Preventing Photomorbidity in Long-Term Multi-color Fluorescence Imaging of Saccharomyces cerevisiae and S. pombe

Obtaining a detectable signal in fluorescent live-cell imaging requires a bright probe and sufficient excitation light. However, excitation light can also influence cell physiology as it is absorbed and excites a variety of cellular molecules (Cadet et al. 2012) in addition to the fluorescent species of interest. For example, short wavelength light (<330 nm) is absorbed by DNA creating mutagenic cyclobutane pyrimidine dimers (Sancar 2016). Microscopists typically prevent such mutagenic damage by using UV blocking filters when illuminating with arc lamps, or by utilizing light sources which do not emit UV light (Frigault et al. 2009). In contrast, UV-A and visible light (>330 nm) can cause cellular damage through indirect mechanisms (Kielbassa et al. 1997; Godley et al. 2005). Most prominently, the cell physiology is altered by creating reactive oxygen species (ROS) originating from the excitation of endogenous (e.g., metabolites) or exogenous (e.g., media components) light absorbing molecules, collectively termed photosensitizers (Cadet et al. 2012). ROS subsequently form adducts with nucleotides, proteins or other metabolites altering or interfering with their function (Godley et al. 2005; Cadet et al. 2012). However, as ROS are generated as a bi-product of normal cell physiology, cells are competent in clearing a limited amount of light-induced ROS damage, and therefore a threshold for visible light dose exists below which no damage is detectable (Tinevez et al. 2012).

Initial attempts measured the mitotic arrest in a tobacco cell line (Dixit and Cyr 2003) and morphological changes and cell death in mammalian cells (Magidson and Khodjakov 2013). These are terminal phenotypes and as such not suited to detect subtle changes in cell physiology. Subtle light-induced effects have been detected in different cell types as changes in cell migration, chromosome movements or mitochondrial membrane potential as well as changes in the proliferation cells or development of embryos (Icha et al. 2017). In Saccharomyces cerevisiae, photodestructive effects were measured using the frequency changes of the nuclear localization of the general stress activated transcription factor Msn2 as a quantitative trait (Jacquet et al. 2003; Logg et al. 2009). As expected, a threshold light dose was found below which Msn2 localization frequency remained unchanged (Logg et al. 2009). However, this assay is very specific to the Saccharomyces sensu stricto and it is unclear if the localization frequency of Msn2 shows an equal sensitivity across the light spectrum. An additional complication of any fluorescent assay is that the sensitivity of the fluorescent readout is directly related to the administered light dose and can be distorted by photobleaching.

How can the threshold light dose be determined? A generic measure of cellular well being of dividing cells is the growth rate (GR). It has been used as a sensitive, quantitative read-out in diverse applications such as chemogenetic profiling (Lee et al. 2014) and systematic mapping of genetic interaction networks (Dixon et al. 2009). Moreover, changes in growth, measured as the timing of cell division, were shown to allow for quantifying subtle, non-toxic effects induced by fluorescence excitation light in C. elegans (Tinevez et al. 2012), mammalian cell culture (Laissue et al. 2017) and also S. cerevisiae (Carlton et al. 2010). We therefore reasoned that the growth rate of dividing cells can be employed as a measure for the absence of photo-destructive effects.

Here, we set out to develop guidelines to avoid the confounding effects of visible light in multi-color long-term fluorescence imaging using the model organisms S. cerevisiae and S. pombe. We aimed to measure ”photomorbidity”, which is defined as the effects of light on cellular well being as measured by decreased growth rate compared to an unstressed control. This is consistent with the definition of ”morbidity” by IUPAC: ”departure, subjective or objective, from a state of physiological or psychological well-being” (IUPAC 2016) and distinct from the more severe term ”phototoxicity”, which is used to describe cell death. We showed that photomorbidity can be described using a classical dose-effect relationship with a characteristic effective dose (ED50) and a no-observed effect level (NOEL) below which confounding effects are not detectable. We found that both measures depend on the cumulative light dose, and within practical boundaries are independent of the imaging interval, light intensity and bandwidth. To compare the suitability of different imaging channels and fluorescent proteins for live-cell imaging, we determined the signal-to-noise ratio (SNR) obtained with fluorescent fusion proteins at the NOEL of their specific excitation wavelength. Our results demonstrate how photomorbidity and cellular autofluorescence, in addition to fluorescent protein brightness, can influence the performance of fluorescent proteins in live-cell imaging. We found that photomorbidity of combined wavelengths in multi-color imaging was additive and could be avoided by limiting light doses to the NOEL. In addition, our findings highlight how particular combinations of fluorescent reporters enable artifact free multi-color time-lapse imaging.

Materials and MethodsYeast culturing

S. cerevisiae cells were precultured in either synthetic minimal medium (Smin containing 1.7 g/l yeast nitrogen base without amino acids or ammonium sulfate (BD Biosciences, Germany), 5 g/l ammonium sulfate) or YP medium (10 g/l yeast extract, 20 g/l peptone, both BD Biosciences, Germany) while S. pombe cells were grown in Edinburgh minimal medium (6.77 g/l EMM-nitrogen-glucose minimal media (Sunrise Science Products, San Diego, USA), 5 g/l ammonium sulfate) at 30 or 32° in an orbital shaker (Sherman 2002; Moreno et al. 1991). All media were supplemented with 2% (w/v) of the appropriate carbon source. All chemicals were purchased from Sigma-Aldrich Co. (Germany), unless stated otherwise. Cells were innoculated from freshly streaked plates in 5 mL medium one (glucose) to two days (acetate and glycerol) before the experiment. The evening before the start of the experiment, cells were diluted 1:500 (glucose) or 1:50 (acetate and glycerol) in fresh media. For loading of the microfluidic device, the cell concentration was measured using a Z2 Coulter Counter (Beckman Coulter, Nyon, Switzerland). Typical cell concentrations ranged from 0.5*106 to 5*106 cells/ml. 1 mL of culture was transferred to a 1.5 mL tube and spun at 1000 g for 2 min and the appropriate amount of supernatant was removed to yield a final cell concentration of 1*107 cells/ml. The cells were resuspended in the remaining media using a Vortex Genie 2 (Scientific industries, New York, USA) for 10 sec at speed setting four.

Plasmid construction

All plasmids were constructed using classical DNA manipulation techniques. Restriction enzymes and T4 DNA ligase were purchased from New England BioLabs. DNA sequences were amplified using Phusion High-Fidelity DNA Polymerase (Thermo Scientific). Yeast codon optimized fluorescent protein sequences were derived from pKT139 (Citrine A206K) (Sheff and Thorn 2004), used in a previous work (sfGFP) (Roberts et al. 2016) or synthesized from GeneArt (mTFP1, mAmetrine, mCardinal, mNeptune2, mNeongreen, mTurquoise2, mKOκ, tSapphire, mRuby2, mKate2, tdKOκ). All fluorescent proteins were cloned into pKT102 using PacI/AscI flanking restriction enzymes (Sheff and Thorn 2004). eGFP in the pFA6 backbone was obtained from Addgene #44900 (Lee et al. 2013). All plasmid sequences were checked by Sanger sequencing (Microsynth AG, Switzerland). The plasmids are listed in Table S8.

Yeast strain construction

All S. cerevisiae photomorbidity experiments were carried out using a prototrophic FY4 strain (FRY2032). FRY2032 was obtained by mating of BY4700 and BY4707 (Brachmann et al. 1998) and subsequent sporulation (parent strain FRY2023) and selection on minimal media (Smin + glucose). All S. pombe experiments were carried out using a prototrophic 972 h- strain (FRSP902) (Leupold 1950). For fluorescent protein characterization, all S. cerevisiae strains were derived by endogenous tagging of the prototrophic strain FRY1455 (Gnügge et al. 2016). Tagging cassettes were amplified from the modified pFA6 cassettes (Table S8) by polymerase chain reaction (PCR), using Phusion High-Fidelity DNA Polymerase (Thermo Scientific, USA). Yeast transformations were performed using the Lithium-Acetate method (Daniel Gietz and Woods 2002) and successful integration was checked by fluorescence microscopy and/or colony PCR. Table S9 lists all yeast strains used in this work. Table S7 lists the primers used for tagging/knock-out/colony PCR of Vph1, Cdc12 and Whi5.

Yeast colony PCR

We checked integration at the targeted site by a colony PCR assay. Yeast colonies were boiled in 3 μl 20 mM NaOH for 10 min and DNA was amplified in a 25 μl volume with 1 M betaine, 1x ThermoPol buffer (New England BioLabs), 0.2 mM each dNTP, 2 μM each primer (Table S7) and 1.25 U Taq DNA Polymerase (New England BioLabs), using the following PCR program: 5 min initial denaturation at 94°; 30 cycles of 30 s at 94°, 30s at 58°, 1 min/kbp at 72°; and 10 min final elongation at 72°.

Microfluidic chip fabrication

The microfluidic chip (adapted from (Frey et al. 2015)) was designed in AutoCAD (Autodesk, München, Germany). The chip consists of one layer of polydimethylsiloxane (PDMS, Sylgard 184, Dow Corning Corp., USA), attached to a 150-μm-thick cover glass (24 mm × 60 mm) using vacuum. The PDMS layer was cast from 4-inch silicon wafers patterned using standard photolithography and dry etching processes. In short, the first layer was fabricated by patterning a dry etching mask using ma-P 1240 photoresist (micro resist technology GmbH, Berlin-Koepenick, Germany) according to the manufacturers instructions. Dry etching was performed on an Ionfab 300 (Oxford instruments, Abingdon, United Kingdom). The first layer defines the gap between the glass and the clamping pad and its height was modified for the specific yeast type: 3.8 μm for S.cerevisiae and 4.1 μm for S.pombe. Three layers of SU-8 were patterned on top of the dry etched silicon substrate to complete the microfluidic design. We followed the datasheet of the manufacturer (Microchem Corp.,Westborough, USA) to structure each of the SU-8 layers. In brief, SU-8 was spin-coated at the desired thickness and soft-baked on a hotplate. The SU-8 was then exposed through a transparency mask (Selba S.A., Versoix, Switzerland) using a UV mask aligner (MA/BA8-Gen3 mask aligner, SUSS MicroTec AG, Garching, Germany) and baked on a hotplate for cross-linking. Unexposed SU-8 was removed using mr-Dev600 (micro resist technology GmbH, Berlin-Koepenick, Germany). The molds were coated with trichloro (1H,1H,2H,2H-perfluoro-octyl)silane (Sigma-Aldrich, Switzerland) in a vapor silanization process, for at least one hour. 25g of freshly mixed and degassed polydimethylsiloxane (PDMS Sylgard 184, 10:1 (w/w) monomer:curing agent) was poured onto the mold. After curing in a convection oven at 80° for at least 2 h, the PDMS was released from the silicon wafer. Single chips were cut from the PDMS sheet and access holes were punched at inlet and outlet sites.

Setup of microfluidic experiment

The PDMS device and glass slide were rinsed with acetone, isopropanol, deionized water and dried using a nitrogen gun. The PDMS device was placed with its structured side facing up in a sterile cell culture hood for chip loading. 0.4 μl of the cell solution was dispensed onto each culturing area of the microfluidic device by using a conventional pipette. The cover glass was placed on top of the device such that it sealed the microfluidic channels. The cover glass and device were held together by adhesion. The device was then transferred to the microscope. First, the vacuum channel was attached to the in-house vacuum supply to facilitate a rigorous adhesion of the PDMS chip to the glass slide using microfluidic tubing (Tygon LMT-55, inner diameter: 0.51 mm, wall thickness: 0.85 mm). Next, the inlet and outlet tubing (Tygon LMT-55, inner diameter: 0.51 mm, wall thickness: 0.85 mm) was attached and the flow through the device (10 μl/min) was started. The flow was established by syringe pumps (Nemesys, CETONI GmbH, Korbussen, Germany) using medical glass syringes (PTFE TLL Luer Lock 25 mL syringe, Innovative Labor Systeme GmbH, Stuetzerbach, Germany). Bubbles introduced during loading were removed in less than one hour through diffusion through the PDMS into the vacuum channel which surrounded the fluid channels. The microfluidic device was placed in a custom-made aluminum chip holder, and fixed using nail polish (Maybelline, L’Oreal Suisse S.A, Vernier, Switzerland).

Measurement of microscope excitation light spectra and intensity

To measure the light spectra of the fluorescence and transmitted illumination light source, an Ocean Optics USB2000+ spectrometer (Ocean Optics, Dunedin, USA) was used. The optical fiber of the spectroscope was inserted into a 5mm thick block of PDMS (Sylgard 184, Dow Corning, Midland, USA) plasma bonded to a 170 μm cover slide and brought into focus of the objective. The light from the fluorescence excitation light source (Spectra X Light Engine, Lumencor, Beaverton, USA) was transmitted through a 40x Plan Fluor Oil DIC N2 NA 1.3 objective (MRH01401, Nikon). To avoid stray light influencing the measurement, the aperture diaphragm was used to restrict illumination to the opening of the optical fiber only. To reduce the light intensity, two neutral density filters (Cyto 2.0 ND 200-0065) were mounted into the same filter cube as the dichroic. The respective excitation filter - dichroic combination (see Table S3 & S6) was inserted in the light path and the light intensity setting of the respective LED was set to 100%. For the transmitted illumination light source the glass slide with the PDMS and optical fiber was turned upside down. The condenser of the microscope was used to focus the transmitted lightsource onto the optical fiber and the field diaphragm was closed to illuminate only the fiber opening. Raw spectra were recorded using Spectra Suite (Ocean Optics, Dunedin, USA). The integration time of the sensor was adjusted so that the peak intensity was between 4-6*104 AU (5-200 ms). Dark spectra with the same integration time and the light source turned off were recorded for background correction. Data were exported as comma delimited file and Matlab (MathWorks, Natick, USA) was used for background correction and calculation of the center wavelength.

The light intensity of the epifluorescence excitation light was measured similarly to what was described earlier (Grünwald et al. 2008). In short, the field diaphragm of the epifluorescence light path was adjusted such that the diaphragm was visible in the field of view of the camera, using the 40x Plan Fluor Oil DIC N2 NA 1.3 objective (MRH01401, Nikon Instruments AG, Egg, Switzerland) and a fluorescent Argo-M Standard microscopy slide (Argolight, Talence, France). An image of the field diaphragm was taken and the number of illuminated pixels was determined in ImageJ. The illuminated area in the sample plane was calculated from the number of illuminated pixels, the size of the pixels of the camera (6.5 μm × 6.5 μm) and the magnification of the objective, Next, the objective was removed and an adjustable iris (SM1D12C, Thorlabs, Newton, USA) was mounted in its place. The diameter of the iris was adjusted to restrict the light beam to the diameter of the objective back aperture, to measure only the light that is actually entering the objective. An S170C microscope slide power sensor (Thorlabs, Newton, USA) was placed in the microscope stage holder, and the power was measured using a PM100USB power meter (Thorlabs, Newton, USA). The respective excitation filter - dichroic combination (see Table S3 & S6) was inserted in the light path and the previously determined central wavelength for each filter set was entered in the PM100USB software, to acquire wavelength corrected power measurements. The light power was measured for light intensity settings of the respective LED ranging from 10% to 100% in steps of 10%. The data were exported as a comma delimited file. Matlab (MathWorks, Natick, USA) was used for calculating the light intensities at the sample plane.

Measurement of the exact exposure time of the microscope setup

To measure the actual time a sample is exposed to the excitation light, we used a TSL2561 (AMS-TAOS USA Inc.) light to digital converter on a SparkFun Luminosity Sensor Breakout board (SparkFun Electronics, USA). The sensor was connected to an Arduino UNO SMD R3 (Arduino LLC) via the SDA and SCL channels and powered by the Arduino UNO board internal 3.3V power supply. A self-made holder was used to mount a glass fiber (diameter: 50 μm, D+X Produkte GmbH, Turbenthal, Switzerland) directly above the photo-active area of the TSL2561 using a duplex glass fiber adapter. The other end of the glass fiber was glued into a Luer Lock needle tip. The needle tip with the glass fiber was inserted into a PDMS block (Sylgard 184, Dow Corning, Midland, USA), which was plasma bonded to a 170 μm cover glass. The PDMS/glass sandwich was placed on the microscope and the glass fiber brought into focus of the 40x Plan Fluor Oil DIC N2 NA 1.3 objective (MRH01401, Nikon). The integration time of the TSL2561 light sensor was set to the shortest possible period (13.7 ms). Exposure times of 50, 100, 200, 500, 1000 and 2000 ms were applied five times each using the microscope control software YouScope and the luminosity was recorded continuously. Matlab (MathWorks, Natick, USA) was used to extract the actual exposure times from the luminosity measurements and plot the measured values against the set exposure times. A linear function was fitted to the data (Figure S4). On average, the additional exposure time due to hardware delays was 187 ms and was constant over all set exposure times. The additional exposure time was taken into account when calculating the light doses applied during photomorbidity measurements.

Microscopy

We chose a microscope configuration which allowed for maximum detectability of the fluorescent signal. Unless stated otherwise, we used a 40x Plan Fluor Oil DIC N2 objective with an NA of 1.3 (MRH01401, Nikon Instruments AG, Egg, Switzerland), as this allowed the maximum SNR (SNR ∼ NA4* Magnification−2, (Rines et al. 2011; Frigault et al. 2009)) to be obtained while maintaining subcellular resolution and imaging of the whole cellular volume in a single focal plane.

Experiments were performed on a Nikon Ti Eclipse (Nikon Instruments AG, Egg, Switzerland) inverted fluorescence microscope controlled using YouScope (Lang et al. 2012). The microscope was placed in an environmental enclosure (Life Imaging Services, Switzerland) to maintain the desired temperature. To keep the cells in focus over the time course of the experiment the Perfect Focus System was enabled. Images were recorded using an ORCA Flash 4.0 V1 camera (Hamamatsu Photonic, Solothurn, Switzerland) operated in water cooling mode (20°). Unless stated otherwise, the microscope was equipped with a Spectra X Light Engine fluorescence excitation light source (Lumencor, Beaverton, USA) and a pE-100 brightfield light source (CoolLED Ltd., UK). For multi-color NOEL imaging, hardware triggering between the light sources and the camera was implemented using an Arduino UNO (Somerville, MA, USA). All measurements were run with a diffuser and a green interference filter placed in the brightfield light path.

The optical filters and light intensities used for the measurement of photomorbidity can be found in Table S3. The optical filters and light intensities used for the measurement of the SNR of the fluorescent protein fusions can be found in Table S6. All optical filters were purchased from AHF Analysetechnik AG (Tuebingen, Germany).

Image analysis

For cell segmentation out-of-focus brightfield images were acquired (± 5 AU, Nikon Perfect Focus System). As an input for cell segmentation brightfield images acquired above the focal plane were divided by brightfield images acquired below the focal plane. The division of images leads to elimination of uneven illumination and enhances the membrane diffraction pattern of the individual cells yielding a better segmentation. Image division was carried out using the ”Image Calculator” function of ImageJ. Cell segmentation was performed using CellX (Mayer et al. 2013). Fluorescent properties were extracted using the in-built functions of CellX.

Measurement of photomorbidity

The microfluidic chip was loaded as described earlier. For morbidity measurements the cells were atoned on the chip for 3-4 hr before time-lapse imaging was started. Imaging was performed in five minute intervals, unless stated otherwise. For measuring one morbidity curve, the light dose was altered by adjusting the exposure time and at least six different light doses were tested in each experiment. A control measurement, where cells were not exposed to any epifluorescence excitation light was included in each measurement. For each light dose and the control, five independent positions in the chip were imaged. Hence, each photomorbidity curve consists of at least 35 data points (measured culturing pads). Some photomorbidity experiments were repeated with different exposure times to capture the whole dose-effect relationship. The optical filters and light intensities used for the measurement of photomorbidity can be found in Table S3. During calculation of the applied light doses the hardware delays during exposure were taken into account. Photomorbidity experiments were performed at 30° in Smin + glucose media for S. cerevisiae and at 32° and in EMM + glucose media for S. pombe, unless stated otherwise. Time-lapse imaging was carried out for 495 min (100 iterations). In the resulting images the regions occupied by colonies where cells touch the image border at any time of the experiment were excluded using in ImageJ (National Institute of Mental Health, Maryland, USA) before segmentation. The number of cells was subsequently extracted by segmentation using CellX (see Image analysis for details). The growth rate was calculated as follows:Embedded ImageEmbedded Image(1)The calculation of the growth rate was based solely on the first (CellsEmbedded ImageEmbedded Image) and last (CellsEmbedded ImageEmbedded Image) image of the observation period, and gives the average growth rate over the whole observation time. This facilitates the growth rate measurement in cases where the growth rate changes over time. Also, the average growth rate is in good agreement with growth rates calculated from cell numbers determined every five minutes, where the growth rate was determined by fitting an exponential growth model. For S. cerevisiae the doubling time (doubling time = 60 min/growth rate) is 117 ± 6 min if all datapoints are fitted with an exponential model and 123 ± 7 min if the average doubling time is calculated from the first and last image only. Likewise, for S. pombe fitting of all datapoints yielded doubling times of 130 ± 6 compared to 128 ± 10 min calculated from the first and last image.

The growth rates were plotted against the light dose and the dose-effect model was fitted using the Matlab R2015b ”fit” function (Mathworks, Natick, USA).Embedded ImageEmbedded Image(2)The residuals were minimized using a robust estimate relying on the absolute distance. The 95% confidence interval of the ED50 and τ were obtained from the fit parameters using the ”confint” function. The no-observed effect level (NOEL) was extracted from the fit as the light dose (LD) at which the growth rate is reduced to 98% of the control growth rate (GREmbedded ImageEmbedded Image).

Measurement and calculation of fluorophore brightness and signal-to-noise ratio

Where possible, the SNR is determined from image areas which contain a homogeneous distribution of signal (Embedded ImageEmbedded Image) and background (Embedded ImageEmbedded Image). However, in biological samples signal and background are non-homogeneously distributed. We therefore determined the signal (Embedded ImageEmbedded Image) by measuring endogenously tagged cells and the noise (Embedded ImageEmbedded Image) by measuring a parent strain carrying no fluorescent protein tag. To analyze a large number of cells, S. cerevisiae cells were loaded in the microfluidic chip and grown at 30° in Smin + glucose media for at least 12 h. The intensity of the fluorescence excitation light was adjusted to 2.91 W cm−2 h−1 for all imaging channels. The fluorescent proteins were imaged as endogenous fusions to either Vph1, Cdc12 or Whi5. Images at different exposure times were taken to measure the light dose dependent average signal value Embedded ImageEmbedded Image (Embedded ImageEmbedded Image, Embedded ImageEmbedded Image and Embedded ImageEmbedded Image). A parent strain carrying no fluorescent protein tag was imaged under the same conditions to measure Embedded ImageEmbedded Image. To exclude effects of photobleaching, each position on the chip was imaged only once.

The obtained images were segmented using CellX (see Image analysis for details) and the segmentation mask was applied to the fluorescence images. Embedded ImageEmbedded Image and Embedded ImageEmbedded Image were extracted as the mean intensity value of the pixels which were identified as belonging to cells.

The measured SNR (mSNR) was calculated as:Embedded ImageEmbedded Image(3)For the calculation of the mSNR, the measurements at discrete light doses (LD) were used.

To be able to predict the SNR (pSNR) for any light dose, Embedded ImageEmbedded Image and Embedded ImageEmbedded Image) and their respective standard deviations were plotted against the applied light dose. Linear functions (”ax + b”) were fitted using the ”weightedfit” function of Matlab R2015b (Mathworks, Natick, USA). From the parameters of the fits, Embedded ImageEmbedded Image and Embedded ImageEmbedded Image could be predicted for any light dose. For Vph1 Embedded ImageEmbedded Image (Embedded ImageEmbedded Image) was directly accessible from the images acquired from the respective Vph1-tagged strains. For Cdc12 and Whi5, Embedded ImageEmbedded Image (Embedded ImageEmbedded Image and Embedded ImageEmbedded Image) was estimated based on the signal levels of the Citrine tagged Vph1, Cdc12 and Whi5 reference strains as follows:Embedded ImageEmbedded Image(4)where,Embedded ImageEmbedded Image(5)and,Embedded ImageEmbedded Image(6)where,Embedded ImageEmbedded Image(7)Using the extrapolated values for Embedded ImageEmbedded Image, the predicted SNR (pSNR) for each fluorescent protein fusion at any light dose could be calculated according to equation 3.

The fluorophore brightness relative to eGFP was calculated based on the linear fit parameters, when Embedded ImageEmbedded Image and Embedded ImageEmbedded Image were plotted against the light dose, as described earlier. The brightness calculation was based on the signal from the Vph1-tagged strains (Embedded ImageEmbedded Image):

Embedded ImageEmbedded Image(8)Calculation of growth rates at SNR of four

The SNR-light dose relationships were calculated for each possible combination of the tested fluorescent proteins with Vph1, Cdc12 or Whi5 as described (see Measurement and calculation of fluorophore brightness and signal-to-noise ratio for details). The light dose at which an SNR of four would be reached was calculated using the ”solve” function of Matlab R2015b (Mathworks, Natick, USA). Using the previously determined dose-effect relationships of photomorbidity and light dose, we were able to compute the growth rates which are expected at the respective light dose to reach an SNR of four. In the cases of mRuby2, mKOκ, tdKOκ, mKate2, mCardinal and mNeptune2, the excitation filterset for imaging (Table S6) was not identical with any filterset used to determine photomorbidity (Table S3). In these cases we based our calculations of the growth rate on the measured photomorbidity dose-effect relationship with the closest central wavelength.

The growth rates for detection of mRuby2, mKOκ and tdKOκ were based on the photomorbidity of 542/20 nm light. The growth rates for detection of mKate2, mCardinal and mNeptune2 were also determined based on the photomorbidity of 542/20 nm light, as the dose-effect relationship for red (605/15 nm) light could not be determined in S. cerevisiae. In these cases a conservative estimate of the growth rate is obtained, as photomorbidity for green light (542/20 nm) is more pronounced than that for red light. Extrapolating between different filtersets with comparable central wavelength is feasible, as we have shown that photomorbidity is independent of the excitation filter bandwidth for light with comparable central wavelength.

Measurement of fluorophore photobleaching

S. cerevisiae cells were loaded and cultured in the microfluidic chip as described (see Measurement of fluorophore brightness and signal-to-noise ratio for details). To correct for bleaching of background fluorescence, a strain carrying no fluorescent protein tag was imaged under the same conditions. The light intensity of the fluorescence excitation light source was adjusted for all channels to be imaged with 2.91 mW cm−2 h−1.

To estimate the in vivo photobleaching, cells were imaged every 10 sec with an exposure time of 2 sec for 200 iterations. The resulting images were segmented (see Image analysis for detail) and the mean fluorescence intensity values (Embedded ImageEmbedded Image and Embedded ImageEmbedded Image) were extracted (see Measurement of fluorophore brightness and signal-to-noise ratio for detail). The mean fluorescence intensity was corrected for background fluorescence (Embedded ImageEmbedded Image - Embedded ImageEmbedded Image) and plotted against the accumulated light dose. A two-term exponential decay model (a*eEmbedded ImageEmbedded Image + c*eEmbedded ImageEmbedded Image) was fitted using the ”fit” function of Matlab R2015b (Mathworks, Natick, USA). The light dose at which the fluorescence intensity dropped to 50% of the initial value was determined (Embedded ImageEmbedded Image, by solving the fitted exponential model using the ”solve” function of Matlab R2015b (Mathworks, Natick, USA).

Data availability

File S1 contains guidelines to prevent photomorbidity in live-cell imaging in practice. File S2 contains a time-lapse movie of a growing S. cerevisiae colony inside the microfluidic chip. Plasmids are available from Addgene (see Table S8). Yeast strains are available upon request. The authors affirm that all data necessary for confirming the conclusions of the article are present within the article, figures, and tables. Supplemental material available at figshare: https://doi.org/10.25387/g3.13048073

ResultsMeasuring photomorbidity in time-lapse microscopy

The GR of a population of cells is a well established indicator for both genetic and environmental effects. Due to the magnification and limited field of view in microscopy, the largest population of cells which can be observed are small colonies and even those can suffer from nutrient limitation (Marinkovic et al. 2019). We implemented a population GR assay using a microfluidic chip to provide a continuous nutrient supply where S. cerevisiae or S. pombe cells grow in a single layer (adapted from (Frey et al. 2015)). When a cell is trapped below a PDMS pillar, all of its daughter cells are retained under the pillar and thus will form a colony (Figure 1A, File S2). After loading the microfluidic chip, we allowed the cells to adapt for at least one cell division time before we started time-lapse imaging. The average GR in each field of view can be calculated from all colonies where no cells reached the edge of the observation area (Figure 1B). The sensitivity of the assay is highest over the time period when no or only a few colonies outgrew the field of view. For glucose grown cells, we imaged in five minute intervals and realized that the maximum sensitivity is reached at ∼ 8 hr. The observed doubling time for S. cerevisiae is 123 ± 7 min and for S. pombe 128 ± 10 min (n = 5, mean ± standard deviation), which is in agreement with commonly reported doubling times in liquid culture (Sherman 2002) (Figure S1 & S2).

Figure 1Figure 1Figure 1

Measuring photomorbidity in a microfluidic chip using time-lapse imaging. A: Yeast cells are immobilized by clamping them between a PDMS pillar and a glass coverslip. B: Growth of S. pombe cells which are confined to grow in a single layer over 12 h. The GR is calculated from the number of cells at the beginning and the end of an observation period. Colonies that reach the border of the field of view (orange regions) are excluded from the analysis. C: S. cerevisiae growth at seven different doses of teal excitation light. For each light dose, the cell number was monitored on five independent culture pads (n = 5). Cell numbers from each condition were summed and normalized to 1 for t = 0 h. D: S. cerevisiae cells were exposed to different doses of cyan light. The GR obtained after 8 h was plotte

留言 (0)

沒有登入
gif