Mapping O2 concentration in ex-vivo tissue samples on a fast PLIM macro-imager

O2 PLIM microscopy was employed in various studies, however current platforms have limitations in sensitivity, image acquisition speed, accuracy and general usability. We describe a new PLIM imager based on the Timepix3 camera (Tpx3cam) and its application for imaging of O2 concentration in various tissue samples stained with a nanoparticle based probe, NanO2-IR. Upon passive staining of mouse brain, lung or intestinal tissue surface with minute quantities of NanO2-IR or by microinjecting the probe into the lumen of small or large intestine fragments, robust phosphorescence intensity and lifetime signals were produced, which allow mapping of O2 in the tissue within 20 s. Inhibition of tissue respiration or limitation of O2 diffusion to tissue produced the anticipated increases or decreases in O2 levels, respectively. The difference in O2 concentration between the colonic lumen and air-exposed serosal surface was around 140 µM. Furthermore, subcutaneous injection of 5 µg of the probe in intact organs (a paw or tail of sacrificed mice) enabled efficient O2 imaging at tissue depths of up to 0.5 mm. Overall, the PLIM imager holds promise for metabolic imaging studies with various ex vivo models of animal tissue, and also for use in live animals.

www.nature.com/scientificreports/ PMT-based, but not in the camera-based systems. As a result, it takes either many seconds to produce a complete 2D image of relatively small size (~ 30 s for a 256 × 256 pixels using Pt-Glc probe 22 ) or sub-seconds to image a bunch of pixel-size reference ROI (0.5 s for 12 micro-ROI using Oxyphor 2P 11 ). To achieve maximal acquisition efficiency, water immersion objectives are commonly used. However, when applied for brain imaging through cranial window, this approach may affect the temperature, capillary blood flow of the tissue and its oxygenation state 34 . Imaging of macroscopic samples such as organs and whole animals in PLIM mode is still rare [35][36][37] ; in a standard microscope setup, image area is too small for the analysis of O 2 dynamics on the organ level. Many of these limitations have been addressed by the new PLIM-TCSPC macro-imager 38 based on second generation, intensified Tpx3Cam optical camera 39,40 , which provides single-photon sensitivity, time resolution of 1.6 ns and a readout rate of 80 Mpixel/s 41,42 . The whole imager is compact and flexible like a common photographic camera, and can currently image objects up to 18 × 18 mm with spatial resolution of 39.4 μm 38,43,44 .
In this study, we assessed the analytical performance of this imager with various biological samples, including suspensions of mammalian cells and ex vivo samples of fresh animal tissue and whole organs isolated from sacrificed animals, in conjunction with the well-characterised near-infrared nanoparticle-based O 2 probe NanO2-IR. The aim was to evaluate the suitability of the imager for in vivo studies with live animals and models of human disease. For that, we performed a series of experiments with phosphorescent staining and imaging of ex vivo tissue samples, and measured intensity and lifetime signals, and O 2 distribution in various samples and conditions.
For the probe toxicity testing, cells were seeded on standard 96-well plates pre-coated with collagen IV at 2.5 × 10 4 cells/well, grown for 24 h, and then exposed to NanO2-IR at 1 mg/ml in complete DMEM for different periods. After that, total cellular ATP levels were measured using CellTiter-Glo Kit, white 96-well plates (Greiner Bio One) and Victor 2 reader (PerkinElmer) under luminometry settings 45 .
In the oxygenation experiments, HCT116 cells grown to 90-100% confluence were trypsinised, washed with DMEM, counted and transferred into 0.5 ml qPCR tubes at 0-5 × 10 6 cells per tube. Then the cells were precipitated to the bottom by mild centrifugation at 200 g, old media was aspirated, fresh complete DMEM with 1 mg/ml NanO2-IR was added to each tube, cells were re-suspended and precipitated as above. After 30 min incubation at 37 °C tubes with cells were imaged.
Preparation, staining and treatment of ex vivo animal tissue. All animal procedures were performed under authorisations issued by the Health Products Regulatory Authority (HPRA, Ireland), in accordance with the European Communities Council Directive (2010/63/EU) and approved by the Animal Experimentation Ethics Committee of the University College Cork. Fresh tissue samples were isolated from four-week-old female mice (Envigo, UK). On the day of the experiment, mice were euthanised by decapitation and their whole brain, lungs, fragments of small intestine and colon (large intestine), tail and front paws were quickly dissected, washed with PBS and placed in DMEM media equilibrated at 37 °C.
The selected samples were: tissues with high and low O 2 consumption (e.g. brain vs lung, or resting vs inhibited tissue); inner and outer areas of hollow organs exposed to air (intestine), and; subcutaneous tissue areas imaged through a thick skin layer (tail and paws). These tissue samples varied in shape and size (~ 8-15 mm, see scale bars in Figures). They were quickly excised, washed with PBS and immediately submerged in complete DMEM pre-warmed to 37 °C. Cortical tissue slices (~ 10 × 10 mm, 1 mm thick) were prepared by dissecting the extracted brain after cleaning; the entire process was completed within 15 min from excision. For the surface staining, the brain, lung and cortical tissue slices were transferred into mini-dishes, submerged in 2 ml of complete DMEM containing 1 mg/ml NanO2-IR and incubated for 30 min at 37 °C. Similar protocol was used for the staining of serosal surface of the intestine (1 mg/ml probe, 15 min). For the intraluminal staining, pieces of intestine were transferred to a dry Petri dish and excess of DMEM was removed with filter paper. Then 1 µl of DMEM containing 1 mg/ml of NanO2-IR was injected into the lumen with a Hamilton syringe, and samples were incubated for 15 min or up to 4 h (time lapse experiments). For the intra-tissue injection, 5 µl aliquots of DMEM containing 1 mg/ml NanO2-IR were injected with a Hamilton syringe subcutaneously into mouse tail or paw, which were pre-incubated for 30 min at 37 °C. After 15 min of further incubation, the samples were imaged under the settings described below.
To inhibit cell respiration in surface-stained tissue samples, they were transferred into Petri dishes with DMEM containing 10 µM AntA and 0.1% NaN 3 , and then incubated for 15  www.nature.com/scientificreports/ intraluminal respiration was inhibited by injecting the AntA/NaN 3 together with NanO2-IR (1 µg) and incubating the samples for 15 min at 37 °C. For imaging, all the samples were placed on a fresh and dry Petri dish and any excess liquid was carefully removed with a filter paper. Except for subcutaneous injections, all the experiments were repeated in three times with samples collected on different days. Unless otherwise stated, PLIM experiments were completed within one hour from the start of staining and in less than two hours from excision of tissue samples. PLIM Imaging setup. The wide-field PLIM-TCSPC macro-imager based on the time-stamping Tpx3Cam optical camera (Amsterdam Scientific Instruments, The Netherlands) is described in detail elsewhere 38 . Briefly, the compact and flexible configuration of the imager was provided by the Cricket adaptor with integrated S25 image intensifier, power supply and back-end relay optics (Photonis, France), and a 760 ± 50 nm bandpass emission filter (25 mm, Edmund Optics, York, UK). The two C-type lens mounts on the Cricket coupled it with the Tpx3Cam on the one end and the 50 mm Navitar NMV-50M11" lens on the other. A super bright 5 mm red LED (627 nm, Parts Express, Springboro, OH), synchronised with the camera and two pulse generators, was used for pulsed excitation of the sample from the top. For imaging the vials with media and cells, the imager was used in a horizontal orientation, while for tissue samples it was used in an upright orientation. A dry bath/block heater (Thermo Fisher Scientific, Waltham, MA) was used as a heating stage to maintain sample temperature at 37 °C.
All PLIM images were acquired at LED power 4 V, pulse width 50 ns and integration time 20 s. Sophy software (Amsterdam Scientific Instruments) was used to tune the operational parameters, align and focus the system. A custom-designed software was used to acquire the Tpx3Cam raw data in a binary format, while a C-language code was used to post-process the data. The resulting data matrix for each camera pixel was then fitted with a two-exponential function 46 to determine lifetime values. More details of data processing can be found in 38,47,48 . Calibration of NanO2-IR probe. The NanO2-IR probe was calibrated using a handheld lifetime reader OpTech Platinum (Mocon, MN). Standard O 2 -N 2 gas mixtures were produced by a precision gas mixer (LN Industries, Switzerland). The gas was pumped through a polystyrene tube containing the probe in DMEM and submerged in a circulating water bath (Julabo, Germany) set to 37 °C. The O 2 concentration was increased stepwise from zero to 21% O 2 (corresponding to 0-210 µM dissolved O 2 ). After ~ 20 min equilibration of the O 2 concentration and temperature in the sample, lifetime readings were taken at least 5 times to ensure stable and accurate lifetime values.

Statistics.
All independent experiments were conducted in triplicate to ensure consistency of results. Imaging data for cell and tissue samples was analysed using ImageJ software 49 . Statistical analysis of lifetime values obtained from whole PLIM images and derived O 2 values was performed using R (v3.6.3) and Rstudio (v1.2.5033) software. Lifetime values ranging 19-45 µs were analysed as per the NanO2-IR calibration equation (see Results). Figures were generated using the ggplot2 and patchwork tools. Measurements were aggregated based on experimental treatment and assessed for statistical differences using the Mann-Whitney U test (W-and p values were produced). Given the non-normal nature of the data and their unimodal distribution in all groups, differences in the modes (Δ-mode, µs) was deemed the most appropriate metric to report differences between treatments. In Figs. 3e,f, 4g-j and 5, we used three visualization elements that inform on the differences in distribution of the tested parameter between non-normal data sets, namely: (1) the box-plot (shows the median and interquartile range per batch of observations); (2) the violin plot (shows relative density of all observations in a data set with its wideness); the semi-transparent density plot (shows the total relative density of all observations of the same type). BoxPlotR tool (https ://shiny .chemg rid.org/boxpl otr/) was used to generate 'boxes-andwhiskers' plots of the intensity and lifetime variability (in Fig. 3h). Variability of the intensity signals and lifetime values was also evaluated using the standard equation for average deviation: AveDev = 1 n x − x′∨ . For that, all data points along three representative lines across brain samples were normalised to the average values, which were defined as 1 a.u.

Results and Discussion
Set up for the imaging of tissue O 2 . Recently, we have shown that the resolving power of our PLIM macro-imager, which utilizes a 256 × 256 pixels imaging chip, is 12.7 lp/mm 38 . The corresponding line width of 39.4 μm is comparable with the size of animal cells (typically, 10-40 µm in diameter). Thus, high-resolution O 2 imaging of samples up to 18 × 18 mm in size can be achieved.
To study oxygenation state and O 2 gradients, which reflect respiration activity of cells, several types of samples including cultured cells, tissue and organs isolated from sacrificed mice were stained with the NanO2-IR probe and then imaged on the imager at 37 °C. For each tissue model, optimised staining (time, temperature, amount of the probe) and imaging (LED power, distance to the sample, signal collection time) protocols were used (data not shown).
Phosphorescent probes used to stain tissue samples respond to local O 2 concentration by reversibly changing their phosphorescence intensity and lifetime 6,50  www.nature.com/scientificreports/ NanO2-IR probe belongs to the family of cell-permeable nanoparticulate O 2 probes 51,52 , well characterised and validated with different cell models. The nanoparticle forming RL-100 polymer serves to retain hydrophobic molecules of the phosphorescent dye, provide optimal microenvironment for their quenching by O 2 molecules and shield from unwanted interferences. The polymer enables solubility of the probe in aqueous solutions, efficient staining of cells and tissues, as well as stable and accurate O 2 calibration. The high brightness, near-infrared spectral characteristics of PtBP dye (excitable at 620 nm, emits at 760 nm 53 ) allows the detection of minute quantities of NanO2-IR and its use in high-resolution O 2 imaging applications.
Preliminary experiments. Staining of cultured cells with nanoprobes based on RL-100 polymer is usually performed at concentrations 5-50 µg/mL and incubation times 3-24 h at 37°C 51,52 . In these conditions probe cytotoxicity is undetectable. However, ex vivo tissue samples are unstable and require faster staining 53 , which can be achieved using higher probe concentrations but may result in significant cytotoxic effects. Using the two common cell lines, HCT116 and HEK293a, we examined cytotoxicity of NanO2-IR at 1 mg/ml concentration in the medium. Analysis of total ATP levels over 4 h exposure period revealed no significant changes in cell viability in the first hour or so, followed by a strong decrease in viability from 2 h onwards (Fig. 1a). Based on this data, we aimed to conduct imaging experiments with live cells and tissues within 1 h from the start of their staining.
Next, we imaged O 2 in samples containing growth media with different number of live mammalian cells, which were placed in 0.5 mL Eppendorf tubes designed for quantitative PCR. Cells were pelleted at the bottom of the tubes by gentle centrifugation, and gradients of dissolved O 2 in the surrounding media were generated by subsequent incubation of samples for 30 min at 37 °C. NanO2-IR probe, added at 1 mg/ml final concentration prior to centrifugation, stayed both in the medium and within cell pellet. Figure 1b shows PLIM images of the  From these preliminary experiments we concluded that bright phosphorescent probe NanO2-IR and the new Tpx3Cam-based PLIM imager can be applied for accurate measurement of O 2 in live tissue samples isolated from sacrificed animals, using the once-off calibration function described above.
Imaging of mouse brain and lung tissue surface-stained with NanO2-IR probe. Due to the ethical and technical constrains associated with the use of live animals, in this study we used samples of tissues freshly isolated from sacrificed animals, in which cells were still alive and respiring, though stressed as their normal supply of O 2 through vasculature was interrupted. In the ex vivo settings, O 2 passively diffuses to the tissue samples from air-saturated medium or ambient air, which could impose an additional hypoxia stress to the inner tissue areas. O 2 solubility, quenching efficiency of NanO2-IR and respiration activity of cells are strongly influenced by temperature 6,50 , therefore temperature of the sample stage was maintained at 37 °C during the experiments. With the above limitations, these tissue models are not ideal for physiological studies, however they were deemed appropriate for the assessment of analytical performance and general suitability of the PLIM imager for in vivo applications.
We imaged different types of animal tissue stained by topical application of NanO2-IR probe. Firstly, we stained three specimens of whole brain cortex and lung with NanO2-IR and analysed them by PLIM. For both tissue types, we used the following staining and imaging protocol. Sample 1 was stained for 30 min and then imaged. Samples 2 and 3 were stained for 15 min, imaged, then treated with respiration inhibitors for 30 min and imaged again under the same settings. After that, sample 1 was imaged again as untreated control, to trace possible changes in tissue oxygenation over the time of the experiment. The settings for acquiring PLIM images of these and other tissue types are given in Table S1.
The resulting phosphorescence intensity and PLIM images in Fig. 3a-d reveal efficient surface staining with NanO2-IR for the two tissues. The colours clearly reflect the difference in lifetime values for the two tissue types and, hence, their different oxygenation, which is driven by respiratory activity. In the lung tissue the overall variability of lifetime signals was higher than in the brain due to the elevated values in sample 1, which has no conclusive explanation.  www.nature.com/scientificreports/ As expected, treatment of the brain tissue with AntA/NaN 3 caused a reduction in NanO2-IR lifetime (Δ-mode = 3.48 µs), indicating the change in tissue oxygenation from ~ 150 to 200 µM upon the dual inhibition of mitochondrial cytochrome c reductase and oxidase (Fig. 3b,e, Fig. S1a). In turn, the same inhibition of mitochondrial respiration in lung tissue had little effect on lifetime values (Δ-mode = 1.46 µs, Fig. 3d,f). We attribute this to the intrinsically low respiratory activity of the lungs: the initial O 2 levels in resting tissue were close to air-saturated and could not be further increased by inhibitory treatments. The porous structure of the lung tissue has higher area of gas exchange, thus also contributing to increased tissue oxygenation.
PLIM of different surface-stained samples of the same tissue type (e.g. samples 1, 2, 3 of the brain, Fig. 3e) or the same sample at different time points (sample 1 of the brain, not shown) produced consistent lifetime values with moderate variability, which was lower than the variability of intensity signals, as demonstrated by analysing line profiles across the sample (Fig. 3g,h). By superimposing the data extracted from all the intensity and PLIM images, we observed no correlation between intensity signals and lifetime (Fig. S2). The higher uniformity of lifetime values and their independence from the recorded phosphorescence signals indicate that lifetime was not affected by the variability in tissue staining efficiency, as expected 33 . Bearing this in mind, we conducted PLIM of brain cortex slices (~ 1 cm 2 , ~ 1 mm thick), freshly dissected and surface-stained with NanO2-IR (Fig. 3i). The observed lifetime values (ranging ~ 20-25 µs) suggested that differences in local oxygenation are driven mainly by respiratory activity of particular areas of the tissue.
Based on these results, we concluded that surface staining of the two tissue types with NanO2-IR works reasonably well, and the PLIM imager enables fast visualisation of O 2 levels at tissue/media interface. As expected, they were close to ambient O 2 levels, with the interquartile range ~ 125-175 µM in the brain and ~ 170-200 µM in the lung (Fig. S1a). The magnitude of responses to mitochondrial modulators was proportional to respiratory activity of the tissue.
Deep tissue staining and imaging using NanO2-IR probe. To probe O 2 levels deeper inside tissue, we injected 1 µl of NanO2-IR (1 µg of the probe) into the lumen of excised samples of large and small mouse intestine (Fig. 4). In this case, the probe was shielded from the detector by a 0.2-0.25 mm thick intestinal wall 54 . It was contained inside the intestine and reported on O 2 levels in the lumen and at the surface of mucosa. A separate series of intestinal samples were also surface-stained and imaged as described above.
Comparison of the PLIM images recorded from the mucosal and serosal sides of the small and large intestine revealed pronounced differences. For the intraluminally stained samples, lifetime distribution histogram was markedly shifted towards larger values compared to the surface-stained tissue (for the colon Δ-mode = 10.45 µs, Fig. 4i, Table S2). This indicated, as expected, a lower oxygenation of the inner areas of the intestine (Fig. S1b). The intraluminally stained samples with inhibited mitochondrial respiration showed largely reduced lifetime values, compared to the untreated samples (Δ-mode = 4.45 µs and 3.93 µs for the small and large intestine, respectively, Fig. 4c,e, Table S2). Still, lifetime signals were not uniform and did not correspond to fully oxygenated state (see calibration in Fig. 2), most likely because respiration in the surrounding tissue was not fully inhibited by  www.nature.com/scientificreports/ the slowly diffusing drugs. Non-mitochondrial, AntA/NaN 3 -independent respiration could also contribute to the residual tissue deoxygenation 55 . For the colonic tissue samples shown in Fig. 4e, detailed oxygenation maps were produced (Fig. S3).
Comparing the non-treated intraluminally stained samples of the two anatomically different areas of the intestine, we found larger lifetime values for the colon compared to the small intestine, suggesting deeper colonic deoxygenation (Δ-mode = 10.13 µs, Fig. 4c,e,g,h).
Aiming to examine the stability of lifetime signals and respiratory activity of the intestinal tissue, we also conducted time-lapse PLIM analysis of intraluminally stained colon samples over 4 h. We observed a marked increase in lifetime values between 15 min and 2 h incubation (Δ-mode = 6.09 µs), which reflected the reduction in O 2 levels in the luminal part of the intestine. No further changes in lifetime values / oxygenation occurred between 2 and 4 h of incubation. Indeed, the mode value of ~ 43 µs corresponded to almost complete deoxygenation (~ 1 µM O 2 ), with Δ-O 2 = ~ 140 µM between the serosal and mucosal sides. We think that the lower lifetime values at 15 min reflect partial reoxygenation of the lumen during NanO2-IR injection with a Hamilton needle (Ø = 0.72 mm), which transiently enabled O 2 access to the lumen (Fig. 4b,e,f,j). From the fact that O 2 levels remained very low for at least 4 h, we also concluded that tissue sustained respiration over the whole period of manipulation, including the dissection, preparation, staining, incubation and imaging procedures.
We also applied subcutaneous injection of the NanO2-IR probe into intact tissue of freshly sacrificed mice. The selected parts of the tail and paw were thicker (~ 0.5 mm 56,57 ) and less permeable to light than intestinal wall. As a result, our first attempt with 1 µl injection of NanO2-IR did not produce sufficient signals under the same imaging settings as above. However, when injection volume was increased to 5 µl (5 µg of the probe), NanO2-IR was successfully imaged and PLIM produced high lifetime values both in the tail and the paw (Fig. 5, Table S2). These PLIM results are in agreement with the current knowledge about deep or even complete deoxygenation of tissues with interrupted O 2 supply, as exemplified by the brain tissue after clinical death 58 .

Conclusions
The Tpx3Cam based PLIM macro-imager was demonstrated in fast, high-resolution imaging of O 2 in sizeable biological samples, including liquid samples with respiring cells, live post-mortem animal tissue and whole organs, passively stained with NanO2-IR probe on the surface or microinjected in tissue depth. Tissue staining and PLIM protocols were optimised to ensure no detectable adverse effects of NanO2-IR probe on the live tissue samples during imaging (within 1 h), and tissues respiration remained active for at least 4 h after isolation. The imager allowed accurately measuring probe lifetime values and generating detailed O 2 concentration maps for tissue surface/interface and inner areas, the latter imaged through 0.2-0.5 mm thick tissue layers under lowpower LED excitation. Along with single photon sensitivity, advantages of our platform also include flexibility, transportability, high level of technology integration and relatively low cost (< 30% of the price for standard PLIM systems). The imager was able to reveal an adequate span of measurable O 2 levels (0-210 µM) in ex vivo tissue samples with higher (e.g. the lungs or brain surface) and lower (e.g. colonic lumen or inner areas of the paw and tail) O 2 levels than those typically found in vivo in vascularised tissues. Potential usability of the imager in various physiological in vivo studies is also supported by the fact that it can accurately measure expected changes in probe lifetime signal: decreases upon mitochondrial inhibition or increases after moving deeper into tissue.