Design

Design - selection of receptors

The first step was to identify interesting and functional receptors along with the EDCs that can be sensed by the receptors from a literature search. This step proved very difficult and required us to read many articles relating EDCs to receptors. This sparked the idea for the software tool we have developed: An NLP tool that processes articles for you and creates a database of EDCs and their target receptors. This is explained in depth on our Software page.

We set the following requirements for these receptors:

• They should function as transcription factors directly.
• We need to be able to acquire a natural hormone for the receptor to test the system on.
• They should have multiple known EDCs related to them.

Using the NLP tool, we landed on working with six receptors for this initial system; ERα, ERβ, GR, AR, MR, and PR. Advantageously, these six receptors can be covered by only two different operator sites (HRE and ERE) and five different natural hormones. The system will be used in Saccharomyces cerevisiae (yeast) strain CENPK113-3C, as the receptors have previously been produced successfully in yeast [1]. This is a TRP deficient strain to allow selection for successful transformants.

In eukaryotic cells, these nuclear receptors are anchored to chaperones in the cytoplasm. Upon ligand binding, they dimerize, dissociate from chaperones, and translocate to the nucleus where they initiate transcription (facilitated by co-activators) [2, 3].

Design - interaction of receptor with DNA

This brought us to the second engineering challenge, concerning the interaction between the receptor and DNA. Literature searches showed us that the receptor response elements do function as singular inverted repeats, but that tandem repeats have been proved to have a much higher effect [1]. Previously, in-cell systems have successfully used five tandem repeats of the REs for different human hormonal receptors [1]. Therefore, the final tandem operator site sequences, denoted ERE5 and HRE5, are shown below.

The six receptors, their natural hormones and response elements are outlined below in Table 1:

Design - generating a measurable read-out

To complete the design of the first engineering cycle, we had to consider the measurable read-out that could be used for the in-cell system. We landed on a well-characterized beta-galactosidase enzyme as a reporter for the system, inspired by the protocol by Edwards, T. M. et al. [4]. Using ONPG as a substrate, activity of the β-galactosidase enzyme can be assessed by measuring absorbance at 405 nm.

Build

This system is simply a proof of concept to test the functionality of the receptors and their response elements, and therefore a simple plasmid-based system can be used. The plasmids utilized were purchased as complete plasmids from Addgene, originally deposited by Miller et al. [1]. The plasmids were later sequenced by an external company using nanopore. This allowed us to save time we could otherwise have spent cloning.

The plasmids are low-copy CEN/ARS plasmids containing a β-lactamase and a TRP1 gene for selection of successful transformants in E. coli and yeast, respectively. Additionally, they carry one of the hormonal receptors listed in Table 1 transcribed from a GAL1,10 or GPD promoter [1]. Finally, they contain a LacZ gene transcribed from a minimal cytochrome C (CYC) promoter. The receptor response element is located upstream of the CYC promoter. The full plasmid map is shown in Figure 1 below.

The S. cerevisiae strain CENPK113-3C was transformed with the plasmids using a heat-shock method (according to this protocol), which disrupts the cell membranes using LiAc. After the addition of the plasmids, the yeast was grown on a YNB medium without tryptophan to select for successful transformation carrying the TRP1 gene.

The design of the plasmid-based in-cell system is shown in Figure 2 below.

Test

To assess whether the system worked, a protocol was designed inspired by the protocol by Edwards, T. M. et al. (2018) [3]. As shown in Figure 2, the β-galactosidase enzyme was used as a reporter of EDCs, and ONPG was used as a substrate, which upon cleavage will absorb at 405 nm. For comparisons between samples, the protocol was based on the principle of measuring the OD610 and OD405 to be able to calculate values independent of the cell density. The complete protocol can be found in our Experiments page.

For data handling, the OD values were compared with appropriate media controls (without cells) and vehicle controls (without EDC). The final LacZ value was calculated as the relationship between the OD405 and the OD610 as shown below:

The initial tests consisted of creating dose response curves using the natural hormones for the receptors (shown in Table 1). The LacZ values were converted to a percentage of the maximum value for easy comparison between the curves. Unfortunately, we were not able to obtain progesterone to test the PR, and when testing the MR we saw unrealistically high standard deviations and no significant response. The four functional dose-response curves are shown in Figure 3 below.

Finally, the system was tested on four water samples. We tested two different types of bottled water, tap water, and a lake sample from a local lake. The data is shown in Table 2 below. As a positive control, the natural hormones (diluted in 50% ethanol) were added to the tap water sample to show that we can replicate the data from the dose-response curves. These positive controls are shown in Figure 3 and fit the dose-response curves nicely.

None of the measurements displayed in Table 2 are significantly different from the lowest value on the corresponding dose-response curve in Figure 3. Therefore we can conclude that none of these water samples contain EDCs binding the tested receptors above our measurement threshold.

If we had some results that showed a significant difference, the procedure would be to calculate sample hormone equivalents (EHQs) by fitting the dose-response curves to a logistic model and using the model to interpolate EHQs [1].

Learn

These experiments lead us to conclude three important things. First of all, the system functions, providing an intermediate proof of concept of the in-cell system. Importantly, four of the yeast-produced receptors functioned, and the design with tandem response elements (HRE5 and ERE5) functioned as expected.

Secondly, as seen in Figure 3, the biosensor sensitivity differs a lot between receptors. This is however not too important, as the sensitivity for the cell-free assay should be different.

Another conclusion was that while working with β-galactosidase as a reporter is reliable, it is also a slow and tedious process involving a lot of pipetting. Therefore, in the cell-free assay, another reporter method will be utilized.

Finally, while working with β-galactosidase as a reporter is reliable, it is also a slow and tedious process. Therefore, in the cell-free assay, another reporter method will be utilized.

Design

As mentioned, we faced many uncertainties in the design of the cell-free biosensor. We got a lot of answers from our stakeholders, but other uncertainties still persisted - the most relevant being that the human receptors, to our knowledge, were not fully tested in a cell-free environment. We therefore had the challenge of reducing the complexity of the environment of the receptors, designing a cell-free system around them.

For the structure of the biosensor, we took inspiration from the ROSALIND cell-free biosensor [5], modifying their design to match our goals. We kept the general idea of having a Transcription Factor (TF) altering the activity of a RNA polymerase, and the output signal as a consequence. We tailored the ROSALIND concept by selecting specific custom transcription factors (TFs) as receptors and designing unique operator sequences to serve as responsive elements. The exact parts are described on our Parts page.

Design - Working Mechanism of the Biosensor

When no EDC is present, the receptor won't bind the DNA, thus the T7 RNAP is free to transcribe the Broccoli aptamer. The aptamer binds to DFHBI-1T and enables fluorescence, by absorbing light at 472 nm and emitting it at 507 nm.

When an EDC is present, it binds the hormone receptor and induces a conformational change that will allow it to bind the receptor response element. Once the receptor is bound to the DNA, it will act as a repressor, suppressing the transcription.

Figure 4: Mechanism of the cell-free system. Output is shown when no EDC is present and when EDCs are present. Not to scale.

Build

DNA Templates: Synthesized and Assembled

In the interest of saving time, we decided to combine the minimal responsive elements with the single Broccoli aptamer, as it could be de novo synthesized and used. On the other hand, we coupled in our design the responsive elements with multiple binding sites (HRE5 and ERE5) with the more efficient 3WJdB double Broccoli aptamer (dB), as these templates needed to be manually assembled.

From the parts listed above (see Figure 4), we created and assembled four different DNA templates to be used in our EndoSense biosensor.

T7-HRE5-dB-T & T7-ERE5-dB-T

The DNA templates denoted T7-HRE5-dB-T and T7-ERE5-dB-T, carrying the same five-tandem repeat of response elements as in the in-cell assay, along with the more efficient double Broccoli aptamer (3WJdB), could not be ordered due to the presence of multiple repetitive sequences in both the response element region and the 3WJdB. As a result, these templates had to be manually assembled. Their structure is represented in the Figure 5 below:

Figure 5: Overview of the HRE5 and ERE5 DNA fragments used for the cell-free system. T7 promoter, response elements, aptamer parts and terminators are shown. Not to scale.

We used USER cloning to assemble the T7-HRE5-dB-T and T7-ERE5-dB-T. For the design of the assembly we used the AMUSER tool. The USER assembly consists of 2 steps - PCR and the USER reaction.

PCR

First, we ran PCR reactions using primers from the AMUSER tool for the following parts:

• pUC19-T7-3WJdB-T (Addgene)
• HRE5 (pRR-PR-5Z template)
• ERE5 (pRR-ERalpha-5Z template)

Each of the primers used to amplify a given response element part (HRE5, ERE5) was equipped with a specific USER overhang, complementary to the overhangs produced in plasmid backbone (pUC19-T7-3WJdB-T(Addgene)). For the PCRs, a Phusion U Hot Start polymerase that tolerates uracil bases was used. As a result, we obtained PCR products ready for USER cloning procedure.

USER cloning

USER cloning is a uracil-based excision technique that utilizes USER (Uracil-Specific Excision Reagent) enzyme to create specific 3’-overhangs on a DNA template. PCR products with double-strand USER overhangs (each containing uracil base) are subdued to the activity of USER and DpnI enzymes, resulting in an assembly of complementary overhangs and ligation of the templates.

In our case, each of the responsive elements (HRE5, ERE5) was cloned into a backbone plasmid pUC19-T7-3WJdB-T (Addgene). As a result, we produced two plasmids:

• pUC19-T7HRE5-3WJdB-T (pUC19_HRE5)
• pUC19-T7_ERE5-3WJdB-T (pUC19_ERE5)

Each plasmid was transformed into and amplified in E. coli strain DH5-α. The final ROSALIND templates (T7-HRE5-dB-T and T7-ERE5-dB-T) were obtained by PCR of the target sequences containing only the DNA parts necessary for Cell-Free Transcription System protocol out of the novel pUC19 plasmids.

Validation

To check the effectiveness of the assembly, we performed gel electrophoresis using HindIII and SspI restriction enzymes on the templates. The procedure produced bands of expected length for pUC19_HRE5, which sequence we later confirmed by sequencing the plasmid with Nanopore technology.

In the end, we were unable to confirm the proper assembly of pUC19_ERE5 plasmid.

T7-HERmin-sB-T & T7-EREmin-sB-T

The ready-to-be-synthesized DNA templates named T7-HREminimal-sB-T and T7-EREminimal-sB-T, carrying only one single binding site and the single Broccoli aptamer flanked by two scaffold, are shown in the image below (Figure 6):

Figure 6: Overview of the T7-HREminimal-sB-T and T7-EREminimal-sB-T DNA fragments used for the cell-free system. T7 promoter, response elements, aptamer parts and terminators are shown. Not to scale.

These DNA templates were ordered as G-blocks from IDT and were amplified with PCR.

Test & Learn

To test the created parts, we performed two iterations. Firstly, we tested and optimized the fluorescence output without the presence of any receptor or ligand (Test & Learn I), to ensure that the design at its basic level works properly.

Secondly, we proceeded by testing the biosensor on its whole with the receptor and the ligands (Test & Learn II).

For the testing process, we followed the protocols from our Experiments page.

Test & Learn I: Cell-free Transcription Optimization

The aim of this first iteration of testing is to ensure that all the build parts are behaving properly in the absence of both receptor and ligand, as well as optimize the output signal.

In our testing, we also used as a reference the T7-3WJdB-T template, without any responsive element, to use it as a benchmark in fluorescence readings.

Adjusting Wavelength and Plate reader setting

Wavelength and Plate reader setting

Rationale: As the signal for the first experiments was erratic and sometimes incoherent, we tried to improve the reading settings.

Result: Higher fluorescence output was yielded by:

• Using the wavelength couple 488/530 nm.
• Reading from the top (instead of from the bottom).

During the experiments so far, we encountered issues with the plate reader that we used. Non-zero values and high peaks were present also in empty or water-filled (blank) wells, and reaction wells had very erratic measurements, as they were below the detection limit of the machine. Once the settings have been changed, the readings were much better.

We recorded the measurements using 2 different couples of wavelengths: 477/507nm (as recommended from the ROSALIND protocol [8] and 488/530 nm (also used to measure the fluorescence of the fluorescein for reference). The latter wavelength couple yielded higher signal:

Figure 7: Wavelength comparison for different DNA templates. Fluorescein sodium salt was used as a reference.

The broccoli aptamer had clearly higher output on using the 488/530 nm couple of excitation/emission.

Buffer test

Buffer test

• Rationale: Optimize the reaction to increase the signal
• Result: The commercial In Vitro Transcription (IVT) buffer was better than the custom-made one.

Initially, for the first transcription test, we used a custom In Vitro Transcription (IVT) buffer recommended from the ROSALIND protocol (where we took the inspiration for the cell-free system). However, we didn’t get any fluorescence emission by using that custom buffer.

When using the commercial buffer, we could see a much higher output signal. The custom buffer clearly is not ideal for the cell-free transcription, while the commercial buffer seems to work much better. A possible explanation for this, is the ionic concentration, which is much higher in the custom buffer (especially NaCl). Either the indicated concentrations were wrong (we in fact acknowledge a mistake in the ROSALIND protocol, as the indicated concentration of the NaCl ion was too high) or we made a mistake in the process of preparing it.

DNA concentration: 10 nM

DNA concentration

• Rationale: Increase the sensitivity of the biosensor, enlarge the limits of detection, and reduce the cost of testing.
• Result: A concentration of 10 nM of DNA was sufficient to yield a substantial signal and the best one to perform the next experiments, also according to the Modeling analysis.

Reducing the amount of DNA enhances both sensitivity and the limit of detection, which are critical factors for our stakeholders. This reduction will ultimately lower the detection threshold, allowing even trace amounts of EDCs in the tested water to produce measurable inhibition. Additionally, it enables us to minimize the use of the receptor, the most expensive component of the biosensor system.

The cell-free transcription was carried out with different DNA concentrations, to assess the lowest one that would lead to an optimal detection:

Figure 8: Test for different DNA concentrations in order to have a readable and measurable signal. Measurements were taken after 1 hour for 20 minutes. Error bars represent standard deviation of triplicates.

We didn’t test for lower DNA concentration, as, for the first iteration, it was not recommended by the initial modeling analysis with the data from literature (see Modeling).

Effects of Interferents & Ligands in the assay

Effects of Interferents & Ligands in the assay

• Rationale: Scout and test the effects of inhibitors and interferents on the signal output.
• Result: Some compounds can negatively affect the aptamer production:
  • Testosterone (dissolved both in water and ethanol at 4µM concentration). We took into account this inhibitory effect when analyzing the data with the receptor present (a more detailed explanation is available in Modeling).
  • Sarkosyl (above 0.25%) [9];
  • Ethanol (above 2.5%) [10].

We tested the hormone dissolved in both ethanol and water to assess the influence of each solvent. Additionally, we examined the effect of sarkosyl, a surfactant included in the buffer of the purchased receptor.

To standardize and make comparable the values from different experiments, we included in each measurement a serial dilution of Fluorescein Sodium Salt (FSS), from the distribution kit, and normalized the values as Micromolar Equivalent Fluorescein (MEF).

These results are summarized in the Figure 9 below:

Figure 9: Effects of inhibitors and interference on the fluorescence output. The T7-HREmin-sB-T DNA template was used where DNA is indicated. The intensity of the signal is expressed as Micromolar Equivalent Fluorescein (MEF), more specifically Fluorescein Sodium Salt (FSS). Error bars indicate standard deviation of triplicates of measurements taken after 1h incubation with the T7 polymerase.

As the hormone itself has an influence on the fluorescence, we took that into account for the next tests. In particular, for testosterone dissolved in ethanol (used in future experiments), the signal was reduced to ~65% when compared to the transcription alone (from a Modeling analysis).

Moving on to sarkosyl, seeing the results prompted us to change the receptor buffer before using it for the reactions.

Learnings from Iteration I: Key Points

1. The commercial RNAPol Reaction Buffer was employed to optimize cell-free transcription.
2. The optimal wavelength for enhanced signal measurement was found to be 488/530 nm. Additionally, measuring from the top, rather than the recommended from the bottom (as recommended by the ROSALIND protocol [8]), resulted in significantly higher signals, improving the detection limit.
3. The lowest and best DNA concentration tested that produced a significant signal was 10 nM of the DNA template.
4. The hormone itself inhibits transcription, a factor that must be considered when comparing fluorescence in the presence and absence of the ligand, especially in the next testing phase, when the receptor is also used in the assay.

Test & Learn II: EndoSense Proof of Concept

For the second iteration of testing, we focused solely on the androgen receptor, as limiting the scope simplified the process (as we learned in the Cycle 1) and additional receptors were unavailable due to failed production and cancelled orders. Only the T7-HREmin-sB-T and T7-HRE5-dB-T, compatible with the androgen receptor, were tested using testosterone (the natural ligand of the receptor).

The main aim of this iteration is to validate the effectiveness of the biosensor in detecting EDCs.

Optimal Receptor Concentration & Validity of Assumptions

Optimal Receptor Concentration & Validity of Assumptions

Rationale: Determine the optimal receptor concentration for the best detection and investigate the mechanism of the designed biosensor, evaluating the assumptions previously made. As the receptor is our limiting resource, we prioritized this test, as being able to use less receptor would mean for us performing more testing.

Results:

Receptor concentration: The optimal concentration determined for the androgen receptor is in the range of 0.75-1 µM. To determine the optimal receptor concentration for the best detection, we measured and compared the fluorescence output at various receptor concentrations, analyzing the signal both in the presence and absence of the ligand.

After a first test with concentrations ranging from 0.025 µM to 2 µM, the most promising results were around 1 µM, so we decided to confine deeper testing around that concentration (see figure 12 below).

In fact, for receptor concentrations above 1.25 µM, there is a complete and irreversible inhibition of transcription, probably due to very stable binding of the receptor to the DNA. On the other hand, for concentrations below 0.5 µM of the receptor, there was no difference between the presence and absence of the ligand (data not shown).

In particular, the difference between the presence and absence of the ligand is the evident when the concentration of the receptor is 1 µM (see Figure 10 below), reason which is why we decided to further the analysis on this concentration.

Figure 10: Comparison of different receptor concentrations after 1h incubation with the T7 polymerase. R - androgen receptor, H - testosterone hormone (4µM in all reactions where indicated). The T7-HREmin-sB-T DNA template (at a concentration of 10 nM) was used where DNA is indicated. The signal has been adjusted considering the negative effect of the hormone on the fluorescence. Error bars, when present, represent standard deviation of triplicates (duplicates only for receptor concentration 0.75 µM).

For all receptor concentrations here tested, there is a noticeable difference between the absence and presence of the ligand, and it is more relevant the higher the receptor concentration (although it needs to be below 1.25 µM). For this reason, we chose to proceed with the 1 µM concentration.

More data is required for the 0.5µM concentration, as it was only performed with a single replicate.

Validity of our assumptions

Firstly, upon adding the receptor, we observed a significant decrease in the signal, invalidating our initial 2nd assumption about the receptor-DNA interaction. It is clear that the receptor binds to the DNA and inhibits the signal even in the absence of the ligand.

However, our first assumption regarding the receptor's interaction with RNA polymerase was confirmed. Even in the presence of the ligand, the signal is lower compared to when the receptor is absent, validating this part of our hypothesis.

The validity of our assumptions can be summarized as follows:

• Absence of Ligand → No inhibition: Invalidated. The receptor binds to the DNA and inhibits the signal even in the absence of the ligand.
• The Receptor is not able to recruit the T7 RNAP: Confirmed. The signal is lower in the presence of the ligand, validating this part of our hypothesis.

Uncertainties	Assumptions	What we learned
Receptor & T7 RNAP: Which interaction takes place between these two entities?	The Receptor is not able to recruit the T7 RNAP	Due to the different conformation, the T7 RNAP will not be recruited by the nuclear receptor as it happens for the eukaryotic RNAP (also because the co-activator proteins are absent). Correct.
Receptor: Which conformation binds the DNA?	Absence of Ligand → No inhibition	The receptor will be able to bind the DNA only after the dimerization induced by the binding to the ligand, as that is what happens in the human cells. Wrong: The receptor can bind to the DNA even when the ligand is not present. No dimerization is needed for the binding.

Revised Working Mechanism: OFF/ON Type

In light of these results, we reconsidered the working mechanism of the biosensor: It behaves as an OFF/ON type, instead of ON/OFF. Which is in fact the opposite of what we were expecting: An increase rather than a decrease in the signal. When the receptor alone is added, it binds to the DNA template, inhibiting the transcriptional activity of the T7 RNAP.

On the other hand, when also the ligand is present, the affinity of the receptor for the DNA seems to be lower, and, as a consequence, the activity of the T7 RNAP is partially restored, resulting in an increase of the signal. This suggests that the receptor is not able to bind DNA when the ligand is present, similarly to what happens in case of the Lac repressor when lactose is added. This behavior is most likely due to a change of affinity for the DNA when ligand is bound to the receptor. This working mechanism is shown in Figure 11 below.

Figure 11: OFF/ON behavior of the EndoSense biosensor. When the water sample does not contain EDCs, the fluorescence signal is inhibited. When EDCs are present, this will result in the transcription of the broccoli aptamer and ultimately a fluorescence signal.

The identification of an OFF/ON working mechanism significantly enhances the robustness of the EndoSense biosensor. Unlike the ON/OFF biosensor, which we discussed during our meeting with DTU Associate Professor Gerd, this design avoids the flaws inherent to the ON/OFF system. In fact, when the signal is increased, this also validates that the assay is working properly, and that there is definitely an EDC in the water sample.

Proof of Concept & Predicted Limit of Detection (LOD)

Rationale: Even though the working mechanism of the biosensor is opposite than designed, we can still see a difference between the presence and absence of the ligand, which we want to further investigate.

Results: The biosensor is able detect with statistical significance the presence of the ligand.

As the difference between the presence and absence of the ligand is more evident from the receptor concentration of 1 µM, we examined it more deeply.

As represented in Figure 12 below, the significance (expressed as the p-value of a t-test between the triplicates in each time point) is below the 0.05 threshold for all measurements, and below 0.01 after 45 minutes the addition of the T7 RNAP.

Test with EDCs

Test with EDCs

Rationale: After having demonstrated that the EndoSense biosensor can detect the natural ligand of the receptor (testosterone), we wanted to test it with some proven EDCs, which we got thanks to our collaboration with the 2024 2024 UCopenhagen iGEM team. In particular, we tested with PCB (polychlorinated biphenyls) and BAC (Benzanthracene).

Results: Unfortunately, we noticed that when the receptor was added, there is no decrease in the signal, regardless of the hormone presence, as can be seen in the Figure 13 below:

Figure 13: Measured signal for two different DNA templates (DNA1: T7-HRE5-dB-T grouped on the left; and DNA2: T7-HREmin-sB-T grouped on the right), addition of androgen receptor (R) and testosterone (H). For the T7-HREmin-sB-T two different EDCs have been tested: Polychlorinated biphenyls (PCB) and benzanthracene (BAC).

From this observation, we concluded that the receptor in this experiment was denatured and not functional, likely due to the buffer exchange process and/or repeated freeze-thaw cycles. Thus, the results regarding the detection are not to be considered in this experiment.

Despite this, the data clearly indicate a negative effect of the tested EDCs on the transcription process. With the receptor's functionality ruled out, it is evident that the EDCs directly influenced the fluorescence output.

This finding suggests that, for future experiments, we should normalize the signal using samples with the EDCs, following the same approach applied with the hormone in earlier tests.