Bacteria Detection at a Single-Cell Level through a Cyanotype-Based Photochemical Reaction

The detection of living organisms at very low concentrations is necessary for the early diagnosis of bacterial infections, but it is still challenging as there is a need for signal amplification. Cell culture, nucleic acid amplification, or nanostructure-based signal enhancement are the most common amplification methods, relying on long, tedious, complex, or expensive procedures. Here, we present a cyanotype-based photochemical amplification reaction enabling the detection of low bacterial concentrations up to a single-cell level. Photocatalysis is induced with visible light and requires bacterial metabolism of iron-based compounds to produce Prussian Blue. Bacterial activity is thus detected through the formation of an observable blue precipitate within 3 h of the reaction, which corresponds to the concentration of living organisms. The short time-to-result and simplicity of the reaction are expected to strongly impact the clinical diagnosis of infectious diseases.


Supplementary information S1: Evaluation of the role of cyanotype precursors.
Due to the importance of understanding each and any interaction between the various reagents of the photochemical reaction, experiments were conducted simultaneously in light and dark and using controls for every combination of reagent, which used at the optimal concentration determined previously. The tests were run for a total of 5 hours, and measurements were taken every hour. In the figures, MH corresponds to Mueller-Hinton culture medium, Bac to bacteria, FeCN to ferricyanide and citrate to iron citrate. The first set of experiments was performed without illumination to evaluate cross-reactivity between reagents, as summarized in Fig. S1. The four graphs on the left show the lack of reactivity between MH culture medium, ferricyanide and iron citrate since these plots did not present changes in the absorbance spectra over time.
Only minor changes were observed in graph G after 4 hours of incubation, which were attributed to some contamination. The four graphs on the right replicated the experiments in the left but including bacteria to study their influence in the reaction. The same spectral variation was observed by the four plots, which was due to bacterial scattering. The impact of biomass scattering on absorbance spectra was very important (i.e. broadband change in the full wavelength range under study) and prevented the observation of additional changes associated to other reactions between reagents. However, due to the similarity between plot B, only containing bacteria proliferating in culture medium, and the other three plots, it may be concluded that it was not Prussian blue formation in any of the experimental conditions previously studied, demonstrating that the light was key for the activation of the photochemical reaction.
The same experiment was performed simultaneously, but with continuous visible light illumination, with the results summarized in Fig. S2. Interestingly, no Prussian Blue formation was observed in the four plots in the left, where all reagents were combined and continuously irradiated in absence of bacteria.
Thus, although the mechanism was intrinsically photocatalytic, the presence of light was not enough to activate the photochemical reaction and Prussian Blue formation. The most remarkable observation in the plots in the left (without bacteria) was that the absorbance peak attributed to ferricyanide (420 nm) slightly decreased over time, probably due to photochemical degradation (graph C and G). When bacteria are present (plots in the right), the spectral response depended on the composition of the medium. Thus, the same increase associated to a conventional bacterial proliferation obtained in the experiments performed in the dark is replicated during irradiation by graphs B, only containing bacteria and MH culture medium, and F, incorporating iron citrate. This indicated that light, bacteria and iron citrate was not enough to activate the photocatalytic formation of PB. Similarly, the graph D, resulting from the reaction of ferricyanide and bacteria in light conditions, presented a bacterial proliferation even lower than in the dark. This confirmed the previous hypothesis suggesting that light may induce some photo-degradation of ferricyanide, probably inducing the release of cyanide ions that killed bacteria and reduce their proliferation. This was in agreement with previous publication

Supplementary information S2: Evaluation of the presence of free iron (III) molecules during the cyanotype-based reaction.
The presence of free iron ions was of key relevance to elucidate the mechanisms of the cyanotype-based photochemical reaction here described, but not possible to attain through spectroscopy directly. For this reason, an additional experiment was performed to evaluate the oxidative state of the iron species produced in the photochemical reaction from the ferric ammonium citrate used as free iron source. The experiment performed is summarized in  Due to impossibility to measure free iron (III) ions, two hypotheses were suggested: free iron (III) were released and rapidly reacted with iron-cyanide complexed to form Prussian Blue.

Supplementary information S3: Study of the proportion and concentration of the precursor solution components.
In cyanotype, ferric ammonium citrate concentrations ranging from 0 to 40 mM were tested against ferricyanide solutions containing between 0 and 40 mM, both in dark and in light. Data belonging to time 0 and after 5 hours of incubation in 5x10 5 CFU/mL of E. coli ATCC 25922 are shown in Fig. S4 and

coli ATCC 25922, after an incubation time of 5 hours Each group of columns refers to a single ferric ammonium citrate concentration. Each colour corresponds to a specific ferricyanide concentration. Data is plotted at a wavelength of 720 nm for A -D. (n=3)
In the absence of light, the lack of PB formation at any of the conditions was apparent. The constant signal for any of the reagent concentrations in the absence of bacteria indicated that the reagents did not react with either the media or between each other. When bacteria were present, no apparent PB formation was observed either. The only exception was when combining 40 mM citrate with a 40 mM FeCN concentration. The high concentration of both reagents in the precursor solution led to the formation of PB, even if it was very little (see the small band at 720 nm shown in Fig. S4 F).
The absorbance in samples with bacteria was higher than without ( Fig. S4 E and F), which was associated to bacterial scattering. Although there were no significant differences between absorbance values at any of the reagent conditions (p<0.05), in the case of the precursor solution containing 40 mM FeCN without citrate, a slight reduction in absorbance was observed. This may be attributed to some toxicity of the ferricyanide when used at high concentrations, as is the case here, which limited bacterial proliferation, reducing the magnitude of bacterial scattering. It was thus evident that a ratio of 8-1 (citrate:FeCN) was optimal for PB formation, which differed from conventional cyanotype processes for photochemical production of PB where a 3 -1 ratio (approximately 760 mM citrate and 240 mM FeCN) is reported as optimal. Also relevant, a minimum of 10 mM citrate was necessary to produce detectable PB concentrations within 7 hours of experiment. All of them presented low cyanotype precursor concentrations and a molar ratio between 4 and 8. From the three, the first one was discarded since the increase was not significantly higher than the biomass itself.

Supplementary information S4: Evaluation of the cross-reactivity of the cyanotype-based reagent with serum and blood components.
This section shows a first step towards testing in real samples, using human serum (obtained from Sigma Aldrich) and pig's blood. The experiments performed in serum (Fig. S6) and blood (Fig. S7) were conducted under the experimental conditions optimized before (2.5 mM citrate -0.625 mM FeCN) as a proof-of-concept. Serum and blood were spiked with bacterial samples and diluted half with the precursor solution (in MH). Since the pH of the solution was crucial for the assay (PB dissolves at basic pH), three separate experiments were performed where the pH of the MH was adjusted to 6.6, 6.1 and 5.5 respectively. Even though three very distinct samples of MH were prepared, the resulting pH after mixing with the serum was 7.0 ± 0.1 due to their buffering effect resulting in no significant differences between experiments. Results from the experiment are plotted in Fig. S6. Both experiments (i.e. with and without PB precursors) showed a conventional bacterial proliferation curve with the lag, exponential and stationary phases clearly visible. The main difference between samples was again the absorbance magnitude, this being higher in the samples containing precursor reagents due to the signal amplification associated to PB formation.
The absence of the sudden jump after 3 -4 hours of incubation in the samples containing PB precursors may be associated to some matrices effects. Serum components may capture some of the components of the assay and therefore, the proportion and concentration of precursors should be optimized in this complex matrix. Anyway, samples did not show cross-reactivity and thus the assay may be conducted directly in serum samples without any pretreatment. Finally, a cyanotype based test was also performed in complete blood samples (Fig. S7). As with serum, all bacterial concentrations followed a conventional proliferation curve with the amplification associated to PB formation. It confirmed that the assay could be performed in complete blood without interference of any of the blood components, opening the possibility for direct bacterial detection in blood samples. One of the aspects to be improved was reagents proportion since, as also observed in the case of serum, the initial sudden increase that reported on the presence of low bacterial concentration, was not obtained in this case. Two more aspects should be remarked from these results. First, as commented in other situations with low bacterial concentrations, the 10 1 CFU/mL sample converted the measurements into a probability assay, resulting in a growth curve with big error bars. Second, the blank line was not completely stable but presented some increase that may be due to some late reaction (after 11h of reaction) between precursor components to produce PB. Therefore, the results did not indicate cross-reaction between precursor reagents and matrix components allowing bacterial detection in complex matrices such as serum and blood directly and without any pretreatment.