Portable live-cell imaging system for low-cost fluorometry. Our live-cell fluorescent imaging system consists of four parts: a portable, compact bench-top incubator, a digital microscope with connection to a portable computational station, a white-light LED ring, and an adjustable petri dish holder (Fig. 1A). All four parts are off-the-shelf components and were chosen for a fast and easy assembly that required only a few modifications to the incubator system. Hence the resultant, low-cost imaging system allows for high reproducibility in a training/classroom setting or a low-resource environment. The bench-top assembly of the incubator is shown in Fig. 1B1 andB2. The incubator has integrated temperature control and the possibility to be upgraded to regulate carbon dioxide (CO 2 ) levels. The digital microscope provides software-controlled switchable blue and yellow light-emitting diodes (LEDs) for 480 nm and 575 nm excitation with an integrated emission filter between 510 and 610 nm. A white-light LED ring was installed at the top of the incubator to add bright field imaging. A representative white-light cell culture image taken with primary cortical neurons is shown in Fig. 1C1, with its inverted version shown in Fig. 1C2. Figure 1C3 shows the corresponding green-fluorescent signal with 480 nm excitation of the Fluo-4 AM loaded neurons. Image contrast of these images can be assessed through histogram plots, which are shown in supplementary data (Fig. S6, Fig. S7, see supplementary files).
The complete live-cell imaging system weighs 8.1 kg, has dimensions of 35 cm × 40 cm × 50 cm, and costs below US$ 2000. Our imaging solution is an improvement of existing long-term imaging systems 5,[8][9][10] , as it combines portability with temperature control and digital calcium fluorometry for low-cost live-cell studies without the need for 3D printing, expertise in computer-automated-design, and skills in assembling. The small footprint and lightweight characteristics of our imaging system make live-cell fluorometry and portability between research and classroom space possible and accessible. Digital fluorescent imaging characteristics. To assess the optical characteristics and limitations of our digital fluorescent imaging system, we used Dragon green fluorescent beads with a diameter of 15.65 µm and compared their optical appearance between the digital fluorescent microscope against a commercially available high-end, fluorescent optical microscope. Figure 2A1 shows the green fluorescent beads in a 24-bit, RGB image taken with a green-emission filter (mercury lamp excitation) under the high-end optical microscope (40×, 2.0 megapixel). In contrast, Fig. 2A2 shows the same beads imaged with the digital microscope in a 24-bit RGB image (220×, 1.3 megapixels). The fluorescent digital microscope provides two-dimensional images from a CMOS camera in a field-of-view with 1280 × 1024 pixel array (1574 µm × 1180 µm), resulting in an imaging resolution of 1.352 pixels per micrometre at highest magnification (220×). The high-end optical microscope provides 1.25 pixels per micrometre (with 10× objective) and 3 pixels per micrometre (with 20× objective). The higher optical zoom, however, reduces the field-of-view for the optical system down to 668 µm × 668 µm. Hence, the imaging resolution of the digital fluorescent microscope is comparable to the high-end optical microscope and, therefore, suitable for fluorometry at the cellular level and similar to image resolutions reported in Yang et al. 11 . The large field-of-view of the digital system furthermore provides an advantage for cell network studies.
Next, we characterized the spatial homogeneity of the optical appearance of the beads within the field of view. For this comparison, 24-bit images were downsampled to 8-bit grey-scale images (Fig. 2B) and single bead images extracted (43 × 43 pixels) for signal-to-noise ratio (SNR), mean-absolute-error (MAE), peak signal-to-noise (PSNR), and root-mean-standard error (RMSE) analysis. The first test compares the spatial bead appearance within the high-end optical microscope image (Fig. 2C1), showing an averaged SNR of 8 dB and an averaged PSNR of 11.5 dB. In contrast, the second test compared images taken with the digital fluorescent microscope. These images have a slightly higher averaged SNR and PSNR of 17 dB and 20 dB (Fig. 2C2), in which a randomized comparison test between the two microscopes also confirms (Fig. 2C3). However, the PSNR and RMSE show lower values for images captured with the digital than with the high-end optical microscope, indicating higher signal uniformity within the digital image acquisition (Fig. 2D1-D3). Furthermore, the high-end microscope showed a more substantial variance in both the SNR and PSNR. This characteristic can be explained through higher sensitivity and imaging artefacts from multiple optical lenses and a longer light path. While high-end optical microscopes have the advantage of picking up signal differences at the subcellular scale, digital fluorometry provides more uniformity and lower noise sensitivity.
Validation of long-term live-cell fluorometry through analyzing temperature-dependent calcium events. Human embryonic kidney cells are known to express endogenous calcium channels 12 . One class of ion channels is called transient receptor potential (TRP) channels and mainly known to regulate the intake of cations under changes in temperature 13 . Within the class of TRP channels, some specific channels are more selective to regulate calcium influx (TRPV5, TRPV6) than others 14 , which is, however, beyond the focus of our study, here. TRP channels sit in the cell membrane pointing towards the extracellular space where they can sense changes in temperature, graphically sketched as a cold blue ball in Fig. 3A. Depending on the type, TRP channels can be either cold-sensitive channel (< 21 °C, TRPM8), or heat-sensitive (> 43 °C, TRPV1) to gate calcium into the cytosol [15][16][17][18][19] . Once activated, TRP channels allow calcium ions to enter the cytosol, which can lead to further downstream processes causing cell apoptosis, changes in metabolic activity, or cell morphology [20][21][22][23][24][25] . Hence, we used HEK cells to demonstrate the ability of our imaging system to monitor temperature-effects in long-term calcium fluorometry. To capture temperature sensation in HEK cells, we performed a 10-h, live-cell fluorescent imaging experiment using Fluo4 AM as a calcium probe in our portable incubator systems either under 37 °C with temperature control, or under room temperature (~ 20 °C, RT). Images were taken every 10 min with 1 s exposure time. From the time-lapse images, we extracted single-cell fluorescent profiles and plotted averaged changes in fluorescent activity from multiple cells over time (Fig. 3B).
HEK cells incubated with 37 °C temperature control maintained regular calcium activity, as represented by an almost constant average intensity profile over the 10 h (Fig. 3C1). Analyzing calcium efflux and influx events, as shown in Fig. 3D1, confirms physiological healthy cell behaviour. We then imaged HEK cells incubated under RT (temperature control was switched off, and the system was cooled to RT) and observed a 67% decay in fluorescence intensity between 5 and 6 h incubation (Fig. 3C2). Between 5 and 6 h, all calcium events can be attributed to calcium efflux, indicating induced cell death to prolonged exposure to temperatures below the physiological temperature level of 37 °C.
Next, we extended our cold-sensation study to primary cortical neurons, which were dissociated from embryonic brain tissues (rat, E18) and grown for 2 weeks. Based on the same calcium fluorometry used in HEK cells, cortical neurons express a higher baseline in normalize calcium signal intensity (ΔF/F max, Fig. 3C1, C2 vs. Fig. 3C3, and 3C4) in their cell bodies with fewer oscillations between calcium influx and efflux events, and lower influx and efflux amplitudes (Fig. 3D1, D2 vs. Fig. 3D3, andD4). Visually comparing the period of oscillatory calcium events between neurons and HEK cells showed lower values for neurons (Fig. 3D1-D4) than for HEK cells. More stunningly, incubating neurons under room temperature for 10 h resulted in a uniform calcium profile. Calcium efflux events were present but not as significant in amplitude as in the HEK-cell experiment (Fig. 3D2,D4). This contrasting behaviour between primary cortical neurons and HEK cells may link to the different members of TRP channels. It may be likely that neurons endogenously express more heat-sensitive TRPV1 than cold-sensitive TRPM8. Variation in calcium activity due to inhomogeneity in cultured neuronal networks or the sample size is shown in Fig. S1 and Fig. S2 and can be excluded as a course of effect (see supplementary file).
Slow, long-term changes in calcium dynamics associated with cell death. One aspect of temperature-mediated processes in neurons is related to calcium signalling and its link to cell death [20][21][22]24,26,27 . Although our experiments suggest that neurons, when imaged at RT, show almost no decrease in average calcium fluores-cent signalling, we carefully analysed our data regarding calcium signals that shown only one spike (a one-time calcium influx followed by calcium efflux) in a cell body over the whole time course. The single spike event may indicate the occurrence of cell death. For calcium spike events that occurred only once, we measured the time delay (ΔT) until the calcium efflux occurred (Fig. 4A). In Fig. 4B1 and C1, we show the distribution of neu- ronal cell bodies where only one spike event was detected, and when influx and efflux events occurred. Under physiological temperature and across 10 h, most of the selected cell bodies seem to die around 5 h (Fig. 4D1). In contrast to RT, 70% of the calcium influx events occurred within 1 h (Fig. 4B2), and 83% of cells also seem to die with ΔT of 1 h (Fig. 4C2,D2). These differences in ΔT between RT and 37 °C point towards a temperaturerelated cell death mechanism.
Validation of fast-scale, low-cost calcium fluorometry to capture spatiotemporal differences in neuronal calcium signalling during nanoparticle interaction. Our fluorescent imaging system has shown the ability to capture long-term changes in calcium signalling based on a 1.7 mHz acquisition rate. The high-speed modalities of the digital microscope, however, can capture images up to 30 Hz, making it an ideal tool to study fast-scale communication patterns in cultured neuronal networks. To validate our system, we chose a commercial suspension of chitosan-coated fluorescent nanoparticles and monitored changes in fluorescent calcium signalling due to nanoparticle endocytosis. The uptake mechanism and behaviour of these nanoparticles with primary cortical neuron cultures have been extensively studied in previous work [28][29][30] , and therefore were chosen as a first validation experiment for our low-cost imaging system. To assess calcium-mediated communication patterns in neuronal networks, we grew dissociated cortical neurons up to 2 weeks and imaged their calcium activity based on Fluo-4 AM. Over an 8h time span, metabolic calcium activity was captured in intervals of 2h for 150 s with a 1 Hz acquisition rate. Representative images of Fluo-4 loaded neurons are shown in Fig. S8 and Fig. S9 (see supplementary file). Using a fully automated cell body segmentation in combination with double-threshold based calcium spike detection gave us spike raster plots for each sample set (Fig. 5A1-A4). From the spike raster plots, the cross-correlation between spike trains was computed using the Sørensen-Dice coefficient and visualized in a neuronal communication graph called connectivity map that shows synchronous calcium firing (Fig. 5A5-5A6) 31 .
Trending in neurobiology is the development of nanotools that can pass the blood-brain barrier, target drugs to specific brain cell types, or act as a local actuator for brain cell stimulation [32][33][34] . Previous studies have specifically explored the effectiveness of magnetic nanoparticle surface coatings on the uptake of the particles into neurons 30,33,[35][36][37] . Understanding the temporal aspect of nanoparticle endocytosis, however, remains elusive as most uptake studies are endpoint measurements. To investigate temporal variations in nanoparticle endocytosis on the synchronicity of calcium firing, we exposed chitosan-coated magnetic nanoparticles to primary cortical neurons grown up to 9 days and imaged changes in calcium activity in our portable imaging system at 1 Hz for 150 s in 2 h time intervals over 8 h at RT and at 37 °C. Representative fluorescent images of Fluo-4 AM loaded neurons are shown for the two temperature conditions in Fig. 5B. Figure 5C1 and C2 show the influence of temperature and nanoparticle endocytosis on the connectivity maps. Without nanoparticle endocytosis, the connectivity maps indicate only minor changes in the location of vertices (active cell bodies), the number of active cell bodies show small cyclic activity remaining overall constant (Fig. 5D1), however, the number of edges reduces about 70% within the first 2 h (lines due to synchronic activity, Fig. 5E1). In contrast, when nanoparticles were administered, synchronic activity increased by about 43%, and a spatial shift of the active region occurred. Between 2 and 6 h, synchronic calcium activity w/NPs reduced by over 64%, before the number of network edges rises again about 120% (Fig. 5E1 andF1).
Comparing different temperatures, the total number of active cell bodies and edges are similar between physiological and room temperature (Fig. 5D1 versus 5E1, w/o NPs) as long as no nanoparticles were administered. The small oscillatory changes in calcium influx and efflux are mirrored in our previous temperature-sensation experiments, where the location of randomly selected cell bodies for a small number of cell groups (< 40 cells) does not impact calcium fluorometry (see supplementary data, Fig. S1 and Fig. S2). Imaging neurons under physiological temperature reveals a trending increase in synchronized calcium signalling with large amplitude oscillation, most-likely correlating with the spatial location of active vertices within the network (Fig. 5C2, D2, E2, F2, w/o NPs). Incubating chitosan-coated nanoparticles with neurons under 37 °C impacted the spatial location of active vertices and the number of edges with an overall decreasing trend between 2 and 8 h. While neurons growing under physiological temperatures showed less synchronous calcium spiking over time (> 2 h) when incubated with chitosan-coated nanoparticles, individual cell spiking activity may still be high or increased as shown in previous studies (see also spike raster plots at 6 h in supplementary data, Fig. S3 and Fig. S4) 29 . Furthermore, we noted that the incubation of chitosan-coated nanoparticles within the first 2 h increased synchronicity independent on the temperature.
The demonstrated capability of our system to assess communication patterns within living neural networks based on fast-scale low-cost calcium fluorometry goes beyond often used cell migration and cell viability assessments 7,8,[38][39][40] . It advances small scale and portable imaging technology 11 and provides an affordable tool to low-resource communities to learn and study more about brain cell communication. Furthermore, it revealed extensive details about how temperature and time interplay with calcium signalling during nanoparticle endocytosis.
With an emerging need to capture time-sensitive aspects of fluorescently labelled proteins and signals in mammalian cells, we have designed a low-cost, portable, live-cell fluorescent imaging system from off-the-shelfcomponents. The described imaging solution is capable of studying calcium dynamics in human embryonic kidney cell lines and primary cultures of embryonic cerebral neurons and can control culture conditions and temperature. Using fluorescent probes and proteins 41,42 in combination with nanoparticle interactions in cells, the low-cost imaging system is suitable to make calcium fluorometry and the study of changes in the cellular macro-(millimetre scale) and microenvironment (sub-micrometre scale) accessible for low-resource environments and provides advanced neuroscience research tools for the classroom 43 . The imaging setup uses digital fluorescent microscopy to capture changes in cytosolic calcium levels across time with up to 30 Hz. The imaging system is a low-tech version of similar live-cell imaging platforms [8][9][10][11]38,[44][45][46][47] . It was purposely kept at low-cost (< 2,000 US$) and required no additional computer drawing skills or access to 3D printing. Although the spatial, optical, and radiometric resolution of our imaging system is lower than that of a high-quality optical microscope, the system can resolve fluorescent signals down to 1.352 pixels per micrometre. The utilized digital imaging modality provided more uniformity and lower noise sensitivity across the imaging sample, which increases the robustness for image signal processing. Furthermore, the imaging system has been demonstrated to provide stable temperature control for cell biology studies. While our imaging system cannot replace super-resolution microscopy to study subcellular transport and interaction of nanoparticles with other organelles, it brings the capability to monitor fast-scale temporal changes within large-scale cell networks. Specifically, we have shown the utility of our system to derive a neuronal graph called a connectivity map that shows synchronous calcium firing. Furthermore, we demonstrated the utility of graph network analysis to derive connectivity maps and applied them to nanoparticle uptake studies, which revealed extensive details about how temperature and time interplay with calcium signalling during the endocytic process.
Overall, we have demonstrated here a portable, low-cost imaging system that allows us to assess communication patterns within neural network going beyond often used cell migration and cell viability assessment and which can be used as an affordable alternative to cost-intense microscopy in low-resource communities to learn and study more about brain cell communication.
Portable, live-cell fluorescent imaging system. An imaging system was designed to maintain constant physiological temperature and humidity control for long-term live-cell monitoring, using off the shelf elements for high reproducibility at a low-cost. The imaging system was assembled based on a small-scale benchtop incubator with digital temperature control (e.g., MyTemp™ mini digital incubator), a digital fluorescent microscope with coloured illumination (LED-based, e.g., Dino-Lite AM4115T-GRFBY), a cell culture sample holder, and a LED-based white light illumination. The digital fluorescent microscope is a glass-lens based mini microscope and can be operated at two different excitation wavelengths of 480 nm and 575 nm to monitor green and red fluorescent probes through a USB-connected desk laptop. The optical sensor in the digital microscope is a CMOS camera with 1280 × 1024 pixels resolution (1.3 megapixels) that can capture up to 30 frames per second (fps). A full comparison of our digital live-cell fluorescent imaging system against other portable, low-cost digital and low-cost traditional optical imaging systems can be found in Table S1 (see supplementary file). Further methods describing the quantification of the digital fluorescent microscope imaging characteristics, and a 48 h long-term live-cell validation experiment based on Normal Rattus norvegicus Kidney (NRK) epithelial cells is also presented in the supplementary data file.
To demonstrate the effect of temperature control of the imaging system, we chose human embryonic kidney (HEK) as they are well known to exhibit endogenous calcium channels 12 and show high sensitivity to non-physiological temperatures 15,18,48 . HEK cells were cultured in mouse embryonic fibroblast (MEF, passage 18) media. When grown to 80% confluency, cells were trypsinated and reseeded into pre-coated 35 mm Petri dishes for the temperature sensation experiment and grown for 2 days. To test the robustness of our imaging system with neurons, we monitored calcium signalling in neural cultures grown from dissociated rat cortical neurons. In neurons, calcium fluorometry is an important imaging methodology to study neuronal cell and network signalling 1,20,21,24,25,[49][50][51][52] . Rat cortical hemispheres were dissected from whole embryonic rat brains (E18, BrainBits) and dissociated with 10% (v/v) papain (Carica papaya, Roche) in Hibernate®-E (BrainBits) at 35 °C for 15 min. Dissociated cortical neurons were seeded at a cell concentration of 1 million cells per ml into PEI pre-coated 35 mm Petri dishes at a cell density of 180 cells/mm 2 and were incubated (95% air, 5% CO 2 , 37 °C) in serum-free Neurobasal with 2% (v/v) serumfree B-27® and 1% (v/v) PenStrep antibiotics, and grown until day 8 in vitro. For calcium fluorometry, Fluo-4 AM with probenecid acid was loaded to the cells (1:1) and incubated for 60 min in a standard incubator (37 °C, 5% CO 2 ) following vendor protocol (ThermoFisher). Somatic calcium dynamics were monitored at 0.1 frames per min, 1 s exposure time for 10 h with cyclic on/off LED-light 480 nm excitation. The digital microscope was set to 160×-220× magnification. Cells were either monitored without heat at room temperature (w/o heat), or at physiological temperature (37 °C, w/heat) in the temperature-controlled live-cell imaging system over 10 h.
Live-cell nanomaterial fluorometry. At 9 days in vitro (DIV), Fluo-4 AM loaded cortical neurons were exposed to chitosan-coated fluorescent magnetic nanoparticles (5 × 10 1111 NP per ml, Chemicell, core: 100 nm, hydrodynamic radius: 190 nm, Fig. S5) and monitored for live-cell fluorescent imaging over 10 h. Extensive characterization of the chitosan-coated NPs can be found in Tay, Kunze et al. 29 . Somatic calcium dynamics were recorded with LED-light 480 nm excitation at 1 fps, 1 s exposure time for 5 min in the incubator system without heat at room temperature (w/o heat), or at physiological temperature (37 °C, w/heat). During a 2-h interval, neurons were left without excitation and imaged again with the same parameters. This process was repeated three times for a total imaging time of 8 h. For control, fluorescent neurons without magnetic nanoparticles were monitored under the same imaging parameters with and without physiological temperature settings.
Grey-scale time-lapse images (8-bit) were analysed by selecting multiple single-cell regions of interests (ROIs). Fluorescence signal distribution was extracted, and relative fluorescence (F rel ) was plotted based on Eq. 1, where F max is the maximal detected fluorescent signal in all images, F is the averaged fluorescent intensity per ROI, and F Bkg denotes the background fluorescent signal. Time-varying changes of somatic fluorescence (F pixel ) were recorded and averaged fluorescent intensity F across each cellular ROI was calculated using Eq. 2. Equation 2shows n as the total number of pixels and F pixel as the intensity value of each indexed pixel per ROI.
Second, F was normalized by the average background (F Bkg ) for each frame resulting in F * as shown in Eq. 3:
Third, a relative fluorescence change ( F * t ), where Δt is the framerate -1 was used for subsequent cal- cium signalling analysis. Calcium spike events were distinguished based on calcium influx and efflux. For both a double threshold analysis was applied based on a static ( �F * �t > ±0.05 ) and a varying threshold ( �F * �t > ±5x standard deviation). A calcium spike event was then set as a calcium influx for a positive amplitude above the highest positive threshold, and as a calcium efflux event for a negative amplitude below the smallest negative threshold. From these calcium events, raster plots and connectivity maps were generated (see supplementary data for more information).
(1)
Chitosan-coated magnetic nanoparticles (NPs) impacts synchronicity in calcium signalling. (A1-A6) Image processing analysis to extract spatially resolved neuronal connectivity map based on primary cortical neurons labelled with Fluo-4. (A1) Time-lapse images show pseudo-coloured calcium activity captured with digital fluorescent microscopy. (A2) Automated image segmentation spatially selects individual regions of interest (ROIs) based on fluorescently active cell bodies. (A3) Changes in calcium activity get recorded over time and (A4) transferred into a calcium spike raster plot based on calcium influx. (A5) A cross-correlation matrix presents high (blue) and low (red) probability of temporally connected calcium spiking activity, which is used in (A6) to connect the ROIs in a connectivity map. (B) Representing fluorescent images of primary neurons incubated for 10 h in the digital imaging system. (C1-C2) Connectivity maps present changes in synchronous calcium spiking activity over an 8-h time span in cortical neurons grown for 9 days in vitro, independent of room temperature (w/o heating), physiological temperature (w/heating) and of incubation with chitosan-coated magnetic NPs (w/NPs). (D1-F2) Network analysis shows (D1-D2) number of active neurons under different temperature conditions, with and without NPs, (E1-E2) number of connections between ROIs under different temperature conditions, with and without NPs, and (F1-F2) active cell normalized count of connections under different temperature conditions, with and without NPs. These four plots indicate the activity of the network and how it changes under the different culture conditions. ◂ 1. Methodology Portable, live-cell fluorescent imaging system. An imaging system was designed to maintain constant physiological temperature and humidity control for long-term live-cell monitoring, using off the shelf elements for high reproducibility at a low cost. The imaging system was assembled based on a small-scale benchtop incubator with digital temperature control (e.g., MyTemp™ mini digital incubator), a digital fluorescent microscope with coloured illumination (LED-based, e.g., Dino-Lite AM4115T-GRFBY), a cell culture sample holder, and a LED-based white light illumination. Wiring of the white light and digital camera were routed through an access port at the back of the incubator and sealed with Parafilm wax. The digital fluorescent microscope is a glass-lens based mini microscope. It can be operated at two different excitation wavelengths of 480 nm and 575 nm to monitor green and red fluorescent probes through a USBconnected desk laptop. The optical sensor in the digital microscope is a CMOS camera with 1280 × 1024 pixels resolution (1.3 megapixels) that can capture up to 30 frames per second (fps). The imaging system is a low-tech version of similar live-cell imaging platforms [1][2][3][4][5][6][7][8][9] . It was purposely kept at low-cost (< 2,000 US$) and required no additional computer drawing skills or access to 3D printing. The live-cell fluorescent imaging system was designed to make calcium nano fluorometry in neurons attractive for low-resource environments and to provide advanced neuroscience research tools for classrooms 10 . A full comparison of our digital live-cell fluorescent imaging system against other portable, low-cost digital and low-cost traditional optical imaging systems can be found in Table S1 (see supplementary file).
To verify the optical properties of the digital microscope at the resolution of single-cell bodies, we imaged Dragon Green fluorescent beads (15.65 µm) diluted in sterile water (1% (v/v)). Imaging characteristics of the digital microscope (220x) were quantified based on shape and brightness uniformity of bead-derived fluorescent signals. They were analyzed using the signal-to-noise ratio (SNR), mean-absolute-error (MAE), peak signal-to-noise (PSNR), and root-mean-standard error (RMSE) using an ImageJ plugin by Sage et al. 11 .
For image comparison, green-fluorescence was captured in a 24-bit RGB format and converted to 8-bit grey-scale. Properties of the digital microscope images were compared with images taken from the beads with an optical fluorescent microscope (Amscope, 20x, and 40x optical magnification with a 6-megapixel camera) as reference. Fluorescent beads were excited with 475 nm light. Batches of ten images of the beads were recorded from both the digital and the optical imaging system for statistical analysis.
Temperature sensation in human embryonic kidney cell culture. To demonstrate the effect of temperature control of the imaging system, we chose human embryonic kidney (HEK) as they are well known to exhibit endogenous calcium channels 12 and to show high sensitivity to non-physiological temperatures [13][14][15] . HEK cells were cultured in mouse embryonic fibroblast (MEF, passage 18) media. When grown to 80% confluency, cells were trypsinated and reseeded into pre-coated 35 mm Petri dishes for the temperature sensation experiment and grown for two days. For calcium fluorometry in HEK cells, Fluo-4 AM with probenecid acid was loaded to the cells (1:1) and incubated for 60 min in a standard incubator (37 ºC, 5% CO2) following vendor protocol (ThermoFisher). Afterwards, HEK cells were gently washed with pre-warmed MEF media and then returned to the incubator for two hours. For the temperature sensation experiment, calcium fluorescent HEK cells were placed into the live-cell fluorescent imaging system, and somatic calcium dynamics were monitored at 0.1 frames per min, 1 s exposure time for 10 h. LED-light excitation at 480 nm was automatically controlled and switched on and off 10 s prior to image capture. The digital microscope was set to 160X magnification. HEK cells were either monitored without heat at room temperature (w/o heat), or at physiological temperature (37 °C, w/ heat) in the temperature-controlled livecell imaging system.
Temperature sensation in primary cortical neuron cultures. Calcium fluorometry is an important imaging methodology to study neuronal cell and network signalling [16][17][18][19][20][21][22][23][24] . To test the robustness of our imaging system with neurons, we monitored calcium signalling in neural cultures grown from dissociated rat cortical neurons. Rat cortical hemispheres were dissected from whole embryonic rat brains (E18, BrainBits) and dissociated with 10% (v/v) papain (Carica papaya, Roche) in Hibernate®-E (BrainBits) at 35 °C for 15 min. The dissociated cortical neurons were centrifuged (6 min, 600 rpm, at room temperature) and seeded at a cell concentration of 1 million cells per ml into 35 mm Petri dishes. Petri dishes were coated with 0.05% (v/v) polyethyleneimine (PEI) for two hours and washed three times with phosphate-buffered solution (PBS) before cell seeding. Cortical neurons were seeded at a cell density of 180 cells/mm 2 and incubated (95% air, 5% CO2, 37 °C) in Neurobasal serum-free with 2% (v/v) serum-free B-27® and 1% (v/v) PenStrep antibiotics and grown until day 8 in vitro. For calcium fluorometry in the neuronal cultures, Fluo-4 AM with probenecid acid was loaded to the cells (1:1) and incubated for 60 min in a standard incubator (37 ºC, 5% CO2) following vendor protocol (ThermoFisher). For the temperature sensation experiment, calcium fluorescent neuron cultures were placed into the live-cell fluorescent imaging system, and somatic calcium dynamics were monitored at 0.1 frames per min, 1 s exposure time for 10 h with cyclic on/off LED-light 480 nm excitation, identical to the HEK cell experiment. Neuronal cells were either monitored without heat at room temperature (w/o heat), or at physiological temperature (37 °C, w/ heat) in the temperature-controlled live-cell imaging system over ten hours.
For both HEK and neuronal cells, 8-bit grey-scale time-lapse images were analyzed in sequence based on selecting multiple single-cell regions of interests. For each single-cell region of interest (ROI), fluorescence signal distribution was extracted, and relative fluorescence (Frel) was plotted based on equation 1, where Fmax is the maximal detected fluorescent signal in all images, 𝐹𝐹 � is the averaged fluorescent intensity per ROI, and 𝐹𝐹 𝐵𝐵𝐵𝐵𝐵𝐵 is denoted as the background fluorescent signal.
Live-cell nano fluorometry. At 9 days in vitro (DIV) cortical neurons were loaded with Fluo-4 AM for 60 min and gently washed as described above. Then chitosan-coated magnetic nanoparticles (5×10 11 NP per ml, Chemicell, core: 100 nm, hydrodynamic radius: 190 nm, Fig. S5) were added to the calcium fluorescent living neurons and placed into the live-cell fluorescent imaging system for further live-cell fluorescent monitoring. Extensive characterization of the chitosan-coated NPs can be found in Tay, Kunze et al. 25 . Somatic calcium dynamics were recorded with LED-light 480nm excitation at 1 fps, 1 s exposure time for 5 minutes in the incubator system without heat at room temperature (w/o heat), or at physiological temperature (37 °C, w/ heat). During a two-hour interval, neurons were left without excitation and imaged again with the same parameters. This process was repeated three times for a total imaging time of 8 h. For control, fluorescent neurons without magnetic nanoparticles were monitored under the same imaging parameters with and without physiological temperature settings.
Calcium spike event detection and synchronous network activity mapping. Fluorescent images acquired in our live-cell imaging system were converted from their native '.wmv' video format to '.tiff' using FFmpeg. All image analysis was performed in MATLAB 2019A, and final graphical data were plotted using Origin Lab 2018b. First, the tiff-based image stack was converted into an 8-bit image stack, and single-cell bodies were segmented and saved as individual regions of interests (ROIs) with their corresponding spatial x, y coordinates. Time-varying changes of somatic fluorescence (Fpixel) were recorded and saved as a time series data with averaged fluorescent intensity values across the pixels in each ROI. Equation 2shows 𝑛𝑛 as the total number of pixels within the ROI and 𝐹𝐹 𝑝𝑝𝑖𝑖𝑚𝑚𝑟𝑟𝑟𝑟 as the intensity value of each indexed pixel in the ROI and 𝐹𝐹 � as the averaged fluorescence intensity per ROI.
Second, 𝐹𝐹 � was normalized by the average background (𝐹𝐹 𝐵𝐵𝐵𝐵𝐵𝐵 ) for each frame resulting in 𝐹𝐹 * as shown in equation 3:
Third, the rate of relative fluorescence change (∆𝐹𝐹 * /∆𝑡𝑡), where ∆𝑡𝑡 is the framerate -1 was used for subsequent calcium signal analysis (spike event detection, signal correlation, and synchronous network activity mapping). Calcium spike events were distinguished based on calcium influx and efflux. For both a double threshold analysis was applied based on a static ( ∆𝐹𝐹 * ∆𝑡𝑡 > ±0.05) and a varying threshold ( ∆𝐹𝐹 * ∆𝑡𝑡 > ±5x standard deviation). A calcium spike event was then set as a calcium influx event for a positive amplitude above the highest positive threshold, and as a calcium efflux event for a negative amplitude below the smallest negative threshold. From these calcium events, raster plots were generated, showing either influx, efflux or both event types.
Pathological calcium events were analyzed separately from transient calcium dynamics. If a cell body exhibited a high cytosolic influx in calcium followed by a substantial efflux, this event might indicate a relation of calcium signalling with apoptosis or necrosis. These calcium events were identified by a systematic scan of the time-varying fluorescent data for a single influx followed by single efflux events. Next, the time delay (∆T) between the influx and efflux event was extracted.
Calcium raster plots were compared for signal correlation and used to derive a connectivity map based on synchronous spiking activity between individual cell bodies. Only cell bodies exhibiting multiple positive or negative peaks were analyzed. Network nodes in our connectivity map were defined by the exact coordinates of segmented active cell bodies (at least one spike event occurred within the segmented ROI). Transient calcium spiking events were compared for a synchronic pair-wise occurrence and mapped onto a cross-correlation matrix in MATLAB 2019A. Calcium signal cross-correlation was determined based on computing the Sørensen-Dice similarity coefficient. For Dice-coefficients between 0.5 and 0.9, cells were assumed to be weakly connected, and a line with a transparency value of 40% and 5 pixels in width was drawn between their corresponding ROIs. For Dice-coefficients larger than 0.9, cells were assumed to be strongly connected, and a line with a transparency value of 40% and 7 pixels in width was drawn between their corresponding ROIs.
Cell network analysis. To compare macro and micro environmental effects on the neural network activity and performance, we further quantified three parameters for each individual network and reported their change over time. The three parameters are the number of active cells, the number of synchronous network connections, and the number of calcium spike events. Cells were counted as active if at least one calcium spike event had been identified from this indexed ROI throughout the recording period. The number of synchronous network connections was summed for each synchronously active ROI pair. Calcium event activity was computed based on the count of network connections normalized by the count of active cell bodies.
To determine toxicity levels upon long-term imaging, we cultured Normal Rattus norvegicus Kidney (NRK, ATCC® CRL-6509™) epithelial cells in the imaging incubator for up to 48 h. Passage 6 NRK cells were detached from tissue culture flasks with 0.25% trypsin, 0.53 EDTA solution (2 min) and dispensed in new culture flasks at a subcultivation ratio of 1:6 (90% Dulbecco's Modified Eagle's Medium, 10% fetal bovine serum (v/v)). Culture flasks were placed in both the imaging incubator and a standard reference incubator for 48 hours. We performed bright field imaging at 0 h, 24 h, and 48 h. Live-dead staining was implemented by adding 4 µL of 3,3′-dioctadecyloxacarbocyanine (DiOC18) in 1 mL of culture medium at 24 h and 2 µL propidium iodide in 1 mL culture medium at 48 h followed by 5 min incubation at 37 °C. Cell viability (CV) was imaged using fluorescent microscopy and computed based on equation 4.
# 𝑔𝑔𝑔𝑔𝑒𝑒𝑒𝑒𝑛𝑛 𝑐𝑐𝑒𝑒𝑐𝑐𝑐𝑐𝑐𝑐 # 𝑔𝑔𝑔𝑔𝑒𝑒𝑒𝑒𝑛𝑛 𝑐𝑐𝑒𝑒𝑐𝑐𝑐𝑐𝑐𝑐 + # 𝑔𝑔𝑒𝑒𝑟𝑟 𝑐𝑐𝑒𝑒𝑐𝑐𝑐𝑐𝑐𝑐
• 100 % 𝑒𝑒𝑒𝑒. 4
Heterogeneity in calcium signalling within cell cultures may impact average calcium intensity plots. Within our data set, the MATLAB algorithm randomly selects single cell bodies. Figure S1 A1 -A2 show the distribution of normalized calcium intensity over time for three different samples with a fixed number of randomly selected single-cell bodies (ncell = 10). The plots only show minor differences in average intensity distribution. When accumulated over time, the data sets are not statistically different (Fig. S1 B). Decreasing the fixed number of randomly selected cell bodies from 10 to 5, as well as increase the number from 10 to 20, resulting in a similar average intensity over time. However, for a fixed number of 40 cells, we observed a significant shift in average signal intensity (Fig. S2). This artefact may be due to differences in sub neuronal cell types, differences in metabolism, or calcium signalling heterogeneity in neuronal networks. To exclude this artefact in our temperature experiments, we compared smaller sample sizes. However, we also used this artefact as a motivation to evaluate calcium signalling synchronicity over time in response to nanoparticle uptake. Figure S3 and Figure S4 show comprehensive calcium data sets which include the rate of fluorescent change over time, extracted raster plots highlighting calcium influx and efflux events, and resulting cell-to-cell cross-correlation matrices from Fluo-4 loaded neurons which were recorded without temperature control (at room temperature, Fig. S3) and with temperature control at the physiological level (Fig. S4).
Incubating primary neuron cultures with nanomaterials has been shown to impact calcium signalling in previous studies [25][26][27] . To validate the capturing of similar calcium events in the digital live-cell imaging system, we incubated primary cortical neurons with chitosan-coated superparamagnetic nanoparticles.
Figure S5 shows the size distribution of the hydrodynamic radius of the utilized nanoparticles measure in Neurobasal.
Histogram plots were used to assess image contrast for monitoring grey-scale, green, and red fluorescent images captured with the digital fluorescent microscope. Figure S6 shows the histogram distribution associated with the representative false-colour and grey-scale images. Figure S8 shows representative fluorescent images extracted from the video recordings done with the digital live-cell fluorescent imaging system to validate long-term image acquisition and portability. HEK cells were monitored in a lab-extern cell culture facility. Primary neuron cultures were monitored in our lab.
Figure S9 shows representative fluorescent images extracted from the video recordings done with the digital live-cell fluorescent imaging system to validate short-term image acquisition using primary cortical neurons.
Figure S10A (right panel) shows NRK cells growing in culture flasks in the portable imaging incubator system at the start (0 h) and after 24 h and 48 h. Comparing NRK cells grown in the portable incubator with a standard incubator system yields no significant differences in their cell density, proliferation rate, morphology and cell viability (Fig. S10A-B). After 48 h, fluorescent imaging displayed cell viability of 98.7% for the established incubation system and 98.6% for the portable incubation system based on Live/ Dead staining (Fig. S10C, Live: green = DiOC18, Dead: red = Propidium Iodide). Hence, the portable imaging incubator can support live-cell imaging in a remote setting over up two days, if needed.
Scientific RepoRtS | (2020) 10:12568 | https://doi.org/10.1038/s41598-020-69412-1