McClean: Integrative Biology Oct. 2013
Details of Strain and Plasmid Construction
Strain yMM1079 (MAT α trpΔ63 leu2Δ1 ura3Δ52 gal1ΔmCherry-caURA3 ) was created by replacing the native GAL1 coding region with mCherry and C. albicans URA3 from pMM066 in yeast strain yMM1146. Oligonucleotides oMM306/307 (Supplemental Table 3) were used to PCR mCherry and caURA3 off of the plasmid, and yeast were transformed with this linear fragment. Transformants were selected on SC-Ura media and checked by colony PCR and sequencing. Strain yMM1079 was transformed with pMM159 (GAL4AD-CIB1) and pMM160 (GAL4DBD-CRY2) to create yMM1081 which was used to test induction of mCherry mRNA and protein in response to blue light pulses (Fig. 3).
Strain yMM1134 (Mat α trp1∆63 leu2∆1 ura3-52 gal1ΔmCitrine-KanMX (PGAL1-mCitrine-KanMX)) was created by deleting the native GAL1 coding region in yMM1146 with ymCitrine from pMM40. Oligonucleotides oMM306/307 were used to PCR mCitrine and the KanMX cassette off of pMM40 and yeast were transformed with this PCR product. Transformants were selected on YPD + 200μg/ml G418 to select for the KanMX resistance marker.
To assay for a stress response in reaction to blue-light, we created strain yMM469 (FY Mat a prototroph HAP1+ MSN2-mCherry-HphMX) by tagging the endogenous MSN2 gene with mCherry in diploid yeast strain yMM391 (FY diploid prototroph HAP1+). The strain yMM391 was transformed with the mCherry-HphMX cassette amplified from pMM145 using oMM32/33. The resulting tagged diploid strain yMM410 was then sporulated and a hygromycin resistant haploid was selected to become yMM469.
The VP16-CIB1 plasmid (pMM281) was created using yeast homologous recombination7. The VP16 activation domain from the GEV artificial transcription factor1 was amplified from yMM1008 genomic DNA using oMM400/401. The pMM159 plasmid was cut with KpnI, which digests the plasmid between the existing SV40NLS and the GAL4AD. The PCR product containing VP16 and the digested plasmid were co-transformed into yMM83 (Mat a his3Δ1 leu2Δ0 LYS2 met15Δ0 ura3Δ0) and positive transformants that repaired the lesion in the plasmid with the VP16 activation domain (replacing the GAL4AD) were selected for growth on SC-Leu media. Plasmids were prepped from yeast and verified by sequencing.
Timecourse of localization measured using the SmartStat
Though we focused on using this technology to control total cytoplasmic protein levels, with the sampling and imaging system we were also able to image and quantify more complicated fluorescence signals including localization.
We demonstrated automatic quantification of protein localization over time by using this device to measure nuclear localization of the artificial GEV transcription factor in response to stimulation with β-estradiol. The GEV transcription factor consists of the GAL4 DNA-binding domain, the estrogen receptor, and the VP16 activation domain. The estrogen receptor is bound by Hsp90, which normally sequesters the GEV transcription factor to the cytoplasm. When β-estradiol is added, it out competes Hsp90 for binding to the estrogen receptor, revealing a nuclear localization signal within the estrogen receptor, and causing GEV to localize to the nucleus. Previously, we characterized the time course of GEV nuclear localization in response to β-estradiol using a microfluidic device and a GFP-tagged version of GEV 1.
To assay GEV localization in response to 1µM β-estradiol in cells growing in the chemostat we grew yeast cells containing a GFP-tagged version of the GEV transcription factor (yMM668 MATα gal4Δ::LEU2 (Pgal10+gal1)ΔloxP leu2Δ0::PACT1-GEV-GFP-KanMX HAP1+) in phosphate-limited chemostat media at steady-state using a 0.15 dilution rate (OD600 ~0.6) at 30°C. The sampling system automatically sampled and imaged cells every 3 minutes. After an initial 2 hours of imaging β-estradiol was added to the 27 mL working volume of the chemostat to a final concentration of 1μM. We then continued imaging for 13 hours. Single cell analysis for nuclear localization was done through custom software employing the particle separation algorithm from the ImageJ API to separate distinct yeast cells. Arrays of pixel intensities from each cell were generated and the top 5% of pixels via intensity was divided by the bottom 95% in order to create a nuclear localization score. All image processing code is available upon request.
The automatic sampling and image processing correctly identifies protein localization within 1 hour, consistent with our previous results 1. GEV stays localized throughout the course of the experiment, presumably because β-estradiol concentrations stay elevated (since the dilution rate is 0.15 giving us a final β-estradiol concentration of 0.1μM (100nM)). This is consistent with 10nM β-estradiol being sufficient to activate GEV-induced transcription 1. We conclude that the fast-sampling aspect of the culturing system, independent of its uses in feedback control to manipulate protein levels, is useful for assaying aspects of cellular physiology such as protein localization. This is promising for monitoring physiological parameters, such as intracellular pH, which are easily perturbed, and hence most easily studied under chemostatic conditions. Transcription in response to blue-light induction
Cells expressing pGAL1-mCherry, pGAL4DBD-CRY2 and pGAL4AD-CIB1 (yMM1081 MAT alpha, trpΔ63, leu2Δ1, ura3Δ52, PGAL1-mCherry-caURA3, pMM159 (pGal4AD-CIB1) pMM160 (pGal4BD-CRY2)) were grown to an OD600 of 0.5 in 10 mL of SC-Leu-Trp-Ura media in the chemostat vessel. This cell culture was exposed to a pulse of blue-light (100% intensity, 5.75 mW) for 5 minutes. Cells were sampled at timepoints T=0’ (before light pulse), T=1’, 3’, 5’ (during light pulse), and T=6’,8’,10’,20’,30’,40’,50’,60’ (after light pulse). To fix cells, sampled culture was incubated immediately with 6.4% paraformaldehyde for 45 minutes at room temperature.
The FISH probes to mCherry transcript were ordered from Biosearch Technologies and designed using the Stellaris™ Probe Designer version 2.0 (http://www.biosearchtech.com/stellarisdesigner/). The probes were ordered conjugated to the Cy3™ alternative dye Quasar® 570. The probe sequences used were: tcctcgcccttgctcaccat, cttgatgatggccatgttat, gcaccttgaagcgcatgaac, ccgttcacggagccctccat, tcgccctcgatctcgaactc, cctcgtaggggcggccctcg, ttcagcttggcggtctgggt, cagggggccacccttggtca, acaggatgtcccaggcgaag, gagccgtacatgaactgagg, ggggtgcttcacgtaggcct, ttcaagtagtcggggatgtc, gaagccctcggggaaggaca, agttcatcacgcgctcccac, gtcaccacgccgccgtcctc, cagggaggagtcctgggtca, ttgtagatgaactcgccgtc, aagttggtgccgcgcagctt, cattacggggccgtcggagg, cccagcccatggtcttcttc, tacatccgctcggaggaggc, cttcagggcgccgtcctcgg, tcagcctctgcttgatctcg, tagtggccgccgtccttcag, ggtggtcttgacctcagcgt, gcacgggcttcttggccttg, acgttgtaggcgccgggcag, ggtgatgtccaacttgatgt, tggtgtagtcctcgttgtgg, gcgcgttcgtactgttccac, gccggtggagtggcggccct, tacttgtacagctcgtccat
Fixed cells were permeabilized and prepared for fluorescence in situ hybridization (FISH) to count single mRNA molecules. Fixed cells were spun at 3000xg for 5 minutes and resuspended in 1.8ml of Buffer B (1.2M sorbitol, 100mM KHPO4 pH 7.5) and transferred to a 2ml eppendorf tube. These cells were then washed three times by pipetting up and down and spinning down at 3000xg for 5 minutes each time. The washed pellet was resuspended in spheroplast buffer with Vanadyl Ribonucleoside Complex (New England Biolabs, #S1402S), lyticase (Sigma #L5263; 25000U/ml), and beta-mercaptoethanol. The total volume is 1002µl (890µl Buffer B + 100µl VRC + 10µl lyticase and 2µl beta-mercaptoethanol). This was incubated at 37°C and the digestion was monitored every 10 minutes. Once digested cells reached ~90% (phase dark cells) the cells were spun for 5 minutes at 3000xg. Cells were washed three times with 1ml Buffer B, spinning at 2000rpm for 5 minutes each wash. To the final pellet, 70% ethanol was gently added and cells were stored over night at 4°C. The following day samples were spun gently at 2000 rpm for 10 minutes and then the 70% ethanol was gently aspirated off of the sample. To the pellet we added 1ml wash buffer (10% formamide and 2x SSC in nuclease free water), and let stand for 5 minutes. This was spun at 2000 rpm for 10 minutes and the wash buffer gently aspirated off of the sample. We added 150 µL Hybridization buffer (10% dextran sulfate, 10% formamide, 2x SSC in nuclease- free water) plus probe(1:500 diluted mcherry probe, Stellaris) to each sample. The hybridization reaction was incubated in a dark humidified chamber overnight at 37°C. The next morning, 1 mL of wash buffer (10% formamide and 2x SSC in nuclease free water) was added to each sample and incubated in a dark humidified chamber at 37°C for 30 minutes. This reaction was spun at 2000 rpm for 10 minutes and the wash buffer gently aspirated off the sample. The pellet was resuspended with 1 mL of wash buffer plus DAPI (10% formamide and 2x SSC and 5 ng/ml DAPI in nuclease free water) and incubated in a dark humidified chamber at 37°C for 30 minutes. This was spun at 2000 rpm for 10 minutes and the wash buffer gently aspirated and the pellet resuspended in 1 mL of 2x SSC, spun again at 2000 rpm for 10 minutes and then the 2x SSC was gently aspirated before resuspending the cell pellet in 1ml of anti-fade buffer without enzymes (0.4% glucose, 10 mMTris-HCl, 2x SSC in nuclease free water). This was incubated for 2 minutes at room temperature, spun at 2000 rpm for 10 minutes and the buffer removed by aspiration, then the cell pellet was resuspended in 100 µL of anti-fade buffer with enzymes (0.4% glucose, 10mM Tris-HCl, 2x SSC in nuclease free water, 1 µL of the glucose oxidase stock (3.7mg/ml) and 1 µL of mildly vortexed catalase suspension). Cells were transferred to a 96-well plate (Nunc #265300; treated with 0.1% poly-L-Lysine for 5 minutes and air dried for 30 minutes then washed with water 3x and air dried for 30 minutes) and incubated at room temperature for 30 minutes. The buffer was then gently aspirated away, leaving only cells affixed to the coverslip bottom. These cells were gently washed with PBS (3X) and 70% EtOH (1X). Mounting media (50µl of Prolong Gold, Invitrogen #P36934) was added to each well.
Imaging was done on an epifluorescence Nikon Eclipse-TI inverted controlled by Nikon Elements software. A Nikon Plan Apo VC 100x/1.40 oil objective was used to image cells and images were collected by a Clara CCD Camera (Andor DR328G; South Windsor, CT) camera. Z-stacks were taken every 0.5µm. Quasar® 570 was imaged at 605nm (70nm bandwidth) upon excitation at 545nm (25nm bandwidth; Chroma Technology 49004 ET-CY3 filter set). DAPI fluorescence was imaged at 460nm (50nm bandwidth) upon excitation at 350nm (50nm bandwidth; Chroma Technology ET-DAPI filter set). Image processing and spot detection were done using custom ImageJ and Matlab scripts which are available upon request. For each channel (Quasar 570 and DAPI) a maximum z-projection of the Z-stack was taken using ImageJ. The DAPI images were used to identify cell boundaries by using rolling ball background subtraction (r=50 pixels), followed by adjusting the threshold and creating a mask (a binary image where pixels inside cells are represented by a 1 and pixels outside of cells are represented as 0). Custom Matlab software then used this mask and the Quasar 570 projection to identify and count spots within each cells. Spots were counted by first using a bandpass filter (bpass) to subtract the background and a peak-finding algorithm (pkfnd) to identify the location of spots in the Quasar 570 projection. Spots within cell boundaries identified in the DAPI mask were then counted. The Matlab algorithms bpass and pkfnd are available here: http://physics.georgetown.edu/matlab/index.html
Blue-Light Stress Response Gene Expression
Blue light has been suggested to cause a stress response in Saccharomyces cerevisiae2. To show that our light induction was not causing a gene expression stress response, we exposed cells to blue-light at the intensity used in the protein concentration control experiments and assayed gene expression using a microarray. The parent strain of yMM1158 (used in the protein concentration control experiments), yMM389 (FY Mat a prototroph HAP1+), was grown up overnight to saturation and 1 ml of overnight saturated culture was inoculated into two 300 mL working volume chemostats filled with minimal phosphate limiting media, KH2PO4 monobasic 10 mg/L (73.5 μM final), dextrose 20.0 g/L (111 mM final), 1x vitamins, and 1x metals 11. Cultures were grown at room temperature (25°C) to saturation in the chemostats in complete darkness. Following saturation, cells grown at a dilution rate of 0.15h-1 were sampled for several days until cultures reached steady state at an OD600 of 0.56. One chemostat was kept in complete darkness and sampled for reference RNA. The other chemostat was exposed to blue-light at 460 nm output and 5.75 mW intensity and sampled at 0, 5 minute, 15 minute, 30 minute, 47 minute, 90 minute and 150 minute intervals. Samples were taken directly from the chemostats by filtration onto a membrane filter and immediately frozen in liquid nitrogen for later RNA preparation. RNA preparation and microarrays were as described in McIsaac, et al 20111. All microarray data is available on the Princeton University Microarray Database (PUMAdb).
Msn2 Localization in Response to Blue Light
Msn2 is a general stress-response transcription factor in Saccharomyces cerevisiae. Msn2 activity is controlled by nuclear localization, making Msn2 nuclear localization an indicator of cellular stress. Indeed, Msn2 has been shown to localize to the nucleus in response to blue-light associated stress2. To assay for Msn2 localization in response to blue light, yeast cells with Msn2 tagged with mCherry (yMM469 FY prototroph HAP1+ MSN2-mCherry-HphMX), were grown overnight to saturation in low fluorescence phosphate limited media (KH2PO4 monobasic 10 mg/L (73.5 μM), dextrose 20.0 g/L (111 mM), 1x low fluorescence vitamins, 1x metals1), without amino acids. Cells were diluted back the following day and grown at room temperature to mid-log phase (OD600 ~0.6) in batch. Samples were split into four separate chemostat vessels and allowed 1 hour to acclimate. Three of the vessels were exposed to 460 nm light at intensities of 5.75 mW, 2.74mW and 1.30 mW respectively. After 2 hours of exposure 900 µl of cells were sampled from each vessel and fixed in 100µl of formaldehyde for 15 minutes. We validated that this protocol maintains appropriate fluorescence and localization of the Msn2-mCherry construct by doing a fixation time course in cells exposed to a 1M sorbitol stress. Cells were centrifuged for 5 minutes at 9300Xg and the formaldehyde mixture was replaced by 200µl of 0.1M KHPO4 solution. Fluorescent images were then taken of each of the samples using the imaging setup described in the Online Methods. As a positive control cells from the culture left in complete darkness were heat shocked at 45°C for 10 minutes and fixed using the protocol above. Localization of Msn2-mCherry was observed in the positive control sample and no localization was observed in negative controls or in the sample kept in complete darkness.
Description of the Feedback Control Software
The software for this control apparatus was coded in a combination of Java, ImageJ, and the Arduino microcontroller language. Main control of the experiment was done in Java, visualized in Supplemental Figure 10. The main Java classes and methods are as follows:
The experiment is run by executing the Java class PumpControllerMain, which connects the computer to each device and then creates a Blue_Light_TC object to start the experiment. The Blue_Light_TC object extends the Timecourse class, and these two classes utilize the InductionLED, Image, and ImageRead Java classes to define the parameters of the experiment and run the data acquisition and analysis.
- Timecourse and Blue_Light_TC
The experiment is managed by the Timecourse method run. This method stores and updates everything happening outside of a single imaging and adjustment step. It maintains the experiment timer and progress (how many imaging events have happened). For each cycle, run waits for a specified flow time, then shuts off the pump. It waits a specified time for the cells to settle, then runs experiment_script to take images and process data. It then turns the pump back on, updates the experiment progress, and repeats until the desired number of cycles have been completed. The Blue_Light_TC method experiment_script manages each time step. It calls the Image class to take images of cells under different fluorescence, and subsequently calls the ImageRead class to analyze and quantify the fluorescence in the images for that cycle. The InductionLED class is then called to adjust blue light induction based on the fluorescence data at that time point and the desired expression profile.
- Image and ImagePlus
The Image class is used to take, store, and return images. This class communicates directly with the camera to set parameters, such as exposure and fluorescence, and to take an image. The image is then converted to the native ImageJ class ImagePlus, which contains a title parameter and is saved as a .tiff file.
The ImageRead class accepts an ImagePlus image as input and is used to analyze the fluorescence information in the image. The getPixelMean and getPixelStdev methods are used to calculate the mean and standard deviation of the pixel intensities over the whole image. Using this information a threshold intensity is defined at 3 standard deviations above the overall mean pixel intensity. ImageRead then calls two native ImageJ libraries, BackgroundSubtractor and ParticleAnalyzer. BackgroundSubtractor removes the background intensity using a rolling ball algorithm and creates a binary mask of pixels that meet the threshold. ParticleAnalyzer then uses this mask to organize each separate area of fluorescence over the threshold intensity into discrete “particles". The method scan_cell is then used to extract the mean fluorescent intensity of the pixels located in these particles in the original fluorescent image. The method is coded to ignore the five largest particles due to frequent instances of glare or other inconsistencies that would skew the data. This mean fluorescent intensity is then both stored in a text file for later review and passed to InductionLED.
The InductionLED class stores information about, and controls the status of, the blue LED used to induce expression. It stores all of the information for the desired expression profile, including the single or multiple target fluorescent intensities, the desired pulse duration (if the duration is not being determined by expression feedback), the current state of the LED, and the current cycle of the profile (if cyclic or oscillatory in nature). The actual bang-bang control scheme for achieving steady state is coded in the very simple method bangbang. This method accepts the latest mean fluorescent intensity as an input and adjusts the status of the blue light based on whether the current intensity is below the target (light turns on) or above the target (light turns off). The oscillate method can be called instead to generate periodic oscillations in expression. This method utilizes bangbang with a shifting target intensity. The oscillate method adjusts both the current target and the light status in response to the current mean fluorescent intensity. If the current expression target is the upper target and the current mean intensity is over that target, the target is switched to the lower intensity target. Alternatively, if the current expression target is the lower target and the current mean intensity is under that target, the target is switched to the higher intensity target. Once the check for this target has been done, bangbang is run using the new or unchanged target. In this way, from a starting intensity below the upper target, the light is turned on until expression reaches the upper target. The light is then turned off until expression falls below the lower target, at which time the process repeats. A limitation of this scheme, however, is that it can lead to some overshooting of the two targets.
The class ComOpener is used to communicate with the custom control board by sending serial commands to turn on or off the LEDs or pumps. This class contains the methods lightOn, lightOff, pumpOn, and pumpOff, which each contain a specific sequence of bits. Each of these methods calls the method write to output this sequence to the control board as a serial command.
Custom Control Board
The computer interfaces with the pump and the LED through an ATmega328p microcontroller running the Arduino boot loader. The board accepts serial commands from the Java software on the computer via an FTDI cable, and sets the appropriate pins on the microcontroller to HIGH (+5V) or LOW (GND) according to these commands. The states of these pins determine the on or off conditions of the microfluidic sampling pump, the blue induction LED, and several indicator LEDs used for troubleshooting.
The microcontroller program also runs a watchdog timer to ensure constant communication with the computer. Any time the computer tells the Arduino to turn the pump on or off, it trips a boolean flag to true, indicating that serial communication has been made. Every ten minutes the watchddog_helper method timer checks to confirm that this flag has been tripped, and resets the flag to false if it has. If it hasn’t, another flag is set to false that keeps the watchdog timer reset from occurring. If the watchdog timer is not reset every 8 seconds, the microcontroller reboots and resets. The watchddog_helper method allows for a longer watchdog timer, as the delays required for cell flow and settling increase the time between serial commands to far beyond the 8 seconds native to the watchdog timer.
The watchdog timer was introduced due to occasional random losses of communication between the computer and the microcontroller, which led to the loss data. When the timer reboots the microcontroller, it is sometimes the case that one image or cycle is lost.