Triplet-Triplet Annihilation Upconversion Based Oxygen Sensor to Overcome the Limitation of Autofluorescence

Abstract

Autofluorescence is one of the many challenges in bioimaging, as it can mask the emission from fluorescent probes or markers – a limitation that can be overcome via upconversion. Herein, we have developed a nanosensor that uses triplet-triplet annihilation upconversion (TTA-UC) to optically report changes in dissolved oxygen concentration. Using a sensitizer-annihilator dye pairing of Platinum (II) octaethylporphyrin and 9,10- diphenylanthracene, we monitored the oxygen consumption (as a proxy for metabolic activity) over time in a biological system - Saccharomyces cerevisiae (brewing yeast). The nanosensors demonstrated good reversibility over multiple cycles, showed good signal and colloidal stability when tested over the course of 7 days, and were sensitive to dissolved oxygen from 0.00 mg/L to 3.17 mg/L O₂. Additionally, there was no signal overlap between the nanosensor emission and Saccharomyces cerevisiae autofluorescence thus underscoring the utility of upconversion as a facile and economical means of overcoming autofluorescence.

Graphical Abstract

Oxygen (O₂) is an essential component for life in all aerobic organisms on earth ^1,2. It plays an important role in mitochondrial respiration, oxidative phosphorylation ¹ and can act as an indicator of microbial activity ³. Therefore, the ability to quantify oxygen concentration is advantageous to various fields including medicine, chemistry, microbiology, environmental science, and food packaging ^2,4.

While there are many ways of detecting and quantifying oxygen, a common approach is via an electrochemical method (Clark electrode) ^5–7. This technique reports the concentration of dissolved oxygen by the chemical reduction of oxygen at an electrode. However, there are a few drawbacks to this method that include consumption of the analyte, spatial and temporal limitations of the measurements, and the need for the probe to be invasively deployed ^6,7. Furthermore, electrodes can be fragile when developed on the microscale, which makes them challenging to deploy into biological systems ⁸. As a result, optical sensors are gaining popularity as an alternative to electrochemical detection approaches. These optical sensors usually incorporate oxygen sensitive dyes whose luminescence is quenched by molecular oxygen and can manifest them-selves in various forms that include, optode films ^9,10, sol-gel matrices ^11,12, fiber optic sensors ^7,13,14, free dye systems or nanosensors ^15–17. Work in oxygen sensing has been reviewed in depth by others ^4,6.

Nanosensors are an exciting tool for in vivo imaging as they overcome some of the limitations of the Clark electrode. Their small sizes (~100 nm diameter) allow for them to be non-invasively deployed in a complex biological matrix and their high surface area to volume ratios allow for enhanced analyte interactions and faster response times than bulk optode films ^18,19. Additionally, they are modular which means their sensing components can be altered to enable detection of an analyte of choice, e.g., sodium, potassium, iron, oxygen ^20–23. However, one of the challenges of biological imaging is autofluorescence. The presence of endogenous fluorophores in the specimen ²⁴ can mask the emission from fluorescent probes or markers ^25–27. Fortunately, there are a few methods of overcoming autofluorescence as outlined in the review by Del Rosal et. al. these include the use of time-gated imaging, near-infrared reporters, persistent luminescence and upconversion ²⁵.

Upconversion works on the principle of an anti-Stokes shift phenomenon, where a low energy excitation source is used to produce a high energy emission ^28,29. In this paper we will be focusing on triplet-triplet annihilation upconversion (TTA-UC), but for a more detailed coverage of other upconversion mechanisms we recommend the review by Zhou et. al ³⁰ and Yanai and Kimizuka ³¹. For TTA-UC to occur energy transfer needs to occur between a sensitizer (donor) and annihilator (acceptor) molecule (see Figure 1) ²⁸. The sensitizers need to be able to absorb low-energy light and must have long excited triplet state lifetimes while the annihilator needs to have a triplet state below the sensitizer’s triplet manifold, and a singlet state higher than that of the sensitizer ²⁸. In general, a larger energy difference between the triplet sensitizer and annihilator is favorable to the energy transfer process. In the TTA-UC mechanism, the sensitizer must be excited such that intersystem crossing (ISC) occurs from the excited singlet state (S₁) to a triplet state (T₁). If oxygen is present, the sensitizer’s luminescence and the TTA-UC process is quenched, but in the absence of oxygen, a triplet-triplet energy transfer (TTET) from the sensitizer to the annihilator occurs (T₁ to T*₁) ^28,32. Upon annihilation, a delayed fluorescence occurs from the now populate singlet state of the annihilator (S*₁) ²⁸. The concept of TTA-UC is not a new one as it has been used in various system such as fiber optic cables, semiconductors, solar cells, and optode films as conjugated dye systems ^33–36. However, the implementation of this technique in nanosensors is still limited. There are a few reports of this mechanism being used to report pH, enzymatic and K⁺ activity ^23,37,38, but to our knowledge reports of nanoparticles that report changes in oxygen concentration are somewhat limited.

Figure 1:

A Jablonski schematic showing the triplet-triplet annihilation upconversion mechanism. This includes the excitation, emission, and quenching pathways for both sensitizer (PtOEP) and annihilator (DPA). Note, * represents the annihilator singlet (S) or triplet (T) states.

As a result, in this work, we leveraged the TTA-UC mechanism to make an oxygen nanosensor. We encapsulated an oxygen sensitive dye system in a hydrophobic nanoparticle core, which is surrounded by an amphiphilic co-block polymer enabling the particle to be suspended in aqueous systems ³. To demonstrate its functionality, the sensor was deployed in a biological system - Saccharomyces cerevisiae (brewing yeast), where the oxygen consumption (used as a proxy for metabolic activity) of yeast was monitored over time.

EXPERIMENTAL SECTION

Materials.

For nanosensor optode fabrication, platinum (II) octaethylporphrin (PtOEP) was purchased from Frontier Scientific (Logan, UT, USA), 9,10- diphenylanthracene (DPA), vitamin E acetate (VEA), and tetrahydrofuran (THF, inhibitor free) were purchased from MilliporeSigma (St. Louis, MO, USA). The poly(styrene)-b-poly(ethylene oxide), (PS−PEG or PS_1.6k-b-PEO_5k) was obtained from Polymer Source, Inc. (Montreal, QC, CA).

For the glucose oxidase assay, α-D-Glucose (G) and Glucose Oxidase from Aspergillus niger (GOx), and Dulbecco’s Phosphate Buffered Saline (PBS, dry powder, modified, without calcium, without magnesium, suitable for cell culture) were purchased from MilliporeSigma (St. Louis, MO, USA). The assay was run in 96-well black-walled, clear bottom optical bottom plates, which was obtained from ThermoFisher Scientific (Waltham, MA, USA).

To conduct the yeast assay, a Saccharomyces cerevisiae (brewing yeast) strain MIP-510 (Kolsch I) and MIP-354 (Kveik Oslo) was obtained from Propagate Labs LLC (Golden, CO, USA). The yeast pack had 200 billion cells per mL in roughly 200 mL of media. A can of Fast Pitch wort was acquired from Northern Brewer (Roseville, MN, USA), 0.22 μm polyethersulfone membrane Luer-lock filters (13 mm diameter, CELLTREAT brand) were obtained from Fisher Scientific (Waltham, MA, USA). Campden tablets (potassium metabisulfate, PMB) were purchased from Crosby and Baker (Westport, MA, USA). 0.5 mL 30 kDa molecular weight Amicon Ultra filters were procured from MilliporeSigma (St. Louis, MO, USA).

For pH testing, 10 N HCl, and 10 N NaOH was obtained from ThermoFisher Scientific (Waltham, MA, USA). Universal buffer was made in accordance with the procedure described by Xie et. al ³⁹. The pH values were verified and adjusted using a Fisherbrand accumet AB150 pH meter and probe.

Nanosensor Fabrication.

Using the flash nanoprecipitation (FNP) method described by Tien et. al. ³, all dyes, core, and shell components were added to a vial with THF solvent to form the optode solution. In short, 0.03 mg of PtOEP, 0.6 mg of DPA, 250 μL of 20 mg/mL VEA in THF (5.0 mg), 250 μL of 20 mg/mL PS-PEG in THF (5.0 mg) were mixed together. Additional THF was added to the vial such that the final volume of THF in the optode was 1 mL, and that there was 5 mg/mL of VEA and PS-PEG each in the optode mixture.

In separate syringes, 1 mL of optode and 1.2 mL of PBS were added to a confined impinging jet (CIJ) mixer (Holland Applied Technologies, Burr Ridge, IL) and fabricated in 8 mL of PBS (quench bath). The sensors were then left to stir for 10 minutes, before being placed in the speed-vac for 40 minutes to extract the THF in the solution. The resultant product was then pipetted into a glass vial and stored in a dry, dark place till further use.

Nanosensor Characterization.

The sensors were excited with a 5 mW 532 nm laser (Z-Bolt Electro-Optics, Happy Valley, OR, USA), with the response being measured on an Avantes StarLine spectrometer, with a 200 μm slit width. 1.0 mL of sensor solution was added to a narrow pathlength screw-top quartz cuvette and sealed with a rubber septa cap. To simulate various oxygen concentrations the gas bubbling system as described by Saccomano et. al. ²¹ was used. In a nutshell, we used two mass flow controllers to vary the flow rate of nitrogen and air, before mixing the two streams in a 50 mL stainless-steel gas mixing chamber. This gas mixture was then bubbled into the sensors in the cuvette via a 15-gauge needle pierced through the rubber septa. Another needle (15-gauge) was added to the cuvette’s headspace to vent any gas build-up from the bubbling process. The total flow rate from the mixing chamber into the cuvette was maintained at 20 mL/min. By changing the ratio of nitrogen to air, we were able to attain various dissolved oxygen (DO) concentrations ranging from anoxic (0.00 mg/L) to atmospheric (6.65 mg/L). To generate the sensor response to various concentration of oxygen, the upconversion data was averaged at 436 ± 2 nm, while the downconversion response of the PtOEP and DPA was averaged at 650 ± 2 nm and 430 ± 2 nm respectively (see Figure S1, for sample emission spectra). This data was then plotted as the ratios of the intensities at a given O₂ concentration (I₀/I, where I is the intensity at a given oxygen concentration, and I₀ is the intensity at 0 mg/L dissolved O₂). Furthermore, a non-linear curve was fit to the data using a quadratic model as developed by Borisov et. al. ³⁶ as highlighted in Eq.15 of their supporting information. We simplified their model and assumed that there was negligible change in the annihilator and sensitizer ground state concentration (see Eq 1. in supporting information). The parameters for K₁ and K₂ in the model were optimized by reducing the sum-squared error between the experimental and modeled data points. The data was then plotted using GraphPad Prism 9 software (Version 9.5.1). Using the same set-up, the sensor’s reversibility was also evaluated by cycling the sample between pure nitrogen (0.00 mg/L DO) and pure air streams (6.65 mg/L DO) at least three times. The reversibility response was quantified by averaging the upconversion luminescence at 436 ± 2 nm.

The absorbance, stability datapoints of our sample, as well as the yeast assay, was obtained on a Synergy H1 microplate reader purchased from Biotek (Winooski, VT, USA). 100 μL of sensors was added to 100 μL PBS in triplicated to a 96-well black-walled non-treated clear bottom optical plate. However, for the upconversion efficiency calculation, the absorbance of our sensors and the Rhodamine 18 reference dye (suspended in ethanol) was quantified on a Genesys 150, UV-Visible spectrometer from ThermoFisher Scientific (Waltham, MA, USA). The particle size distribution and ζ-potential measurements were obtained on a Brookhaven ZetaPALS (Brookhaven Instruments Corporation, Holtsville, NY), by diluting the sensors to 5% by volume in PBS.

pH Sensitivity Test.

The Synergy H1 microplate reader was also used to evaluate our sensor’s pH response from pH 3 to pH 9. The sensors were concentrated down and raised back up to their original concentration in pH-adjusted universal buffer at each pH value. For the oxygenated pH response, 40 μL of universal buffer was added to 160 μL of sensor. While for the deoxygenated pH response, 7 different glucose stocks were made at each pH value with the pH adjusted universal buffer, while the glucose oxidase (GOx) was made at in PBS (pH 7.4). To a well, 160 μL sensor, 20 μL glucose (18 mg/mL) and 20 μL GOx (0.5 mg/mL) were added. In this assay both upconversion and downconversion endpoints of the sensors were measured. This assay was conducted in a 96-well black-walled clear optical bottom well plate.

Yeast Assay Growth, Preparation & Plate Reader Setup.

This assay was performed using a protocol like those previously described ^3,21, with a few exceptions which include changes to the: yeast dilutions, wort preparation, volume ratios in the 96-well plate, and excitation and emission channels (see Figure S2 for plate set-up). Note that the yeast assay was maintained at 30°C and was set to shake in a double-orbital motion before the start of each run. From the yeast stock, three dilution sets of 3.3 % (1:30), 10.0 % (1:10) and 20% (1:5) by volume were made in filter-sterilized PBS for both strains. The yeast stocks and dilution sets were then stored at 4°C till further use. The wort stock was extracted from the can and sterilized through a 0.22 μm sterile vacuum filter. Additionally, the ratios for this assay were modified such that 100 μL wort, 80 μL nanosensors (concentrated to x5 from stock solution), and 20 μL of yeast or 20 μL of PBS was used. For the ‘no nanosensor controls’, 80 μL of filter-sterilized PBS was added instead of nanosensors. Finally, three different optical settings were established for each read cycle. (1). The upconversion channel excited the sample at 532 nm and observed the luminescence at 430 nm and 650 nm at a gain of 100. (2). The downconversion channel excited the sample at 350 nm and 400 nm and observed the emissions 430 nm and 650 nm respectively, with a gain of 60 for both readings. Both channels had 50 measurements per datapoint. (3). Finally, the absorbance or optical density (OD) was recorded at 532 and 430 nm.

RESULTS

We created an oxygen nanosensor that uses the TTA-UC mechanism to produce an upconversion luminescence signal response to oxygen concentrations which overcomes some of the limitations imposed by sample autofluorescence. These nanosensors use a well-known sensitizer-emitter pairing of PtOEP and DPA to produce a response in the 400 – 500 nm region (with excitation at 532 nm). These dyes were incorporated into nanoparticles via a FNP fabrication approach that yielded nanoparticles with an average diameter of 104 ± 8.5 nm, polydispersity of 0.10 ± 0.02, with no significant changes in particle size over 7 days (see Figure S3). The sensors had a ζ-potential of −7.12 ± 1.42 mV (see Figure S3), and an upconversion efficiency of 0.59% at 25°C (see Figure S4A), which changes as a function of ambient temperature (see Figure S4B).

Figure 2A shows the upconversion spectral response of sensor when subjected to oxygen concentrations ranging from 0 mg/L to 6.65 mg/L O₂. The oxygen concentration in the solution was modulated by mixing nitrogen and air streams before bubbling into the solution. As the DO concentration was increased from 0 mg/L to 6.65 mg/L O₂, the sensors’ luminescence decreased as expected due to the quenching effects of molecular oxygen on the TTA-UC process. Furthermore, as seen in Figure 2A, the upconversion signal change between 3.17 mg/L and 6.65 mg/L O₂ was negligible indicating complete quenching. In this case, 6.65 mg/L is equivalent to a 100 % dissolved oxygen (DO) concentration at 5,675 ft. (elevation in Golden, Colorado). To obtain the calibration curve for the sensor, the intensity when there is no quenching occurring (I₀) was divided by the intensity at a given oxygen concentration (I). The intensity values were obtained from the upconversion spectral data at 436 nm.

Figure 2:

A. Normalized upconversion emission spectra of DPA, when excited with a 532 nm light source. Data points were taken at five different oxygen concentrations ranging from anoxic (0 mg/L O₂) to atmospheric (6.65 mg/L O₂) conditions. As the oxygen concentration increases, the sensor luminescence decreases. This data was normalized to the peak intensity at 436 nm at 0 mg/L O₂ (n=3). B. Calibration curve generated from emission spectra, where the sensors have a nonlinear response to increasing oxygen concentration (n=3). A quadratic curve was fit to the experimental data (for model see Eq. 1 in supporting information). Where not visible, the error bar is smaller than that of the data point.

With non-upconversion based oxygen sensors, a linear response or Stern-Volmer calibration curve is seen, however, as seen in Figure 2B the upconversion calibration curve has a non-linear response to increasing oxygen concentration. Additionally, the sensors had a high degree of sensitivity to oxygen and a smaller dynamic range from 0 mg/L to 3.17 mg/L O₂. Previously developed oxygen sensitive downconversion sensors developed by our group have had wider dynamic ranges from 0 mg/L to 6.65 mg/L ^3,16,21. For DO concentrations greater than 3.17 mg/L O₂, there is no significant change in the response of the upconversion-based sensor reported here.

In addition to responsiveness, the sensors need to be reversible to dynamically report the changes in oxygen concentration, and stable in terms of their luminescence and resistance to aggregation. When tested over the course of a week, our sensors show a consistent upconversion and downconversion response in both oxygenated and deoxygenated conditions (see Figure S8). Furthermore, the sensors demonstrated good colloidal stability as the particle size distribution did not change over the course of 7 days post fabrication (see Figure S3). The reversibility of our sensors was tested by bubbling the solution with pure air and nitrogen streams to oxygenate and deoxygenate the system. As seen in Figure 3, the sensors demonstrated excellent reversibility when oxygenated and deoxygenated over three consecutive cycles. When compared cycle 1 the sensors showed a 98.4 % and 98.8% recovery for cycle 2 and cycle 3 respectively (see Figure S9 for emission spectra).

Figure 3:

The sensors show good reversibility when subjected to anoxic (N₂ or 0 mg/L O₂) and atmospheric (air or 6.65 mg/L O₂) conditions, for three consecutive cycles. Where not visible, the error bars are smaller than that of the data points (n=3).

To demonstrate the functionality of these oxygen sensors, we deployed them into a biological system – microvolume beer fermentation with Saccharomyces cerevisiae (brewing yeast). In this assay, the sensors showed that they can be used to report changes in metabolic activity by measuring oxygen consumption and overcome autofluorescence. As the yeast grow, they will consume the oxygen in the well resulting in a higher sensor luminescence. With the addition of PMB, an anti-microbial agent, yeast activity would be inhibited resulting in lower sensor luminescence via atmospheric reoxygenation of the sample. As seen in Figure 4A, the growth of the Kolsch I strain was monitored at three different dilution ratios with 1:5 being the least dilute and 1:30 being the most dilute set. It was observed that the most concentrated sample (Kolsch 1:5) grew the fastest, with its metabolic activity peaking around 12 h, while the most dilute strain’s oxygen consumption peaked around 17 h. The metabolic activity of all three concentrations showed the same profile after 18 h (likely resulting from a balance of oxygen transport from the atmosphere with yeast oxygen consumption). The timeframe of the yeast reaching the stationary phase of their growth was further supported by the optical densities of all three concentrations (see Figure S7B). As seen in Figure 4B, the addition of PMB at 42 h inhibited any microbial activity, as the upconversion luminescence decreased over 6 h until there was no signal change. Whereas the sample without any PMB maintained a steady signal till the end of the experiment, thus indicating a steady state, low-oxygen environment – a trend consistent with similar assays conducted by our group ^3,21.

Figure 4:

Metabolic growth and oxygen consumption of the brewing yeast, Kolsch I, as monitored by the upconversion nanosensors. A. The growth profiles of three different concentrations of the Kolsch I strain were monitored by the nanosensors as function of time. As expected, the most concentrated strain (Kolsch 1:5, green triangle) grew the fastest, with the most dilute (Kolsch 1:30, blue square) growing the slowest. B. The addition of PMB, an antimicrobial agent, (to the Kolsch 1: 10) inhibits the growth of the yeast cells, thus ceasing the consumption of oxygen. As a result, the assay returns to atmospheric conditions as seen by the decrease in the upconversion signal (red circles). C. The yeast by themselves are autofluorescent when excited with 350 nm light (red circles), but not when excited with lower energy 532nm light (blue squares) when observed at an emission wavelength of 430 nm as used with the upconversion sensors. This demonstrates how upconversion can be used to circumvent autofluorescence issues. Errors were emitted for all data points (n=4) for clarity purposes, see supporting information for data with error bars (see Figures S5, S6 and S7A respectively).

As a control, the nanosensors were tested in the absence of yeast with and without PMB. Both datasets showed the same behavior where there was no change in the downconversion signal and upconversion channel over time. This indicates that our wort, PMB and nanosensors are sterile (see Figure S10). When examining a yeast only control (Figure 4C), autofluorescence of the yeast can clearly be seen in the downconversion channel (350 nm excitation, 430 nm emission) with no autofluorescence in the upconversion channel (532 nm excitation, 430 nm emission). The sensitizer’s downconversion channel (400 nm excitation, 650 nm emission) also showed that using downconversion signaling with a large stokes shift is effective at reducing autofluorescence but does not fully eliminate it (see Figure S7A). This trend agrees with the absorption and emission spectra of the yeast itself, where the overall absorbance and autofluorescence is lower at higher wavelengths (see Figure S7C, S7D, S7E)

DISCUSSION

We chose FNP as our nanosensor fabrication technique as it allows for size tunability, excellent dye encapsulation, and scalability ^40,41. Using this technique we observed good colloidal stability, limited batch-to-batch variability in the sensor’s O₂ response (see Figure S11) and a reliable method of loading dyes into the nanosensors that showed limited dye leaching (see Figures S12 and S13). For this sensor, we chose a well-known dye TTA-UC pairing of PtOEP and DPA, that has been used in other systems to monitor enzymatic reactions, pH, and potassium. PtOEP was used as the sensitizer because it is an oxygen sensitive dye whose luminescence is quenched by molecular oxygen, it is soluble in organic phase of our nanoparticles, is a well-studied sensitizer molecule and has a strong absorbance peak in the green region (532 nm). DPA was used as the annihilator, as it too is soluble in our sensor’s organic phase, has a high quantum yield, does not absorb any green light, and emits in the blue region ⁴². Furthermore, the triplet state of PtOEP is higher than the triplet states of DPA, but lower than its singlet state, thus allowing for the TTA-UC to occur ²⁸. This phenomenon is not just limited to PtOEP and DPA, as highlighted in Singh-Rachford et. al. in their review, there are sensitizer-annihilator dye pairings capable of demonstrating TTA-UC. While our sensor emits in the blue region (400 – 500 nm), with other dye pairings it is possible to achieve emissions in the 550 – 560 nm range ^28,43. However, it is important to carefully chose a sensitizer-annihilator pairing, as not all dye pairings are capable of upconversion based sensing of oxygen. While a dye pairing of ruthenium diphenyl phenanthroline and perylene demonstrated an upconversion response to oxygen concentrations, a dye pairing of the same ruthenium diphenyl phenanthroline and coumarin 343 did not (see Figure S14).

The spectral response (PtOEP/DPA sensors) shows that the luminescence of both the upconversion (DPA) and downconversion (PtOEP) signal changes as a function of oxygen concentration (see Figure S15 and Figure S16, respectively). At low oxygen concentrations, the sensitizer’s downconversion signal does not show a significant signal change until 3.17 mg/L O₂, after which the sensitizer’s luminescence is quenched by the increasing oxygen concentration (0.00 – 6.65 mg/L O₂). A possible explanation for this could be that the PtOEP emission is quenched by both oxygen and the TTA-UC mechanism³⁶ which could result in the sensitizer having a non-linear profile to increasing oxygen concentration. While the PtOEP or downconversion response is responsive to oxygen concentrations up to 6.65 mg/L O₂, the upconversion response has a much smaller dynamic range (0.00 – 3.17 mg/L O₂) and non-linear response too. A similar upconversion non-linear response was observed by Borisov et. al. in their sensing foil system ³⁶ using a different dye pairing. It was suggested that this non-linear response arises due to nonradiative deactivation pathways being more likely to occur than TTA with the annihilator triplets ³⁶. Furthermore, the high oxygen sensitivity of these sensors could potentially be attributed to the high sensitivity of PtOEP in comparison to our traditional nanosensors made from PtTFPP, another oxygen sensitive dye ².

The sensors were deployed in a yeast assay to monitor the metabolic activity of Saccharomyces cerevisiae (brewing yeast) developed by Saccomano et. al. ²¹. In this assay, the oxygen concentration (used as a proxy for metabolic activity) of the yeast cells was observed over 60 hours. Figures 5A and 5C, shows the lag phases for the first 6 hours of the experiment, where there is no change in oxygen consumption. During this time the yeast are building up food reserves for the fermentation process ^44,45. At the 6-hour mark, the yeast cells being to grow as indicated by the rapid consumption of oxygen and the sharp increase in the optical density. As expected, the wells with the highest yeast density (Kolsch 1:5) demonstrated a shorter lag time and more rapid onset of deoxygenation in comparison to the most dilute set (Kolsch 1:30). Once the concentration of oxygen in the well stabilized, PMB was added to inhibit the growth of the yeast at the 42-hour mark. When the PMB is added to aqueous media, it dissolves and produces sulfur dioxide and other by-products that can inhibit cell metabolism and trigger cell death ^21,46.

This was observed as the upconversion signal rapidly decreased back to baseline over the course of 6 hours. However, the cells without PMB continued to metabolize till the end of the experiment as indicated by the continually high luminescence signal. A potential confounding factor is the acidification of the fermentation medium as the yeast grow, but our sensors are not pH sensitive (see Figure S17). This reinforces the finding that the signal decrease was indeed caused by a decrease in the metabolic activity of the yeast, and not a pH dependency. The functionality of these sensors is demonstrated in Figure 4C, where the fluorescence of the yeast sans nanosensor was monitored. When excited with a low energy light source (532 nm, upconversion excitation) there is no autofluorescence observed, but when excited with a high energy blue light (350 nm) a large autofluorescence signal is observed from the yeast by itself. It is worth noting that in addition to molecular oxygen, factors such as viscosity ⁴⁷, sensing matrix ³⁶, temperature ⁴⁸ and annihilator concentration ⁴⁹ can influence the upconversion efficiency, as highlighted by Schmidt and Castellano ⁴⁹ in their in-depth analysis of factors impacting TTA-UC kinetics. Given that our yeast assay uses both a viscous medium (wort) and high temperature (30°C), it is important to assess their impact on sensor response. Using a glycerol-PBS mixture we modulated the viscosity of the bulk solution from 1.00 mPa·s to 3.70 mPa·s, where the upconversion response was not impacted by solution viscosity (see Figure S18). A possible explanation for this observation is that unlike in traditional TTA-UC systems where the dyes interface with the bulk solution directly and dynamics are directly impacted by solution viscosity, our dyes are housed in our hydrophobic core, protecting it from the bulk solution. Therefore, changes to the hydrophobic core material and viscosity are more likely to impact the sensor response than the bulk solution. Additionally, as seen in Figure S4 B, temperature does impact upconversion efficiency. Given that this assay was conducted at 30°C, it is possible that some of the high signal to noise ratio could be attributed to the higher upconversion efficiencies that occur at higher temperatures. Nonetheless, the key findings of this assay highlights and reinforces the advantages of upconversion as a facile and economical means of overcoming autofluorescence.

While these sensors are stable and able to detect changes in oxygen, there is room for improvement. Firstly, they are not ratiometric. This means it is difficult to discern if a change in signal is occurring due to changes in analyte or confounding factors such as changes in sensor concentration ^18,50. While not a problem in well plate-based assays such as that presented here, this is a key limitation for e.g., in vivo applications where sensor migration can cause changes in sensor concentration. As a result, future works can look at adding a reference dye system, to enable ratiometric sensing. Another challenge that needs to be overcome is of scattering and penetration depth when imaging thick samples. Our sensors use a short-wavelength excitation light source and while it is effective in overcoming autofluorescence, it is still impacted by light scattering in tissues and light depth penetration issues ²⁷. A possible solution to this problem could be to develop near infrared (NIR), TTA-UC based sensors. Finally, the upconversion efficiency of our sensors could be improved. This would result in lower sensor concentrations being needed for imaging, or a lower powered light source which reduces the likelihood of sample damage or photodegradation. This can be attained by looking at other sensitizer-annihilator ratios or combinations.

CONCLUSION

In this work we used the triplet-triplet annihilation upconversion mechanism to develop an oxygen nanosensor to overcome the limitations of biological autofluorescence as a background signal. By using a sensitizer-annihilator dye pairing of PtOEP-DPA, we were able to detect and report oxygen concentration using upconversion from 0% to 10% dissolved O₂ saturation in solution. The sensors demonstrated high sensitivity to oxygen, good signal and colloidal stability, reversibility, and a non-linear oxygen response - a phenomenon also reported by Borisov et. al. ³⁶.

To demonstrate the functionality of this sensor in a biological system, we deployed the nanosensors in a high-through put metabolic assay assessing the oxygen consumption of yeast over time. In this assay, we monitored the growth of a brewing yeast strain (Kolsch I) over 60 hours. The sensor’s upconversion response was in accordance with trends observed with other oxygen sensors developed by our growth. This assay also demonstrated that by using upconversion, the effects of autofluorescence were eliminated. This makes these sensors an exciting tool for researchers aiming to study analytes in biological systems with high degree of autofluorescence. Future works can involve enhancing the upconversion efficiency of the sensor, exploring other dye pairings to cover larger areas of the visible spectrum and adding a reference dye to enable ratiometric sensing.

ABBREVIATIONS

TTA

triplet-triplet annihilation

UC

upconversion

FNP

flash nanoprecipitation

PtOEP

Platinum (II) octaethylpor-phyrin

DPA

9,10- diphenylanthracene

PS–PEG

polysty-rene- block-polyethylene glycol