The field of redox biology has always been a significant challenge for analytical chemistry. This is because many redox active molecules demonstrate extreme reactivity and, therefore, cannot be isolated from cells or tissues and subjected to traditional assays without introducing pronounced measurement artifacts. One way to avoid this is to focus on the detection of oxidative stress biomarkers in the samples that are the secondary products in the reactions between oxidants and cellular components. However, this strategy is not simple. An ideal biomarker must meet a number of requirements, among which are relative stability and selectivity. Unfortunately, various oxidants with completely different biological sources, production regimens, kinetic properties, and metabolic functions modify similar compounds to the same final products. For example, the thiol group (-SH) of free cysteine is an important target for a wide set of oxidants, including H2O2 and HOCl, for which the rate constants of this reaction differ by as much as 7 orders of magnitude [1,2]. However, sulfenic acid (-SOH) is the final product in both cases, since -SCl groups are prone to fast hydrolysis. Thus, it is often impossible to reconstruct the exact chemical nature of oxidative stress experienced by a biological sample unless sophisticated kinetic modeling is involved. It seems that mono- and di-chlorinated tyrosine residues are the main (not the only) selective markers of HOCl-mediated damage, but the corresponding kinetic constant is about 6 orders of magnitude lower than for regular cysteine residues, making this reaction relatively unexpressed at low oxidant concentrations [2].

Considering the above, it is easy to see why the scientific community is putting so much effort into the search for reporter compounds that specifically interact with different redox agents and are compatible with in vivo implementation. And here comes another question about the method for their registration. While omics approaches are developing at extraordinary speed and they provide a unique opportunity to analyze enormous amounts of information in combination with artificial intelligence approaches, the scaling of such techniques to time-resolved experiments is extremely labor-intensive. However, this is not the case with optical spectroscopy. Fluorescence imaging has a long history usage in the field of redox biology. Even Otto Warburg and his colleagues who described the chemical structures of NAD+ and NADH, investigated their optical properties and found that the 340 nm absorption band is a characteristic feature of the reduced form [3]. Ultraviolet spectroscopy of NADH became an important technique for studying mitochondrial metabolism in 1951, after Theorell and Bonnichsen found that the NADH spectrum changes with the addition of alcohol dehydrogenase [4]. Unfortunately, due to its low sensitivity, this method was applicable only to cell suspensions and very thin tissue slices until 1957, when Duysens and Amesz described NADH fluorescence spectrometry in baker's yeast [5]. In later years, Britton Chance's research group significantly improved the described technique, which allowed them to study many interesting aspects of energy metabolism. In particular, in 1962, they, for the first time, implemented NADH fluorescence imaging in vivo to visualize redox processes in the brain and the kidney in anesthetized rats [6]. More information on this topic can be found in several excellent review papers [3,7]. One more important step in the field was taken in 1992 by Schneckenburger and König, who measured fluorescence decay patterns of the NAD(P)H pool in living Saccharomyces cerevisiae cells [8]. During the early 2000's, commercial equipment for fluorescence lifetime imaging microscopy became widely accessible, which led to widespread interest in this variant of the method over the past decades [9].

Obviously, most redox compounds do not demonstrate pronounced fluorescence and, therefore, cannot be measured via direct optical registration. By the end of the 20th century, the scientific community was extremely interested in obtaining fluorescent tags that could be implemented for in vivo imaging of biological processes. For example, Noble prize winner Roger Tsien's group developed a probe for cAMP registration based on protein kinase A and fluorescein/rhodamine energy transfer couple [10]. This instrument was used in a number of interesting experiments, allowing visualization of cAMP gradients in individual neurons of Aplysia during memory formation [11]. However, this sensor was not fully genetically encodable, because it required the purification of the detector domains, their conjugation with a low molecular weight dye, and subsequent microinjection of the product into cells. At that time, the scientific community did not know of any auto-fluorescent proteins. The genes encoding several phycobiliproteins had been cloned before 1990 [12], but their protein products relied on external chromophores for functioning. Their practical implementation, therefore, required the delivery of the dyes into cells or heterologous expression of the genes required for their synthesis, which, with the level of development of molecular biology at that time, was an even more difficult task. The situation changed dramatically in 1992, when Douglas Prasher and his colleagues cloned the gene encoding the green fluorescent protein, from a complementary DNA library based on Aequorea victoria mRNA [13]. In the original paper, the authors hypothesized that its chromophore is formed via post-translational modification of the polypeptide in the absence of any organism-specific cofactors. A little later, experimental proof of this idea was obtained by the Noble prize winner Martin Chalfie, as well as by Inouye and Tsuji by heterologous expression of the protein in nematodes and bacteria [14,15]. Since at that time ideas about creating genetically encoded sensors were already in the air, there was not a long wait for the development of the first instruments. It should be noted that redox sensors were among the first to emerge. For instance, a nitric oxide indicator based on thionein [16] and the first redox sensitive fluorescent protein, rxYFP [17], were published in 2000 and 2001, respectively, just three and four years after Cameleon [18], which is considered the first genetically encoded sensor.

Since then, dozens of new instruments have been described and they cover a wide range of target analytes, optical parameters, and biochemical properties. A standard genetically encoded sensor consists of two main parts: a fluorescent protein derived reporter domain and a sensory interface, located either on the reporter domain, or on a separate module, which in this case is referred to as a sensory domain (Fig. 1A). The principal role of the sensory interface is to interact with the physicochemical parameter of interest via a so-called molecular switch. Then the information about successful interaction is transmitted to the reporter domain affecting its optical properties. Genetically encoded sensors, therefore convert microscopic molecular events into a macroscopic signal, which can be registered using traditional equipment. There are a huge number of molecular switches that can be integrated into sensory interfaces, so it is possible to develop an indicator for almost any desired analyte. Some examples include substrate binding, phosphorylation, formation of a disulfide bond, structural rearrangement due to molecular tension, and membrane depolarization. The same applies to cross-domain communication. Usually, this communication is achieved via conformational coupling, but other, more exotic, mechanisms exist. Among which are subcellular redistribution, changes in protein lifetime, and static quenching.

Over the past two decades, genetically encoded sensors have revolutionized our picture of redox biology, as well as the general methodology in the field. Among the many opportunities they offer, the most important is the ability to measure biochemical parameters in the cellular context, which allows artifacts caused by sample preparation to be avoided. When redox active fluorescent proteins became part of the researchers' toolkit, it became clear that the normal glutathione redox pool is much more reduced than had been previously suggested [19]. Furthermore, the sensory interfaces that biosensors use are often derived from natural proteins that have been shaped over millions of years of evolution. In this light, it is no wonder that genetically encoded probes usually provide unique sensitivity. Given the above, they are capable of unravelling complex metabolic processes in which various redox species play different roles. As an example, elevated hydrogen peroxide concentrations and glutathione oxidation in tissues do not always coincide during fruit fly development, which once again emphasizes the imperfection of the concept of “general redox state” [20]. Upon further consideration, this paves the way for revealing the mechanistic details of biochemical programs in vivo. Counterintuitively, it turned out that increased longevity in superoxide dismutase overexpressing Caenorhabditis elegans is not related to elevated antioxidant capacity and in fact results from stimulation of the unfolded protein response [21]. Another study challenged a well-known astrocyte-neuron lactate shuttle hypothesis, according to which astrocytes supply neurons with lactate as an energy source during elevated neurological activity. NAD(H) monitoring with a genetically encoded probe showed that inhibition of monocarboxylate transporters did not affect the redox landscape of neurons [22]. Finally, there was a study dedicated to deciphering how NAD(H) redox state couples cytosolic and mitochondrial calcium levels in ribbon synapses [23].

Traditional approaches tend to perform poorly in cases where sample heterogeneity is pronounced, since this significantly blurs real analyte dynamics. At the same time, genetically encoded probes are of polypeptide nature, so they can be targeted to various cell types or organelles via a coordinated framework of expression systems and protein sorting machinery. First, this allows one to distinguish between subpopulations of cells in the organism. For example, it was shown that murine leukemia initiating cells fall into two categories on the basis of NADH/NAD+ ratio, which demonstrate different niches and regimens of division [24]. Second, it paves the way for the observation of heterogeneity on a single cell level. Thus, in murine peripheral axons, hydrogen peroxide levels are higher in mitochondria that reside near nodes of Ranvier, and they additionally increase during firing [25]. Third, non-equal steady-state concentrations or temporal dynamics can be measured for different compartments. In one work, the authors monitored the glutathione redox pool and hydrogen peroxide levels in the endoplasmic reticulum and cytosol of C. elegans during development. They found that redox homeostasis in these compartments was coupled, but the parameters studied often shifted in opposite directions [26].

Genetically encoded sensors are compatible with almost any model system; therefore, they have penetrated numerous fields of research. With their use, various teams have visualized redox processes during embryogenesis from the single cell stage, as well as during further development and ageing. The accumulated data sheds light on the role of hydrogen peroxide in zygote symmetry breaking [27], reveals oxidative oscillations that accompany early divisions [28], and captures redox waves over the course of life [29]. Inflammation and regeneration are another popular direction of investigation. It was genetically encoded indicators that proved the epithelial source of hydrogen peroxide during zebrafish tail fin amputation [30], its role in leukocyte recruitment, and the role of myeloperoxidase in its elimination [31,32]. Similar processes were observed for tadpole tail and zebrafish heart regeneration [33,34]. However, it was shown that hydrogen peroxide apparently does not participate in neutrophil migration towards the inner ear of larvae infected with Pseudomonas aeruginosa [35]. In another interesting work, an interplay between peripheral nerves, the Sonic Hedgehog pathway, and hydrogen peroxide was identified, which orchestrates tissue regeneration [36]. This finding suggests that redox metabolism is a key factor in the development of chronic wounds in neuropathies. The implementation of genetically encoded probes is not limited to animal research. In plants, these tools have helped to investigate pollen germination [37], stomata functioning [38], interaction with either pathogens [39,40], or symbionts [41], and signal exchange between plastids, nuclei, peroxisomes, and mitochondria [[42], [43], [44], [45]]. The scope of their use is no less wide in the context of studying microorganisms from various Kingdoms of life. Thus, genetically encoded probes have taken their rightful place in the toolkit aimed at understanding redox metabolism under different growth and stress conditions, during pathogen-host interaction, and after drug treatment. Research in this direction regularly bears fruit. Redox active fluorescent proteins reveal a non-equal ability of subcellular compartments in baker's yeast to recover from oxidative stress [46], as well as a non-equal response of cytosol and mitochondria to hypoxia with subsequent reoxygenation [47]. Moreover, it seems that these compartments have different access to NADPH produced in the pentose phosphate pathway. Another study identified a molecular pathway that regulates adaptation to light in yeast and includes hydrogen peroxide as a key messenger [48]. Genetically encoded probes confirmed that pathogenicity island-2 in Salmonella Typhimurium operates by providing evasion mechanisms for reactive oxidants [49]. Such probes allowed different redox subpopulations of mycobacteria that help each other in adaptation to antibiotics to be found [50,51]. Above we have listed only some interesting examples among many remarkable studies carried out using biosensors. In 2020, we published a comprehensive paper on in vivo research with genetically encoded redox indicators and corresponding model organisms, to which an interested reader can refer [52]. While these instruments have become a hot topic for a series of other thorough reviews [19,[53], [54], [55], [56], [57]], in recent years the scientific community has observed rapid growth in the number of novel tools, which has inevitably led to obsolescence of information. We therefore feel that there is a strong demand for systematization, description, comparison, and, finally, reassessment of genetically encoded redox sensors. We believe that an up-to-date review that can serve as a guide for potential biosensor users, is required.

One of the main trends in modern analytical biochemistry is the move towards multiparameter visualization [58]. This approach not only increases the amount of information obtained during a single experiment, but allows to compare the dynamics of different analytes more precisely, since this can be difficult to achieve using the traditional settings due to the variance characteristic of biological systems. Thus, we can observe in the literature that much effort is put into “recoloring” successful genetically encoded sensors, so they can be combined with each other. Besides, more and more research is driven towards testing novel readouts, which can be more fruitful than traditional intensiometric signal registration. In this field, fluorescence lifetime imaging microscopy, which is not dependent on the concentration of the probe or sample thickness, and requires a relatively small optical window, is one of the brightest examples. Finally, research is emerging in which genetically encoded sensors are combined not with each other, but with completely different optical techniques.

For instance, Raman spectroscopy is a powerful tool in the field of redox biology that allows the study of conformational and redox changes in molecules in various environments. It is based on the registration of photons scattered on the electronic clouds of atoms in molecules after non-elastic interaction of the incident light with the atoms (Fig. 1B). The energy difference between the incident and the scattered photons corresponds to the energy gap between vibrational states of the molecule after and before interaction with the excitation light. Since the properties of the molecule and its close environment affect the vibrational states, it is possible to judge various changes in the molecule according to the intensity and the energy of this non-elastic (or non-Rayleigh) light scattering. The pioneering experimental studies of molecules using the non-Rayleigh scattering method, later called Raman spectroscopy, were published in 1928 [59,60]. However, application of Raman spectroscopy to research on biomolecules was demonstrated later due to the absence of appropriate laser sources. Since the 1960's, laser-excited Raman spectroscopy has been used in the investigation of various living systems [[61], [62], [63]]. The combining of Raman spectrometer with confocal system and microscopes resulted in the rapid development of Raman microspectroscopy, which provides opportunity to record Raman spectra and Raman images of various biological preparations with the objective-dependent lateral resolution and z-resolution of approximately 1 μm or less. This allows to obtain Raman spectra from different regions of cells/tissues representing Raman scattering of all molecules located in the registration spot (Fig. 1C and D). Depending on the molecule concentration, its Raman cross-section, and laser light absorption it is possible to achieve higher Raman scattering of certain cell components and to estimate relative amount, redox, and conformational properties of the molecules of interest. At the moment there are many reviews on the application of this method and its subtypes in medicine and life sciences [[64], [65], [66], [67], [68]].

The aim of the current paper is to provide an update on the modern optical approaches in the field of redox biology. It consists of two main chapters. In the first, we describe the existing collection of genetically encoded fluorescent redox indicators and their biochemical properties, which underlie the advantages and disadvantages of these tools. We also discuss fluorescence lifetime imaging microscopy as a powerful technique for the measurement of their signals and the possibilities that open when combining redox sensors with other imaging approaches like NAD(P)H autofluorescence registration. In the second chapter, we focus on the applications of Raman spectroscopy in redox biology to cover its usage in the studies of the molecules participating in redox reactions, of O2 or NO• transfer, and in the detection of oxidative stress markers.