Various in vivo and in vitro microscopy techniques are available for investigating spatial heterogeneity in phototrophs, each characterized by distinct physical principles and applications. In vitro methods, notably AFM and EM, are widely employed in the field (see e.g., (Bussi et al., 2019; MacGregor-Chatwin et al., 2019; Zhao et al., 2020; Zhao et al., 2022; Garty et al., 2024). AFM operates by visualizing samples through the measurement of forces between a sharp tip and a sample surface whereas EM relies on the interaction of an electron beam with the specimen. These methods offer superior resolutions, that can reach up to 1 nm or better (Hoogenboom, 2021), compared to in vivo techniques like confocal microscopy employed in Live-Cell Imaging (LCI). Confocal microscopy, based on fluorescence detection, yields resolutions from approximately ∼250 nm (in x-y) under conventional conditions and 20–120 nm in super-resolution mode depending on the method. Unlike AFM and EM based approaches the various LCI methods tend to not necessitate additional sample preparation but more importantly provide a wealth of complimentary information that address dynamic responses to the environment. In the following paragraphs of this review, we primarily focus on rapid 2D/3D live-cell imaging methods, with some exceptions (e.g., nanoSIMS, Fluorescence In Situ Hybridization–FISH). For further insights into in vitro imaging methods offering the highest possible resolution for phototrophic cells/membranes, we recommend consulting recent leading papers in the field of AFM, EM and its alternatives such as cryo-electron tomography and focused ion beam milling cryo-electron tomography (Bussi et al., 2019; Zachs et al., 2020; Zhao et al., 2020; Zhao et al., 2022; Wietrzynski and Engel, 2023).

The application of standard CM and SM in vivo methods have already shed light on many open questions in photosynthetic research; for instance how and where are the TM proteins assembled into the thylakoid membrane (Sun et al., 2019a; Huokko et al., 2021), and evidence of mosaic membrane protein nanodomains (Johnson et al., 2014; MacGregor-Chatwin et al., 2017) and microdomains (Strašková et al., 2019). LCI methods also revealed slow dynamics of photosynthetic protein complexes inside (Liu et al., 2012; Gutu et al., 2018; Rast et al., 2019; Krynicka et al., 2023) and outside of TMs in the form of proteins/complexes (Savage et al., 2010). LCI methods together with fluorescence tagging also helped to explore dynamics of membrane-less organelles containing RNA Helicase (Whitman Brendan et al., 2023), spatial and temporal dynamics of cyanobacterial chromosome (Chen et al., 2012; Liao and Rust, 2021), localisation of RNAs molecules during protein synthesis (Mahbub et al., 2020) and the role of several proteins (“structural determinants”) important for cyanobacterial morphogenesis, shape and cell division (see references in (Springstein et al., 2020)). Additionally, 2D and 3D confocal imaging has been able to visualise in vivo TM architecture in cyanobacteria (Strašková et al., 2019) as well as in chloroplasts (Iwai et al., 2016; Bykowski et al., 2021). LCI has facilitated the visualization of changes in TMs caused by various controlling factors like carotenoids (Bykowski et al., 2021), the role of Mg²⁺ ions on chloroplast structure (Rumak et al., 2010) and kinetic changes in the membrane architecture (Iwai et al., 2014; Iwai et al., 2016). These standard CM methods are limited in their resolution (in x-y ca. 250 nm), however, this weak point can be overcome by application of super-resolution methods that bypass Abbe’s diffraction limit (Baddeley and Bewersdorf, 2018; Schermelleh et al., 2019), reaching a resolution of up to 100 nm in x-y, such as commercial Structural Illumination Microscopy (SIM; MacGregor-Chatwin et al., 2017; Masakazu et al., 2018) and the Airyscan detector added to the Zeiss Confocal microscope (see the recent application in (Kaňa et al., 2023)). Regarding the physical principles, SIM methods enhance the resolution by projecting structured light patterns onto samples, while the Airyscan method is based on increasing the number of detectors (hexagonally packed detector array) together with image deconvolution (Huff, 2015). Additionally, there are a newly emerging methods named Expansion Microscopy, based on special sample preparation (Wassie et al., 2019; Bos et al., 2023) and Super-resolution Confocal Live Imaging microscopy (SCLIM) that is based on the combination of high-speed spinning-disk confocal scanning, ultrahigh-sensitivity detection and data processing (Iwai et al., 2016). Another useful method is re-scan confocal microscopy (RCM–see (De Luca et al., 2013)), which increases the resolution of standard confocal microscopy with an optical (re-scanning) unit that projects the image directly on a CCD-camera (see application for phototrophs in (Simonovic Radosavljevic et al., 2021)). Recently, a new super-resolution method named Single Pixel Reconstruction Imaging (SPiRI) was developed (Streckaitė et al., 2022) which represents a promising approach as the images are obtained simply using a classical epifluorescence microscope equipped with a sensitive detector and a precisely focused laser beam. The technique has been successfully applied recently in vivo in chloroplasts (Messant et al., 2023) and cyanobacteria (Chenebault et al., 2020). SPiRI and Airyscan (Figure 1) together with SIM, SCLIM and Re-scan confocal microscopy (RCM) represent some of the most promising techniques in the field of photosynthesis with resolution reaching ∼120 nm.

Other SM methods can achieve even higher spatial resolutions, up to 10–50 nm, then the methods described in the previous paragraph. These include PAINT (Point Accumulation for Imaging in Nanoscale Topography), PALM (Photoactivated Localization Microscopy), STORM (Stochastic Optical Reconstruction Microscopy) methods. However, their application on phototrophs is still rather challenging due to their special methodology connected with stochastic optical reconstruction (Ovečka et al., 2022). This approach can interfere either with high-pigment content or/and with the natural process of excitation energy transfer in photosynthetic light-harvesting antennae that is stochastic in nature (e.g., random nature of molecular interactions and fluctuations in pigment excitations/emissions). The PAINT method relies on the stochastic process of binding/unbinding of organic fluorophores to the target molecules of interest. The “blinking” effect observed in PAINT images allows then increase in resolution up to ∼30 nm compared to conventional dye-based strategies (Farrell et al., 2022). PALM/STORM methods are based on the stochastic activation (photoswitching) and localization of sparse subsets of fluorophores to achieve high-resolution imaging (∼10–30 nm) (Betzig et al., 2006; Shroff et al., 2008). Finally, a partially different approach is then applied in STED that relies on the overlap of two light beams in the focal region to deplete the fluorophores’ excited state around the focal point (Willig et al., 2006). It creates sub-diffraction-sized areas of emission and significantly increases resolution (20–50 nm). These three types of SM methods with resolution below ∼50 nm–PAINT, PALM/STORM, and STED–potentially have promise for future research in phototrophs however they need to overcome specifics of photothrophic samples (Bierwagen et al., 2010) to surpass Ayriscan, SPiRI, RCM, or SIM methods (resolution reaching ∼120 nm) that currently dominate in the application for photothrops (Iwai et al., 2016; Chenebault et al., 2020; Kaňa et al., 2023). Additionally, further progress in the field can be expected when all these SM methods will be combined with other specific microscopy techniques employing some additional principles (e.g., spectral detection, antiStokes microscopy, fluorescence kinetics, FLIM, Raman microscopy, cryo-confocal microscopy, etc.–see Section 4 - Special Microscopy Methods).

Studying the dynamical rearrangement of the TM structure and proteins/lipid trafficking is crucial for understanding the regulation of photosynthesis. For instance TM architecture and protein co-localization have been recognized as significant players regulating light-harvesting and electron transfer processes (see the review (Johnson and Wientjes, 2020)). Additionally, they are also important in protein assembly (Herbstova et al., 2012; Sun et al., 2019a; Huokko et al., 2021) and in functional re-shaping of the membrane architecture (see, e.g., review (Pribil et al., 2014)) that is one of the key questions in the field. Traditionally, the protein/lipid dynamics in TMs has been studied by means of the Fluorescence Recovery After Photobleaching method (FRAP; see the reviews of (Mullineaux, 2008; Kaňa, 2013; Kirchhoff, 2014). The FRAP method relies on photobleaching of fluorescently labelled molecules in a specific cell/membrane area followed by monitoring of the recovery of fluorescence in that region over time. This method allows researchers to estimate mobility parameters (e.g., diffusion coefficients in of lipids/proteins in TM see (Kaňa, 2013; Kirchhoff, 2014) or proteins binding dynamics (see, e.g., (McNally, 2008)) within different cellular environments. The method however has several pitfalls and possible artefacts (e.g., internal photo-physics effects in PBS (Liu et al., 2009; Gwizdala et al., 2018)) or in fluorescence proteins like GFP (Mueller et al., 2012). Additionally, FRAP measures mobility processes only at the mesoscopic scale (Mullineaux, 2008) whereas nanoscale protein trafficking visible by SM (Kaňa et al., 2023) is invisible for FRAP. Therefore, a perspective approach requires a combination of FRAP with other “single-pixel-based” methods like Single particle tracking (based on tracking single fluorescence particles, see, e.g., (Consoli et al., 2005) or with the more common method called microscopic Fluorescence Correlation Spectroscopy - FCS (Iwai et al., 2014; Janik et al., 2017; Crepin et al., 2021; Kaňa et al., 2021; Crepin et al., 2022). FCS is a semi-single molecule method that relies on detection of fluorescence fluctuation in time inside of the focal volume and on a subsequent time-correlation analysis of the obtained fluorescence kinetics. Typically, the focused laser beam illuminates only a few fluorescent molecules (between 10–100 particles) in a very small focal volume (∼250 nm in x – y). The time course of fluorescence changes in FCS reflects then all processes that change fluorescence intensity/yield inside of this focal volume. Microscopic FCS can measure molecule diffusion through the focal volume, protein-protein interactions, absolute proteins concentration and other parameters (see, e.g., (Digman and Gratton, 2011). The other microscopic correlation methods, Raster Image Correlation Spectroscopy (RICS) and/or Spatial-Temporal Correlation Spectroscopy (STICS) (Di Rienzo et al., 2013) are then built upon the foundational principles of single-spot FCS and extend it by moving the focal point alongside the sample. Therefore, they allow analysing the spatial and temporal dynamics of fluorescent molecules (correlation pixel by pixel) leading to creation of a detailed “map” depicting various measured parameters, such as molecular diffusion, binding kinetics, concentration gradients, particle movement directionality, dynamic alterations in cellular structures, and others. The future adaptation of these time-space correlative microscopy methods for photothrophs will allow us to resolve the nano-scale spatial variance in the protein/lipid dynamics in thylakoids (with, e.g., 100 nm resolution) that is currently only known with sub-micrometre resolution (see, e.g., difference in protein mobility between grana and stroma TM in (Kirchhoff et al., 2013).

It needs to be noted that only a combination of semi-single molecule microscopic FCS with mesoscopic FRAP will be able to address both types of proteins/lipids movement in TM, faster free diffusion and slower restricted diffusion (Im et al., 2013), because FCS can address only the fast processes (with characteristic time τ_char ∼ 1–1000 ms) and FRAP is suitable only for the slower processes (τ_char ∼ 1–300 s - protein-protein interactions (Wachsmuth et al., 2008)). This is one of the reasons why the diffusions of TM proteins obtained by FCS (Iwai et al., 2014) are so different from those obtained by FRAP (Kirchhoff et al., 2013): each method (FCS or FRAP) “observes“ different protein fractions (Wachsmuth et al., 2008). Their combination will better reflect the mosaic of TM proteins representing a complex combination of a stable organization (e.g., grana/stromal-like thylakoids in cyanobacterial microdomains (Strašková et al., 2019)) with fast protein (see super-resolution time-lapse imaging (Iwai et al., 2014; Iwai et al., 2016; Kaňa et al., 2023) and possibly also lipid trafficking (Sarcina et al., 2003). In fact, the importance of specialized lipid-based membrane microdomains is known in bacterial membranes as regions of increased fluidity (Strahl et al., 2014; Gohrbandt et al., 2022) or special “Functional Membrane Microdomains” (bacterial rafts-like domains) representing a mix of lipids and a specific set of proteins (Lopez and Koch, 2017). Therefore, the study of interaction between the polymorphic TM lipids (Garab et al., 2017) and the small structural proteins (e.g., Vipp, Curt, Flotilins (Siebenaller and Schneider, 2023) represent a future direction in the field of photosynthesis research.

At a larger scale, phototrophic cells or filaments can form multicellular tissues (e.g., higher plants) or characteristic cell colonies (certain cyanobacteria such as Microcystis or Trichodesmium) or microbial mats representing associations of different cyanobacterial and bacterial species. The microbial heterogeneity in these systems (scaling from hundreds of micrometres to a few mm) can be analyzed in vivo by different sensor-based approaches. These include traditional microelectrodes as well as microfiber-based optical sensors. The classical microsensors allow spatially resolved measurements (with point measurements at 100 to ∼10 µm resolution) of various parameters connected with photosynthetic or respiratory processes like O₂ concentration, pH, carbonate ion concentration, variable chlorophyll fluorescence, or irradiance. Recent developments in sensor technology have further improved spatial resolution; for instance minimum tip sizes of Clark-type O₂ electrodes are close to 3 µm (Weits et al., 2019), and smaller than 0.5 µm for carbon-fibre based electrodes (Alova et al., 2019). Also various new sensor types have been developed, including sensors for H₂O₂ (Ousley et al., 2022), CO₂ (Revsbech et al., 2019) and total dissolved inorganic carbon (Steininger et al., 2021). Additionally, planar optodes and optode micro/nanoparticles can simultaneously provide temporal and spatial distribution (2D or 3D) of pH, CO₂ or O₂ (Moßhammer et al., 2019; Elgetti Brodersen et al., 2020). The combination of multiple sensors applied on the same sample together with machine-learning approaches (Zieger and Koren, 2023) shows promise in understanding the complex interactions and feedback mechanisms between biological processes in microbial microenvironments (Lichtenberg et al., 2017; Wangpraseurt et al., 2017; Haro et al., 2019).

Most microscopy methods are based on detection of autofluorescence from the abundant photosynthetic pigments (e.g., chlorophylls, phycobilins), the fluorescence of lipid dyes (Strahl and Errington, 2017), protein tagging by fluorescent proteins (Yokoo et al., 2015; Cui et al., 2023; Zhang et al., 2023), or imunofluorescence staining (Trigo et al., 2017; Whitman Brendan et al., 2023). Chlorophyll a-based methods are very useful to detect photosynthetic activity and photosynthetic efficiency in vivo (Lazar, 2015). They target either the faster kinetics of fluorescence lifetimes (in 0.1–10 ns range—Fluorescence Lifetime Imaging (FLIM)), or the slower fluorescence kinetics (e.g., fluorescence kinetics microscopy (FKM) that detects activity of photosystem II (PSII) either by epifluorescence microscopy (Setlikova et al., 2005; Komárek et al., 2010), or by confocal microscopy (Storti et al., 2023)). FLIM represents a powerful technique that acquires the fluorescence lifetimes by photon-counting pixel-by-pixel with the spatial resolution depending on the microscopy method (see previous Section 2.). In contrast to steady-state fluorescence microscopy or slower FKM-based methods, FLIM is, by definition, concentration independent and is governed by the excitation states dynamics. Therefore, it can resolve and co-localize chromophores/proteins with the same emission band, that are however different in their fluorescence lifetimes (e.g., highly quenched emission from photosystem I (PSI) vs. mildly quenched light harvesting antennae of PSII). Therefore, FLIM has been widely used to study different processes in phototrophs with applications including spatial and temporal information about the distribution of the photosynthetic complexes (Iwai et al., 2010; Iermak et al., 2016; Nozue et al., 2016), regulation of light-harvesting processes and photoprotection (Pascal et al., 2005; Holub et al., 2007; Bhatti et al., 2021), chromatic adaptation in cyanobacteria (MacGregor-Chatwin et al., 2022) and to study proteins redistribution (Verhoeven et al., 2023).

The much slower kinetics of FKM-like methods are based chlorophyll autofluorescence (from microseconds to minutes) and they are used as a marker of photosynthetic activity (see e.g. (Kupper et al., 2000; Setlikova et al., 2005; Komárek et al., 2010; Storti et al., 2023)). Additionally, these microscopy methods are sometimes combined with kinetic detection of the whole fluorescence spectrum (Grigoryeva and Chistyakova, 2020). These slow FKM-like methods require commercial and custom-made measuring systems combining chlorophyll a fluorimeters with a camera and/or microscope (e.g., WALZ Imaging PAM, Microscopy PAM). They include systems suitable for sequential multicolour variable fluorescence imaging (RGB-Microscopy-IPAM) allowing to distinguish the photosynthetic activity of cells with different pigmentation in natural samples (e.g., epiphytes on seagrass leaves; see (Brodersen and Kühl, 2023)). Measuring the slow kinetics of variable chlorophyll fluorescence has already provided insights into the cell-to-cell variability in the regulation of photosynthesis in nitrogen-fixing filamentous cyanobacteria (Kupper et al., 2004; Ferimazova et al., 2013; Nozue et al., 2017; Colussi et al., 2024). Besides the epifluorescence-based FKM (Komárek et al., 2010) other systems have been used including confocal (Grigoryeva and Chistyakova, 2020; Storti et al., 2023), two-photon excitation (Kumazaki et al., 2007; Kumazaki et al., 2013) as well as a hyper-spectral confocal microscope that allows the detection of variability in the fluorescence emission or absorption spectra inside a single cell (Vermaas et al., 2008; MacGregor-Chatwin et al., 2017) or chloroplast (Kim et al., 2015). Additionally, there are some other less-known methods like excitation spectral microscopy (Zhang et al., 2021), cryo-electron microscopy (see (Vacha et al., 2007; Shibata et al., 2014; Steinbach et al., 2015)), and an Anti-Stokes fluorescence microscopy that is driven by thermally activated intramolecular vibrations, effective in selective imaging of PSI in different organisms (Nozue et al., 2016). We recommend those interested in these specialized methods to read the recent review on optical spectroscopy/microscopy by (Zhang et al., 2023).

Apart from spectroscopy methods based on visible (VIS) light, infrared (IR) absorption and Raman spectroscopy are two important techniques providing useful information on the photo-physics and photochemistry of photosynthetic organisms at the microsocpic level. In contrast to VIS methods (Section 2.1.), these methods analyze the frequencies of molecular vibrations, providing insights in to the chemical composition. IR absorption has limited applications in microsopy in aqueous environments due to the strong interference of the absorption band of water. In contrast, in vivo Raman microscopy requires minimal sample preparation (Bec et al., 2020) and it does not require the introduction of artificial staining or labelling to detect different chemical components, simultaneously, with a sub-cellular resolution. It has revealed spatial details about photosynthetic pigments, and other organic and inorganic compounds in various algae (Collins et al., 2011; Moudríková et al., 2016; Mojzes et al., 2020; Moudríková et al., 2021; Oka et al., 2021). Very few studies have addressed photosynthetic heterogeneity using Raman microscopy. Nevertheless it has been used to follow changes in thylakoid membranes by probing chlorophyll a-associated photoluminescence and carotenoid/phicobilin Raman signals in heterocyst cells (Tamamizu and Kumazaki, 2019). Raman applications need to cope with strong autofluorescence, long acquisition times and/or high laser intensities that can cause local heating of the sample (Butler et al., 2016). Some of these difficulties can be overcome by using near-IR excitation wavelengths (Heraud et al., 2018; Tamamizu and Kumazaki, 2019) or by coupling a spectrometer to a light-sheet illumination (Müller et al., 2016). An efficient way of increasing the Raman signal is to tune the excitation wavelength to closely match an electronic transition of the studied molecule and this is called resonance Raman (RR). Under these conditions, the Raman signal can be enhanced by orders of magnitude, allowing selective observation of the molecule of interest in a highly complex medium, thereby negating most of the negative aspects described above (Llansola-Portoles et al., 2022). For example, RR spectroscopy has been applied macroscopically to pinpoint signals of the pigments involved in photoprotective energy dissipation in intact chloroplasts and whole leaves (Ruban et al., 2007). Further advances in the field include the development of a high-resolution fluorescence-resonance Raman microscope based on the SPiRI methodology described in Section 2.1.

Spatial mapping of chemical elements up to a resolution of 50 nm can be achieved using nano-scale Secondary Ion Mass Spectrometry (nanoSIMS) (Musat et al., 2016). It is a powerful analytical technique used for high-resolution imaging and quantification of stable isotopes and elemental composition at the nanoscale. By bombarding a sample surface with a focused primary ion beam, nanoSIMS induces the emission of secondary ions, which are then detected and analyzed using a mass spectrometer. Therefore, it is a destructive method that however enables investigation of spatial distribution and heterogeneity of elements and their stable isotopes within biological samples with an unparalleled sensitivity and resolution. Indirectly, nanoSIMS can be used to adress physiological activity, such as spatial heterogeneity in the carbon/nitrogen assimilation processes (see Figure 2A), when it is combined with stable isotope incubations (Polerecky et al., 2021a). The method has been succesfully applied to phototrophs, revealing cell-to-cell heterogeneity in carbon and nitrogen assimilation in nitrogen-fixing filamentous (Eichner et al., 2017; Nieves-Morión et al., 2021) and unicellular cyanobacteria (Schreiber et al., 2016; Masuda et al., 2020; Polerecky et al., 2021b)). Its informative value can be further enhanced by correlative imaging involving, e.g., transmission and thin section EM, immunolabelling, X-ray fluroescence or Fluorescence In Situ Hybridization (FISH) for mRNA, rRNA or DNA (Martínez-Pérez et al., 2016; Krueger et al., 2018; Decelle et al., 2019; Loussert-Fonta et al., 2020; Polerecky et al., 2021b). The molecular cytogenetic technique FISH is a very useful tool to address RNA localization in cells hybridized with fluorescently labelled probes. The mRNA FISH technique has already provided important insights into the localization of TM protein biogenesis in cyanobacteria (Mahbub et al., 2020; Mahbub and Mullineaux, 2023), chloroplasts (Uniacke and Zerges, 2009; Schottkowski et al., 2012; Sun et al., 2019a) and the spatial localization of various physiological processes (e.g., carbon assimilation (Savage et al., 2010), respiration (Liu et al., 2012), DNA localization (Chen et al., 2012; Ohbayashi et al., 2019; Liao and Rust, 2021).

The most common methods that address spatial heterogenity in physiological processes in phototrophs are immunofluorescence approaches (Trigo et al., 2017) and genetically-based tagging of non-fluorescent membrane/cysosolic proteins with fluorescent proteins (FPs) ((Yokoo et al., 2015). To avoid potential artifacts or misinterpretations when employing FPs in highly pigmented phototrophs, several possible side-effects and important factors need to be considered: 1) FPs can interfere with the fuction of the tagged protein by affecting the natural photosynthetic energy transfer (fluorescence quenching, energy transfer from FPs to photosynthetic antenna, etc.); 2) Spectral crosstalk: the FPs may exhibit spectral overlap (in absorption and fluorescence) with other fluorophores or autofluorescent cellular components like chlorophylls/phycobilins (Yokoo et al., 2015); 3) pH sensitivity of the FPs: a low pH, as present in the lumen may quench the fluoresecence of FPs and it should be consider when selecting the probe (Shinoda et al., 2018); 4) Photostability and photophysics of the FPs: the effect of photostability on fluorescence blinking should be considered for particular microscopy methods (see, e.g., effect of GFP photoswitching in FRAP (Mueller et al., 2012), or natural blinking of the PBS protein (Gwizdala et al., 2018); 5) Background autofluorescence: it is crucial to include control strains (without FPs) in case of FPs tagging of low-abundant proteins. (5) Aggregational and structural artifacts: FPs tagging may cause unnatural filamentous structures (e.g., MreB proteins after YFP tagging (Swulius and Jensen, 2012)) or oligomerisation artifacts (Petersen et al., 2020).

Hence, several control expertiments are necessary to validate (unusual) fndings obtained using FP tagging. This includes application of label-free methods (e.g., EM), comparing protein tagging with different FPs (the most suitable FPs for the phototrophic “spectral window” between 500–600 nm are eGFP, YFP, mClover, TFP), and localization by immunogold labeling (Petersen et al., 2020). Another key control experiment is verifiying the proper assembly of the protein-complexes tagged by the FPs (CLEAR-Native native gels), testing the physiological function of the new strain (see, e.g., (Strašková et al., 2018; Strašková et al., 2019) and also the use of special methods that can address changes in energy transfer on the microcospic level (e.g., FLIM).