Fluorescence saturation imaging microscopy: molecular fingerprinting with a standard confocal microscope

Molecular specificity in fluorescence imaging of cells and tissues can be increased by measuring parameters other than intensity. For instance, fluorescence lifetime imaging became a widespread modality for biomedical optics. Previously, we suggested using the fluorescence saturation effect at pulsed laser excitation to map the absorption cross-section as an additional molecular contrast in two-photon microscopy [Opt. Lett. 47(17), 4455 (2022).10.1364/OL.46560536048677 ]. Here, it is shown that, somewhat counterintuitive, fluorescence saturation can be observed under cw excitation in a standard confocal microscopy setup. Mapping the fluorescence saturation parameter allows obtaining additional information about the fluorophores in the system, as demonstrated by the example of peptide hydrogel, stained cells and unstained thyroid gland. The suggested technique does not require additional equipment and can be implemented on confocal systems as is.


Introduction
Confocal fluorescence microscopy is a technique widely used for molecular imaging in cells and tissues.The development of various fluorescent probes allowed imaging of microenvironment parameters (viscosity, pH, temperature etc.) or certain molecules (e.g., using genetically encoded probes) with submicrometer resolution in real time.Fluorescent probes provide enhanced specificity and sensitivity of detection, as well as increased detection depth in case of tuning their spectral parameters to the optical transparency window [1,2].Although, in most cases, endogenous fluorescence (or autofluorescence) of cells and tissues does not possess specificity like exogenous fluorescent probes, it can be used for fundamental biochemical studies and biomedical diagnostics in numerous applications.For instance, endogenous fluorescence of NAD(P)H and flavins in cells is the basis of optical metabolic imaging [3][4][5].
An increase in specificity and sensitivity of optical molecular imaging can be obtained by measuring parameters of fluorophores other than fluorescence intensity.For instance, fluorescence lifetime imaging (FLIM) technique allows mapping of fluorescence decay parameters and can be combined with confocal microscopy.The advantages of FLIM are the independence of fluorescence decay parameters on excitation intensity and concentration of fluorophores, and high sensitivity of excited state lifetime to conformational changes of a fluorophore or alterations of its microenvironment.Thus, FLIM is a powerful modality for imaging and sensing in live cells and tissues [5][6][7][8].
Recently, we suggested another approach to mapping of fluorophores parameters with confocal microscopy based on the fluorescence saturation effect [9].At high excitation intensity I the dependence of a fluorophore's emission intensity Fluo on I starts to deviate from linear law due to ground state depletion (i.e., when the rate of a fluorophore's excitation from the ground to excited states exceeds the deactivation rate of the excited state), intersystem crossing, or other effects.This non-linear behavior of fluorescence on the excitation intensity is called saturation and has been studied for decades [10][11][12][13].The practical usefulness of the fluorescence saturation is that it can be described by a set of equations, which account for photophysical parameters of a fluorophore (absorption cross-section, fluorescence lifetime, relaxation rates) as parameters.Hence, these parameters can be obtained from fitting the fluorescence saturation curve Fluo(I) to the developed model of photophysical processes.At two-photon excitation, which is routinely used for FLIM, fluorescence excitation is high enough to induce saturation due to short (∼100 fs) laser pulses.The possibility to perform mapping of the absorption cross-section of endogenous fluorophores in living cells from a set of images measured at different excitation intensity in a two-photon regime has been demonstrated in [9].However, the question of whether implementation of Fluorescence Saturation Imaging Microscopy (FSIM) is possible with a standard confocal microscope in a continuous-wave excitation regime remains open.
The diagram describing the dynamics of the fraction of the fluorophores in the excited state (n e ) in two-level system is shown in Fig. 1(A), and more details can be found in Supplement 1, Section S1.The switch between the linear and saturation regimes is controlled by a dimensionless parameter characterizing the ratio of the excitation and deactivation rates of fluorophores Fστ, where F = I/ℏω is the photon flux density, σ is the excitation cross section, and τ is the fluorescence lifetime of a fluorophore, respectively.At Fστ ≪ 1 the fluorescence intensity depends linearly on F, while at Fστ≳ 1 fluorescence saturation is observed.Accounting for the fact that typical absorption cross section of fluorophores is of the order of σ∼10 −16 cm 2 and the fluorescence lifetime τ∼10 −9 -10 −8 s, one can expect that under photon flux densities of F ∼ 10 24 -10 25 cm −2 s −1 saturation effect can be observed.Despite the seemingly high values, such photon flux densities are routinely achievable even in "cw-mode" in standard confocal microscope systems.Indeed, estimating F ∼ P/(ℏωS), where P is the laser power, ℏω is the energy of the excitation photon, S is the beam area in the focal waist, and at P = 1 mW, λ = 405 nm, waist diameter d = 0.81 λ/NA, where NA = 1.4 is the numerical aperture of the objective, F is of the order of ∼3 × 10 24 cm −2 s −1 .
The fluorescence saturation effect in confocal microscopy has been analyzed previously in the scope of its influence on spatial resolution [13][14][15][16], yet the possibility of using this effect to determine molecular-specific contrast has not been considered.Indeed, by analyzing fluorescence saturation at different pixels it is possible to map molecularly specific parameter στ over the whole image.This procedure only requires capturing fluorescence images at different excitation intensities in the corresponding excitation intensity range (Fig. 1(B)).Hence, the aim of this work was to present a direct evidence of fluorescence saturation in cells and tissues in confocal microscopy experiments and to prove that implementation of FSIM results in additional modality that can be used for molecular imaging experiments.

Fluorescent dyes and quantum dots solutions
To evaluate the possibility of observation of fluorescence saturation effect with standard confocal microscopy setup, we first utilized various fluorescent molecular dyes and fluorescent quantum dots solution varying by the fluorescence lifetime and absorption cross-section.We used the water-soluble fluorescent dye thioflavin T (ThT), which can significantly change its fluorescence decay time (hence, the Fστ parameter) when bound to protein aggregates and fibrillar hydrogels [17][18][19][20].To test the hypothesis, samples of an aqueous ThT solution, ThT bound to insulin amyloid fibrils, and ThT bound to Fmoc-diphenylalanine (Fmoc-FF) hydrogel were used.
Aqueous solution of ThT (Sigma-Aldrich, Germany) was prepared using double distilled water with ThT concentration of 10 µM.The preparation of amyloid fibrils from insulin protein was carried out according to standard procedures and is described in detail in [21].Briefly, human recombinant insulin (PanEco, Russia) was used to prepare amyloid fibrils.Insulin was dissolved in pH 2 buffer to a final protein concentration of 34 µM with the addition of 2 µM ThT at room temperature, and then, to induce fibril formation, the solution was incubated at 62 °C for ∼24 hours.

Hydrogels preparation
To observe the fluorescence saturation effect on a structure with a heterogeneous distribution of the saturation parameter, we used hydrogels prepared from N-fluorenylmethoxycarbonyl diphenylalanine peptide (Fmoc-FF, GL Biochem, China) stained with ThT using a method similar to our previous works [22].The Fmoc-FF hydrogel was prepared using the solvent-switch method by diluting a dimethyl sulfoxide (DMSO) stock solution of Fmoc-FF with aqueous solution of ThT (30 µM) to a final concentration of peptide of 0.1 wt% (1.9 mM), the final concentration of DMSO in solution was 1%.This procedure resulted in the formation of a turbid Fmoc-FF suspension, which became optically transparent after a ∼2 hours at room temperature.After that the hydrogel was measured using confocal microscopy.For the convenience of further measurements using the microscopy method, hydrogel sample was prepared in a confocal Petri-dish.

Thyroid cells sample
To assess the heterogeneity of the distribution of the fluorescence saturation parameter in cell nuclei, thyroid cells stained with Hoecst-33342 dye were used.The specimens were collected from thyroid tumor (follicular thyroid adenoma).Fresh tissue in the nodule was scraped by one of the study pathologists using a sterile surgical blade to collect cells for optical measurements.Cells were then incubated with aqueous Hoechst-33342 (ThermoFischer, USA) for 40 min.Following staining, the cell layers were washed 3 times with PBS to remove excess dye, according to standard staining protocol, recommended by manufacturer.

Thyroid tissue samples preparation and analysis
Thyroid tumor tissue samples were obtained during surgical treatment of patients at the Endocrinology Research Center (Moscow, Russia).Autofluorescence measurements were performed within 3 hours after the surgery.To exclude drying, the sample was placed in the confocal Petri-dish filled with phosphate buffer saline (pH 7.4).
For histochemical analysis, the thyroid samples were fixed in 10% buffered formalin, processed in the histological staining system of Leica ASP200, and embedded in paraffin.Subsequently, paraffin sections with a thickness of 3 µm were cut from the paraffin-embedded tumor tissue samples using a microtome and applied to slides treated with poly(l-lysine).The slides were then stained with hematoxylin and eosin following the standard procedure.
Immunohistochemical analysis of tumor tissue sections was performed using the standard technique with a peroxidase detection system with DAB on an automatic Leica BOND III IHC staining system using Leica reagents.Antibodies to thyroid transcription factor-1 (TTF1, Leica) were used.All histological slides were scanned using a Leica Aperio AT2 system at 20× magnification for further analysis.

Fluorescence lifetime measurements measurements via fluorescence up-conversion
To evaluate ThT fluorescence lifetimes from fluorescence decay curves with a subpicosecond time resolution we used fluorescence up-conversion measurement system FOG-100 (CDP Ltd., Moscow, Russia).General principles of the method are reviewed in [23].Briefly, this setup uses the excitation of a femtosecond TiSa laser TISSA-50, generating pulses with a duration of less than 100 fs with a repetition rate of 86 MHz at a wavelength of 800 nm and an average power of 500 mW.Laser radiation at the fundamental frequency is partially converted into a second harmonic signal (average SH power -40 mW), which was used to excite the fluorescence signal of the sample, and the other part of the radiation at the fundamental frequency was used for optical gating of the fluorescence signal due to the generation of a sum frequency signal on a nonlinear crystal.The fluorescence signal with emission wavelength of 500 nm was detected at the sum frequency of fluorescence signal and gating pulse (800 nm) at the wavelength of ∼307.7 nm using a dual monochromator and a PMT (R1527, Hamamatsu, Japan) which has sensitivity from 185 nm to 680 nm.

Fluorescence saturation imaging confocal setup
Confocal imaging was performed using an Olympus Fluoview-3000 confocal microscopy system.Imaging was performed using a 60× oil-immersion objective with a numerical aperture of NA = 1.42 (UPLXAPO60XO, Olympus, Japan), using a continuous laser excitation sources with the excitation wavelength of 405, 488, 561 and 640 nm (OBIS405-50LX, OBIS488-20LS, OBIS561-20LS and OBIS-640-40LX, respectively, Coherent Inc., USA) and maximum output power of ∼1 mW after the objective (for all excitation wavelengths).A pinhole value of 200 µm was used, scanning over the field of view was carried out using a resonant scanner, providing dwell-time per pixel less than 0.1 µs.The fluorescence response was detected using a GaAsPphotomultiplier tube (PMT, FV31-HSD, Olympus, Japan) in the spectral ranges, individually adjusted for different fluorescent dyes and samples, ensuring low values of the fluorescence signal at high excitation intensities in order to avoid the effect of signal oversaturation at the detector.
To obtain maps of fluorescence saturation parameters, scanning of the same field view was carried out at different excitation intensities: from 0 to 0.5 mW with 0.05 mW step, first in the direction of increasing the excitation intensity, and then in the direction of decreasing the excitation intensity.
Spectrally resolved autofluorescence intensity maps were obtained in the "lambda-scan" mode of the Fluoview-3000 system using FV31-SD spectral detector, using which the fluorescence intensity of the same field of view was detected within multiple spectral channels with spectral width of 10 nm covering 430-700 nm range.

Fluorescence saturation imaging microscopy analysis
To estimate the fluorescence saturation parameter at different pixels of an image, the following procedure was used.Images acquired for the same field of view at different intensities were combined into one 3D image stack, containing information on the fluorescence intensity at given spatial position x, y and excitation intensity I.The fluorescence signal in each image was preliminary binned within a 13 × 13 window, i.e. the fluorescence signal in each pixel of an image was averaged over the 13 pixels adjacent to it.
After the preprocessing, the dependency of the fluorescence signal Fluo(I, x, y) in each pixel with coordinates x, y on the excitation intensity I was fitted using the model: where parameter A accounted for the dependence of fluorescence on excitation intensity in "linear" regime, parameter B took into account fluorescence saturation at cw excitation, and C accounted for background noise of the detector.Parameters A, B and C were estimated using non-linear least squares method.A detailed rationale for this formula is presented in the SI.The parameter BI max , where I max is the maximum excitation intensity used in the experiments, is similar to the dimensionless parameter F max στ, was used for further mapping of the fluorescence saturation in the obtained confocal images.

Fluorescence spectral imaging microscopy analysis
To visualize the results of spectrally resolved autofluorescence confocal imaging, the following procedure was applied.First, spatial binning of the signal was performed over the spectral stack of images in a window of 5 × 5 pixels, after that the average-weighted fluorescence emission wavelength λ avg (x, y) was calculated for each pixel using the formula (2): where Fluo(x, y, λ) is the fluorescence signal detected for a pixel with spatial coordinates x, y in a spectral channel with a central emission wavelength λ.The obtained maps of λ avg (x, y) were used to visualize spectral differences in the fluorescence image.All data analysis was performed using custom-written Python 3 scripts using the Numpy, Pandas, Matplotlib, Scikit-image, and LmFit packages.

Evidence of fluorescence saturation in confocal microscopy
We first investigated the fundamental possibility of observing linear and saturation regimes with a confocal microscopy setup in a cw-mode for several molecular fluorescent dyes and quantum dots solutions, varying by their absorption cross-section and fluorescence decay time τ, which determine the dimensionless parameter Fστ responsible for the saturation effect.
As a first system we chose the water-soluble dye thioflavin T (ThT).ThT is a fluorescent probe widely used for staining of amyloid fibrils and tracking protein aggregation kinetics in solution [17][18][19][20].Being a molecular rotor, in a free state (i.e., when not bound to a protein or a protein aggregate) ThT has a fluorescence decay time of ∼1 ps, while upon binding to a protein aggregate the fluorescence decay time of ThT increases up to ∼1000 ps [24].Thus, upon binding, the fluorescence decay time of ThT, concurrently with the fluorescence saturation parameter Fστ, exhibits an increase by three orders of magnitude.Hence, the tested hypothesis was that for free ThT molecules linear regime would be observed, while for bound ThT fluorescence saturation is possible.
Using the fluorescence up-conversion technique that allows measuring fluorescence decay curves with a ∼100 fs resolution, the decay kinetics of free and protein (insulin) aggregate-bound ThT were measured (Fig. 2(A)).Fitting of the obtained curves to monoexponential decay model yielded fluorescence lifetimes of 1 ps for free ThT and ∼1000 ps for bound ThT.Fluorescence saturation curves for free and bound ThT solutions were measured with a confocal microscope with a cw excitation at 405 nm using a 60× objective with NA = 1.42 and laser power ranging from 0 to 0.5 mW (corresponding to photon flux density of ∼1.6 × 10 24 cm −2 s −1 ) with a step of 0.05 mW.We mapped the microscopy field of view at different excitation intensities and then averaged the fluorescence signal of the entire field of view in the case of an aqueous solution of ThT, or over the region corresponding to the protein aggregate for bound ThT.It was found that in solutions of free ThT, the fluorescence decay curves in the range of photon flux densities up to ∼1.1 × 10 24 cm −2 s −1 are linear (Fig. 2(B)), while in the case of protein aggregates saturation of fluorescence was observed at photon flux densities exceeding ∼0.5 × 10 24 cm −2 s −1 (Fig. 2(C)).To ensure that the observed deviations of fluorescence intensity from the linear law are not caused by photobleaching, we used a resonant scanner providing short pixel dwell time of no more than 0.1 µs minimizing the excitation dose per pixel, and measured the fluorescence signal first in the direction of increasing the excitation intensity, and then in the direction of decreasing intensity (Fig. 2(B),(C)).In the case of significant influence of photobleaching, it was expected that the curves measured in both directions would not be the same, with the curve measured in reverse direction showing lower fluorescence intensities.However, this was not observed in the experiment (Fig. 2(B),(C)).Additional considerations on the influence of photobleaching on the measured fluorescence saturation curves are presented in Supplement 1 (Section S2, Fig. S2).
Using the model describing the dependency of fluorescence signal on the excitation intensity (Eq.( 1), Supplement 1 Section S1) accounting for saturation, we estimated the saturation parameters F max στ for both test systems.In the case of free ThT, the parameter was negligibly small (Fig. 2(B)), while for ThT bound to protein aggregates this parameter was F max στ ∼ 1.2 (Fig. 2(C)), confirming the fundamental possibility of observing fluorescence saturation with a standard confocal microscope and single photon excitation.
Figure 2(D) demonstrates fluorescence saturation curves for different dyes obtained at 405 nm excitation with the saturation parameter F max στ varying from negligibly small (e.g. for rhodamine-B) to high values of the order of ∼5-10 (for CdTe quantum dots).Measurements at different excitation wavelengths also allowed us to compare whether the observed fluorescence saturation effect is indeed governed by the absorption cross section.We exmined the molecular dye AlPc(SO 4 ) 4 and CdTe dots for which saturation was clearly observed at different excitation wavelengths and compared the ratio of the saturation parameters at different excitation wavelengths to the values of the absorption cross-sections estimated by spectrophotometry (Figs.2(E),(F)).As can be seen, the ratios of absorption cross sections at excitation wavelengths (dashed lines in the upper panels) and saturation parameters are close within relative error of ∼20%, corroborating the hypothesis on the mechanism of the observed fluorescence saturation effect.Fluorescence saturation curves measured for the specified dyes both upon sequential increasing and decreasing of excitation photon flux density obtained for the same field of view are presented in Figure S3 in Supplement 1.

Fluorescence saturation imaging microscopy: the case of hydrogels
The next step of FSIM verification comprised mapping of the fluorescence saturation over heterogeneous structures stained with exogenous fluorescent labels.The first system under investigation was the hydrogel obtained as the result of Fmoc-FF peptide self-assembly stained by ThT.Upon self-assembly, Fmoc-FF aggregates into structures, which bind ThT with high specificity, and fluorescence lifetime of ThT in Fmoc-FF hydrogels is ∼1 ns, and could be heterogeneously distributed over different filaments in hydrogel upon different stages of gel formation [22].Hence, saturation of ThT fluorescence can be expected at excitation parameters similar to that used in the case of insulin aggregates with saturation parameter, following the fluorescence lifetime, inhomogeneously distributed over the image.
A special care was taken to prove that the decrease of fluorescence at high excitation intensity is not caused by photobleaching.For this, the system of Fmoc-FF hydrogel-ThT was measured at maximum laser power (P = 1 mW, F ∼ 3.2 × 10 24 cm −2 s −1 ) multiple times for the same field of view (Fig. 3(A)).The dependence of the integral fluorescence intensity on the scan number is presented in Fig. 3(B).We observed that at high laser powers only a subtle photobleaching of the order of 0.5% per scan is observed under the experimental conditions used.In the case of significant fluorescence saturation (Fig. 2(C)), the observed deviations from the intensity expected in the absence of the saturation effect are ≈20%, corroborating that the associated effect is not caused by photobleaching.
Having confirmed that the effect of photobleaching was insignificant, we performed mapping of the saturation effect over the hydrogel-ThT system.Previously, it was shown that the structure of Fmoc-FF hydrogel is heterogeneous, and it contains both fibers and spherical structures formed at intermediate stages of hydrogel formation [22].Upon binding to fibers and spheres ThT exhibits different fluorescence lifetime τ, varying approximately ∼1.5-2 times, thus, it could be expected that saturation parameter Fστ would be different for these structures.The confocal fluorescence images of the hydrogel were measured at different excitation power ranging from 0 to 0.45 mW (photon flux density ∼1.4 × 10 24 cm −2 s −1 ) in both directions, i.e., upon increase and decrease of intensity, to control for possible photobleaching.Then, fluorescence saturation curves Fluo(F) for each pixel were fitted using the fluorescence saturation model (Materials and Methods, Section 2.4).
Representative image of the ThT-stained hydrogel and the corresponding map of the saturation parameter F max στ are presented in Fig. 3(C) and 3(D), respectively.The saturation parameter F max στ exhibits variation in the range from 0 to 2.5 units across the image.Figures 3(E),(F) show fluorescence saturation curves for the areas highlighted by red and blue rectangles on the maps in Figs.3(C),(D).As can be seen, these regions have almost the same fluorescence intensity at maximum excitation intensity, while the fluorescence saturation parameter F max στ differs almost twofold.This fact means that two areas of the sample can not be distinguished in terms of different ThT photophysical parameters if only fluorescence intensity is measured (Fig. 3(C)), while measuring fluorescence saturation allows for fingerprinting of ThT molecules with different microenvironment, which influences the ThT fluorescence lifetime and absorption cross-section, thus, the fluorescence saturation parameter F max στ.

Fluorescence saturation imaging microscopy of stained cell nuclei
Having observed fluorescence saturation in Fmoc-FF hydrogels with ThT, we further assessed the possibility of mapping fluorescence saturation in cells stained with dyes different from ThT.We used cells isolated from follicular adenoma of the thyroid gland and stained them with the Hoechst-33342, a dye that incorporates into cell nuclei and emits fluorescence when excited at 405 nm.The purpose of the experiment was to assess the extent to which the saturation parameter of exogenous dyes can be used as a source of information additional to fluorescence intensity, assess heterogeneity of its distribution within cells' nuclei and evaluate whether the distribution of this parameter will be of interest for additional studies in future.Previous findings have shown that the fluorescence lifetime of common "nucleus" dyes can indicate chromatin condensation regions due to local viscosity alterations within nucleus [26].The fluorescence lifetime of the 4',6-diamidino-2-phenylindole (DAPI) dye in nuclei allowed the classification of chronic lymphocytic leukemia B cells [27].The interaction of Hoechst-33342 and 5-bromo-2-deoxyuridine enables the visualization of cell cycle phases by modulating the Hoechst dye's lifetime [28].Through fluorescence lifetime imaging with specialized dyes DNA conformational substructures like G-quadruplexes can be precisely localized within cell nucleus [29].Thus, it can be expected that the fluorescence saturation parameter, following the fluorescence decay time, will exhibit heterogeneous distribution within cell nuclei, potentially offering biological insights through variability of this photophysical parameter.
Figure 4(A) illustrates a fluorescence saturation map of Hoechst-33342-stained thyroid cells, accompanied by representative saturation curves from distinct regions displayed in Fig. 4(B).Similar to the approach used for saturation mapping in hydrogels and dyes solutions, we controlled the photobleaching process by acquiring saturation stacks with both increasing and decreasing excitation intensity.We found, that, indeed, the saturation parameter is non-uniformly distributed within each nucleus with the parameter F max στ varying from 0.5 to ∼1.5.Furthermore, this parameter exhibited variations in median values across different cells' nuclei spanning from ∼0.7 to ∼0.9 Histograms depicting the distribution of the saturation parameter in segmented regions of cells manually delineated in the image are presented in Fig. 4(C) (segmentation regions depicted in Fig. S4).This observed heterogeneity holds promise for characterizing the "biological" state of cells and potentially aiding in their differential diagnosis.

Fluorescence saturation imaging microscopy of endogenous fluorophores in thyroid gland ex vivo
The results obtained on the model systems of Fmoc-FF hydrogels with ThT and thyroid cells with Hoechst dye demonstrate that FSIM provides molecular-specific information with spatial resolution.This concept is similar to that of the FLIM method, which is based on mapping of fluorescence lifetime.Hence, we applied FSIM to biotissues ex vivo to verify whether different structures can be highlighted by the values of their saturation parameter similar to what is obtained with FLIM.The experiments were performed on thyroid glands ex vivo, which are known to contain multiple fluorophores [30].These fluorophores include NADH in cells, fluorescent cross-links in collagen [31], and fluorophores in the granules of the thyroid follicles, whose origin is debatable [32,33].Figures 5(A), B demonstrate the typical structure of the thyroid gland region with visible follicles containing thyroid hormone surrounded by thyrocytes, observed with hematoxylin and eosin staining (Fig. 5(A)), as well as with immunohistochemical staining on thyroid transcription factor-1 (TTF-1), highlighting thyrocytes in brown (Fig. 5(B)).Figure 5(C) shows an autofluorescence image of the thyroid gland with clearly visible follicles containing thyroid hormone and granules surrounding the follicles obtained with 405 nm excitation and emission detection in the 500-600 nm spectral range.Applying FSIM, we observed that the fluorescence saturation parameter for a thyroid gland was heterogeneously distributed, both in granules and inside follicles (Fig. 5(D)).For comparison, we mapped the fluorescence emission spectra of a given tissue region over the range of 430-700 nm at similar excitation and colored the image using the weighted average fluorescence emission wavelength (Fig. 5(E)).Spectrally resolved microscopy showed that the fluorophore localized inside the follicles has a shorter wavelength emission with λ avg = 520-540 nm, while the granules surrounding the follicles have a longer wavelength emission with λ avg = 550-570 nm.At the same time, we discovered that granules that are homogeneous in terms of the position of the fluorescence emission maximum turn out to be heterogeneous in terms of the saturation parameter F max στ ranging from 0.5 to 2.0, and vice versa, some of the pixels in the image can be clustered according to their saturation parameter, but not according to their spectral band shape.Hence, fluorescence saturation parameter and fluorescence spectral band shape are both molecularly specific parameters which complement each other (Fig. 5(F)).
Interestingly, it was found that endogenous tissue fluorophores, namely, collagen cross-links, also exhibit heterogeneity of fluorescence saturation parameter (Fig. 6).As can be seen, different microscopic regions are observed in collagen, which are characterized by different saturation parameters F max στ, varying in the range from 0.2 to 1.5.The reason for this heterogeneity may be due to the different origin of fluorescent cross-links [34,35].

Discussion
Increasing detection specificity is a central task in molecular imaging, as it allows in-depth analysis of processes in a studied system.For this, specific probes and molecular sensors are designed, which cover selected areas of application.A more general approach is based on obtaining additional information by implementing multimodal detection, for instance, by combining different techniques -fluorescence and Raman [36], OCT and fluorescence [37], fluorescence and AFM [38] etc -or by measuring photophysical parameters different from fluorescence intensity.These include spectral parameters [39], photobleaching rates [40] and fluorescence decay parameters (FLIM) [7].Observation of fluorescence saturation in confocal microscopy experiments free from evident photobleaching extends the possibilities of molecular imaging by adding saturation parameter to the set of molecular-specific indicators, which can be used as a contrast.We previously demonstrated this approach for two-photon microscopy, where it can be combined with FLIM, while in this paper it is shown that photon flux densities high enough to induce saturation can be achieved at cw excitation.Thus, novel information can be obtained about spatial distribution of fluorophores without the need for additional equipment.
We firstly demonstrated that saturation contrast is indeed observed in the cw mode for standard confocal microscope in different fluorescent molecular dyes and quantum dots that differ both in fluorescence lifetime (case of free and protein-bound Thioflavin T, Fig. 2(A)-(C)) and in fluorescence cross-section (molecular fluorophores and quantum dots (Fig. 2

(D)-(F)).
Next, to show that fluorescence saturation indeed provides novel information on the studied object, the following examples were considered.Firstly, imaging of the hydrogel stained with ThT was performed.During the time course of Fmoc-FF self-assembly, intermediates with high affinity towards ThT form [22], and as the structural organization of aggregates in the system evolves, fluorescence lifetime of ThT also changes, making it possible to track major transitions in hydrogel formation.The photophysics of ThT is governed by local microviscosity, which modulates mutual rotation of its fragments and leads to variation of non-radiative deactivation rate [18,19].Hence, different photophysical parameters of ThT reveal differences of its microenvironment.Notably, radiative relaxation rate and absorption spectra of ThT are only slightly modulated by microenvironment [18][19][20], and changes in the saturation parameter F max στ are likely to correlate solely with ThT fluorescence lifetime τ.The obtained maps of the saturation parameter for Fmoc-FF hydrogels (Fig. 3) suggest that FSIM can be used to monitor the self-assembly process similar to FLIM.
The utilization of FSIM also presents a novel avenue for investigating saturation parameter within cell nuclei.Our examination of thyroid gland follicular adenoma cells stained with Hoechst-33342 revealed a heterogeneous distribution of the saturation parameter across the nuclei (Fig. 4).While further experiments are warranted to elucidate the precise nature of this heterogeneity, it is conceivable that akin to the heterogeneous distribution of fluorescence decay times observed for different dyes within the nucleus [26][27][28][29], variations in the saturation parameter could potentially signify nuances in DNA distribution density, cell cycle progression, or other biologically pertinent information.This intriguing finding underscores the potential of FSIM as a valuable tool for unraveling intricate cellular dynamics and warrants future exploration into its applications in cellular imaging.
Finally, for demonstrating FSIM capabilities in molecular imaging thyroid gland was investigated (Fig. 5).Thyroid is known to exhibit rich autofluorescence originated from a multitude of fluorophores [30].It was shown that, indeed, fluorescence saturation parameter exhibits high heterogeneity for thyroid, thus revealing heterogeneity of fluorophores and their environment.Thyroid fluorophores can be also nicely separated with spectrally-resolved imaging (Fig. 5), which demonstrated fluorescent granules (presumably residing in thyrocytes within the follicles) characterized by red shifted emission.These granules exhibit similar spectral properties, suggesting identical fluorophores within them.On the other hand, fluorescence saturation parameter for the granules showed rather high variability with lack of correlation with average fluorescence emission wavelength.As a hypothesis, we suggest that the saturation parameter στ can be an additional independent source of information, which allows one to independently identify different tissue areas based on the saturation parameter while its variability can be caused by biological complexity of the sample or by more complex model of photophysical parameters in the case of concentrated fluorophores than the one shown in Fig. 1.Namely, at high concentration bimolecular processes can take place, such as singlet-singlet annihilation [41] -the larger the concentration, the larger the saturation parameter.This hypothesis however needs additional testing.
Heterogeneity of fluorescence saturation parameter was also observed for collagen bundles (Fig. 6) that can be due to differences in spatial distribution of fluorescent cross-links and posttranslational modifications.Altogether, the presented demonstrations show that FSIM provides new independent information of cells and biotissues, which could be useful for biomedical diagnostics.

Conclusion
The concept of fluorescence saturation imaging microscopy with a standard confocal microscope has been suggested and tested on exemplary objects -hydrogel and thyroid gland.The possibility of mapping molecular-specific parameter, the fluorescence saturation parameter F max στ, was demonstrated.Our results suggest that FSIM could find applications in bioimaging as an additional modality as it does not require any upgrade of confocal systems, although multiplexing FSIM with FLIM would benefit the informativity of data obtained.

Fig. 1 .
Fig. 1.A) Schematic description of the fluorescence saturation effect for single-photon excited fluorescence due to the ground state depletion using the model of a two-level system.The differential equation is given for the fraction of fluorophores in the excited state n e , where I is the intensity of the excitation, ℏω is the energy of the excitation photon, τ is the fluorescence decay time of the fluorophore.B) The scheme of fluorescence saturation parameter analysis: acquisition of fluorescence images at different excitation intensities and analysis using the fluorescence saturation model.

Fig. 2 .
Fig. 2. A) Fluorescence decay curves of an aqueous solution of free Thioflavin T (blue curve) and Thioflavin T bound to a protein aggregate (red curve).B,C) The dependence of fluorescence intensity on the excitation photon flux density for free Thioflavin T (B) and Thioflavin T bound to insulin aggregates (C).D) Fluorescence saturation curves for different molecular dyes and quantum dots obtained with fluorescence excitation at 405 nm.E, F) Top panels: normalized absorption spectra; bottom panels: fluorescence saturation curves for different excitation wavelengths for sulfonated aluminum phthalocyanine (AlPc(SO4)4) (E) and CdTe quantum dots (F).Depicted ratios provide an estimate of the absorption cross-sections estimated by spectrophotometry and via fluorescence saturation analysis.

Fig. 3 .
Fig. 3. A) Fluorescence image of the Fmoc-FF hydrogel stained with ThT mapped to assess the effect of photobleaching.B) Dependence of the average intensity, normalized to the average image intensity for the first scan, on the number of scan (excitation power P = 1 mW, photon flux density ∼3.2 × 10 24 cm −2 s −1 ).S,D) Fluorescence intensity (C) and fluorescence saturation parameter F max στ (D) maps for the Fmoc-FF hydrogel stained with the ThT dye (F max ∼ ∼1.4 × 10 24 cm −2 s −1 ) E,F) Fluorescence saturation curves for the regions of the maps depicted with the red (E) and blue (F) rectangles in panels C,D.

Fig. 4 .
Fig. 4. A) Fluorescence saturation parameter map for thyroid cells stained with Hoechst-33342 dye at 405 nm excitation.B) Representative fluorescence saturation curves from regions differing in saturation parameter depicted in panel A as red rectangles.C) Histogram of the distribution of the fluorescence saturation parameter across regions, corresponding to individual cell nuclei in the image (F max ∼ 1.1 × 10 24 cm −2 s −1 ).

Fig. 5 .
Fig. 5. A,B) Hematoxylin & Eosin (A) and thyroid-transcription factor-1 (B) stained tissue of the thyroid tissue cancer sample.C, D, E) Map of fluorescence intensity (C), fluorescence saturation parameter (D) and intensity weighted average fluorescence emission wavelength map (E) of the tyroid gland section.F) Correlation of weighted average fluorescence emission wavelength and fluorescence saturation parameter for different pixels in the images shown in panels D and E (F max ∼ 1.4 × 10 24 cm −2 s −1 ).