Team:Leiden/Measurement
Measurement
As our project depends on reliable detection of bacterial cell stress using reporter strains, we spent much time on creating reproducible measurement protocols. We decided to use Green Fluorescent Protein (GFP) for quantitative measurements after Read more about our consultation on our Integrated Human Practices page consultation with potential users of our stress-detection system. However, agar plate-based applications of our system do not require detailed quantitative information on gene-expression, and would benefit from a cheap and simple readout. Therefore, we also designed a novel method for naked-eye detection of stress responses in E. coli using visual reporters based on chromoproteins. These reporter cell lines will change colour when exposed to stress. In the following chapters, our journey to obtaining a reliable measurement framework for both systems is explained, including interpretation of data and controls.
There are several ways to measure the production of GFP in organisms. Popular methods include microscopy, flow cytometry and plate reader measurements[1]. For analysis of our system, predominantly flow cytometry and confocal microscopy was used. In the sections below, the choice for these two techniques will be explained. Furthermore, our journey to optimizing the experimental setup will be shown and data interpretation will be discussed.
Evaluating measurement methods
Two of the most common methods of quantifying GFP expression in E. coli are plate reader and flow cytometry measurements[1]. Using plate readers is an easy and accessible setup for relatively high-throughput GFP measurement. Therefore, we aim to implement such measurements in our Read more about our design process on our Product Design page final product design. However, plate reader measurements give very little information about the cells apart from an optical density (OD) and fluorescence. With flow cytometry, we can obtain more details such as the extent of cell death and whether all cells display increased fluorescence or whether multiple are populations formed. Furthermore, plate reader measurements must be normalised to OD. In our project we routinely used substances that inhibit growth, creating a great variability in OD between samples. This has the potential to skew the data after normalisation of fluorescence to OD. Because of the aforementioned reasons, we chose to validate our strains with flow cytometry. For visual understanding and extra confirmation of the effect of antibiotics and other stressful substances on our strains, we used confocal microscopy.
Starting validation
For flow cytometry, we started by testing whether this technique would be able to measure a GFP signal in E. coli. Hence we compared a wild-type E. coli DH5α strain and the GFP-expressing InterLab positive control (BBa_I20270) in flow cytometry experiments. As can be seen in figure 1, the fluorescence intensity is much higher for the InterLab strain indicating that this technique could successfully detect bacteria expressing GFP.
For the initial validation of our system we stressed our strains with a You can read more about the stressors we used on our Results page range of stressors of which the mode of action is known. These stressors potentially have drastic influences on the detected GFP signal in flow cytometry by for example altering the morphology of the cells. For instance, cell elongation due to nalidixic acid treatment may cause an increase in detected fluorescence intensity regardless of a change in expression, leading to misinterpretation of the results[2]. We performed a series of experiments to test the influence of these stressors on autofluorescence and GFP expression of E. coli. For this purpose, we performed experiments with E. coli DH5α to establish the levels of autofluorescence with and without stressor treatment (Figure 2). As a second control experiment, we took the positive control provided in the InterLab studies (BBa_I20270) and repeated the experiment to understand the influence of antibiotics on measured GFP expression (Figure 3).
The data in figure 2 shows that stressor treatment had no significant influence on the autofluorescence of the DH5α wild-type strain.
As can be seen in figure 3, no stressor lead to a significant increase in detected GFP expression or autofluorescence. Therefore we concluded that false positives as a result of a change in morphology or another effect caused by stressor
treatment
could not occur. We did not test for ampicillin as the plasmids used for validation were cloned in ampicillin resistance backbones.
Several antibiotics lead to a decrease in the GFP signal, as seen in figure 3. One might object with the argument that we will never detect a signal when we use these antibiotics.
First of all, it should be noted that we are not necessarily looking at the strength of the signal, only at hits that exceed the negative control. A slight decrease in signal will therefore not be an issue.
For hydrogen peroxide, chloramphenicol, zeocin, vancomycin and rifampicin we believe that the observed decrease is small enough to be overcome by stress-related induction of the promoters. This assumption is confirmed by our results. For
instance, our pSoxS-GFP construct (BBa_K2610031) showed increased GFP expression for treatment with hydrogen peroxide and zeocin, and
pKatG-GFP (BBa_K2610018) displayed an upregulation in response to hydrogen peroxide treatment.
Tetracycline, streptomycin and kanamycin are protein synthesis inhibitors (PSIs), which explains the marked decrease in detected GFP signal[3]. Ideally, we would like to use a concentration where cells are stressed and no
protein synthesis inhibition occurs. However, this is not realistic as it is unlikely stress occurs when no protein synthesis inhibition occurs. Therefore, we chose to keep using the same concentrations of these stressors while being
careful with
interpretation of results, using other PSIs as controls (to verify whether the response was specific) and by including extra controls when we suspected protein synthesis inhibitors caused stress-related promoter regulation. For example, in
our Read our Results page for more information Results pagewe discuss pCspA-GFP (BBa_K2610037), a reporter which appeared to be specifically downregulated by protein synthesis inhibitors. To
test whether this was a regulatory effect and not simply inhibition of GFP production, we
performed a confocal microscopy experiment where we included a pSoxS-mCherry strain as a positive control for protein synthesis. If the response was specific to pCspA, only GFP production would be inhibited and mCherry production would
still be observed. This is indeed what we saw (Figure 4).
Good measurement practices and reproducibility
After simple proof-of-concept GFP detection with the flow cytometer, we started validation of our GFP-expressing strains in response to stress. To increase reliability, every flow cytometer experiment was done either in duplo or in triplo
on the plate, and experiments were repeated over the course of different days. Negative controls were blank LB medium, non-transformed E. coli DH5α and a reporter strain without the addition of a stressful compound. To prevent
the additional production of GFP over the course of the experiment and to eliminate interference by particles that interfere with the flow cytometry measurements (e.g. salt crystals), cells were kept on ice and washed two times with cold
PBS.
For confocal microscopy, only pictures taken on the same day under the same conditions were compared with each other. A wild-type DH5α was used as negative control to measure the autofluorescence, while a GFP reporter strain without
any additional compounds was used to visualise the basal expression.
More details of the stress-inducement procedure as well as technical details and the equipment and settings used can be found on the protocols page.
Data analysis
There are several ways to analyse flow cytometry data. One way is to gate the cloud which are presumably cells and to then only use data from that gate (Figure 5A). Alternatively, one can remove measurements that yielded no fluorescence, as all cells display a basic level of autofluorescence[4] (Figure 5B).
We analysed data from an experiment with both data interpretation techniques, the results are shown in figure 6. We can see that the absolute values differ, but the overall pattern is similar and the same conclusions can be drawn from both methods. Considering that the second method of removing measurements is more time efficient - especially with the large amount of experiments we had to analyse - we chose that method for further analysis.
In terms of statistics, it was decided to represent the data with the geometric mean fluorescence intensity (MFI). It must be noted that MFI is sometimes used as an abbreviation for mean or median fluorescence intensity. Because we are generally dealing with very asymmetric populations, outliers - both low or high - will have a large influence on the mean intensity. We considered using the median intensity, which is common in flow cytometry experiments to represent a ‘typical’ member of the population[5]. However, in our results, it is more relevant to assess the entire population. Because of that, we decided to represent our data in geometric mean fluorescence intensity.
Fluorescence units
The units of the detected GFP signal in flow cytometry are arbitrary and may differ greatly between repeat experiments. As such, absolute values of different measurements cannot be compared. This can be mitigated using calibration of fluorescence units to known standards, as performed in the View our InterLab page for more information InterLab study. However, for our experiments this is not of great importance because we only considered the difference in detected signal compared to a negative control, which was always included in the same experiment. This difference can be quantified in fold changes, which is the numeric value we used for comparisons between measurements.
References
[1]: Soboleski, M. R., Oaks, J., & Halford, W. P. (2005). Green fluorescent protein is a quantitative reporter of gene expression in individual eukaryotic cells. The FASEB Journal : Official Publication of the Federation of American Societies for Experimental Biology, 19(3), 440–442.
[2]: Renggli, S., Keck, W., Jenal, U., & Ritz, D. (2013). Role of Autofluorescence in Flow Cytometric Analysis of Escherichia coli Treated with Bactericidal Antibiotics. Journal of Bacteriology, 195(18), 4067–4073.
[3]: McCoy, L. S., Xie, Y. and Tor, Y. (2011), Antibiotics that target protein synthesis. WIREs RNA, 2: 209-232.
[4]: Herzenberg, L. A., Tung, J., Moore, W. A., Herzenberg, L. A., & Parks, D. R. (2006). Interpreting flow cytometry data: a guide for the perplexed. Nature immunology, 7(7), 681.
[5]: Tanqri S, Vall H, Kaplan D, Hoffman B, Purvis N, Porwit A, Hunsberger B, 0 TV; on behalf of ICSH/ICCS working group. Validation of Cell‐based Fluorescence Assays: Practice Guidelines from the ICSH and ICCS–Part III–Analytical Issues. Cytometry Part B 2013:84B:291–308.
Besides GFP reporters, an important part of our project was the use of chromoproteins to create visual reporters. Similar to the stress-responsive GFP strains, careful thought was put into measuring and validating our chromoprotein strains. In the next section, a short explanation on our measurement methods, optimization of the experimental setup and analysis is provided.
Evaluating methods
The idea to use chromoproteins for the detection of bacterial cell stress was prompted by conversations about Streptomyces overlays. For these experiments, a visible readout on agar plates is crucial. For initial validation of agar-plate based reporters, we employed disk diffusion experiments because they allowed for greater control of the antibiotic used and concentration thereof. In addition, we performed experiments in liquid culture to assess whether we could observe a dose-dependent response of our chromoprotein strains.
Starting validation
Initially, to confirm that chromoproteins would be visible on agar plates and in liquid cultures, we created constructs that constitutively express blue and pink chromoproteins with our novel pGapA promoter (BBa_K2610040): pGapA-AmilCP (BBa_K2610041) and pGapA-pink (BBa_K2610042) respectively. Expression was visible in both conditions, but the blue chromoprotein has a more pronounced contrast on agar plates (Figure 7). Therefore, we chose to create our first visual reporter with the blue AmilCP chromoprotein.
We took one of our most promising stress promoters, pSoxS (BBa_K2610030) - which showed a strong response to nalidixic acid - and created pSoxS-AmilCP (BBa_K2610033). Using Read more about our modeling this on our Modeling page theoretical modeling, we estimated the nalidixic acid concentration in the petri dish to help understand the stress-response of the chromoprotein strains. Knowing details about the concentration of nalidixic acid allowed us to compare the dose response of pSoxS-AmilCP (BBa_K2610033) to pSoxS-GFP (BBa_K2610031).
Good measurement practices and reproducibility
For all experiments conducted with chromoproteins, we used a wild-type DH5α as a negative control. For the pSoxs-AmilCP (BBa_K2610033) strain, negative controls are inherent in the experiment as there will always be unstressed regions on the agar plate on which the experiment is carried out. As mentioned before, pGapA-AmilCP (BBa_K2610041) and pGapA-spisPink (BBa_K2610042) were used as positive controls. All experiments were performed multiple times across multiple days to decrease the chance that our results were pure coincidence.
Analysis
Analysis of the success of our chromoprotein strain was mostly based on visibility to the naked eye. Therefore, we always included a negative control in pictures to ensure we did not take the wrong conclusions.
Our project aimed to improve antibiotic discovery by detection of cellular stress in E. coli using reporter strains. GFP expression can be measured with plate readers or flow cytometry. For initial characterisation we chose flow cytometry as it provides more information about the treated population of bacteria. Confocal microscopy was used for further validation. We also developed visual reporters that produce coloured chromoproteins when exposed to stress. Comparison to positive and negative controls clearly show that this novel system works. All that remains is amplification of the signal to obtain a faster and better readout.