoF YedQ). This suggests that the mutated DGC WspR leads to the greatest increase in polysaccharide production and, consequently, the most significant enhancement in biofilm formation.

In contrast, the mutated variants of DGC PisoF_00565 (DGC PisoF R196A and DGC PisoF R240A) did not show the same increase in fluorescent signal as the other DGCs. While DGC R240A displayed a modest increase in fluorescence compared to the controls, it was not as distinct as in the other tested DGCs.

Interestingly, DGC PisoF R196A showed no measurable increase in fluorescence. This result contradicts the findings from the c-di-GMP assay, where the DGC PisoF R196A led to an increase in intracellular c-di-GMP levels.

Throughout different experiments, we observed that the strains carrying different DGCs had reduced growth efficiency when induced with rhamnose. We have two hypotheses of why this could be the case. First, it could be that rhamnose is toxic to bacteria in high concentrations. Alternatively, the overexpression of a DGC, leading to the overproduction of c-di-GMP, might be costly for the bacteria and thus inhibits their growth. This could also explain why the strains containing different DGCs did not show a significant increase in polysaccharide production. During the biofilm staining assay, we were unable to measure the OD of each strain, meaning that we don’t know how well each strain grew on the plate. As a result, we could not normalize the polysaccharide production based on an OD measurement.

To further analyze the results, we applied the same linear model used in the previous experiments, with bacterial strain as the explanatory variable (10 categories) and mean fluorescence as the response variable. The model showed a statistically significant increase in fluorescence for the following strains:

As noted, the strains containing the mutated forms of DGC PisoF_00565 (R196A and R240A) did not lead to a significant increase in fluorescent signal.

The adjusted R-squared for the model is 0.9733, with an F-statistic of 78.35. The explained variance from the bacterial strain variable is 18’046.4, while an unexplained variance of 259.1 remains, with an overall p-value of 2.062e-07.

In conclusion, the data from these biofilm staining assays confirm that certain DGCs, particularly WspR and its mutated form WspR R242, significantly enhance polysaccharide production. WspR R242 demonstrates the greatest increase in biofilm formation overall. Surprisingly, the mutated forms of DGC PisoF_00565, R196A and R240A did not lead to a notable increase in polysaccharide production.

These results confirm that introducing certain diguanylate cyclases (DGCs) can significantly increase and therefore promote polysaccharide production, which represents an enhanced and stronger biofilm formation. Elevating polysaccharide production through the introduction of a DGC provides an efficient and straightforward approach to boost biofilm formation, without the need to focus only on specific genes in the genetic pathway to produce one or two polysaccharides.

2.3 Swimming Motility DGCs

As DGCs drive biofilm formation by producing c-di-GMP, we expect to see a shift in our DGC strains from a motile to a more sessile state in which biofilm-formation can take place.

Figure 5 shows that the DGC WspR and the mutated form WspR R242A show the highest motility (indicated by the swimming motility diameter) when the DGC is not induced by rhamnose. The high motility levels suggest that these strains may be in a state with less biofilm formation. However, in the presence of rhamnose, their motility drastically decreases in the first assay (the effect is not as drastic in the second one). This indicates that when c-di-GMP is further increased due to DGC activity, motility decreases as expected, which aligns with an increased potential for biofilm formation. The sharp drop in motility suggests that WspR likely contributes heavily to biofilm formation when expressed.

Our mutations DGC PIsoF R240A and R196A show lower motility in both conditions, yet the rhamnose condition shows higher motility than no rhamnose condition. This suggests that this DGC may not be as important as the WspR in decreasing the motility for example. Our controls P. sp. IsoF no plasmid, empty plasmid and PDE overexpression show higher levels than our rhamnose induced DGCs.

These results confirm that for some DGCs, such as DGC WspR and DGC WspR R242A, the relationship between DGC activity, motility, and the biofilm formation process is strongly correlated. While it is not clear why our DGC mutations R196A and R240A show higher motility with rhamnose, this is also seen when rhamnose is added to the YedQ DGC. This may suggest that rhamnose could promote bacterial motility in other metabolic contexts, separate from DGCs.

Plant experiments

---

Raw data from the plant experiment was obtained on the date of inoculation - time point 0, and two weeks later.

The plants were divided in 4 groups that underwent different treatments: PisoF with an empty pBBR1MCS5 plasmid, PisoF YedQ, PisoF PDE overexpression and LB medium. Half of the plants were regularly watered, in the other half we induced drought. For more information about the set up of the plant experiment and the processing of the plants go to the Experiments Page.

Time point 0

One plant out of each treatment for the dry and wet conditions was taken out and processed on the first day after 30 minutes of inoculation. The initial measurements are summarized in the following graph:

Figure 7 represents the time point 0 measurements that serve as a baseline for our analysis. Based on these values we can make observations about the growth that the rest of the plants in the same treatment group exhibit.

Wet weight Measurement after 14 days

We conducted the second measurement after 14 days. The remaining plants from each treatment in the dry or wet conditions respectively were processed. The data points are depicted in Figure 8:

As expected, the drought conditions result in an overall lower mass of the aerial part of the plant than the normal conditions. However, under drought conditions there is barely a difference in the mean values between the treatments. This contradicts our hypothesis that the bacterial treatments would result in a higher biomass than the LB medium treatment. We think this might be due to the dry treatment being too harsh, possibly resulting in conditions too dry for the bacteria to have any beneficial effect. In addition, the variance in the data points we obtained from the LB treatment was too high for us to effectively draw a conclusion from it. The same applies to all the data points obtained from the wet measurements.

One notable observation is that the PDE overexpression treatment exhibits lower biomass values than the PisoF empty pBBR1MCS5 plasmid and the PisoF DGC overexpression plasmid in both drought and normal conditions. This is consistent with our expectations, since the PDE overexpression plasmid leads to more c-di-GMP degradation, less biofilm and therefore less plant growth.

In order to be able to draw conclusions about the plant growth under different treatments and conditions, we took the average biomass of all plants measured at time point 0 and compared it to the mean values from each box plot in the graph above. In the drought condition there is no growth observed. In fact, the wet weight of the biomass decreases after 14 days of drought. This is likely due to the fact that many plants were severely dehydrated, which suggests that water loss is the cause of the decreased measured weight. In normal (non-drought) conditions, on the contrary, one can observe growth in all treatments. However, the variation between the individual data points we obtained within each treatment is high. Therefore, it is difficult to draw precise conclusions about the effect of the treatments on the weight of the biomass, especially since sample size was rather low. If this experiment is repeated in the future, we would need to make sure to include more plants in the experiment for a more reliable statistical analysis. Overall though we can observe the highest increase in growth after the DGC overexpression treatment, which is consistent with our hypothesis.

Dry weight Measurement after 14 days

After we processed the wet weight, we kept the plants in the oven for a few days until they were completely dry. Then we measured them again to obtain the dry weight of all plants. These measurements are illustrated in Figure 9 below:

There is a substantial difference between the dried weight of the plants from the drought condition and those grown in normal conditions. This most likely arises due to the severe dehydration of the plants grown in drought. The mean values of the dry plants are almost identical, which suggests that the treatments with our bacterial strains did not necessarily prevent the drought stress imposed on the plants.

Additionally, the results in the group of plants treated with bacteria under normal watered conditions are also inconclusive. Visually, one can note that the highest mean value is at the PisoF DGC YedQ treatment, which is consistent with our expectations. PisoF DGC YedQ is known to form more polysaccharides and an induced biofilm. The PDE overexpression plasmid treatment which we use as a negative control also exhibits a mean value, lower than the two other bacterial treatments. Against our expectations, the strain with an empty pBBR1MCS5 plasmid shows the same mean value as all plants processed at Time point 0.

However, there is high variability between the individual data points,therefore no meaningful observation can be made about the effect of the different treatment groups. For future experiments and better statistical analysis, we would need to include more plants in our treatment groups.

CFU count

We conducted a colony forming unit count from the rhizosphere of each of the treated plants. With the aim of establishing the number of P. sp. IsoF colonies 14 days after the treatment implementation we took a sample of the rhizosphere and performed a dilution series. Those were plated and counted. Different amounts of soil were stuck to the roots in each sample, we adjusted the count for 1g of rhizosphere.

Figure 10 shows that the DGC YedQ and PDE overexpression strains have the highest number of surviving colonies, followed by the PisoF pBBR1MCS5 plasmid and LB. The plates we used had a gentamicin resistance to help us select for our strains that carry the plasmids we introduced with a gentamicin resistance. This aligns with our expectations and proves that P. sp. IsoF can survive in conditions mimicking the environment and retain the plasmids that have been introduced to it. The difference between the wet and dry conditions can be explained by the fact that some of the bacteria might have been washed away when the wet plants were watered.

In addition, the data we obtained 14 days after inoculation represented in figure 11 shows more relevant results.

When it comes to a difference between the drought and normal conditions, we can observe the same trend as in the plants processed at Time point 0. There are less CFUs in the wet rhizosphere potentially due to the bacteria being washed away in the process of watering. The plants in the drought condition show that the bacterial treatments also result in a higher CFU count in the rhizosphere compared to the mean of the LB treatment. On the other hand, the variance within the LB-treated samples is too high for us to make a notable observation.

For both the drought and the regularly watered condition the difference in the mean CFUs of the different bacterial strain treatments is barely recognizable. This suggests that the genes being expressed on the pBBR1MCS5 backbone barely cause more strain to the bacteria than the presence of the empty backbone and the colonies are able to survive despite being genetically engineered.

Killswitch

---

In order to finetune every part of the killswitch, first, the toxin expression has to be optimized. We needed to find the sweet spot of toxin production: high enough to kill the cell, but low enough for the antitoxin production to override the toxin’s effect. Expression levels are adjusted using three different strengths of ribosome binding sites (RBSs).

Toxin Assay 1: OD Measurement

Toxin assay 1 in E. coli SY327 (1)

The toxin test did not work, as there was no significant difference in cell growth between the cells induced with rhamnose and the ones that grew without rhamnose. If anything, the cells treated with rhamnose seemed to grow better, potentially using rhamnose as an extra source of energy.

Before testing our toxin again, we needed to rule out that the issue lies with the rhamnose promoter.

Rhamnose promoter characterisation in E. coli SY327

To test whether the promoter was the issue and at which rhamnose concentration its expression would be the strongest, we conducted a promoter characterisation. We did so using a plasmid that had our rhamnose promoter with RFP as a reporter gene.

We performed a preliminary test which, after an overnight incubation with 1% rhamnose, exhibited a red color, confirming that the rhamnose promoter was active.

To quantify the promoter’s activity at different rhamnose concentrations and the potentially toxic or nutritious effect of rhamnose on the cell, we performed an overnight growth assay, measuring OD600 and fluorescence at excitation 530±25 nm and emission 590±35 nm.

The promoter was detectably activated after 2hrs of inoculation with rhamnose, regardless of concentration. The highest fluorescence per OD and thus the strongest promoter activity was at 1% rhamnose, followed by 1.5%. This means the promoter’s maximum activity with the minimum negative effect of rhamnose was at 1%.

From this test we conclude that the issue does not lie with the promoter and that the toxin test needs to be repeated.

Toxin Testing in E. coli SY327 (2)

In this second toxin test on E. coli SY327, we again observed no significant difference in the strains’ growth between the ones with an induced toxin production and the ones without. Generally, the growth did not differ between the toxin strengths and the controls, but rather reflected the changes in rhamnose concentrations. Generally, most strains that grew the most were the ones with no rhamnose added, followed by 1% and then 1.5%. We concluded that rhamnose is affecting the cells’ growth negatively and more so than the toxin, regardless of its strength. To account for that, we subtracted the growth of the empty plasmid control (pJUMP29-1D(sfGFP)) from the strains tested at the respective rhamnose concentrations. This way, we would see the difference in growth caused by our toxin plasmid alone - expecting at least some slowed growth.

At first, the medium toxin strength seemed to present the slowest growth in the first three hours. On second examination this is purely due to the fact that the medium strength strains already started off with an OD600 0.06 lower than the control at the respective rhamnose concentrations. Secondly, there is no significant difference between 0% and 1% or 1.5% rhamnose, indicating that the variability is not due to a difference in toxin production. Regardless, ccdB is expected to fully inhibit the cell’s growth, as it is a potent gyrase toxin, and would show a more drastic difference than seen above.2

We concluded from this that the most likely cause for our toxin’s lack of effect is a gyrase mutation in E. coli SY327. Upon further research and consultation with Zaira we found that E. coli SY327 is based on E. coli S17-1, which carries a gyrA462 allele - a gyrase mutation making cells immune to ccdB toxins3. This allele has been deliberately transferred to several strains derived from E. coli S17-1, though we could not find evidence to support it was also transferred to E. coli SY327. Still, we found this to be reason enough to assume our current E. coli strain had a gyrA mutation, wielding it immune to the ccdB toxin and thus explaining our toxin’s lack of effect. We therefore had to transform our plasmids into a new strain and repeat the toxin test. After some brief research and consulting which strains were readily available by our host lab, we chose to go with E. coli DH5α.

Toxin assay 1 in E. coli DH5α

Once more, the cell’s growth seems to be unaffected by the ccdB toxin, as no significant difference was found between the control, the different toxin strengths, or their induction with rhamnose. We revisited our plasmid construct to eliminate all possibilities of a design error. While in our original research we found no gyrA mutations in E. coli DH5α, an AddGene blog entry by Mathew Ferenc shed light on a gyrase mutation (gyrA96) mutation carried by E. coli DH5α which, though not proven to provide resistance to ccdB, may affect the toxin’s effectivity.4 While this finding was frustrating, it taught us how unorganized and hidden scientific information can be - the lack of one universal database for bacterial strains just being one example. We used Ferenc’s blog to screen for other potentially sensitive E. colistrains and noted the strains which our lab had available. We decided to continue the testing of our toxin in HB101, MC1061 and CSH50.

Toxin assay 1 in E. coli HB101, E. coli MC1061 and E. coli CSH50

In all three E. coli strains tested above (HB101, MC1061 and CSH50), the cell’s growth seems to be unaffected by the ccdB toxin, as no significant difference was found between the controls, the different toxin strengths, or their induction with rhamnose. There were still some differences between the tests such as E. coli MC1061 strains having more variability, E. coli HB101 overall growing to higher concentrations and the positive control in E. coli CSH50 seemingly growing slightly despite the presence of Kanamycin. However, none of these differences in results affect the conclusion that our ccdB toxin system does not kill or inhibit any of our cell’s growth. As at this stage of our project we were starting to run out of time, we decided to perform a last toxin test in Pseudomonas species IsoF for the sake of thoroughness.

Toxin Testing in P. sp. IsoF

As expected, the P. sp. IsoF cells responded to the ccdB toxin very similarly to all previously tested E. coli strains. The cells, regardless of toxin strength or rhamnose concentration, grew to OD600 ~1 and stagnated thereafter. There was no significant difference between any of the strains. By now, it was evident that the issue with our toxin was neither in the plasmid design nor the strains, but rather had to lie in our way of testing. Alternatively, the amount of transcription induced by our rhamnose promoter could simply be insufficient to produce enough toxin to have an effect. Upon revisiting the iGEM parts registry, we discovered that, for testing similar constructs, the plates themselves can also be induced with Rhamnose - something we hadn’t previously implemented. We’ve decided to incorporate this method into our next experiment.

TOXIN ASSAY 2: rhamnose induced plates

Toxin assay 2 in P. sp. IsoF

As mentioned earlier, we adjusted our toxin testing protocol by incorporating rhamnose into the plates, with the goal of performing a CFU (colony-forming unit) assay.

Figure 21 shows the plates after letting the bacteria grow over the weekend. Unfortunately, we did not prepare a dilution series, making it difficult to accurately analyze the results. Due to time constraints, we were unable to repeat the experiment with a proper dilution series. However, based on visual inspection of the plates, we were not satisfied with the results and could not conclude that our toxin had a clear impact on reducing CFU.

Sensing

---

To assess the promoter's sensing capability, we conducted a fluorescence measurement experiment over a 24-hour period, during which we recorded fluorescence and optical density (OD) at various time points. Unfortunately, we were unable to detect any significant increase in fluorescence compared to the control samples at any time. After troubleshooting the issue, we identified a design error in our construct and are currently awaiting new results.

Future Steps

---

Sensing:

We will perform an inverse PCR to include the operator that was missing an the reason our sensing assay failed. As a backup plan we will order the operator as oligonucleotides and perform a digestion ligation after oligo annealing to implement the operator into the plasmid. In case one of those cloning methods is successful, we will perform the sensing assay described in experiments again.

Biofilm components:

Since our controls didn't show the expected results, further experiments should be performed to confirm our findings.

Kill Switch:

We will conduct another toxin experiment in which we directly expose colonies to the ccdB protein. If the bacteria are killed under these conditions, it would support our suspicion that the promoter may not be strong enough to produce enough toxin. Instead of redoing another toxin experiment with the CcdB/CcdA system, that has initially been established for E. coli strains, we could switch to another toxin-antitoxin system, that is established in Pseudomonas species.

Final construct:

If we had more time we would have recloned the level 2 plasmids and introduced them into P. sp. IsoF to test our genetic circuit.

We would first research the concentration of xylose exceeded by plant roots. We could then induce the strain with different xylose concentration, one of them being the exceeded concentration and again measure the c-di-GMP concentration, polysaccharides production and bacterial motility to see if an increase in xylose would lead to an enhanced biofilm formation. If this is the case we would know our construct works as intended and if the xylose concentration in the rhizosphere would be enough to trigger our system.

Further, we could remove the xylose source again by washing the culture with NaCl and adding new LB medium to let them recover. We could then plate the strains and see how many colonies survived. In case our construct works as intended we expect to not have any colonies on our plate, since the bacteria should be killed when it is not in proximity of plant roots. Removing the xylose by washing the cells would mimic the distancing of the bacteria from the rhizosphere of the plant root. Lastly, we could test the full construct again on plants under drought and normal conditions to see whether our system would also function in vivo.