Here we have gathered the steps leading to experimental evidence for the functionality of the NANOFLEX system. Learn more about its design here, the experimental details of previous stages here and about the background of the stochastic model here.

Introduction

In the presence of the target antigen, the cells containing the plasmids pSB3K3+DBD-CaFF and pSB1C3+pCadBA-mRFP1 should induce the expression of the selected reporter. Only once the basic functionality is tested it is reasonable to propose improvements for the system, like using a different nanobody or a reporter gene of faster response such as lacZ. We want to be able to detect caffeine at very low concentrations to prove that the system is functional and useful in all relevant conditions. The main questions that need to be answered are:

• Is NANOFLEX reporter expression induced by the targeted antigen?
• Does the device affect cell growth in any way?
• How much time does it take to observe the induction?
• Can we model and predict NANOFLEX behavior?
• Does the system still work as expected while using a complex sample?
• Can the system be transported and stored for later use?

Results

Dimerization of DBD-CaFF upon caffeine binding triggers mRFP1 expression

The cell growth and fluorescence signal upon exposure of the NANOFLEX device to the different caffeine concentrations was measured over time when using the caffeine induction assay. You can read more about this assay here (Pages 1 and 2 of the notebook). The growth curve was similar at different caffeine concentrations, showing that these caffeine concentrations were not toxic to the cells and that the stationary phase is reached at 3.5 hours of growth (Figure 1A-B).

The engineered strain of E. coli DH5α reacts to the presence of caffeine in different concentrations (Figure 1C). When 1 μM caffeine is present (blue), the levels of fluorescence are significantly higher compared to the background signal generated in the absence of caffeine (black). Higher caffeine concentrations (light blue, purple, apricot, and red) cause an earlier appearance of the signal distinguishable from the background and higher fluorescence levels later on. No increase in the fluorescent levels can be seen when the results of 20, 50, and 100 μM caffeine concentrations are compared, which testifies about the saturation of the system at 20 μM. During measurements, cells were cultivated in the humidity chamber of the Spark 10M microplate reader (Tecan), at 37 ºC, using Nunclon 96 Flat Black plates (Thermo Fisher Scientific). Fluorescence was measured by measuring emittance at 611 nm after excitation at 585 nm, and normalized by the optical density measured at 700 nm after subtraction of the LB media fluorescence values. Each dot represents the average of two biological replicates and error bars are respective standard errors. After the culture reaches the stationary phase, the cells show induction of mRFP1 reporter expression from concentrations as low as 1 μM, reaching saturation at 20 μM (Figure 1D).

Cells were cultured and induced with caffeine on the same conditions as mentioned before but on a higher volume of 5 mL. The cells were pelleted and the supernatant discarded. The pellet in all samples showed similar weight, showing that the amount or size of cells does not account for the increase in fluorescence (data not shown). The cells containing only the plasmidpSB1C3+pCadBA-mRFP1 do not seem to express the fluorescent protein at all, and the gradient of mRFP1 expression at different concentrations can be seen in Figure 2. This proves that DBD-CaFF is necessary for the expression of the reporter gene and that the presence of caffeine regulates the strength of the expression.

The experimental results are encouraging, but it is problematic that every time we change one of the components of the system we need to titrate the sensitivity of the biosensor. Some of the variables and equilibrium constants of the system are known experimentally or can be estimated after fitting the model with the experimental data. A model that allows us to predict the level of reporter expression reliably would be invaluable to functionalize the NANOFLEX system in biosensor applications and to accelerate the development of improvement on the system. We have used a model using stochastic differential equations (SDEs) for the simulation of mRFP1 production shown in the previous section of the proof-of-concept. The stochastic model formula can be seen below.

Stochastic model formula

• R_dimer_pro_syn:
  dimercaff + free > dimerpromoter

  k1*dimercaff*free

• R_dimer_pro_deg:
  dimerpromoter > dimercaff + free

  k2*dimerpromoter

• R_mRNA_syn:
  dimerpromoter > mRNA + dimerpromoter

  kmRNA_syn*dimerpromoter

• R_mRNA_deg:
  mRNA > $pool

  kmRNAdeg*mRNA

• R_Prot_syn:
  mRNA > Protein + mRNA

  kProtsyn*mRNA

• R_Prot_deg:
  Protein > $pool

  kProteindeg*Protein

• R_MatureProtein_syn:
  Protein > MatureProtein

  kProtmat*Protein

• R_MatureProtein_deg:
  MatureProtein > $pool

  kProtmat_deg*MatureProtein

The formulas above describe the stoichiometries and rates of each reaction in the model.

We can regulate the estimated amount of caffeine in the model by changing the dimercaff variable since the Kd of dimerization is much higher than the Kd of caffeine binding (2). The leakiness of mRFP1 expression is accounted for by maintaining a background minimum level of dimercaff.

A limited number of promoters was set, which are categorized into free promoters (variable free) and DBD-CaFF-bound promoters (variable dimerpromoter). In the case of the backbone pSB1C3, the literature shows that it has a copy number per cell of between 100 and 500. We assume in our model the amount per cell as 100 promoters in total.

Values of 4, 472, and 4720 copies per cell of caffeine were chosen because they correspond to the concentrations tested in the lab, 0.1µM, 1µM, and 10µM.

The simulations were run at values of K1 of 0.001, 0.01 and 0.1 because this variable is not known experimentally (Figure 3). Only a few of them will be presented here as relevant to the proof of concept. At concentration 4 units of caffeine per cell, the K1 value determined the maximum level of mature protein until a maximum at K1 = 0.1 with about 600 copies per cell. There is a reduction in that result at K1 = 1 with 500 copies per cell (data not shown). At 472 and 4720 the system was not responsive to changes in the value of K1, indicating saturation of the system. The results at K1 = 0.1 mirrors the behavior observed in the empirical data within an order of magnitude for the concentration that causes saturation of induction.

NANOFLEX can reliably respond to a complex sample containing caffeine in a predictable manner

The caffeine assay was also performed using actual coffee in order to prove the detection of caffeine within a complex sample (Pages 3 and 4 of the notebook). Decaffeinated coffee was also used since it is not completely caffeine-free but it contains a lower concentration and thus should give a lower level of induction using the same amount of sample. According to Livsmedelsverket, there is about 44.5 mg of caffeine per 100 mL of coffee (1). Thus, one cup of coffee has a concentration of around 2.5 mM caffeine (C), that added using the induction assay protocol gives a final caffeine concentration of 50 μM (see notebook). Decaffeinated coffee have an approximated cup concentration of 0.1 mM (dC), and a final assay concentration of 2 μM. According to the data in Figure 1, using this concentration in the induction protocol is sufficient for the induction of mRFP1 expression. The coffee used for the experiment was Lindvalls Kaffe Tricole, and Gevalia Mellanrost Koffeinfritt was used as the decaffeinated control. The samples did not affect growth in any significant way (Figure 4A-D). The plots were organized as separate biological replicates to allow for easier understanding. However, the sample of replicate B that has no coffee sample at all has an extremely high standard error and technical variation, and therefore it was taken into a separate graph so that its error bars do not block all the others (Figure 4E).

Induction of expression of the reporter was observed according to the expected results (Figure 5A-B). The samples in absence of C or dC had the lowest levels of fluorescence while the other samples saw an increase in fluorescence (data not shown). The induction using a similar dilution of 1/50 C (1μM) was higher than for the dilution 1/50 dC (0.04 μM), but almost the same as dC (2 μM). The induction using the normal concentration of C was higher than that of dC. The induction of the second gave a similar value to the expected one, but it wouldn’t be unexpected if the result was different due to the fact that the nanobody that we use also bind to other molecules of the caffeine pathway like theobromine, albeit with a lower affinity (2). In any case, these results only confirm that the system is responsive to caffeine and sensitive to the concentration of caffeine in the sample.

Lyophilization of NANOFLEX is possible for potential storage and delivery

Conclusions

Induction of the Nanoflex system in the presence of the target antigen was observed. The results were replicable, and the induction was responsive to the concentration of caffeine. Moreover, the system maintained the expected responsiveness against a complex sample that contains many different compounds. These are encouraging results that indicate the sensitivity and specificity of the system is mainly determined by the nanobody attached to Nanoflex, with few off-target binding. The ASSURED principles are generally used for evaluating the effectiveness of diagnostic components (3), and here we have proven two of those requirements.

The stochastic model was capable of predicting the concentration of caffeine that saturates mRFP1 maximum amount within one order of magnitude, which suggests that the system can be accurately modeled as long as reliable empirical data on the binding affinity of new nanobodies, promoter strength, and other biochemical parameters are provided.

After cell lyophilization, cell growth and mRFP1 expression was achieved after 1 week of storage. According to the literature, cells in these conditions maintain viability and functionality even after 1 year of storage. The possibility of long-term storage and transport is necessary to deliver any potential biosensor derived from Nanoflex to the end user.

Overall, many of the desired targets of NANOFLEX have been achieved, but there are still improvements to be done. It’s important to suppress the leaky expression in the absence of the target antigen in order to avoid false readings from a user. The reporter gene should be one that gives a stronger and faster signal than mRFP1 for fast and clear understanding from a non-expert. Multiple nanobodies of medical, economical or of other relevant interests have to be tested to prove the validity and usefulness of NANOFLEX.

References

1. Livsmedelsverket (2020). Koffein. https://www.livsmedelsverket.se/livsmedel-och-innehall/kosttillskott/amnen-i-kosttillskott/koffein?AspxAutoDetectCookieSupport=1 (Accessed 26 October, 2020)
2. Sonneson, G. J. and Horn, J. R. (2009). Hapten-induced dimerization of a single-domain VHH camelid antibody. Biochemistry. 48(29), 6693-6695
3. Smith, S., Korvink, J. G., Mager, D. and Land, K. (2018). The potential of paper-based diagnostics to meet the ASSURED criteria. RSC advances. 8(59), 34012-34034
4. Prévéral, S., Brutesco, C., Descamps, E. C., Escoffier, C., Pignol, D., Ginet, N. and Garcia, D. (2017). A bioluminescent arsenite biosensor designed for inline water analyzer. Environmental Science and Pollution Research. 24(1), 25-32