Imaging oxygen in neural cell and tissue models by means of anionic cell-permeable phosphorescent nanoparticles

Abstract

Cell-permeable phosphorescent probes enable the study of cell and tissue oxygenation, bioenergetics, metabolism, and pathological states such as stroke and hypoxia. A number of such probes have been described in recent years, the majority consisting of cationic small molecule and nanoparticle structures. While these probes continue to advance, adequate staining for the study of certain cell types using live imaging techniques remains elusive; this is particularly true for neural cells. Here we introduce novel probes for the analysis of neural cells and tissues: negatively charged poly(methyl methacrylate-co-methacrylic acid)-based nanoparticles impregnated with a phosphorescent Pt(II)-tetrakis(pentafluorophenyl)porphyrin (PtPFPP) dye (this form is referred to as PA1), and with an additional reference/antennae dye poly(9,9-diheptylfluorene-alt-9,9-di-p-tolyl-9H-fluorene) (this form is referred to as PA2). PA1 and PA2 are internalised by endocytosis, result in efficient staining in primary neurons, astrocytes, and PC12 cells and multi-cellular aggregates, and allow for the monitoring of local O₂ levels on a time-resolved fluorescence plate reader and PLIM microscope. PA2 also efficiently stains rat brain slices and permits detailed O₂ imaging experiments using both one and two-photon intensity-based modes and PLIM modes. Multiplexed analysis of embryonic rat brain slices reveals age-dependent staining patterns for PA2 and a highly heterogeneous distribution of O₂ in tissues, which we relate to the localisation of specific progenitor cell populations. Overall, these anionic probes are useful for sensing O₂ levels in various cells and tissues, particularly in neural cells, and facilitate high-resolution imaging of O₂ in 3D tissue models.

Electronic supplementary material

The online version of this article (doi:10.1007/s00018-014-1673-5) contains supplementary material, which is available to authorized users.

Introduction

Optical bio-imaging has evolved as a powerful tool for studying biological processes on a cellular level, providing single cell and even single molecule resolution [1, 2]. However, targeted delivery, intracellular distribution and operational performance of many bio-imaging probes are still far from optimal, especially for live-cell imaging applications. Nanoparticle (NP) probes offer significant advantages including increased brightness, photostability and shielding of the indicator dye from unwanted interferences, but their chemical modification and successful intracellular penetration remain difficult to achieve [3]. NP probes often use complex endocytic delivery mechanisms that limit their specificity to a narrow range of cell and tissue types [4], have limited stability in biological media (e.g. resulting in aggregation or leaking of cargo [5]) and display batch to batch variability. In order to increase the cell permeability of NP probes they are generally small in size (<100 nm), positively charged and contain cell targeting groups located on their surface [4, 6–8].

Recently, the ability to sense and image molecular oxygen (O₂) by phosphorescence quenching has shown significant advances. A number of NP probes based on Pt(II)-tetrakis(pentafluorophenyl)porphyrin (PtPFPP) and Ru(dpp)²⁺₃ dyes have been described; these probes possess increased brightness, photostability and compatibility with different detection modalities including ratiometric intensity and lifetime imaging under one and two-photon excitation modes [9–14]. Successful results using such probes have been demonstrated with 2D and 3D cell and tissue models (multi-cellular aggregates, spheroids [15]), in vivo imaging [8, 16], ex vivo cultures [17–19], analysis of cellular O₂ gradients [20], as well as hypoxia and cellular responses to drug treatment [15, 21, 22]. For applications involving gross anatomical features it is sometimes sufficient to stain the surface of cells and tissue samples, rather than deliver the probe into the cytosol or mitochondria [15, 23]. Alternatively, intracellular localisation of bio-imaging probes is beneficial for high-resolution O₂ mapping of meso- and microscopic objects (i.e. individual cells, multi-cellular aggregates, tissue explants) and allows for advanced investigations into detailed cellular processes. For instance, there is currently a debate regarding whether significant intracellular O₂ gradients exist in respiring cells and tissues, and the extent to which these molecular gradients may play regulatory roles [20, 24–28]. The dynamics of changing O₂ levels are also important for a comprehensive understanding of the physiology, metabolism and bioenergetics of neural tissue [29–32]. Oxygenation of neural tissue models has traditionally been analysed with micro-electrodes or extracellular probes [33, 34], which provide limited spatial information regarding O₂ levels. In current research, many model systems (e.g. models of stroke, neurodegenerative disorders, nervous system development, microcirculation, tumorigenesis, and drug delivery) are now shifting from 2D cell culture [35] towards 3D cell and tissue models (e.g. spheroids, organotypic slices, ganglions, organoids) [36–38] and in vivo models [14, 39, 40]; therefore, novel, high-resolution O₂ imaging probes are becoming increasingly necessary. To date, cell-penetrating O₂ probes have been largely used in “proof-of-principle” studies within specific model systems [7, 41–46], and their narrow range of cell specificity as well as general incompatibility with other models greatly limits their use [47–49]. Ultimately, very few O₂ imaging probes are suitable for detailed physiological studies with neural cell and tissue models [16, 19, 50].

Here we describe two new cell-penetrating probes based on anionic NPs of acrylic co-polymer, poly(methyl methacrylate-co-methacrylic acid); these polyacrylate NP (PA) probes are highly suitable for imaging of neural cell and tissue models. We evaluated their cell specificity, transport mechanisms, and cytotoxicity, and found them to be superior in many aspects when compared with the cationic NP probes currently in use. Further, we demonstrated their analytical performance in O₂ sensing and imaging experiments using various model systems, including 2D cultures of primary neural cells, multi-cellular aggregates (3D spheroids) exposed to metabolic stimulation and cultured organotypic brain slices. The relation between local O₂ levels and development of neural progenitor cells in developing tissue slices was also examined, using multi-parametric imaging combining live and subsequently fixed samples.

Results and discussion

Design and characterisation of PA probes

In previous work with the positively charged probes NanO₂ and MM2, we observed poor intracellular staining of neuron-like PC12 and SH-SY5Y cell lines as well as neurospheres from primary cells [17, 51], with ‘patchy’ patterns that were not ideal for high-resolution O₂ imaging such as confocal or light-sheet microscopy. We therefore searched for alternative polymers harbouring charged groups and allowing for easy fabrication of O₂ sensors by nanoprecipitation techniques [52]. Although better cell penetration is usually achieved with cationic nanoparticle structures [4, 53], we found that anionic poly(methyl methacrylate-co-methacrylic acid) (PMMA-MA) had moderate oxygen permeability [54] and formed stable NPs in aqueous solution, leading to very promising results. Two NP structures (PA probes) were prepared from this polymer: PA1 containing a single phosphorescent dye PtPFPP, and PA2 with PtPFPP and a fluorescent antenna and reference conjugated polymer dye PFO (Fig. 1a). PA1 displayed a smaller particle size and narrower distribution than PA2, with Z _AV = 72 nm, PDI = 0.041 and Z _AV = 95 nm, PDI = 0.096 (Fig. 1b). For all probes and pure polymer the Z-potential was in the range of −18 to −35 mV (depending on the particle size, see “Methods” section). Spectral characteristics of the PA probes were similar to their cationic analogues NanO₂ and MM2 [42, 55]. At O₂ levels of 21 and 0 %, the emission lifetimes were 39 and 68 µs for PA1 and 31 and 70 µs for PA2, respectively (PBS, 37 °C); these levels are well suited for O₂ sensing in the physiological range (0–200 µM). The PA2 probe showed efficient FRET and more intense emission of PtPFPP than PFO (Fig. 1c). It produced a ratiometric response to O₂ comparable to MM2 (~2.5 times for I _0 %/I _21 %, Fig. 1d), and displayed similar imaging properties under two-photon excitation (see below).

Fig. 1.

Initial evaluation of PMMA-MA-based nanoprobes with neural cells. a Chemical structures of nanoparticle-forming components; b size distribution of the two probe samples obtained on a Zetasizer analyzer; c luminescence excitation and emission spectra of PA2 compared with MM2, measured at 0 % O₂ at 37 °C. d Corrected emission spectra of PA2 at air saturation and anoxic conditions (λ _exc 390 nm, 25 °C). e Confocal images of neural cells (E18, 6 DIV) stained with MM2 and PA2 (1 µg ml⁻¹, 16 h, red) and co-localised with cell bodies by Calcein Green staining (1 µM, 0.5 h, green). Right: localisation of PA2 in mouse primary neurons (E16, 8 DIV) at higher magnification. Optical sections are 0.4 µm thick. f Wide-field images of neural cells stained with PA2 (1 µg ml⁻¹, 16 h): PtPFPP (red), PFO and DAPI (blue) signals and counter-staining (green) with antibodies against β-III-tubulin (mature neurons) and GFAP (astrocytes). Scale bar unit is µm

Cell staining with PA probes

PA1 and PA2 were first evaluated with cultures of primary neural cells from rodent brain (rat and mouse cortical neurons of E16–18, 3–9 DIV), which represent mixed populations of developing neurons and astrocytes (Fig. 1e). When used at concentrations of 1–10 µg ml⁻¹ in the culture medium, the PA probes stained cell bodies and neurites, co-localising with the cytosolic tracer Calcein Green and producing punctuated patterns with a small number of extracellular aggregates. The PA2 probe co-localised with immunohistochemically stained mature neurons (β-III-tubulin) and astrocytes (GFAP) (Fig. 1f), thus reflecting efficient staining efficiency of two dominant but diverse neural cell types. In contrast, the previously described MM2 and NanO₂ showed negligible internalisation by neural cells and prominent formation of extracellular aggregates (Fig. 1e, S1).

We also tested fibroblast (MEF, Fig. S2), neuron-like PC12 cells and cortical neurons from mouse brain (E18) (not shown) and observed efficient intracellular accumulation of PA1 and PA2, with endosomal and lysosomal patterns of localisation, different from staining of mitochondria (TMRM). PC12 cells showed linear concentration dependence of staining up to 25 µg ml⁻¹ (3, 20 h), while in MEF cells saturation was observed at 2.5 and 10 µg ml⁻¹ (Fig. S2b–c). This suggests a slightly lower staining degree of fibroblast cells compared with NanO₂ [55] and a different internalisation mechanisms for PA probes in different cell types. Nonetheless, all tested cells produced sufficient phosphorescent signals for O₂ monitoring experiments (exceeded 30,000 cps threshold on the TR-F reader [55]). Total ATP assay revealed no significant toxicity of the PA2 probe in PC12, MEF cells, primary cortical or cerebellar granule neurons at concentrations 0–25 µg ml⁻¹ and staining times ≤20 h (Fig. S3a–c). This is in agreement with reported low or no toxicity of PMMA-MA nanoparticles [56, 57].

By modifying the synthesis procedure (see “Methods” section), we also produced 135 nm NPs and tested the size dependence of NP internalisation on primary rat neural cells (Fig. 2a). The larger NPs showed less internalisation and a high number of aggregates outside the cells (indicated by blue arrows in Fig. 2a), while NPs <100 nm were localised largely intracellularly (as indicated by Calcein Green counter-staining). Staining patterns for PC12, MEF and mixed population of primary neurons and astrocytes suggested that internalisation occurred via complex mechanisms involving endocytosis. Indeed, under ATP depletion conditions, we observed a significant decrease of PA2 uptake by neural cells (Fig. 2b). Inhibitors of endocytic pathways (CPZ for clathrin-mediated endocytosis, EIPA for macropinocytosis and MβCD for lipid-raft/caveolae-dependent endocytosis) [58] also decreased probe uptake markedly, but not completely (Fig. 2c). We conclude that PA NPs utilise mixed endocytic, energy-dependent mechanisms of cell entry, similar to the cationic RL100-based probes [55].

Fig. 2.

Evaluation of mechanisms of cell entry for PA2 with cultured rat primary neural cells (E18, 6 DIV). a Effect of NP size on cellular uptake. Confocal images of PA NPs (2 µg ml⁻¹, 16 h) and Calcein Green (1 µM, 0.5 h) probes, yellow arrows indicate intracellular localisation, blue-aggregates. b Effect of ATP depletion on cellular uptake of PA2. Cells were incubated for 1.5 h in glucose-free medium, then treated with oligomycin (10 µM, 0.5 h) and PA2 (10 µg ml⁻¹, 3 h) and analysed. c Effect of inhibitors of endocytosis on cellular uptake of PA2. Cells were treated for 2 h with CPZ (50 µM), EIPA (50 µM) or MβCD (5 mM), then stained with PA2 (10 µg ml⁻¹, 3 h) and analysed. Average intensity signals for multiple ROIs are shown. Scale bar in µm

Lifetime-based sensing of cellular O₂

We found that the PA NPs demonstrated efficient cell penetration and photostability when compared with previously described cationic NP O₂ probes (Fig. 3a); therefore, these PA probes are suitable for measurements of intracellular O₂ in neuron-like cells on a TR-F plate reader (high throughput analysis of cell populations) and by phosphorescence lifetime imaging microscopy (PLIM). Using the TR-F plate reader, we calibrated the PA2 probe in solution and in stained mouse cortical neurons in which cellular respiration was blocked by treatment with antimycin A (AA), and exposed the cells to different levels of atmospheric O₂ (Fig. 3b–c). The PA probes displayed non-linear Stern–Volmer relationships, similar to other reported O₂ probes [10], which reflects the heterogeneous environment within the nanoparticle space [59, 60] (Fig. 3c). However, the calibrations were not affected by the local environment, and no drift of the lifetime signal was observed in stained cells cultured over periods of up to 36 h (not shown). This is important since such drifts would indicate leakage of the dye(s) or degradation of the NPs due to normal cellular processes (e.g. in lysosomes). Notably, both aqueous solution and “in cell” calibrations were identical, indicating that there was a minimal effect of the local environment on the O₂ sensing (PA2 inside or outside the cells). Additionally, the calibration data fitted well with the “two-site” modified Stern–Volmer model [11, 59].

Fig. 3.

O₂ monitoring experiments with neural cells and PA2. a Photobleaching under continuous illumination on a microscope in comparison with other NP probes (650 nm emission). b Example of phosphorescence intensity decay shown with single-exponential decay fitting curve (red). c O₂ calibration and Stern–Volmer (SV) relationship produced with mouse cortical neurons (7 DIV, “cells”) and in solution (“no cells”) on a TR-F reader. Fitting of Stern–Volmer relationship accordingly to two-site model is shown in red (SV fit). d Oxygenation profiles of a confluent monolayer of mouse cortical neurons (7 DIV) stained with PA2 (2.5 µg ml⁻¹; 16 h) and measured on a TR-F reader. After baseline stabilisation, cells were treated with FCCP (500 nM) or DMSO (mock), followed up KCl (50 mM) or AntimycinA (AA, 10 µM), as indicated by red lines (top of the graph). N = 3. e Single optical section of a PC12 cell aggregates stained with PA2 (green) and Calcein Green (red), and merged image. Scale bar in µm. f PLIM images of PC12 aggregates (shown above) at rest and after sequential treatment with FCCP and AA. g Corresponding O₂ levels calculated for selected ROIs within the aggregates, N = 5. Dashed line shows air-saturated O₂ level

We also monitored local oxygenation of a confluent monolayer of mouse cortical neurons stained with PA2 (Figs. 3c, S3d) and observed a basal O₂ level of approximately 175 µM at rest. This level was markedly decreased upon mitochondrial uncoupling and subsequent depolarisation of the plasma membrane after application of FCCP and KCl, respectively (Fig. 3d), which are used in a two drug model with known mechanisms of action in this regard [24, 61]. In contrast, the mitochondrial inhibitor Antimycin A (AA) reverted cellular O₂ to air-saturated levels (Fig. 3d). The magnitude and timescale of the responses demonstrate a moderate respiration rate for this cell type.

For demonstration of the PA2 probe performance in PLIM we used cultured PC12 cell aggregates, which possess active respiration and display local O₂ gradients [19, 24]. Staining with PA2 overlapped with Calcein Green, thus confirming its intracellular accumulation (Fig. 3e). The aggregates located within cells showed rapid responses (2–4 min) to stimulation with FCCP and AA (Fig. 3f), similar to neurons cultured on microplates (Fig. 3d). Randomly selected ROIs (1–4) within the same aggregate showed different responses (Fig. 3f), with a stronger magnitude response corresponding to higher respiratory activity. Subsequently, AA brought back local deoxygenation to initial (resting) levels (Fig. 3g). These data demonstrate that O₂ levels in neuron-like cells can be reliably measured with the PA2 probe in plate reader [24] and PLIM [19, 55] formats.

Analysis of cultured brain slices

Two-dimensional neuronal cultures often lack the cyto-architectural context of 3D tissues and organs, and therefore have limited translational potential. We tested a common 3D tissue model of live organotypic rat brain slices (400 µm thick explants) maintained under ambient O₂ [19, 62] with the PA2 probe. The presence of the PFO antenna in the PA2 probe allowed us to perform two-photon imaging at depths >100 µm. Both control and PA2-stained (25 µg ml⁻¹ for 3–24 h) samples showed a comparable (or lower) degree of staining with the membrane integrity probe CellTox Green (Fig. 4a), indicating that the PA2 probe had no effect on cell viability. Slices cultured for 1–3 days also demonstrated positive staining with TMRM (functional mitochondria) and cholera toxin (neuronal cells), confirming the presence of a high number of viable cells in cultured slices (Fig. 4b). This was consistent with 2D culture data (Fig. S3a–c).

Fig. 4.

Staining of live rat cortical slices with PA2. a E16 slices cultured 3 DIV and stained with PA2 (25 µg ml⁻¹, 3 h) and CellTox Green dye (0.05 %, 3 h). b Analysis of 3 DIV slices with live cell markers Hoechst 33342 (HXT, 2 µM), TMRM (20 nM) and CTX-Alexa 488 (1 ng ml⁻¹). c Two-photon images of brain slices (P7) stained with MM2 and PA2 (100 µg ml⁻¹, 24 h): 4 µm thick optical sections for PFO (420–460 nm) and PtPFPP (605–680 nm) dyes (excited at 780 nm). d One-photon confocal single optical sections showing staining of brain slices of different age with CTX and PA2 (25 µg ml⁻¹, 3 h). e Average staining intensities calculated from (d). f, g One-photon O₂ PLIM images of P21 slices stained with PA2, and imaged at different depths (0–50 µm). f Average phosphorescence intensities of PA2 deep in the tissue, depending on scanning depth. B/G: background counts. g Combined PLIM images and calculated profiles of lifetime and O₂ concentration for selected ROI. Scale bar in µm

Next, we compared PA2 with the MM2 probe in imaging experiments in brain slices (P7 slices). Both probes were clearly visible but demonstrated different patterns of staining (Fig. 4c): MM2 probe showed marked aggregation and poor penetration into tissue, and the PA2 probe showed diffuse uniform staining across the tissue volume, which penetrated at least 50–100 µm into tissue (not shown). Staining for 3–4 h was sufficient for positive labelling with the PA2 probe and neuronal tracer Cholera toxin (CTX), while longer incubation gave no improvement (Fig. S4a). This short sample preparation time facilitates the analysis of ex vivo brain slices, particularly in conjunction with sensitive confocal PLIM systems [14, 17].

The developing brain is a highly heterogeneous structure with many different cell types and pronounced structural changes upon maturation. Therefore, to test the performance of the PA probe with diverse tissue cyto-architectures we examined rat brain cortical slices ranging in age from embryonic stages (E12) to adult (P21). We observed efficient staining with PA2 (3–4 h incubation) for all samples, however, postnatal tissue (especially P21) showed higher probe aggregation and reduced probe penetration (Fig. 4d). Quantitative analysis of ROIs free of aggregates confirmed improved staining of embryonic samples E12–E16 (Fig. 4e). In contrast, the MM2 probe produced mostly ‘patchy’ patterns in tissue of all ages (Fig. S4b). In-depth deoxygenation of stained tissue samples was analysed by confocal PLIM, by exciting the PA2 probe at 405 nm and collecting emission at 650 nm (Fig. 4f, g). We found that the PA2 staining was bright and photostable with a uniform tissue distribution, and produced high phosphorescence signals and accurate lifetime measurements from under 405 nm laser excitation (Fig. 4f). Postnatal (P21) samples showed more than twofold reduction in O₂ across a depth of 50 µm, which reflects the diffusion limitations of O₂ transport to actively respiring cells. Spatial heterogeneity across the 50 × 50 µm² region was also high as evidenced from the calculated profiles: ±25 µΜ O₂ (Fig. 4f, right).

Multiplexed analysis of development of neural progenitor cells

To further demonstrate potential applications of PA2, we carried out multiplexed analysis of cultured brain tissue using live cell imaging techniques combined with immunofluorescence. We transfected a specific population of progenitor cells in cultured embryonic brain slices (E12) using plasmid DNA where GFP expression is driven by a brain lipid-binding protein (BLBP) promoter (BLBP-GFP); this plasmid is used as a specific marker of developing radial glial cells [62, 63]. Slices were then cultured for 1–3 days in vitro (DIV). Subsequently, confocal microscopy confirmed that the distribution of GFP-positive cells co-localised with markers of neural progenitor cells (SOX2, nestin) and actively proliferating cells (Ki67) (Fig. 5a). This co-localisation gradually decreased with the age of slices in culture (Fig. 5b), indicating that active differentiation and developmental processes were ongoing in the tissue slices. Therefore, BLBP-GFP transfection provided visualisation of a population of developing neural progenitor cells in this 3D tissue model.

Fig. 5.

Analysis of O₂ influence on development of BLBP-GFP-positive cells. a Visualisation of BLBP-GFP-positive, progenitor cells (SOX2) and proliferating cells (Ki67) in an organotypic brain slice (E12, 3 DIV) by immunofluorescence (IF). b Co-localisation of BLBP-GFP, SOX2 and Ki67 at 1 and 3 DIV showing that a large portion of the GFP expressing cells are actively dividing (Ki67+) and remain in an undifferentiated state (SOX2+). c Examples of images of regions with overlapping BLBP-GFP and PA2 distributions (intensity, lifetime and overlay) showing ROIs used for quantification. d Profiles of BLBP-GFP expression and corresponding O₂ levels on 1 and 3 DIV, determined for such overlapping regions. Scale bar in µm

We then analysed the distribution of live neural progenitor cells in relation to local O₂ levels. BLBP-GFP-transfected slices (1 or 3 DIV) were stained with the PA2 probe (3 h) and analysed by multi-parametric imaging and PLIM (Fig. 5c). We selected regions of tissue (20–50 µm²) with overlapping PA2 and BLBP-GFP staining (Fig. 5c, left panel) and compared 10 ROIs after 1 DIV (at this stage 25 % of GFP labelled cells were actively proliferating cells and 40 % of GFP labelled cells remained a progenitor population, i.e. SOX2-positive) and at 3 DIV (at this stage GFP labelled cells exhibit decreased proliferation and expression of progenitor cell markers). We found a large variation in O₂ levels within slices, ranging 50–200 µM (Fig. 5b–d), however this range was similar at 1 DIV and 3 DIV. This suggests that the O₂ requirements for development and differentiation of organotypic slices is heterogeneous, but is not necessarily dependent on an absolute change in O₂ levels over this developmental time period; although the actively proliferating population of BLBP-GFP labelled cells was larger at 1 DIV, this did not correspond to an increased/decreased level of overall tissue oxygenation. Importantly, the range and average value of tissue oxygenation in the slices were unchanged after 3 DIV, indicating that the culture environment did not result in significant deterioration of tissue slices due to deoxygenation over this time period, further validating this 3D model system.

Although the range and average of O₂ levels in slices did not differ over this time period, the high spatial heterogeneity of O₂ within live tissue slices (Figs. 4f, 5d) could be due to the presence of micro-regions or cell types with different O₂ demands. This is suggested by the correlation between O₂ levels and the distribution of specific cell populations. For example, this can be seen in the two highlighted ROIs in Figure S5: ROI1 contains a region with GFP expressing cells that are Ki67-negative, indicating that they are of radial glial cell lineage but are no longer actively proliferating, and this specific region displayed high O₂ levels (150–200 µM); in contrast, ROI2 contains no GFP labelled cells but does contain nestin-positive fibres and faintly positive Ki67 staining, indicating the presence of recently proliferating neural stem cells, and this region had twofold lower O₂ values (Fig. S5). These findings suggest that meaningful relationships between specific cell populations and O₂ levels can be determined using these PA probes, however further studies are needed to determine the full implications of changes in oxygenation in these, and other, cell types.

As with other fluorescence microscopy imaging methods, O₂ imaging results can potentially be affected by measurement artefacts, such as: (1) non-ideal optical sectioning on one-photon confocal PLIM microscope due to limited tissue penetration of excitation light, scattering effects, limited photon “budget” available for precise calculation of phosphorescence lifetimes, and unwanted recording of out-of-focus phosphorescence [64–66]; (2) uneven distribution of the NP probes across the tissue volume can potentially affect the imaging precision [49] (3) artifacts related to photoinduced O₂ consumption [67] and others [15]. However, PLIM method will be among the methods that are most minimally affected by these factors and its reliability in O₂ measurements has been demonstrated in many physiological studies [26, 40, 68, 69].

The functional imaging of O₂ levels in relation to specific cell populations in living brain tissue in this study represent intriguing “proof of concept” results and demonstrates the potential to investigate dynamic changes in O₂ levels in relation to ongoing developmental processes occurring in on a cellular as well as a whole tissue level. Future studies optimizing both the distribution of O₂ probes and live detection of specific cell types in organotypic tissue slices are required to allow for the specific correlation of O₂ levels with neuronal and neuroglial activity in living and developing neural tissues.

Conclusions

The PA1 and PA2 nanoparticle probes (<100 nm) based on anionic PMMA-MA polymer demonstrate good staining efficiency and intracellular penetration across a range of cell types and neural tissues. We attribute this to their lower aggregation and non-specific adsorption [70] and more specific interaction with membranes of neural cells. Similar to other nanoparticles [4, 53], their internalisation was strongly inhibited by endocytosis inhibitors and ATP depletion, suggesting their utilization of energy-dependent cell entry (Figs. 1, 2, S2). These probes showed lysosomal patterns of localisation and photo-physical properties similar to cationic NP probes [42, 55] including high brightness and photostability provided by PtPFPP, two-photon antenna and reference PFO dye, and stable and reproducible O₂ calibration (Fig. 3). Therefore, these PA probes offer significant advantages as they allow for similar applications but on a broader/different range of cell and tissue models which include primary neural cell culture, as well as 3D models such as cell aggregates and brain tissue slices. In contrast with recently developed red and infra-red molecular probes [19, 71] PA2 and related nanoparticles provide multi-modal imaging capability with efficient two-photon excitation, this can be very beneficial for ex vivo and in vivo models.

Monitoring of O₂ in 2D cultures produced metabolic responses similar to those observed with existing probes; however, with 3D models (aggregates and tissue explants) the PA2 probe resulted in improved distribution across tissue volumes and ultimately better imaging performance. Probe penetration into brain tissue (Fig. 4) was comparable to small molecule conjugates (e.g. Pt-Glc) [19], with the added advantage of increased brightness and the option of two-photon and ratiometric detection. Using the PA2 probe, BLBP-GFP plasmid DNA transfection to identify specific cell populations, live-cell imaging, and immunofluorescence techniques (Fig. 5) we performed multi-parametric functional analysis, relating GFP labelled developing progenitor cell populations with local O₂ levels in cultured embryonic brain tissue over a 3-day developmental period. Heterogeneity of tissue oxygenation was shown and its possible effects on various cell types discussed-highlighting the potential applications for future studies with these PA probes. The results of this study and subsequent advancement in the knowledge of cell-penetrating behaviour of anionic PMMA-MA nanoparticles can be utilised for designing future sensor probes and drug delivery systems for a variety of tissues, importantly including neural tissue [57].

Methods

Materials

Platinum(II) meso-tetra(pentafluorophenyl)porphine (PtPFPP) was from Frontier Scientific (Inochem Ltd, Lancashire UK), poly(methyl methacrylate-co-methacrylic acid), 9:1 (PMMA-MA) was from Polysciences (Warrington, PA, USA), poly(9,9-diheptylfluorene-alt-9,9-di-p-tolyl-9H-fluorene) was prepared as described previously [42]. MitoImage-NanO₂ and MM2 probes were from Luxcel Biosciences (Cork, Ireland). Secondary antibodies labelled with Alexa Fluor 488, 555, 594, 633, recombinant cholera toxin, subunit B-Alexa Fluor 488 conjugate (CTX), B27 serum-free supplement, Calcein Green AM, tetramethylrhodamine methyl ester (TMRM) and Pro-long Gold anti-fade mounting medium were from Life Technologies (BioSciences, Dublin, Ireland). Goat anti-SOX2 antibody was from Abcam (Cambridge, UK). Rat anti-Ki67 antibody and Fluoromount-G mounting medium were from eBiosciences (San Diego, CA). Viability Assay kits CellTiter-Glo™ and CellTox™ Green were from Promega (MyBio, Ireland). Mouse anti-nestin antibody was from Millipore (Cork, Ireland). Anti-GFAP (G-A-5) antibody, bis-Benzimide Hoechst 33342 and all the other reagents were from Sigma-Aldrich (Dublin, Ireland).

Standard cell culture grade plasticware was from Sarstedt (Wexford, Ireland) and Corning (VWR, Ireland), glass bottom mini-dishes were from MatTek (Ashland, USA), glass bottom multiwell slides were from Ibidi (Martinsried, Germany). Polystyrene scaffold membranes Alvetex™ were from Reinnervate (Amsbio, UK). Plasmid DNA BLBP-GFP [63] was kindly provided by Dr. E.S. Anton (University of North Carolina).

Preparation of PMMA-MA nanoparticles

PA1: 200 mg of PMMA-MA and 3 mg of PtPFPP were dissolved in 45 mL of tetrahydrofuran and the solution was diluted to 100 mL with acetone. The solution was transferred into a 1 L glass beaker and 600 mL of deionised water was added within 3–5 s under vigorous stirring. Using a rotary evaporator, organic phase was removed and the nanoparticles were concentrated to 10 mg ml⁻¹. The solution was cleared from aggregates by centrifugation and then stored at 4 °C.

PA2: nanoparticles were prepared similarly, by using 2.4 mg of PtPFPP and 20 mg of poly(9,9-diheptylfluorene-alt-9,9-di-p-tolyl-9H-fluorene). Particle size and charge were determined on a particle size analyzer Zetasizer Nano ZS (Malvern Instruments, Germany), UV–Vis and luminescence spectra as described before in [42, 52]. Phosphorescence lifetimes were measured in air-saturated and deoxygenated (5 mg ml⁻¹ KH₂PO₄, 5 mg ml⁻¹ Na₂SO₃) conditions. Large size NPs bearing PtPFPP and PFO dyes (135 nm, −24.5 mV) were produced when water was added with slower speed.

Cell and primary neural cultures

All the procedures with animals were performed under a licence issued by the Irish Government Department of Health and in accordance with the EU Directive 2010/63/EU. Embryonic tissue from time mated Sprague–Dawley rats (Biological Services Unit, University College Cork, Ireland) at embryonic day (E)12 to postnatal day (P)21 were used.

Mouse embryonic fibroblast (MEF) and rat pheochromocytoma (PC12) cells from ATCC (Manassas, VA, USA) were handled as described before [51, 72]. Mouse primary cortical (E16) and cerebellar granule neurons (P7) were obtained as described before [73]. The culture of rat primary cortical cells (E18) was prepared and handled under differentiation conditions (DMEM F12 Ham medium, supplemented with 2 % B27, 1 % FBS, 1 % penicillin–streptomycin solution) for 5–7 DIV, adhered onto poly-d-lysine-coated glass or plastic dishes as described before at 75–100 % confluence [17]. The staining with probes was achieved by incubation at indicated concentration (2.5–25 µg ml⁻¹) in regular culture medium for 3–24 h, followed by 2–3 washes.

PC12 cell aggregates were grown in RPMI-1640 medium supplemented with 10 % horse serum (HS), 5 % FBS, 10 mM HEPES-Na, pH 7.2 and penicillin–streptomycin. For cell staining kinetics, leakage and ATP assays, MEF and PC12 cells were seeded on collagen IV-coated 96-well plates and cultured for 48 h to reach 75–100 % confluence. Metabolic stimulation was performed in 35 mm Cell+ mini-dishes (Sarstedt) in 2 mL of Phenol Red-free DMEM supplemented with NGF (100 ng ml⁻¹) at 37 °C and 20 % O₂, adding 10× stock solutions of drugs (1/10 of the volume) to the samples.

In the studies of cell entry mechanism, cells grown on multi-well coverslips (Ibidi) were pre-treated with inhibitors of endocytosis (as indicated), incubated with PA2 (10 µg ml⁻¹, 3 h) and analysed on a wide-field microscope [17]. ATP depletion was performed as described before [72]. Intensity signals (collected under 390 nm LED excitation and 635–675 nm emission filter, background corrected) from 6 to 7 randomly selected ROI were used for quantification of drug effects.

Brain slice culture

Tissue slices were obtained and handled as described previously [19]. Pregnant rats (E12–E18) were anesthetized with isoflurane followed by decapitation, embryos were collected by laparotomy and placed into ice cold sterile HBSS. For postnatal ages (P1–P21), rat pups were anesthetized with isoflurane followed by decapitation and the brain was dissected out. Tissues were embedded in 4 % low melting agarose (prepared in PBS, supplemented with 27 mM glucose). Transverse sections (400 µm) were sliced in ice cold L15 dissecting media using a vibratome (VT1200, Leica Microsystems). Using the interface method [74] slices were positioned on Millicell inserts (0.4 µm, 30 mm diameter, Millipore) or pre-wet polystyrene scaffold membranes Alvetex™ (Reinnervate, Amsbio, UK) in phenol red-free DMEM supplemented with 25 % HBSS, 10 % FBS, 1 % penicillin–streptomycin, 10 mM Glc, 20 mM HEPES-Na, pH 7.2 (“Slice medium”) at 20 % O₂, 5 % CO₂, 37 °C in a humidified incubator. Electroporation with BLBP-GFP plasmid [62, 63] was carried out directly on culture inserts using a pair of gold-plated “genepaddle” electrodes on BTX ECM 830 Square Wave Electroporator (Harvard Apparatus, UK), using 2 µl of 1.0 µg µl⁻¹ DNA under previously optimised settings (60 V, 5 pulses at 50 ms pulse duration, 950 ms interval) [62, 75]. 3–5 h after dissection slices were analysed or cultured for 1–3 DIV. Tissue staining was achieved by incubation with the O₂ probe for 3–24 h at 25–100 µg ml⁻¹, or with CTX (1 ng ml⁻¹) in Slice medium, followed by one wash.

Immunofluorescence

Brain slices were fixed in 4 % paraformaldehyde solution (4 h, 4 °C), washed with PBS (0.5 h, 5 times), blocked in 10 % HS supplemented with 0.4 % Triton X100 in PBS (1 h, room temp.), incubated with primary antibodies diluted (1:500) in 2.5 % HS, 0.4 % Triton X100 in PBS (48 h, 4 °C), washed with PBS, incubated with appropriate Alexa Fluor-conjugated secondary antibodies (18 h, 4 °C). They were then washed in PBS, placed between two coverslips with a 400 µm thick slice of agarose to act as a spacer, mounted in Fluoromount-G medium (eBiosciences) and imaged on a confocal laser-scanning microscope Olympus FV1000 [42]. Maximum intensity projections of an imaged z-stack were processed in Olympus Fluoview (FV10-ASW 4.0 Viewer) and brightness and contrast were adjusted accordingly in Adobe Photoshop and Illustrator software.

Two-photon microscopy

This was performed on an upright microscope Olympus FV1000 MPE equipped with Mai Tai DeepSee Ti:Sapphire laser (Newport Corporation, CA), XPLN 25x NA:1.05 water immersion objective and temperature (temperature set at 37 °C)/CO₂ (5 %) control, using excitation at 770 nm (2 % laser power) for PA2 and MM2 probes [17].

One-photon confocal PLIM

This was performed on a system based on an upright fluorescent microscope Axio Examiner Z1 (Zeiss UK) equipped with 20× NA:1.0 W-Plan-Apochromat and 40× NA:1.1 LD-C Apochromat W Korr. objectives, heated stage (temperature set at 37 °C) with motorized Z-axis control, DCS-120 confocal scanner (Becker and Hickl), R10467U-40 photon counting detector (Hamamatsu Photonics K·K.) and TCSPC hardware (Becker and Hickl) [17]. Phosphorescent probe-stained cells were excited with a 405 nm picosecond diode laser BDL-SMC (Becker and Hickl) and emission was collected at 635–675 nm. Fluorescent dyes Calcein Green, Alexa Fluor 488, BLBP-GFP, CellTox Green and CTX were excited with a picosecond supercontinuum laser SC400-4 (Fianium, UK) at 488 nm with emission collected at 512–536 nm; TMRM was excited at 540 nm and emission collected at 565–605 nm; LysoTracker Red and MitoTracker Red were excited at 590 nm and emission collected at 635–675 nm. DIC images were recorded with D5100 digital camera (Nikon) connected to the microscope. Phosphorescence lifetimes were calculated in SPCImage software (Becker and Hickl) using mono-exponential decay fits. 2D matrices of lifetime data for each measurement were then correlated with intensity and DIC images, and selected ROIs were converted in O₂ concentration using the following calibration function produced on TR-F microplate reader Victor2: [O₂] = 87,270.1 × exp(−τ/5.07006) (mono-exponential fitting) or [O₂] = (0.57575/(τ/69.86 − 1 + 0.57575) − 1)/0.0861 (two-site model, r ² = 0.997, Ksv = 0.0861 µM⁻¹, τ ₀ = 69.86 µs).

Plate reader measurements

These were performed on Victor2 (Perkin Elmer) TR-F plate reader placed into hypoxia Glove Box (Coy lab products, Michigan, USA) as described before [42]. Briefly, cells stained with PA2 probe treated with antimycin A (10 µM) were exposed to different levels of atmospheric O₂ at 37 °C, and measured lifetime values were used to produce the calibration function by fitting. Calibration without cells was performed in PBS.

Cell viability and toxicity assessment

This was performed by measuring changes in total cellular ATP in 2D cell culture, or by staining brain slices with live cell markers (TMRM, CTX, HXT). In the ATP assay, MEF, PC12 or primary neural cells grown on 96-well plates were stained with 0–25 µg ml⁻¹ of PA2 for 3–24 h, washed with fresh medium, lysed with CellTiter-Glo reagent (Promega) and analysed for total luminescence. Membrane integrity was assessed by staining with CellTox Green (Promega) reagent [17]. Briefly, slices were incubated with reagent (0.05 %), washed with medium, imaged on the confocal microscope. Stained nuclei were counted and compared with control.

Data assessment

Plate reader data represent averaged values with error bars (standard deviation) for >6 replicates. Cell loading studies were performed on wide-field microscope in triplicate, averaging intensity values from 6 to 10 randomly selected ROI. Experiments with brain slices were performed in duplicate, with analysis of 3–5 independent ROI. Phosphorescence lifetime images (2D matrices) were generated in SPCImage software (Becker and Hickl) by single-exponential fitting. They were converted into ASCII format and processed in Microsoft Excel to produce O₂ concentration values using calibration function. For BLBP-GFP and O₂ correlation, photon counts per pixel for different spectral channels were compared in MS Excel software, then co-localising regions were selected and analysed. To ensure consistency, all the experiments were repeated at least for 2–3 times, as indicated in figure descriptions.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

3D

Three-dimensional

BLBP

Brain lipid-binding protein

CPZ

Chlorpromazine

CTX

Cholera toxin, subunit B

DIV

Days in vitro

DMF

N,N-dimethylformamide

DMSO

Dimethyl sulfoxide

EIPA

5-(N-ethyl-N-isopropyl)amiloride

GFP

Green fluorescent protein

HBSS

Hanks balanced salt solution

HXT

Hoechst 33342

iO₂

Intracellular O₂

MβCD

Methyl-β-cyclodextrin

NP

Nanoparticles

PA

Polyacrylate NP

PBS

Phosphate buffered saline

PDL

Poly-d-lysine

PDT

Photodynamic therapy

PLIM

Phosphorescence lifetime imaging microscopy

PMMA-MA

Poly(methyl methacrylate-co-methacrylic acid

PFO

Poly(9,9-diheptylfluorene-alt-9,9-di-p-tolyl-9H-fluorene)

PtPFPP

Pt(II)-tetrakis(pentafluorophenyl)porphine

ROI

Region of interest

RT

Room temperature

TBST

Tris-buffered saline, tween 20

TCSPC

Time-correlated single photon counting

TR-F

Time-resolved fluorescence