Team:Manchester/Measurement

UoM iGEM | Project Cutiful

Measurement

"Is Cutiful viable?" You are right to ask yourself that question, and Manchester iGEM 2019 is proud to answer that: "yes, it is!". We have shown on the Colour and Fragrance pages the ability of DH5⍺ to successfully secrete chromophores and odorants, and here we want to prove to you that, not only does DH5⍺ attach to hair, it is also able to survive in saline and chlorinated environments - such as found in pools, or the sea - and resist a shampoo wash. After all this, as the Brits say: "Bob's your uncle"!

Our achievements …

In 7 bullet points

1.

Achievement of quantitative, reliable and reproducible data that provides extensive characterisation of our multiple constructs in all three aspects of our project: colours, fragrance and decapeptides.

2.

Use of well-established precise techniques and instrumentation including: fluorescence microscopy, microplate readers, spectrophotometers, ultra-performance liquid chromatography coupled with tandem mass spectrometry (UPLC-MS/MS) and gas chromatography-mass spectrometry (GC-MS) among others.

3.

Provided evidence of the importance of data normalisation in fluorescence spectrometry measurements, allowing sufficient comparison between different laboratories without the need for a detailed interlab study.

4.

Highlighted the limitations encountered when using fluorescence spectrometry and suggested the use of fluorescence microscopy in parallel to support and validate data.

5.

Confirmed evidence that OD600 gives false increase in apparent cell density in RFP-expressing bacterial cultures. Hence, we suggest future teams to directly estimate cell density of these genetic engineered bacteria with a discrete measurement of at least 660 nm (or even higher to 700 nm).

6.

We proved that discrete measurements at 588 nm of AmilCP gave a false increased readout. We suggest for future teams to use specific spectra of 500 nm to 650 nm.

7.

By using highly specific and sensitive GC-MS and UPLC-MS/MS, we detected and quantified limonene and new-to-iGEM intermediates in the vanillin biosynthesis pathway respectively.

Measurements-where we Excel

Or: in which there is a stick, pretty cool pieces of equipment, and we aren’t allowed to press the big red button. Also: reproducibility, regrets, late notices, and a lot (read: a lot) of data analysis done late at night.

Prologue

“Three thousand” - a new measuring unit of love, implemented by Morgan Stark. Avengers: End Game, Marvel, 2019.

A big challenge that we encountered throughout Cutiful was the idea of measuring the product of our designed constructs after bacterial production. We wanted to achieve quantitative, reliable and reproducible data that could allow other teams around the world to use our methods.

Colour constructs required quantification through the measurement of fluorescence and absorbance depending on the selected fluoroprotein or chromoprotein. This was performed using both spectrophotometers and high-quality microplate reader instruments. Additionally, we wanted to achieve qualitative results through the use of fluorescence microscopes.

Fragrance constructs required precise techniques such as ultra-performance liquid chromatography-mass spectrometry (UPLC-MS/MS) and gas chromatography-mass spectrometry (GC-MS) respectively.

Repair constructs were designed with the addition of a histidine tag for detection purposes. Therefore, traditional Western Blots with specific well-characterised antibodies would allow their detection

This page aims to show our main measurement techniques: we have decided to focus on fluorescence, absorbance at different wavelengths (colours) and mass spectrometry (fragrance) measurements and how these have helped our overall project. However, it is important to note that for each technique we only highlighted one example (full details in Colour, Fragrance, and Protocols), we used these techniques for all of our constructs, with biological and technical replicates for each. We have used well-established, robust and reliable techniques that have allowed Cutiful to not only generate data proving that a hair care alternative could be a replacement in the future; but also to provide further detailed characterisation to previously registered iGEM parts. Hence, we believe we have been able to contribute with high-quality data to the iGEM community in order to aid future teams.

Act I: Detecting chromoproteins - a world of colour (unicorns vomiting rainbows)

“It’s not a unicorn — it’s a horse with a sword on its head that protects our measurements page.” - Adapted from: Tad Quill & Gabrielle Allan, “My Unicorn”, Scrubs S4 E11, 2004, Dr. Dorian

The colour branch of our project required the production and characterisation of selected chromoproteins (visit the colour tab of our wiki). We wanted to have a wide range of hair dyes that were visible under normal and UV light. Therefore we selected proteins that absorb colour, like AmilCP, and proteins that absorb and emit light, making them fluorescent (sfGFP and mRFP1). The selection process and the modelling behind it can be found in our modelling page.

Scene 1: Detecting and characterising fluorescence chromoproteins (mRFP1 and sfGFP)

In this section we have highlighted the experiments that provided us with a further insight into the techniques of fluorescence spectrometry and microscopy.

Fluorescence spectrometry through the use of a microplate reader:

One of the techniques that we used in order to detect our bacterial-based chromoproteins was fluorescence spectrometry. This consists of exciting a molecule with a given wavelength of light, and measuring the corresponding emitted wavelength. Fluorescence intensity is not an absolute value, and it is commonly given as relative fluorescence units (RFU) (Fluorescence Intensity Measurements | BMG LABTECH, no date). One of the biggest challenges regarding fluorescence measurements is the variance that RFU values possess. There have been several attempts to standardise measurements for different variants of fluorescent proteins, however these require a large inter-lab study for each specific protein. This is a time consuming process that requires involvement of several labs simultaneously. As we are unable to provide that amount of standardised data, we decided to address this by presenting RFU values as a ratio of their OD600 measurements. This is an appropriate tool because it makes results comparable to one another, when the bacterial culture is at the same growth phase and therefore should be expressing approximately the same amount of protein.

Data normalisation provides a more accurate comparison between samples. Whereas RFU values cannot be compared with each other, as expression is dependent on growth; normalised RFU by OD data does make direct comparison possible. Normalised data can allow comparison even across different laboratories or within different strains (as suggested by our results performed in two different E. coli).

Here we provide an example of our measurement technique with our sfGFP-producing bacterial constructs.

For this experiment we used three different constructs. These were: sfGFP with OmpA, an N-Terminal secretion signal (BBa_K2906001) sfGFP with HylA, a C-Terminal secretion signal (BBa_K2906002), and sfGFP on its own (BBa_K2906000). These secretion signal tags altered the expression of sfGFP, therefore we needed an appropriate technique to quantify them. Therefore, we used the microplate reader to measure RFU and OD600 for 13 hours after induction.

In order to correctly normalise the obtained data, an appropriate negative control needs to be used. In our case, we selected a chloramphenicol-resistant E. coli TOP10 as it does not produce colour. Therefore, any background fluorescence can be blanked by these values. For a full description of different experiments and specific details please visit plate reading protocol and Colours page.

Bacterial cultures were grown and induced. Then, they were placed in a CLARIOstar® (BMG Labtech) microplate reader at 30°C overnight, following their respective microplate protocols. After overnight growth, data was collected and analysed. Data was initially plotted individually, with separate graphs for OD against time, and RFU against time to note for any major noise or issue in the data. Then, individual wells were processed by dividing the RFU value by its respective OD value at every time-point recording. These were then blank-corrected with LB (for OD) or E. coli TOP10 (for fluorescence) and averaged. The mean was plotted ± SEM. An example graph of processed RFU/OD data can be seen below:

Figure 1: The plot shows the mean RFU/OD ± SEM from three replicates of sfGFP constructs expressed in E. coli DH5⍺ and BL21(DE3). The OD was measured at 600 nm and GFP fluorescence was measured at Ex ƛ 485, Em ƛ 510, every 15 minutes for 13 hours. The RFU values were normalised by the OD600 and the triplicates averaged. OD values were blank corrected with LB and RFU values were blank-corrected with the negative control, E. coli TOP10 because it does not express any colour. A total of 52 recordings were made per well, with three technical replicates per construct.

From the graph above we can see how RFU values are comparable between constructs and between strains. They show a good comparison and the plot shows high quality, reliable data. The plot also shows clearly that sfGFP constructs (BBa_K2906000) express significantly more fluorescence than sfGFP with a secretion signal tag. Only sfGFP + OmpA (BBa_K2906001) shows limited levels of fluorescence, but invisible to the naked eye. All experiments done for RFU/OD were repeated at least twice in separate moments (data not shown).

This is just one example which we have selected to justify the use and the importance of normalisation when performing fluorescence measurements. In reality we used this technique for every single fluorescent construct, three for sfGFP and three for mRFP1, as well as characterisation of iGEM parts, with several biological and technical replicates to validate our results. This highlights how we can now compare results and provide an appropriate ratio of fluorescence by bacterial growth, generating data that can be used and compared by other labs even with different machinery.

Fluorescent Microscopy

Fluorescence microscopy is an essential tool to measure cell physiology. It allows researchers to monitor live events with spatial and temporal resolution. It works by directing a light of a specific wavelength to a sample, and measuring the emitted wavelength from the fluorophore. Background light is separated through a spectral emission filter and therefore the microscope can resolve fluorescent molecules in bacteria (Sanderson et al., 2014).

For this technique we used the same sfGFP samples as the ones stated above. We noticed the colour production of some of our constructs (sfGFP + HylA (BBa_K2906002)) was almost negligible (barely above the negative control as seen in Figure 1). Therefore, we decided to test if this was due to limitations in the fluorescence spectrometry measurements where we were obtaining a degree of background, or if the constructs are actually producing some fluorescent molecules. In order to further characterise our colour parts we decided to perform fluorescence microscopy (Nikon eclipse TE2000-U). This was carried out on sfGFP untagged (K2906000) and sfGFP + HylA (BBa_K2906002) since this last construct showed limited fluorescence in the microplate experiment described above. As a negative control we used E. coli TOP10 since it does not express any fluorescence.

For all the constructs we captured a phase contrast image and a GFP-filter image, these were then overlaid using ImageJ software. By doing this we could resolve and see specific bacterium that were producing the fluorescent proteins. Results for sfGFP untagged (BBa_K2906000) were in accordance to the microplate values, as it shows a high RFU/OD ratio and can be seen to have high density and fluorescence under the microscope. However, sfGFP + HylA construct (BBa_K2906002) does not appear to show fluorescence in the microplate results, but the fluorescent microscope images below seem to differ. They show that the bacteria are indeed expressing the protein, some even suggest that the protein is aggregating in the poles of the E. coli cells.

Figure 2. Fluorescent microscopy overlaid images of phase contrast and GFP filter taken from three different constructs. a) Negative control (E. coli TOP10) showing very slight background readings for fluorescence. b) sfGFP untagged construct (BBa_K2906000) showing significant levels of fluorescence in most bacteria. This supports the RFU/OD values obtained from the microplate. c) sfGFP + HylA (BBa_K2906002) showing limited fluorescence in some bacteria. Background GFP can still be captured suggesting that fluorescence is not as intense.

The figure above shows three constructs expressing different levels of GFP fluorescence. The last construct, sfGFP + HylA, shows how using only fluorescence spectrometry values can be misleading, and using a second technique to validate the results is necessary. However, only using fluorescence microscopy means that constructs cannot be easily compared to one another, and cannot be quantified appropriately.

This is just an example to show that using only fluorescence spectrometry has limitations on the conclusions that can be determined from its RFU/OD measurements. Hence, using only one technique for fluorescence characterisation is insufficient and using at least a second technique can be useful for validation of the results. This technique was widely applied to all of our sfGFP constructs in both E. coli DH5a and BL21(DE3), as well as all of our mRFP1 constructs in both E. coli strains. Finally, all of these methods show how different techniques, applied in tandem, create robust, reliable, good-quality results that are repeatable and shareable by other labs in the wider community. We have provided substantial evidence of this throughout our project (see colour for mRFP1 and complete sfGFP).

Spectrophotometry

Spectrophotometry is a well-established technique in biology. Spectrophotometers provide a quantitative measurement of the reflection or transmission properties of a liquid, material or substance as a function of wavelength (Spectrophotometry - Chemistry LibreTexts, no date). For our experiments, the use of spectrophotometers was destined not only at the detection of the Optical Density (OD) in relation to bacterial growth but also it was required for the quantification of our selected non-fluorescent chromophore (AmilCP).

However, we soon encountered two main obstacles:

1) The absorbance measurements of mRFP1 (or any other RFP-expressing bacteria) at OD600 have been shown to falsely increase apparent cell density. This effect has been reported to affect measurements with an increase as high as 10% (Hecht et al., 2016).

2) Our selected non-fluorescent blue chromophore, AmilCP, has a reported absorbance maximum of 588 nm (Alieva et al., 2008). However, we noticed that this value was situated remarkably close to 600 nm (typically used to estimate the concentration of bacteria and determine its growth). Therefore, with our experiments we were able to determine how making discrete measurements at 588 nm could not reliably allow the distinction between AmilCP-expressing bacteria as growth of the cultures was substantially increasing the background.

Therefore, we wanted to address these two concerns: first by proving if the theoretical concerns exposed above translated to our laboratory experiments. Then we wanted to find and provide a solution to those issues by making appropriate corrections to normal protocols.

Addressing our concerns:

Measuring optical density of RFP-expressing bacteria: Data and full characterisation shown in contribution page.

Initially, we performed RFP measurements of BBa_J23102, BBa_J23106 and BBa_K199118 (For full growth and induction conditions go to contribution page). When possible (except for T7 regulated BBa_K199118), this experiment was performed in two different E. coli strains DH5a and BL21(DE3) respectively.

Once samples were induced (when required- only BBa_K199118) they were placed on a microplate reader (CLARIOstar®, BMG Labtech). The microplate was set-up at 30°C and was set to record RFP fluorescence and absorbance at 600 nm every 15 minutes for 13 hours (full microplate protocol).

After overnight incubation in the microplate reader, data was collected and analysed. The data was corrected by LB blank (for OD values) and with our negative control, E. coli TOP10 (for RFU values) since this strain does not produce any colour. Then, for each well, the RFU value was divided by the OD value at the same time point. Since the constructs were loaded in triplicates, these were averaged and the standard error of the mean (SEM) calculated. Below we can see RFU/OD600 values for each evaluated construct in E. coli DH5a and BL21(DE3).

Data between OD600 and OD660 apparently varies between the two different tested E. coli strains. In DH5⍺, measurements at 600 nm are lower than measurements at 660 nm, compared to BL21(DE3). Therefore we can conclude that a potential false increase in apparent cell density was only significant for the BL21(DE3) strain and not in the DH5⍺. This means that from our results, measuring OD at 660 nm was only significantly more appropriate for E. coli BL21(DE3) cells. However, maybe a higher optical density such as 700 nm could allow a more accurate estimation of cell abundance in RFP-expressing bacteria (Hecht et al., 2016).

Detecting absorption of the non-fluorescent chromoprotein AmilCP.

Published literature has suggested that the maximum absorbance of the coral chromotein (AmilCP) is 588 nm (Alieva et al., 2008). We noticed that this value was particularly close to OD600 which is typically used to estimate cell density in bacterial cultures. Therefore, we hypothesised that measuring discrete wavelengths could result in a false detection of the chromoprotein (caused by background of bacterial growth).

Initially we decided to test our hypothesis by measuring discrete 588 nm wavelengths. In order to perform this experiment, the AmilCP coding sequence was cloned into the pBbB2c vector using type IIS assembly (for more information go to colours tab). Within pBbB2c (origin of replication pBR1) expression of the chromoprotein was regulated by the pTet promoter. Therefore, bacterial cultures needed to become induced with anhydrotetracycline in order to produce the chromoprotein. The AmilCP-containing plasmid was transformed into two different E. coli strains DH5a and BL21(DE3) respectively. This experiment was performed twice allowing the validation of our results and proving high reproducibility. The two performed replicates were treated the same, however they were induced at two different OD600 of ~0.4 (replicate 1) and ~0.6 (replicate 2). Below (Figure 7) we show the two replicate experiments which measured absorbance at 588 nm for the AmilCP construct. Additionally, different AmilCP-containing constructs were also tested throughout our experiments which allow to further confirm our hypothesis (data not shown, provided as Figure 25 and 26 in colours tab). As a control we used E. coli with pBbB2c-GFP in DH5a only uninduced to not produce any colour. The Cary 60 Uv/Vis (Agilent Technologies) was used for spectrophotometry without fluorescence.

Figure 7. E. coli replicate 1 (a) and replicate 2 (b) absorbance at 588 nm which is the absorbance maxima of AmilCP. Measurements were made at 5 different time points. The negative control was pBbB2c-GFP uninduced because it did not produce any colour. The plots show how the constructs do not produce a clear difference whether they are producing AmilCP (Colour Alone) or not (Negative control).

Through our experiment we were able to confirm our hypothesis. We noticed that, when measuring OD588 (Figure 7) all bacterial samples seemed to suggest AmilCP-expression as no significant difference is seen between the AmilCP-transformed bacteria and the negative control. This suggests that bacterial growth itself was causing interference in the readout, since its measurement is done at 600 nm, close to 588 nm. Therefore, from Figure 7 we were able to determine that measurements of AmilCP-expressing E. coli cultures at 588 nm are unable to provide an accurate representation of the chromophore ́s production.

We then decided to explore and try to find a solution to this phenomenon. To the best of our knowledge, the best way to measure AmilCP production is by spectral scan in order to see the relative increase in the absorbance at 588 nm compared to its surroundings. An example of a spectral scan from 300 nm to 800 nm can be seen below. The same bacterial samples as the above discrete measurements (shown in Figure 7) were used to perform spectral scans simultaneously. Additionally, the provided negative control was also the same. This allowed us to compare the spectral scan of AmilCP-producing bacteria, with bacteria that do not produce any colour. The data was retrieved and plotted.

Figure 8. Whole-spectra scans of replicate 1 a) and replicate 2 b) after overnight incubation of all different constructs in DH5⍺ and BL21(DE3). All of the overnight scans overlayed show that there is a mean absorbance peak at 588 nm (arrow). The negative control was pBbB2c-GFP uninduced because it does not produce any colour. The plots show how AmilCP-producing E. coli have a clear absorbance maximum around 588 nm. The mean absorbance maxima between the expressing construct was calculated and shown to be 588 nm. Analysis was performed with a spectrophotometer (Cary 60, Agilent technologies) from 300 to 800 nm.

By the use of spectral scans, as seen in Figure 8, we were able to detect AmilCP production, unlike with discrete 588 nm measurements (Figure 7). The scans were able to properly distinguish colour expressing bacteria from controls. This experiment was also repeated twice (with induction at ~0.4 and ~0.6 respectively), again this provides extensive, robust and reproducible data to prove that spectral scans are, to the best of our knowledge, the best way to detect AmilCP production in bacterial cultures. With these results we are also able to suggest a more specific spectral scan, from 500 nm to 650 nm since based on our results, that is the range where AmilCP maximum absorbance is located. The suggested specific spectral scan was determined taking into account the variations obtained with the different AmilCP versions tested (data not shown here; go to colours Figure 13 to 21).

Scent detection: Mass spectrometry (MS)

In the fragrance module of our project, we decided to engineer our dye-secreting bacteria to produce the volatile vanillin or limonene molecules to cover the unpleasant scent of the bacteria (see Fragrance). To detect these compounds and their intermediates, we decided to make use of the highly specific analytical technique, mass spectrometry (MS) which measures the mass-to-charge (m/z) ratio of ions typically presented as mass spectra (Maher et al., 2015).

UPLC-MS/MS

The metabolic intermediates in the vanillin biosynthesis pathway (Figure 9), p-coumaric acid, caffeic acid, ferulic acid, are phenolic acids. We used ultra-performance liquid chromatography (UPLC) coupled with tandem mass spectrometry (MS/MS) to detect and quantify these molecules. Compared to other analytical techniques, such as high-performance liquid chromatography (HPLC) , UPLC has shorter retention time, higher separation efficiency and enhanced peak resolution. This is because UPLC uses columns packed with smaller silica particles (~1.7 μm diameter) and requires higher operating pressure (Dass, 2007, Yu et al., 2006). Moreover, HPLC, and other analytical techniques, require complex extraction, pre-concentration and hydrolysis and/or derivatization procedures for quantification that often lead to oxidation or degradation of phenolic acids (Nardini et al., 2002).

Figure 9. Vanillin biosynthesis pathway from L-tyrosine consisting of five enzymes.