Editors’ Choice—Luminescent Oxygen Sensors: Valuable Tools for Spatiotemporal Exploration of Metabolism in In Vitro Systems

A common biological theme on Earth is the importance of oxygen, regardless of an organism’s metabolic capabilities. This commonality makes the quantification of O2 essential in understanding life as we know it. There are many sensing methods that enable researchers to measure this important analyte, but not all sensors are compatible with every system. This perspective highlights common O2 sensing formats (and recent innovations) with the goal of guiding the reader towards a sensor choice for their desired application. We emphasize the importance of exploring unfamiliar metabolic processes, commercializing new sensors, and establishing collaborations for maximizing innovation and accelerating discovery.

The 17th century alchemist and philosopher, Michael Sendivogius (1566-1636 theorized the "food of life" was within the expelled gas of heated saltpetre. 1,2 Unknowingly, the polish thinker reduced nitrate to nitrite, producing molecular oxygen that would get officially recognized as a chemical element by Antoine Laurent Lavoisier nearly 170 years later. 3,4 Since the discovery, the biological importance of O 2 has come into the spotlight: from the realization of mammalian physiology to microbial metabolic processes that occur in the utmost extreme environments-truly fulfilling the "food of life" hypothesis by Sendivogius. While the percentage of O 2 in the atmosphere is ca. 21% (160 mmHg), it varies greatly in the human body from ∼100 mmHg in pulmonary arterial blood, to ∼34 mmHg in the brain-even hypoxic levels (< 15 mmHg) are required in certain physiological processes. 5 On the microbial front, the range of metabolic capability is much larger than that of higher order eukaryotes, but the extent of O 2 variability in microbial environments is similar. Oxygenic photosynthesis of many algae and cyanobacteria enables microenvironments to exceed atmosphere equilibrated O 2 concentrations (DO) while some bacteria can only tolerate ⩽0.5% O 2 in the gas phase. 6 Overall, there is a common theme in biology on Earth: that is, regardless of an organism's metabolic capability, O 2 is of utmost importance.
Methods that allow for quantification of O 2 are, therefore, essential in the path to better understanding the biological processes in and around us. Identifying key metabolic parameters like O 2 diffusion rates, biochemical O 2 demands (BODs), and dissolved O 2 gradients in biological samples require a slew of innovative sensing methods which tend to diverge from a "one-sensor-fits-all" model. No matter the method, evaluating O 2 fluctuations in biological regimes often does require a continuous and reversible sensing mechanism. In the broadest sense, we can categorize the bulk of O 2 sensors as either optical or electrochemical.
The prime example of an electrochemical O 2 sensor is the Clark electrode which functions by electrochemically reducing O 2 at the surface of a platinum electrode after applying a voltage. While popular, this method has disadvantages in metabolic monitoring applications that include: 1) consumption of O 2 during acquisition, 2) limited spatial throughput (i.e., the user can only take point measurements), and 3) the invasive nature of the electrode tip (typical tips are ca. 100 μm, but the user is still limited even with the 10 μm tip sizes).
Optical O 2 sensing has been recognized as an alternative to electrochemical methods for biological applications. 7 An optical readout enables remote sensing and minimal invasiveness.
Specifically for O 2 , this mode overcomes many of the restrictions above and can be superior for a variety of reasons. Optical probes do not consume O 2 , are fully reversible, and can be used in imaging and microscopy applications. Additionally, the modularity of this approach and plethora of O 2 sensitive dyes allows for measuring in a wide range of concentrations. That said, the optical route is not always the answer. Adding an optical component may overcrowd your spectral space; integrated sensors may impact the growth of your cells; in situ measurements can be extremely difficult if the specimen has an unexpected and excessive amount of autofluorescence or scattering. If the metabolic analysis has issues with an optical approach or the user only needs point measurements of their system, the Clark electrode and related approaches are the goldstandard.
While there are many optical transduction modes that can be used to quantify pO 2 in a wide range of settings, most approaches are not compatible or very practical for real-time metabolic analysis in vitro. For example, pulse oximetry functions on the basis of hemoglobinbased O 2 transport and is, therefore, mostly limited to in vivo mammalian studies. 8,9 Absorptiometric indicators for metabolic analysis are impractical due to high absorption of biological samples, and most are irreversible with a few exceptions. 10 Conversely, luminescent sensors that are quenched in the presence of O 2 are compatible with a much wider range of applications. This sensing mechanism has gained traction in a variety of fields and has proven to be applicable for both in vitro studies of cell metabolism and in vivo applications for physiological monitoring with implantable or wearable transcutaneous devices. While transcutaneous devices for physiological O 2 monitoring have made significant progress in recent years, [11][12][13][14][15] this perspective will mainly focus on the in vitro use of luminescent sensors that are quenched by O 2 with the main goal of guiding the reader towards a logical sensor choice for their desired application. As such, we are not attempting to review the entire field of O 2 sensing for metabolic evaluation, but rather to provide the reader with examples of what is possible to help guide their research and facilitate innovation in this domain.

Fundamentals of Optical Oxygen Sensing
Nearly all spectroscopic methods have been used for sensing O 2 , but luminescence spectroscopy is the most common where the detection relies on collisional quenching of a phosphorescent indicator molecule (Fig. 1A). 10 For other optical readout mechanisms, molecular indicators for O 2 , and molecular mechanisms of the photochemistry involved, see established literature. 10,16 The collisional quenching process that occurs between the phosphorescent z E-mail: kcash@mines.edu probes and molecular O 2 can be described by the Stern-Volmer equation (Eq. 1): where I 0 and I are the luminescence intensity in the absence and presence of O 2 , τ 0 and τ are the phosphorescent decay times of the probe in the absence and presence of O 2 , and K sv is the Stern-Volmer constant, which is a product of the bimolecular quenching constant, k q and τ . 0 Figure 1B shows a theoretical Stern-Volmer calibration curve. The K sv is a parameter used to describe the sensitivity-an important analytical parameter when considering the sensor of choice in metabolic monitoring.
In many cases, the indicator is dissolved in a polymer host which may give rise to heterogenous microenvironments-a phenomenon that often leads to non-linear Stern-Volmer plots (Fig. 1B, dotted line), for which a two-site model is used. 16

Current Sensor Formats
There are many classes of dyes that are quenched by O 2 to various degrees. Because it is a collisional quenching process, the degree of dye quenching is probabilistic and impacted by the dye's phosphorescent lifetime, its local environment, and O 2 permeability into that environment. 10,17 This multifactorial process is highly modular, and the sensitivity and dynamic range can be tuned based on these factors. The most common indicators typically incorporate transition metals like ruthenium, osmium, iridium, platinum, or palladium and have been used to provide both time-dependent and intensity-based measurements in a range of O 2 concentrations and emission wavelengths. 10 In this section, we will give a brief overview of the sensor formats that incorporate common O 2 probes and have been used to monitor O 2 flux in biological systems.
Small molecular probes.-Hundreds of O 2 sensitive probes have been developed for a variety of biological applications. 10 Ruthenium polypyridyl complexes are very popular, as they were amongst the first to be used as O 2 indicators and are widely available from commercial vendors (e.g. Sigma Aldrich, Frontier Scientific). Many Ru(II)-chelates are water soluble (e.g., Ru(Bpy) or Ru(dpp[SO 3 ]) without any conjugation required and creative advancements have been made to yield a ratiometric conjugate probe for intracellular quantification. 18,19 The hydrophobic counterpart Ru(II) tris(4,7diphenyl-1,10-phenanthroline) (or Ru(DPP) 3

2+
) is more commonly embedded in a polymeric support. The metalloporphyrins, due to their negligible water solubility, are less common as stand-alone probes unless hydrophilic functional groups are added. [20][21][22][23] The "bare-bones" modifications are carboxylated metal-porphyin chelates that are commercially available; albeit, have mediocre solubility in aqueous solvents. Recent advancements in metalloporphyrin modifications are done by Vinogradov and coworkers with their development of various Oxyphors ( Fig. 2A). By tuning the properties of the dendrimeric shell around the porphyrin core, their group has achieved O 2 probes with great aqueous stability and biocompatibility. [20][21][22][23]26,27 Arginine rich modifications of platinum, palladium, or other metal porphyrins have been used for intracellular 26,27 measurements whereas polyglutamate or poly(arylglycine) followed by a polyethylene glycol shell has been used for intravascular and interstitial pO 2 quantification. [20][21][22][23] Oxyphors have been deployed in vivo and in deep tissue applications from tracking O 2 distribution and consumption in xenographed tumors 20,21 to monitoring O 2 dynamics in a mouse brain following a micro-stroke. 22 Functionally, these sensors have pushed the boundaries of monitoring O 2 metabolism in murine models and have tremendous potential for cellular studies. So, what's the caveat? While innovative, most of the molecular probes in this category are synthesized in house and further innovation may be outside realm of expertise for some.
Optodes-the "tried and true".-Optode sensors have been implemented in a wide range of disciplines-from medical and environmental research to materials sciences. 28 Doping polymer matrices with O 2 indicators provides a protective, nonpolar environment for the probe and for O 2 to favorably partition into. The optode matrix provides shielding from signal interference, prevents loss of the indicator to the sample, and makes it easier to exchange the indicators for a different sensitivity range or wavelength. 10,28 While confined to two-dimensional analysis, they come in many different sizes (from μm 2 to cm 2 dimensions) and physical characteristics that enable fine tuning towards the metabolism of interest. Optodes are, in essence, the "tried-and-true" in the realm of optically monitoring O 2 metabolism. Comparatively, optodes are not only more facile to make in the lab than the Oxyphors but are also widely commercially available which provides ample opportunity for use.
Microbiologists and environmental scientists have taken advantage of the planar optode to measure the formation of gradients and heterogeneities caused by the consumption or production of O 2 in soil. [29][30][31][32] In recent applications, the integration of an O 2 optode on to the walls of an in vitro soil column has allowed for researchers to characterize rhizospheres (Fig. 2B), 24,33 implicate disinfestation in microbial mediated greenhouse gas emission, 31 and link millimeterscale O 2 heterogeneity to post-fire biogeochemical responses. 30 Not only can the optode identify hotspots (represented by O 2 heterogeneities in 2D), but the reversibility of the molecular probe inside the host matrix can allow for identification of hot moments in soil over longer time scales towards the goal of monitoring long term nutrient dynamics. 34 The technology has also been used to characterize the extremely demanding metabolism of cable bacteria inside a custom "trench slide," showing the adaptability of the sensor membranes towards custom culture techniques. 35 Additionally, O 2 optodes have proven useful to explore the physiology of ammonia oxidizing archaea, Nitosopumilus maritimus at nanomolar concentrations of O 2 -it was discovered that the microbe produces O 2 in the dark and uses it for ammonia oxidation. 36 The optode provides microbiologists a tool to measure 2D O 2 dynamics in a range of metabolic capabilities that thrive in environments with either high or low O 2 .
The O 2 optode format also benefits the biomedical realm with the integration of the sensors into a variety of cell culture techniques from microfluidic devices to culture dishes. [37][38][39][40] The Lockett group has used optodes to investigate tumor metabolism and cell localization dynamics inside their 3D culture devices. 38,41,42 Their approach allowed them to identify cancer cells expressing an O 2 independent, invasive phenotype that are likely responsible for metastasis. 38 In a similar setting, O 2 optodes helped map tumor microenvironments and identify how hypoxia is related to spatial chemotherapeutic resistance. 43 Another approach used an O 2 optode adhered to a hydrogel "ramp," allowing researchers to define a gradient from the bottom of a culture flask to the media-air interface-giving a proxy for metabolic activity relative to O 2 diffusion into the sample. 44 For the user that needs similar measurements to the Clark electrode, but prefers an optical readout, fiber optic sensors are an appropriate tool. 45 Fiber optics largely benefit those working with microvolumes by not consuming the analyte while allowing for a spatial resolution of ∼50 μm. These tools function the same way as an optode, but the excitation source and detector are linked to a micro membrane at the end of a needle. This sensing format allows the user to take measurements in the field to monitor samples in situ, which has obvious implications in environmental work. 46 While this approach does not chemically alter the sample (relative to the Clark electrode); it does introduce a physical perturbation by piercing the sample surface akin to the electrode counterpart. Most companies that offer optodes also offer fiberoptic coupled O 2 sensors.
Micro/nanoparticles.-The transition from the optode membrane to a micro or nanoscale particle has recently become more appealing. This sensor format has increased throughput and made high resolution, spatiotemporal O 2 monitoring less demanding.
Quantifying O 2 can be obtained with either intensity or lifetimebased measurements with particles, although the intensity route can be much more streamlined for the end user with a plate reader, confocal microscope, or basic spectrometer setup. By colocalizing an O 2 insensitive fluorophore into the particle matrix, a ratiometric measurement can be obtained to correct for particle concentration within media or on a heterogeneous surface. 47 Most particles are made of materials that provide similar benefits as the planar optode but are coated in an amphiphilic surfactant layer that stabilizes the hydrophobic matrix in an aqueous environment (Fig. 2C). Particle sensors have been made with various core materials like plasticized PVC, polystyrene, silica, tocopherol acetate, and many more. [48][49][50][51] For an extensive workup on the particle materials and methods for particle sensor synthesis, see review by Koren and coworkers. 52 Functionally, the micro/nanosensors interact with the analyte in the same way as described in previous sections (via collisional quenching), but the miniaturization introduces advantages over the planar counterpart. First, the particle format allows the user to monitor rapid O 2 dynamics as the response times are much faster than the planar optode due to the increased surface-to-volume ratio. Second, 3-dimensional quantification is possible in samples that are spatially complex. For example, our group embedded nanosensors into biofilms of Pseudomonas aeruginosa PAO1 to show altered 3D O 2 consumption in response to antibiotic treatment. 25 By growing the biofilm in nanosensor-doped media, the particles get embedded within the extracellular polymeric substance and allows for gradient identification from the boundary layer to the biofilm center. Third, the particle format allows for a variety of application approaches to the sample of interest because it overcomes the need to be in contact with a 2D sensor membrane/foil. This allows complex sample geometries to be analyzed. The user can, for example, grow biofilms around the particles to monitor gradients, add a dimension to rhizosphere analysis, 53,54 disperse them throughout a batch culture for continuous monitoring of O 2 metabolism, 48 or spray paint the particles on to complex surfaces to characterize respiration and photosynthetic activity in 3D. 55 Like some dendrimeric porphyrins, 26 nanoparticles have the potential to push the realm of O 2 sensing into intracellular settings, albeit with various difficulties. Obtaining intracellular measurements via endocytic routes requires tedious optimization of the particle physicochemical characteristics and culture parameters like media ionic strength, temperature, and others that may impact nanoparticlemembrane interactions. 56 Even with these optimizations, cytosol localization and overcoming signal background from lysosome encapsulated sensors becomes a huge challenge. For this reason, nanoparticle based, intracellular quantification O 2 are limited to few studies in recent years. 50,57,58 Overall, nanoparticles are more appropriate for extracellular measurements in 3D. If the reader is interested in measuring inside cells, we recommend seeing Small molecular probes section on molecular probes and Refs. 26, 59-61.

Authors Perspective on Applying O 2 Sensors in Your Research
The reader most likely noticed that, while investigating the previous sections, there are multiple sensor types that are compatible with their system. There are many things to consider when finding the optimal sensor to measure O 2 in a biological system. Essential questions are outlined in this section with information aimed to direct the reader towards the analytical tool of choice.
What is the O 2 sensing regime? Here, the reader should consider probe sensitivity and dynamic range that is compatible with system of interest. These sensor parameters are tunable by the choice of dye and the characteristics of the housing matrix. For example, Kraft et al. used an optode composed of the highly quenched PdTFPP embedded in an extremely O 2 permeable, fluorinated polymer to make sensitive measurements at nanomolar concentrations. 36 If the reader is needing trace O 2 measurements, a similar sensor design (i.e., a dye with a relatively long luminescence lifetime (τ 0 = 1.65ms 10 ) coupled to a very O 2 permeable matrix) would be optimal, but would not allow for measurements at higher concentrations. On the other hand, Ding et al., was able to quantify O 2 in a wider range from 0 to 40 mg l −1 using RuDPP (τ 0 = 6.65 μs 10 ) encapsulated in mesoporous silica. 50 Their design was used to monitor HeLa cell O 2 metabolism but would also be very useful in photosynthetic systems that exceed 160 mmHg.
Are there other fluorescent sources in the system? At what wavelength? The wavelength of cell autofluorescence or common cell stains often determine the O 2 dye that can be used. Spectral separation between the sensor signal, reference signal, and the cell or stain emission is essential for quantitative applications. Fortunately, most O 2 sensitive dyes have emissions that do not overlap with the common nuclear stain, DAPI (359 nm/457 nm), making it easier to identify cells and monitor flux O 2 , simultaneously. However, as you approach red wavelengths with cell stains, there is a higher risk for peak overlap with the O 2 indicator. In general, multiplexing with stock microscopes can be difficult task, so introducing a sensor signal for quantification can often stymie researchers. While luminescence lifetime microscopy (FLIM or PLIM) or two-photon microscopy can help overcome limitations in spectral space, these approaches can be expensive and/or hard to design. Overall, preliminary analysis of the biological system is needed before choosing the sensor type and dye for wavelength selection and spectral separation for O 2 quantification.
Do I need high spatial resolution? 2 or 3 dimensions? This question addresses only the choice of sensor type. Fiber optics (or electrodes) are optimal for point measurements or quantification in 1 dimensional point measurements. The planar optode is optimal for 2 dimensions and flat sample geometry like the bottom of a culture plate-they are also used on the wall of sediment columns. Nano/ microparticles are a great choice for complex sample geometry and quantification in 3D as done in biofilms and coral surfaces. Imaging depth and scattering also becomes a factor here-probe wavelength should be considered when working in 3D. For intracellular work, Oxyphors and other custom small molecules have proven useful.
What is the time scale of the process that I need to monitor? This can influence the dye used and choice of sensor format. While most O 2 sensitive dyes are less prone to photodegradation than classic fluorophores, taking recurring measurements over extended periods can still diminish signal. Metal-porphyrins tend to be the very photostable over long periods. 10 Particle sensors may limit the timescale by eventual settling and both fiber optics and planar optodes tend to have a degree of fouling depending on how long they are in contact with the sample.
What equipment do I have available to take measurements/ images? The available spectrometer/detector set up can govern the dye, the sensor type, and whether you have the means to monitor what you need to. If you do not have a spectrometer, fiberoptic options often come with handheld meters from vendors. A generic fluorescence plate reader setup is compatible with particle formats. Fluorescence microscopes with a 405 nm laser can excite many O 2 sensitive dyes but the detector may limit the user to visible range emissions. Planar optodes are often imaged with CCD cameras but less expensive camera systems have also been used. 62 Without the ability to measure luminescence lifetime with either a spectrometer or imaging system, a ratiometric approach is recommended to correct for artifacts like sensor/dye distribution in the sample.
Is there a commercially available sensor that will fit my needs? In many cases, commercially available sensors can help solve the biological problem. Pyroscience, 63 PreSens, 64 and EasySensor 65 all sell planar optodes and fiberoptic sensor devices for O 2 . These are a viable option for many applications. Various Oxyphors are sold by OxygenEnt. 66 Nano/microparticles are available at Pyroscience 63 and Colibri Photonics. 67 The Future-What Boundaries Should Be Pushed?
In previous sections, we highlighted a few applications with the goal of portraying the potential of optical O 2 sensors in making novel discoveries in metabolic monitoring. Generally, recent efforts have been directed at developing new sensors for measuring O 2 -all of which require impressive engineering, but often provide little benefit over already existing technology. In our opinion, more effort should be allocated towards using the already developed sensors in more complex biological settings. These collaborative efforts between biologists and sensor engineers have the potential to make impactful discoveries with measurements that matter. Sensing, by default, is an interdisciplinary field and those deploying the sensors into unique settings should embrace that attribute by collaborating. Many goals will need to be met to bring this into fruition. First, the current innovations in sensor design must be made commercially available. Collaborations, while extremely important, can be difficult to navigate and increasing commercialization will allow new O 2 sensors to reach the hands of the biologist more readily. The sensor community should also capitalize on this opportunity for other analytes. Another major hurdle for most analytical tools is the reproducibility of measurements, especially in the case of optical transduction (with scattering and penetration depth varying significantly across sample types). Exploring avenues of different wavelengths, measurement techniques, and deconvolution methods should allow for better reproduction across samples. Additionally, if a novel sensor is to be commercialized, the reproducibility and standardization of manufacturing, storage, and shipment will be a challenge for the sensor engineers to conquer. Another important emphasis that should be placed on sensor design is the stability of the probe after deployment. To make long term measurements with either a planar optode in soil or particles deployed in vivo, the sensing unit needs to function sufficiently within the time domain of the phenomenon that is being observed. As many O 2 sensitive molecules are notorious for being photostable, innovations in this domain should revolve around housing material and its stability and biocompatibility.

Conclusions
Nearly 400 years ago, a hypothesis about the "food of life" in the air helped ignite experimentation that led to the chemical revolution. Indeed, life on Earth is largely impacted by the same gas produced in the experiments by Michael Sendivogius. 1 Since the early 17th century, researchers have determined that O 2 can be essential to some and detrimental to others; some produce O 2 and some consume it; some are extremely particular and some are unconcerned. The recent development of analytical tools of all shapes and sizes has allowed these researchers to ask increasingly complex questions in metabolic analysis. In truth, a "one-sensor-fits-all" model does not exist in the realm of monitoring O 2 metabolism. Particle sensors will not solve every problem and nor will optodes, small molecular dyes, or dendrimeric porphyrins. The key is choosing the optimal sensor for the biological problem. In this perspective, we aimed to highlight a range of applications in optical O 2 monitoring to aid the reader in making that choice. In the future, more emphasis should be placed on exploring new biological systems and monitoring unfamiliar metabolic processes with the tools at hand. At the same time, the commercialization of various sensor designs will inherently inform the field of subpar sensor performance thus accelerating further optimization. But most importantly, collaborations between biologists and sensor engineers should be prioritized to maximize innovation and discovery about the biology in and around us.