Measuring mRNA copy-number in individual Escherichia coli cells using single-molecule fluorescent in situ hybridization (smFISH)

Abstract

We present a method for measuring the absolute number of mRNA molecules from a gene of interest in individual, chemically fixed Escherichia coli cells. A set of fluorescently-labeled oligonucleotide probes are hybridized to the target mRNA, so that each mRNA molecule is decorated by a known number of fluorescent dyes. Cells are then imaged using fluorescence microscopy. The number of target mRNA is estimated from the total intensity of fluorescent foci in the cell, rather than from counting discrete “spots” as in other currently available protocols. Image analysis is performed using an automated algorithm. The measured mRNA copy-number distribution obtained from many individual cells can be used to extract the parameters of stochastic gene activity, namely the frequency and size of transcription bursts from the gene of interest. The experimental procedure takes 2 days, with another 2-3 days typically required for image and data analysis.

INTRODUCTION

Development of the protocol

Fluorescent in situ hybridization (FISH) has been used to detect individual mRNA molecules of a gene of interest and measure their number in individual eukaryotic cells^1-3. This procedure is referred to as single molecule FISH (smFISH). Different variants of the method exist^4-6. In one particularly popular protocol, introduced in⁴, ~20-base long nucleotides are used as probes. Each probe is labeled with a single fluorescent dye molecule. A set of ~50 different probes are hybridized to the mRNA of interest. As a result, a single mRNA molecule produces enough signal to be easily detectable as a diffraction-limited “spot” under the fluorescence microscope. Counting these spots provides an estimate of mRNA copy-number in the cell^3,4,6,7.

We recently adapted smFISH for measuring the number of mRNA copies from a gene of interest in individual Escherichia coli cells^8,9. Examining the copy-number statistics in a population of cells then allows us to extract the parameters of stochastic gene activity, namely how often transcription bursts occur (burst frequency) and how many mRNA molecules are produced within each burst (burst size)^8-10. This procedure can be repeated for different endogenous genes, under different growth conditions and expression levels⁹. Our protocol is derived from the one of⁴ in terms of probe design and biochemical procedures. However, we diverge from other smFISH protocols^4,6 in two important aspects.

The first difference is that the estimation of mRNA number in the cell is not achieved by counting discrete spots, but instead relies on quantifying localized fluorescence. Due to the optical properties of a standard fluorescence microscope, a single mRNA molecule creates an image of size ~250 nm in the horizontal plane^11,12. Thus, two molecules that are closer than that distance will overlap each other and will not be detectable as separate spots. This distance is equivalent to a concentration of ~10 nM, or ~10 molecules in one E. coli cell. For comparison, the induced lactose promoter produces ~50 mRNA molecules per cell⁹. Thus, counting spots will not allow us to reliably measure mRNA levels for a highly expressed gene in E. coli because many of the apparent spots will consist of more than one mRNA. Our solution is instead to measure the number of bound probes, based on the total fluorescence intensity (photon flux) of the spots, without requiring that individual mRNAs appear as separate spots. By performing a calibration step, the total intensity of spots in the cell can then be converted to the number of target mRNAs. This procedure is inspired by the method we previously developed for counting mRNAs in live cells using the MS2-GFP labeling scheme^10,13. It involves the development of automated image and data analysis algorithms, as described below.

A second difference from most previous protocols^5,6,14 is that all biochemical steps (fixation, permeabilization, washes and hybridization) are performed in test tubes rather than on microscope slides. We reasoned that quantitative biochemical measurements require perfect mixing and uniformity of conditions. In contrast, cells attached to a slide are subject to nonuniform conditions, sometimes resulting in spatially-inhomogeneous labeling¹⁵. Uniformity is especially critical when aiming to accurately quantify cell-to-cell variability, as one must avoid increasing any experimental heterogeneity. We therefore developed the tube-based protocol presented here.

Applications of the method

The protocol allows measuring the absolute number of endogenous mRNA molecules from a gene of interest in individual E. coli cells. The dynamic range of the measurement is from <1 to ~100 molecules/cell⁹ (Fig. 1a-e). The estimated precision of the measurement is <1 (i.e. single-molecule resolution) at low mRNA levels (Fig. 1d). Under the assumption that the labeling and detection of individual mRNA molecules are statistically independent³, this translates to an error of a few percent at the higher end of the measurement range. While other factors may increase the measurement error, the low error estimation is supported by the good agreement that smFISH data shows with quantitative (“real time”) PCR (qPCR)^8,9 as well as with theoretical predictions^8,9. In a typical experiment, mRNA numbers from >1,000 individual cells are measured for a given sample, and the population mean $\langle n\rangle$ and variance σ² are estimated^8,9. The copy-number histogram is fitted to a simple model of transcription kinetics^2,16,17. The parameters of the fit are used to calculate the frequency and size of transcription bursts^8-10. We have previously used the procedure above to characterize the activity of 20 promoters under different growth conditions^8,9.

Figure 1. Experimental procedure.

(a) Image acquisition. Phase contrast and fluorescence images are acquired for each sample. Multiple focal planes (z positions) are imaged to allow precise fluorescence detection throughout the depth of the cell. (b) Cell and spot segmentation. The positions of individual cells and fluorescent foci (“spots”) are identified using custom MATLAB codes. Cell recognition uses as input the phase contrast images, and is performed using the Schnitzcells program (see MATERIALS, Equipment) [Copy Editor: is this correct?]. Fluorescent spots are identified from the stacks of fluorescence images, using the Spätzcells program developed in our lab. (c) Discarding false positives. False-positive spots, which are the result of probe binding to non-target RNA, are discarded following examination of histogram of peak height (spot intensity maximum) in a negative sample. (d) Identifying fluorescence intensity of one mRNA. The spot intensity corresponding to a single mRNA molecule is identified by examining the histogram of single spot intensities in a low-expression sample, where individual mRNAs are spatially separable. (e) Converting fluorescence intensity into mRNA numbers and extracting kinetic parameters. The one-mRNA intensity value is used to convert the total spot intensity in any cell to the number of target mRNA molecules. By measuring mRNA numbers in >1,000 cells per sample, the population mean and variance are estimated. The copy-number histogram is fitted to a simple model of transcription kinetics. The parameters of the fit are used to calculate the frequency and size of transcription bursts. Scale bars, 1 μm.

A number of variations and extensions to the application above are possible. The mRNA numbers for multiple genes can be measured simultaneously by using probe sets labeled with fluorophores of distinct spectra. We have used multicolor smFISH to examine the co-expression of two genes in E. coli⁸. Detecting 3 mRNA species simultaneously has been reported in eukaryotic systems⁴. This should also be achievable in bacteria following the same considerations regarding probe design and optical setup^4,6,12. Achieving an even higher degree of multi-gene combination would require more advanced spatial or spectral multiplexing methods^12,18.

In addition to counting mRNA, smFISH has been used in eukaryotes to examine the spatial organization of RNA molecules in the cell^6,19. A similar investigation can be performed in bacteria. For example, the site of active transcription from the gene can be distinguished from the individual, freely diffusing mRNAs based on the morphological properties of the corresponding fluorescent foci (not shown). Characterizing the positions of mRNA molecules in the cell can shed light on their spatiotemporal life history, a subject of ongoing debate^20-24.

Finally, we have successfully applied the image analysis component of this protocol—namely spot recognition and calibration of fluorescence to mRNA numbers—when performing smFISH in higher organisms using the protocol of Raj et al.⁴. The same image analysis algorithms can also be used, mutatis mutandis, for estimating mRNA numbers from live cell images where mRNA is labeled using the MS2 coat protein fused to a fluorescent protein^10,13,25.

Comparison with other methods

In terms of quantifying relative expression levels (i.e. comparing two samples), smFISH-based measurements show good agreement with the results of qPCR over ~3 orders of mRNA levels^8,9. The comparison between the two methods can be extended to absolute numbers by calibrating the qPCR data, using either external (in vitro transcribed RNA) or internal (e.g. ribosomal RNA) controls^10,26. However, we have not performed such a comparison. In higher organisms, smFISH data has also been compared to results obtained by RNA-sequencing (RNA-seq)²⁷, but again we are not aware of a similar comparison in bacteria.

In terms of single-cell measurements of mRNA numbers, two other smFISH protocols (both performed on microscope slides rather than in test tubes) have been reported in bacteria. Maamar et al.¹⁴ used smFISH to quantify mRNA levels in individual Bacillus subtilis cells. Taniguchi et al.⁵ labeled mRNA in E. coli using a single probe, carrying a single fluorophore, per mRNA target. In both these protocols, mRNA numbers were estimated by spot counting. This approach likely limits the measurement accuracy at high mRNA numbers. Consistent with this hypothesis, no mRNA levels higher than ~10 molecules per cell were reported in these studies. This is in contrast to our work, were such higher values were repeatedly measured⁹.

Finally, mRNA can also be labeled in individual live cells using a fusion of fluorescent protein to an RNA binding protein (such as the MS2 coat protein), and fusing an array of the corresponding binding sites to the gene of interest^20,28,29. We previously used this approach to follow the kinetics of mRNA production in E. coli¹⁰. This labeling scheme is highly perturbative to gene function¹⁰. Thus, direct comparison of expression level with genetically unperturbed cells is not possible. Nevertheless, the typical size of transcription bursts detected using live cell imaging is similar to that found using smFISH^8-10.

Experimental design

In terms of probe design and biochemical procedures, our protocol is directly based on that of⁴. A set of ~50 probes, each ~20 bases long, are designed against the transcript of interest (see REAGENT SETUP). The probes can be purchased pre-labeled, with a single fluorescent dye molecule on the 3’ end of each oligonucleotide. Alternatively, amine-modified oligos can be purchased and then fluorescently labeled in the lab and purified by ethanol precipitation (PROCEDURE Steps 1–11). We have successfully used as few as 48 and as many as 72 probes per gene.

Each experiment is performed using the strain of interest, a low-expression control sample and a negative control sample (i.e., a strain lacking the target mRNA). The cells to be studied are grown overnight (Step 12). The next day, cells are grown to mid log phase (Steps 13, 14), fixed (Step 16) and permeabilized (Step 17). The cells are then mixed with the labeled probes and hybridized overnight (Steps 19–23). The next morning, the cells are washed to remove nonhybridized probes (Steps 24–30), and finally resuspended in imaging buffer (Step 31). These steps are all performed in tubes, to guarantee that cells all are experiencing a uniform environment and to promote perfect mixing.

To acquire data, cells are placed between a coverslip and a thin agar slab (Fig. 2, Step 32) and imaged using both phase contrast and epifluorescence microscopy (Fig. 1a, Steps 33–34). Images are acquired using a high-quantum yield, cooled CCD camera. Multiple positions on the coverslip, providing data for >1,000 cells from each biological sample, are taken. Imaging is performed at multiple focal planes (z positions) to allow high resolution coverage of the cell depth (~1 μm).

Figure 2. Preparation and use of agarose pads.

(a) Stack five microscope slides on a leveled surface. (b) Pour the molten agarose solution onto the slides. (c) Cover the agarose with the remaining slide, placing a weight on top. Let solidify for 45 min at room temperature. (d) Remove the four slides from the sides of the agarose pad, leaving the top and bottom slides for easy storage and handling. (e) Remove the excess agarose from the slides with a razor blade. (f) For use in imaging, carefully move the slides exposing the agarose, and excise a 1×1 cm agar pad with a razor blade. (g) Pipette 2 μl of the cell suspension onto the center of a 24×50 mm coverslip. (h) Lay the agarose pad slowly on top of the cell suspension droplet with the razor blade. (i) Cover the pad with a 22×22 mm coverslip.

Next, images are analyzed using custom MATLAB codes, to obtain the positions of individual cells and fluorescent foci (Fig. 1b, Steps 35–37). Cell recognition uses as input the stacks of phase contrast images, and is performed using the Schnitzcells program³⁰. Fluorescent foci (“spots”) are identified from the stacks of fluorescence images, using the Spätzcells program, developed in our lab for that purpose.

The conversion of fluorescent foci intensity to mRNA numbers is achieved in a few steps (Fig. 1c–e and Steps 38–40). First, false-positive spots, which are the result of probe binding to non-target RNA, are discarded following examination of spot statistics in a negative control sample. Next, the spot intensity corresponding to a single mRNA molecule is identified by examining the histogram of spot intensities in a low-expression control sample, where individual mRNAs are spatially separable. Finally, this single-mRNA intensity value is used, in a sample of an unknown expression level, to convert the total spot intensity in each cell to the estimated number of target mRNAs.

By measuring mRNA numbers in >1,000 cells, the population mean and variance are estimated. The copy-number histogram is fitted to a simple theoretical model for stochastic gene activity (Fig. 1e, Step 41). The fitting parameters are used to estimate the rate and size of transcription bursts from the gene. mRNA half-life, which is required for this calculation, can be measured using standard methods^8,9.

Limitations

As described above, the purpose of the protocol is to obtain a precise estimate of the number of mRNA molecules from a gene of interest in individual cells, and to use copy-number statistics from a population of cells to extract the underlying parameters of stochastic gene activity. Achieving this goal can be hindered by a number of factors. First, the calibration of fluorescence intensity to the number of mRNA molecules requires the use of a completely negative (i.e. no mRNA of interest present) control sample in order to discard false positive spots, as well as a low-expression control sample, in which most mRNAs are discernible as individual spots. There may be cases where either of these controls is not available, for example if the gene of interest is essential and cannot be deleted. In those cases, calibration is harder to perform and may result in lower accuracy.

We also note that the accuracy of the measurement is estimated mainly using internal controls, namely by assessing the error in identifying the single-mRNA peak in the spot intensity histogram (Fig. 1d). Additional external controls are potentially very useful. In particular, comparing spot intensity to the fluorescence of individual probes in order to estimate the probe hybridization efficiency⁶ or comparing smFISH-based mRNA levels with the results of qPCR^8,9. However, at least in our hands these added measurements are more technically challenging than the smFISH measurements themselves and harder to quantitate, and thus are somewhat limited as standards against which to compare the smFISH data.

The estimation of gene activity parameters—frequency and size of transcription bursts—is performed using the mRNA copy-number histogram in a population of cells. The validity of this procedure depends on an assumption of steady-state, i.e. that mRNA production and degradation in the cells are balanced^2,8,9,16,17. This assumption is most easily fulfilled during exponential cell growth. In some cases, however, gene activity outside steady state is of interest. For example, the transient response to an external stimulus such as a drug^31,32. In that case, mRNA numbers can be measured at samples taken at different time points, with a temporal resolution of ~1–2 minutes (limited by sample handling times). The analysis of cell-to-cell variability in that case is more complicated than in the steady-state case¹⁷.

Another challenge in converting mRNA statistics to gene activity is the presence of multiple gene copies in the single cell. An E. coli cell growing in rich medium at 37 °C may contain up to 8 copies of a chromosomal gene³³, with different cells in the population having different copy-numbers. The observed number of mRNA molecules in a given cell reflects the combined stochastic activity of each of these gene copies. Neglecting such dosage effects will lead to distorted parameter estimation⁹. To avoid that, cells can be grown at a slow growth rate (>80 minutes generation time³⁴). Under such conditions, a chromosomal gene will only replicate once in the cell cycle, and cells having one versus two gene copies can be discriminated based on cell length⁹.

MATERIALS

CRITICAL Use RNase- and DNase-free materials whenever possible.

REAGENTS

• 6-carboxytetramethylrhodamine, succinimidyl ester (6-TAMRA; Life Technologies/Invitrogen, cat. no. C6123)

• Sodium hydroxide (Fisher Scientific, cat. no. BP359-500)

• Dimethylsulfoxide (DMSO; Fisher, cat. no. BP231-100)

• Sodium Bicarbonate (Fisher Scientific, cat. no. BP328-500)

• Sodium Acetate Anhydrous (Fisher Scientific, cat. no. BP333-500)

• Glacial acetic acid (Fisher Scientific, cat. no. BP2401-500) ! CAUTION Glacial acetic acid is extremely volatile and can cause severe burns if in direct contact with skin or eyes, or if inhaled. Handle with protective clothes under a fume hood.

• Ethanol (Decon Labs, cat. no. 2716) ! CAUTION Flammable material.

• Tris-EDTA (TE; 1×; Fisher Scientific, cat. no. BP2473-100)

• FISH DNA probes modified with a 3’ amine group, or alternatively, already labeled with a fluorescent dye (Biosearch Technologies). To study the lacZ mRNA, we designed 72 antisense probes. Their sequence, from 5’ to 3’, are the following (see REAGENT SETUP for design details):

GTGAATCCGTAATCATGGTC	TCACGACGTTGTAAAACGAC
ATTAAGTTGGGTAACGCCAG	TATTACGCCAGCTGGCGAAA
ATTCAGGCTGCGCAACTGTT	AAACCAGGCAAAGCGCCATT
AGTATCGGCCTCAGGAAGAT	AACCGTGCATCTGCCAGTTT
TAGGTCACGTTGGTGTAGAT	AATGTGAGCGAGTAACAACC
GTAGCCAGCTTTCATCAACA	AATAATTCGCGTCTGGCCTT
AGATGAAACGCCGAGTTAAC	AATTCAGACGGCAAACGACT
TTTCTCCGGCGCGTAAAAAT	ATCTTCCAGATAACTGCCGT
AACGAGACGTCACGGAAAAT	GCTGATTTGTGTAGTCGGTT
TTAAAGCGAGTGGCAACATG	AACTGTTACCCGTAGGTAGT
ATAATTTCACCGCCGAAAGG	TTTCGACGTTCAGACGTAGT
ATAGAGATTCGGGATTTCGG	TTCTGCTTCAATCAGCGTGC
ACCATTTTCAATCCGCACCT	TTAACGCCTCGAATCAGCAA
ATGCAGAGGATGATGCTCGT	TCTGCTCATCCATGACCTGA
TTCATCAGCAGGATATCCTG	CACGGCGTTAAAGTTGTTCT
TGGTTCGGATAATGCGAACA	TTCATCCACCACATACAGGC
TGCCGTGGGTTTCAATATTG	ATCGGTCAGACGATTCATTG
TGATCACACTCGGGTGATTA	ATACAGCGCGTCGTGATTAG
GATCGACAGATTTGATCCAG	AAATAATATCGGTGGCCGTG
TTTGATGGACCATTTCGGCA	TATTCGCAAAGGATCAGCGG
AAGACTGTTACCCATCGCGT	TGCCAGTATTTAGCGAAACC
AAACGGGGATACTGACGAAA	TAATCAGCGACTGATCCACC
GGGTTGCCGTTTTCATCATA	TCGGCGTATCGCCAAAATCA
TTCATACAGAACTGGCGATC	TGGTGTTTTGCTTCCGTCAG
ACGGAACTGGAAAAACTGCT	TATTCGCTGGTCACTTCGAT
GTTATCGCTATGACGGAACA	TTTACCTTGTGGAGCGACAT
GTTCAGGCAGTTCAATCAAC	TTGCACTACGCGTACTGTGA
AGCGTCACACTGAGGTTTTC	ATTTCGCTGGTGGTCAGATG
ACCCAGCTCGATGCAAAAAT	CGGTTAAATTGCCAACGCTT
CTGTGAAAGAAAGCCTGACT	GGCGTCAGCAGTTGTTTTTT
TACGCCAATGTCGTTATCCA	TAAGGTTTTCCCCTGATGCT
ATCAATCCGGTAGGTTTTCC	GTAATCGCCATTTGACCACT
AGTTTTCTTGCGGCCCTAAT	ATGTCTGACAATGGCAGATC
ATAATTCAATTCGCGCGTCC	TGATGTTGAACTGGAAGTCG
TCAGTTGCTGTTGACTGTAG	ATTCAGCCATGTGCCTTCTT
AATCCCCATATGGAAACCGT	AGACCAACTGGTAATGGTAG

• E. coli strains expressing the gene of interest, as well as a negative control strain where the gene of interest is deleted. For example, to study the expression of lacZ from the lac promoter (P_lac), we use the strain TK310 (ΔcyaA ΔcpdA ΔlacY³⁵). As a negative control, we use the strain BW14894 (ΔlacIZYA³⁶).

• Bacterial growth medium. Defined media are preferred over rich media (such as LB) because they typically provide more reproducible results. For example, to study the expression of lacZ from P_lac, we used a defined medium (M9CAgluc, Teknova, cat. no. M8010, composed of 1× M9 salts + 1% glucose + 0.1% casamino acids + 0.5 μg/ml thiamine + 0.2 mM magnesium sulfate + 0.1 mM calcium chloride).

• Chemical inducers. For example, to induce P_lac, we use Isopropyl β-D-1-thiogalactopyranoside (IPTG; Sigma, cat. no. 16758-1G) and Adenosine 3’-5’-cyclic monophosphate (cAMP; Sigma, cat. no. A9501)

• PBS, RNase-free (10×; Life Technologies/Ambion, cat. no. AM9625) SSC, RNase-free (20×; Life Technologies/Ambion, cat. no. AM9763)

• DEPC-treated water (Life Technologies/Ambion, cat. no. AM9922)

• BSA, RNase and DNase-free (50 mg/ml; Life Technologies/Ambion, cat. no. AM2616)

• Formamide, deionized, nuclease-free (Life Technologies/Ambion, cat. no. AM9342) ! CAUTION Formamide is highly toxic and a known teratogen. Avoid contact with the eyes or skin, inhalation or ingestion. Handle under a fume hood while wearing a lab coat and protective gloves.

• E. coli tRNA (Sigma, cat. no. R4251-500UN)

• Dextran Sulfate Sodium Salt (Sigma, cat. no. D8906-10G)

• Ribonucleoside Vanadyl Complex (VRC; 200 mM; New England Biolabs, cat. no. S1402S)

• Formaldehyde (37%, Fisher Scientific, cat. no. BP531-500) ! CAUTION Formaldehyde is highly toxic and a known carcinogen. Handle under a fume hood wearing a lab coat and protective gloves.

• 4',6-Diamidino-2-phenylindole, hydrochloride (DAPI; Fisher Scientific, cat. no. PI-46190)

• Low melt agarose, RNase and DNase-free (Fisher Scientific, cat. no. BP160-100)

EQUIPMENT

• Microcentrifuge tubes (National Scientific, 1.5 ml, cat. no. CN1700-BP; 2 ml, cat. no. CN2000-BP)

• PCR tubes (200 μl; Eppendorf, cat. no. 951010006)

• Table-top centrifuge (Eppendorf, cat. no. 5424)

• Vortex (Fisher Scientific, cat. no. 02-215-365)

• NanoDrop 2000 Spectrophotometer (Thermo Scientific)

• Aluminum foil (Reynolds wrap, Heavy Duty) •

Luer Lok syringes (10 ml; BD Biosciences, cat. no. 309604; 30-ml, cat. no. 309650)

• Syringe filters (0.22 μm; Millipore, cat. no. SLGP033RS)

• Kimwipe (VWR, cat. no. 82003-820)

• Parafilm M (VWR, cat. no. 52858-000)

• Falcon round-bottom polypropylene tubes (14 ml; BD Biosciences, cat. no. 352059)

• Orbital air shaker (Thermo Scientific, MaxQ4000, cat. no. SHKE4000). Alternatively, use Orbital water shaker (Thermo Scientific, MaxQ7000, cat. no. SHKE7000).

• Baffled shaker flasks (250 ml; Fisher Scientific, cat. no. 09-552-34)

• Motorized pipet filler (Fisherbrand, 03-692-164)

• Serological pipettes (Corning; 10 ml, cat. no. 4488; 50 ml, cat. no. 4490)

• Spectrophotometer (Bio-Rad, SmartSpec Plus Spectrophotometer, cat. no. 170-2525)

• Spectrophotometer cuvettes (BioRad, cat. no. 233-9955)

• Conical-bottom centrifuge polypropylene tubes (50 ml; Corning, cat. no. 430828)

• Centrifuge (Thermo Scientific, Sorvall Legend XTR, cat. no. 75004521)

• RNAase-free pipette tips (FisherBrand; 0.2–2.5 μl, cat. no 02-707-442; 10–100 μl cat. no. 02-707-431; 100–1250 μl, cat. no 02-707-404)

• Nutating mixer (VWR, cat. no. 82007-202)

• Water bath (Fisher Scientific, Isotemp, cat. no. 15-462-20)

• Incubator (Fischer Scientific, Isotemp, cat. no. 11-690-637-D)

• Microscope slides (Fisherbrand, cat. no. 12-550-A3)

• Microscope cover glass (Fisherbrand; 24×50-1, cat no. 12-544-E; 22×22-1, cat no. 12-541-B)

• Kimax-35 glass media lab bottle (100 ml; VWR, cat no. 16171-004)

• Plastic wrap (Saran)

• Razor blades (Garvey, no. 9, cat. no. 40475)

• Test weights (200 g, McMaster-Carr, cat no. 1777T28)

• Optical table (TMC, Breadboard, cat. no. 78-30551-01; 4-post support, cat. no. 63-36186-02)

• Wide-field inverted fluorescence microscope (Nikon, Eclipse-Ti)

• High numerical aperture oil immersion objective (Nikon, 100× Plan Apo 1.4 NA)

• Immersion oil A (Nikon, cat. no. MXA20233)

• Band-pass filters to separate fluorescence signals (for 6-TAMRA: Nikon, HC HISN Zero Shift, cat no. 96365; for DAPI: Nikon, UV-2E/C, cat. no. 96310)

• Mercury lamp (Nikon, Intensilight C-HGFIE)

• High sensitivity EMCCD camera (Photometrics, Cascade II 1024)

• Motorized optical shutter (Sutter Instruments, SmartShutter)

• Motorized stage controller (PRIOR, H31 ProScan III)

• Universal specimen holder (Nikon, cat. No. H473XR)

• Microscope management software (Nikon, NIS-Elements)

• MATLAB (The MathWorks)

• Schnitzcells program (Elowitz lab³⁰, http://cell.caltech.edu/schnitzcells/) [Copy Editor: is this referred to correctly?]

REAGENT SETUP

CRITICAL Use RNase- and DNase-free materials whenever possible.

Probe design and labeling

Probe design is based on the protocol described in^4,37. In brief, a set of 48 to 72 oligonucleotide probes is designed to bind to the target RNA. As guidelines for the design, we set the probes to be between 17 and 22 (typically 20) nucleotides long, keeping an inter-probe separation of at least 2 nucleotides and a GC-content as close as possible to 45%. To design such a set we use the algorithm Probe Designer, developed by Arjun Raj (available at http://www.singlemoleculefish.com/). To study lacZ expression, we designed 72 probes spanning the whole coding sequence of the lacZ gene (see REAGENTS). Custom probes can be ordered from several providers. We typically use Biosearch Technologies (Novato, CA, www.biosearchtech.com/). Each probe is ordered with a 3’ amine group, which allows covalent modification with NHS-ester derivatives of fluorescent dye molecules (for example 6-carboxytetramethylrhodamine, succinimidyl ester, see REAGENTS). Alternatively, the probes can be ordered pre-labeled. However, we have typically found in-house labeling to be cheaper (per experiment). We order 10 nmol per probe, purified using a reverse phase cartridge. We order our oligos in 96-well plates, diluted in ~100 μl water, giving a probe concentration of ~100 μM. Upon arrival, we let the probes thaw and then centrifuge the 96-well plate at 12,000g for 2 min at 4 °C. We then transfer each oligo solution to a 1.5 ml tube. The oligo solutions can be stored at –20 °C for several years.

Sodium hydroxide (1 M)

Dissolve 2 g of sodium hydroxide in 50 ml distilled water. Sterilize the solution by passing it through a 0.22 μm syringe filter. The solution can be stored at room temperature (~22 °C) for up to a year. ! CAUTION Sodium hydroxide is corrosive. Handle wearing a lab coat and protective gloves.

Sodium bicarbonate (1 M, pH 8.5)

Dissolve 4.2 g of sodium bicarbonate in 50 ml distilled water. Adjust the pH to 8.5 with 1 M sodium hydroxide. Sterilize the solution by passing it through a 0.22 μm filter. CRITICAL Prepare the solution fresh before use. The FISH probe labeling reaction is extremely sensitive to pH, and sodium bicarbonate solutions change their pH upon long storage. ! CAUTION Sodium hydroxide is corrosive. Handle wearing a lab coat and protective gloves.

Sodium bicarbonate (0.1 M, pH 9.0)

Dissolve 0.42 g of sodium bicarbonate in 50 ml distilled water. Adjust the pH to 9.0 with 1 M sodium hydroxide. Sterilize the solution by passing it through a 0.22 μm filter. CRITICAL Prepare the solution fresh before use. The FISH probe labeling reaction is extremely sensitive to pH, and sodium bicarbonate solutions change their pH upon long storage. ! CAUTION Sodium hydroxide is corrosive. Handle wearing a lab coat and protective gloves.

Sodium acetate (3 M, pH 5.2)

Dissolve 123 g of Sodium acetate in 400 ml distilled water. Adjust the pH to 5.2 with glacial acetic acid. Bring to a final volume of 500 ml. Aliquot into glass bottles (100 ml/bottle) and sterilize by autoclaving. The solutions can be stored at room temperature for up to a year.

IPTG (500 mM)

Dissolve 0.595 g of IPTG in 4 ml distilled water, and adjust the volume to 5 ml. Sterilize the solution by passing it through a 0.22 μm filter. Aliquot into 1.5 ml tubes (1 ml/tube). The solutions can be stored at –20 °C for up to a year.

DAPI (10 mg/ml)

Dissolve 10 mg DAPI in 1 ml deionized water. Aliquot into 200 μl PCR tubes (100 μl/tube) and wrap in aluminum foil. The solutions can be stored at –20 °C for up to a year.

Growth media

To study the expression of lacZ from P_lac, we use M9CAgluc medium with IPTG and cAMP as inducers. First, add 115 mg cAMP to 35 ml M9CAgluc medium to make a 10 mM cAMP M9CAgluc solution. To prepare the high-expression medium, combine 30 ml of this solution with 60 μl of 500 mM IPTG (final concentrations 10 mM cAMP and 1 mM IPTG). To make the low-expression medium, combine 30 ml M9CAgluc, 90 μl of 10 mM cAMP M9CAgluc solution and 60 μl of 500 mM IPTG (final concentration 0.03 M cAMP and 1 mM IPTG). Sterilize both solutions by passing them through a 0.22 μm filter. CRITICAL Prepare fresh before use.

Ribonucleoside Vanadyl Complex (VRC)

Before opening the bottle, reconstitute the solution by incubating in a 65 °C water bath for 10 minutes. Aliquot into 200 μl PCR tubes (100 μl/tube). The solutions can be stored at –20 °C for several years.

1× PBS, 3.7% (vol/vol) Formaldehyde

Combine 800 μl DEPC-treated water, 100 μl 10× PBS and 100 μl of 37% formaldehyde in a polypropylene tube. Scale up as necessary. ! CAUTIONFormaldehyde is highly toxic and a known carcinogen. Handle under a fume hood while wearing a lab coat and protective gloves and dispose according to safety regulations. CRITICAL Prepare fresh before use.

Formamide

After opening the bottle, aliquot into 2 ml tubes (~1.9 ml/tube, to keep the tube as air-free as possible) and wrap the caps with parafilm. The aliquots can be stored at –20 °C for several years. ! CAUTION Formamide is highly toxic and a known teratogen. Handle under a fume hood while wearing a lab coat and protective gloves.

Wash Solution (40% (w/vol) formamide, 2× SSC)

In a polypropylene tube, combine 100 μl 20× SSC, 353 μl formamide and 547 μl DEPC-treated water. Scale up as necessary. Wrap the cap with parafilm and store at 4 °C or on ice during the protocol. ! CAUTION Formamide is highly toxic and a known teratogen. Handle under a fume hood while wearing a lab coat and protective gloves. CRITICAL Prepare fresh before use.

Wash Solution with DAPI (40% (w/vol) formamide, 2× SSC, 10 μg/ml DAPI)

In a polypropylene tube, combine 100 μl 20× SSC, 353 μl formamide, 547 μl DEPC-treated water and 1 μl DAPI 10 mg/ml. Wrap the cap with parafilm and store at 4 °C or on ice during the protocol. ! CAUTION Formamide is highly toxic and a known teratogen. Handle under a fume hood while wearing a lab coat and protective gloves. CRITICAL Prepare fresh before use.

Hybridization Solution (40% (w/vol) formamide, 2× SSC)

This recipe is adapted from http://www.einstein.yu.edu/uploadedFiles/LABS/robert-singer-lab/mammalian_insitu.pdf, see also⁶. Add 5 ml of DEPC-treated water to a 50 ml polypropylene conical bottom tube. Add 1 g dextran sulfate to the water and dissolve by vigorous vortexing. Mix on an orbital shaker until bubbles disappear (~30 min). Meanwhile, take the formamide out of the freezer and bring it to room temperature before opening. Add to the solution 3530 μl formamide, 10 mg E. coli tRNA, 1ml 20× SSC and 100 μl of 200 mM VRC. Bring the final volume to 10 ml by adding DEPC-treated water and nutate the solution until it is homogeneous. Sterilize the solution by passing it through a 0.22 μm filter. Aliquot into 1.5 ml tubes (500 μl/tube). The solutions can be stored at – 20 °C for up to a year. ! CAUTION Formamide is highly toxic and a known teratogen. Handle under a fume hood while wearing a lab coat and protective gloves.

Agarose pads

Wash six microscope slides with 100% ethanol and rinse with distilled water. Dry the surface with a Kimwipe. Stack five slides on a leveled surface (Fig. 2a). Add 20 ml 1× PBS and 0.3 g of low-melt agarose to a 100 ml Kimax-35 bottle. Dissolve by heating in a microwave at low power for 5 min, swirling the solution every 1 min. Pour the molten agarose solution onto the slides (Fig. 2b). Cover the agarose with the remaining slide, placing a 200 g weight on top (Fig. 2c). Let solidify for 45 min at room temperature. Remove the four slides from the sides of the agarose pad, leaving the top and bottom slides for easy storage and handling (Fig. 2d). Remove the excess agarose from the slides with a razor blade (Fig. 2e). For use in imaging, carefully move the slides exposing 1 cm of the agarose pad, and excise a 1×1 cm agar pad with a razor blade (Fig. 2f). The slide-encased agarose pads can be wrapped in Saran wrap and stored at 4 °C for up to 24 h. CRITICAL Use slow heating to help dissolve the agarose without evaporating too much water, which may cause the agar to be viscous and hard to handle. Bubbles in the agarose pad can be avoided by placing the top slide with one edge first, then laying the slide down to cover the liquid agarose.

EQUIPMENT SETUP

Microscopy setup

For imaging, we use a conventional inverted epifluorescence microscope (Nikon, Eclipse Ti) equipped with a cooled EMCCD camera (Photometrics, Cascade II: 1024) and motorized stage control (Prior, Proscan III). A mercury lamp is used as the light source (Nikon, Intensilight C-HGFIE). A fast motorized optical shutter (Sutter Instruments, SmartShutter) is used to control the fluorescence illumination exposure time. Band-pass filter cubes (Nikon) are used for spectral separation. A 100× N.A. 1.40 oil immersion phase contrast objective (Nikon, Plan Apo 100× / 1.40 Oil) is used with an additional 2.5× lens in front of the camera. The specimen is mounted on a Universal specimen holder (Nikon, cat. no. H473XR). The microscope is installed on an optical table (TMC, Breadboard and 4-post support) to dampen mechanical vibrations. Microscope management software (Nikon, Elements) is used to control the microscopy setup.

Many alternatives exist to our setup above. When choosing a microscopy system for quantitative imaging, the most important feature to consider is photon collection efficiency. This can be achieved by combining a high quantum efficiency CCD camera, a strong light source and a high-NA (>1.3) objective. Alternative equipment we have used successfully with this protocol include: an Eclipse TE2000-E microscope (Nikon), Cascade 512 and Evolve 512 cameras (Photometrics), and the Metamorph software (Molecular Devices).

PROCEDURE

CRITICAL Use RNase- and DNase-free materials whenever possible.

Labeling FISH probes TIMING overnight

1 Take equal volumes of each of the 100 μM oligo solutions and pool in a 1.5 ml tube to give a final volume of 360 μl. Add 40 μl of 1 M sodium bicarbonate. Mix thoroughly by pipetting. For example, to label our lacZ probe set, we take 5 μl of each of the 72 oligo solutions to give a final volume of 360 μl.

2 Weigh 1 mg of succinimidyl-ester modified dye (for example, 6-TAMRA) in a 2 ml tube and dissolve in 2.5 μl DMSO. Add 25 μl 0.1 M sodium bicarbonate and mix thoroughly by pipetting.

CRITICAL STEP Succinimidyl-ester dyes are highly unstable upon dilution in DMSO. Dissolve the dyes fresh before use and continue with the next step immediately. Wrap the tube in aluminum foil to protect the solution from the light.

3 Add the oligo solution from Step 1 to the dye solution from Step 2. Mix thoroughly by pipetting. Wrap the tube in aluminum foil. Incubate in the dark overnight at 37 °C.

CRITICAL STEP The optimal pH for the probe labeling reaction is in the range 8.5–9.3. Using a buffer with a pH outside this range may decrease the probe labeling efficiency (^4,6 and manufacturer's instructions). If desired, a larger or smaller amount of probes can be labeled. In that case, scale the amount of reagents so that the final concentration of probes, dye, sodium bicarbonate and sodium acetate (see below) remains the same.

Ethanol precipitation TIMING 1 - 1.5 d

4 Add 47 μl of 3 M sodium acetate to the labeled oligo solution from Step 3. Mix thoroughly by pipetting.

5 Add 1180 μl of 100% ethanol. Mix thoroughly by pipetting. Place at –80 °C for at least 3 h and up to overnight.

6 Centrifuge at 15,000g for 30 min at room temperature in a table-top centrifuge. Decant the supernatant. Remove the remaining liquid using a Kimwipe, avoiding touching the pellet.

7 Dissolve the pellet in 45 μl DEPC-treated water. Add 5 μl of 3 M sodium acetate. Mix thoroughly by pipetting.

8 Add 125 μl of 100% ethanol. Mix thoroughly by pipetting. Place at –80 °C for at least 3 h and up to overnight.

9 Repeat Steps 6 to 8 once more.

10 Resuspend the pellet in a total of 250 μl 1× TE to make the 10× probe stock solution. Transfer 50 μl of this solution to a 1.5 ml tube and add 450 μl 1× TE to make the 1× probe stock solution. Aliquot the 1× stock into 1.5 ml tubes (50 μl/tube) and wrap the 10× and 1× stocks in aluminum foil. The stock solutions can be stored at –20 °C for several years.

CRITICAL STEP Aliquoting small volumes (50 μl/tube) of the 1× stock helps avoiding probe degradation caused by repeated freeze-thaw cycles. Keep the tubes on ice while using the stocks.

Measuring probe labeling efficiency TIMING 30 min

11 The probe labeling efficiency (LE) is defined as LE = [dye (μM]/[probe (μM]. With a spectrophotometer, measure and [DNA (μg/ml)] and [dye (μM)] for the 1× probe stock solution. Then, calculate [probe (μM)] using the formula [probe (μM)] = (1,0000/ MW_DNA) * [DNA (μg/ml)], where MW_DNA is the approximate molecular weight of a single stranded DNA, given by MW_DNA = #nucleotides * 303.7 (g/mol). To make these measurements, we use the NanoDrop 2000 under the “Microarray” application, which allows the simultaneous measurement of [DNA (μg/ml)] and [dye (μM)]. The DNA concentration of the 1× stock solution should be 10–16 μM. We typically obtain probe labeling efficiencies higher than 90%. Independent measurements of the probe labeling efficiency using HPLC yield similar results, so we generally do not HPLC-purify our probes. However, if significantly lower probe labeling efficiencies are obtained, HPLC purification is advisable^37,38.

? TROUBLESHOOTING

Growing cell cultures TIMING ~18–22 h

12 Add 2 ml of defined medium to two 14 ml polypropylene falcon tubes. Inoculate the cultures with single colonies of E. coli strains to be used in the experiment. Grow the cultures overnight (12–16 h) in an orbital shaker at 265 rpm and at 37 °C.

13 The next morning, for each of the samples, add 30 ml of defined medium to a 250 ml baffled shaker flask. To study P_lac, we use at least three samples: the negative control strain BW14894 grown in M9CAgluc medium, the positive strain TK310 grown in low-expression medium, and the positive strain TK310 grown in high-expression medium (see REAGENT SETUP for growth media).

14 Dilute the overnight cultures 1:1,000 into their respective growth medium. Begin growing the overday cultures in an orbital shaker at 265 rpm and at 37 °C.

CRITICAL STEP It is advisable to measure the growth rate of the bacterial strains before performing the smFISH experiment. Depending on the growth rate, dilute the strains between 1:20,000 and 1:250 so that they reach OD₆₀₀ ~0.2 at approximately the same time. To study P_lac, we dilute the overnight cultures of BW14894 by 1:20,000 and TK310 by 1:250. The cultures take 5–6 h to reach OD₆₀₀ ~0.2.

Fixation and permeabilization TIMING 2 h

15 ~1 h before harvesting the cells, prepare an ice-water bath and pre-chill 50 ml centrifuge tubes for each of the cultures.

16 Fixation of cells can be performed by either harvesting and resuspending them in formaldehyde (“standard method”, option A below), or by adding formaldehyde directly to the cell culture (“direct method”, option B below). We generally use the standard method, but the direct method may be used if one wishes to measure mRNA levels over multiple time points, or if one wishes to increase the throughput of the protocol.

(A) Standard method.

(i) Measure the OD₆₀₀ of the overday cultures in a spectrophotometer every ~1 h. When the OD₆₀₀ is 0.2–0.4, harvest a volume of culture having the same number of cells as 15 ml of culture at OD₆₀₀ = 0.4 (this volume can be calculated as ice-cold V (ml) = 6/OD₆₀₀). Transfer the harvested culture to an50 ml centrifuge tube. Keep the tubes in the ice-water bath while harvesting the cultures. Centrifuge the tubes at 4,500g for 5 min at 4 °C.

CRITICAL STEP It is important that all samples have approximately the same number of cells at the time of harvesting. This is achieved by adjusting volume of harvested culture to compensate for differences in OD₆₀₀.

(ii) Decant the supernatant and tap the inverted tubes on paper towels to remove the remaining liquid. Resuspend each pellet in 1 ml of ice-cold 1× PBS, 3.7% formaldehyde. Transfer to 1.5 ml tubes. Mix gently for 30 min at room temperature using a nutator. ! CAUTION Formaldehyde is highly toxic and a known carcinogen. Handle under a fume hood while wearing a lab coat and protective gloves and dispose according to proper safety and environmental regulations.

(iii) Centrifuge the cells at 400g for 8 min at room temperature. Pipette out the supernatant and discard it.

(iv) Wash the cells twice in 1 ml 1× PBS. Each time add 1 ml 1× PBS, resuspend by pipetting, centrifuge at 600g for 3.5 min at room temperature, pipette out the supernatant and discard it.

(B) Direct method

(i) Measure the OD₆₀₀ of the overday cultures in a spectrophotometer every ~1 h. When the OD₆₀₀ is 0.2–0.4, add 37% formaldehyde directly to the cell culture, to a final concentration of 3.7% formaldehyde. Transfer a volume of culture having the same number of cells as 15 ml of culture at OD600 = 0.4 (this volume can be calculated as V (ml) = 6/OD₆₀₀ to a 50ml polypropylene tube and incubate for 30 minutes at room temperature using a nutator. ! CAUTION Formaldehyde is highly toxic and a known carcinogen. Handle under a fume hood while wearing a lab coat and protective gloves and dispose according to proper safety and environmental regulations.

(ii) Centrifuge the cells at 400g for 8 min at room temperature. Pipette out the supernatant and discard it.

CRITICAL STEP The pellet may be very loose. Be careful when pipetting out the supernatant.

(iii) Resuspend each pellet in 1 ml 1× PBS. Transfer to 1.5 ml tubes. Centrifuge at 600g for 3.5 min at room temperature. Pipette out the supernatant and discard it.

(iv) Add 1 ml 1× PBS, resuspend by pipetting, centrifuge at 600g for 3.5 min at room temperature, pipette out the supernatant and discard it.

17 Permeabilize the cells. Resuspend the pellet in 300 μl DEPC-treated water. Add 700 μl 100% ethanol and mix thoroughly by pipetting. Mix gently for 1 h at room temperature using a nutator.

CRITICAL STEP Dissolving the pellet in water before adding ethanol prevents the formation of cell aggregates.

PAUSE POINT After mixing at room temperature for 1 h, the cells can be kept in 70% ethanol at 4 °C for up to 7 days without affecting the results.

18 While the cells are being permeabilized in ethanol, prepare 40% wash solution. Keep on ice.

Hybridization TIMING overnight

19 Centrifuge the cells at 600g for 7 min at room temperature. Carefully, pipette out the supernatant and discard it.

CRITICAL STEP The pellet may be very loose, or may be spread over the tube wall. Be careful when pipetting out the supernatant.

20 Resuspend the pellet in 1 ml 40% wash solution. Mix gently for 5 min at room temperature using a nutator.

21 In a new 1.5 ml tube, add 50 μl per sample of 40% hybridization solution. Add 1× probe stock solution to a final concentration of 1 μM. For example, if the concentration of the 1× stock is 15 μM, add 3.6 μl per sample. Mix thoroughly by pipetting, avoiding bubbles.

CRITICAL STEP For multicolor smFISH, add 1× stock solutions of each probe set to the hybridization solution, so that the final concentration of each probe set is 1 μM.

22 Centrifuge the cells at 600g for 7 min at room temperature. Pipette out the supernatant and discard it.

23 Add 50 μl of the hybridization solution with probes (Step 21) to the cell pellet. Resuspend the cells thoroughly by pipetting, avoiding bubbles. Incubate overnight (~14 h) at 30 °C.

CRITICAL STEP The formamide concentration of the hybridization- and wash- solutions is the main parameter that controls the stringency of probe binding. Low formamide concentration (low stringency) favors non-specific binding of probes. Increasing the formamide concentration (high stringency) reduces the non-specific binding, but eventually also reduces the binding of probes to the target mRNA. Our protocol has been optimized for oligos that are 20 nucleotides-long with a GC content close to 45%. If the probe set departs from that standard, adjust the stringency of the hybridization and wash solutions by changing the concentration of formamide.

PAUSE POINT Cells in the hybridization solution can be stored at 4 °C for months and the experiment resumed from this point without visible decrease of signal quality.

Washing TIMING 2 h

24 Transfer 10 μl of the samples in hybridization reaction to a 1.5 ml tube. In the case that additional imaging sessions are needed, resume the protocol from this step using 10 μl of the remaining hybridization reaction.

25 Add 200 μl of 40% wash solution to the tube. Mix thoroughly by pipetting. Centrifuge at 600g for 3.5 min at room temperature. Pipette out the supernatant and discard it.

26 Add 200 μl of 40% wash solution. Mix thoroughly by pipetting. Incubate for 30 min at 30 °C.

27 Mix thoroughly by pipetting. Centrifuge at 600g for 3.5 min at room temperature. Pipette out the supernatant and discard it.

28 Repeat Steps 26-27 once more. While the cells are being washed, prepare the agarose gel for imaging (see REAGENT SETUP), turn on the microscope, and prepare the 40% wash solution with 10 μg/ml DAPI.

29 Add to the samples 200 μl of 40% wash solution with 10 μg/ml DAPI. Mix thoroughly by pipetting. Incubate for 30 min at 30 °C.

30 Mix thoroughly by pipetting. Centrifuge at 600g for 3.5 min at room temperature. Pipette out the supernatant and discard it.

31 Resuspend the cells in 10 μl of 2× SSC.

CRITICAL STEP Pipet the cell suspension up and down multiple times to separate any cell aggregates before assembling the slides for imaging.

Microscopy setup and recording TIMING 2–4 h

32 Assemble the slides for imaging. For each sample to be imaged, take a 24×50 mm coverslip and pipette 2 μl of the cell suspension onto the center of the coverslip (Fig. 2g). Cut a 1×1 cm agarose pad with a razor blade (Fig. 2f). Lift the agarose pad from one corner with the blade and lay the pad slowly on top of the cell suspension droplet (Fig. 2h). Cover the pad with a 22×22 mm coverslip (Fig. 2i).

CRITICAL STEP Be sure to use coverslips of thickness #1. Other commonly used coverslips (e.g. thickness #1.5) show a higher fluorescent background. We note that other smFISH protocols call for adding anti-fading reagents to the imaged sample^4,6.

33 Set the imaging parameters. To characterize the optimal imaging conditions for your experimental setup, first image the sample with the highest expression level. Find the best focal plane (z-position) in the phase contrast channel, and acquire images from the Fluorescence channel. Repeat this procedure for different exposure times and stage positions (fields of view), recording the maximum pixel value of the fluorescent foci. We have found that a good rule is to choose exposure times that produce foci pixel values no higher than 60% of the maximum pixel value of the camera (maximum value of 65,535 for a 16 bit camera). For our imaging setup, exposures in the range of 250–500 ms fulfill this criterion. If the signal is still low, we also increase the electromutiplier gain of the CCD camera, as longer exposure times may lead to increased photobleaching. Additionally, check that the chosen conditions allow the visualization of dim foci (coming from a single probe or a small set of overlapping probes) in the negative control sample. Repeat the same procedure for the DAPI channel.

CRITICAL STEP For a multi-color smFISH experiment, the filters used have to be optimized to minimize the crosstalk between channels. To check for the presence of crosstalk, image single-color labeled samples under all the different filter sets.

? TROUBLESHOOTING

34 Acquire the images at different focal planes (z-slices) for all channels (Fig. 1a). We typically image the cells in 9 successive z-slices at 200 nm spacing. Repeat the procedure for different stage positions (fields of view) either manually or by using an automatic acquisition mode if available. Acquire enough images so that ~1,000 cells are imaged per sample, typically 10–40 image positions.

CRITICAL STEP A systematic scanning of the slide prevents imaging areas that have been illuminated previously. Check that the illuminated area is only slightly bigger than the field of view by changing the aperture of the fluorescent light diaphragm. Avoid areas where the cells are too dense, because cell recognition routines work better on images in which the cells are not in close contact.

? TROUBLESHOOTING

Data analysis TIMING 2–3 d

35 Convert the data stacks to TIFF format. Microscope controller software such as Nikon Elements or Metamorph save each field of view as a multidimensional data stack (.nd or .nd2 format). Convert these image stacks into standalone .tif images by using the built-in utilities inside the microscope controller software.

36 Perform automatic cell segmentation. We have used both Schnitzcells³⁰ and MicrobeTracker³⁹ software, but any program that generates cell segmentation masks (label matrix, readable by MATLAB, where non-cell pixels have a value of zero and cell pixels have integer values corresponding to the cell identification number (Fig. 1b)) from phase contrast images could be used.

37 Perform automatic spot recognition using the Spätzcells program (Fig. 1b). We developed Spätzcells to identify and measure the properties of fluorescent foci (“spots”) across multiple focal planes in image stacks. A copy of the program and the accompanying documentation are available upon request. The algorithm works as follows: Spätzcells first identifies 2D local maxima of fluorescence intensity, with height above a predefined “spot detection threshold”. These maxima are then classified as “spots” only if they appear in multiple adjacent image planes (z positions). Finally, for each spot, the fluorescence intensity profile (at the focal plane where the spot is in focus) is fitted to a 2D-Gaussian function, and features such as the position, peak height (amplitude of the Gaussian fit), and the integrated fluorescence intensity, are recorded. In the case that other spots are present in the vicinity of the spot being fitted, a 2D-multi-Gaussian fit is performed.

CRITICAL STEP To accurately characterize mRNA spots, the spot detection threshold should be set low enough such that some false positive spots are recognized in the negative control sample (~1 spots/cell) and all spots are recognized in the positive sample.

? TROUBLESHOOTING

38 Select a “false positive threshold” to discard the spots resulting from non-specific binding of probes. The low spot detection threshold used to recognize spots in the previous step, ensures that all genuine spots (i.e. spots corresponding to target mRNA) are recognized, at the price of increasing the number of false positives. To discard such false positives, compare the peak height distributions of the spots in the negative control sample to the ones in the low-expression control sample. Select a “false positive threshold” in peak height that separates the population of false positives from the population of genuine spots in the low-expression control sample (Fig. 1c). Spots with peak heights lower than this false positive threshold are discarded from the subsequent analysis of all samples.

39 Define the fluorescence intensity of a single mRNA. After discarding the false positives, use the remaining spots in the low-expression control sample to construct a spot intensity histogram. This histogram should show a predominant species, corresponding to a single mRNA molecule (Fig. 1d). Use MATLAB, or another data analysis software, to fit the complete histogram to a sum of Gaussians with increasing peak positions and decreasing peak heights, corresponding to one, two, three, etc. mRNA molecules per spot. Each Gaussian in this sum has a mean that is an integer multiple of the first Gaussian and a variance that scales with the mean, reflecting the statistical independence of labeling and detecting individual mRNAs³. Estimate the single-mRNA intensity as the mean of the first Gaussian.

40 Convert the total fluorescence of the spots inside each cell into the number of mRNA molecules in the cell. Sum the spot intensities from all the spots in a cell and divide this number by the intensity of a single mRNA molecule. Round to the closest integer.

41 Use the population statistics to estimate transcription kinetics parameters. We model transcription as a two-state process: the gene switches on and off, producing multiple mRNAs (a “burst”) during the ‘on’ period. Use MATLAB, or another data analysis software, to fit the histogram of mRNA copy-numbers per cell to a negative binomial distribution (predicted by the two state model⁴⁰) (Fig. 1e):

$$P\left(n\right)=\left(\begin{array}{c}\hfill n+r-1\hfill \\ \hfill n\hfill \end{array}\right)p^r{(1-p)}^n$$

Where P(n) is the probability of observing n mRNAs in a cell. r and p are fitting parameters. The values of r and p can be used to estimate the frequency, f, and size, b, of transcription bursts as follows: f = r/τ_RNA, where τ_mRNA lifetime (measured separately using standard methods^8,9), and b = (1 – p)/p. An alternative method for estimating f and b is using the relations $b={\sigma }^2∕\langle n\rangle$ and $f={\langle n\rangle }^2∕\left(({\sigma }^2-\langle n\rangle ){\tau }_{RNA}\right)$, where $\langle n\rangle$ and σ² are the mean and variance of the mRNA copy number in the sample. The two procedures typically yield very similar results⁹ (for more details of these mathematical relations, see ^8,9).

CRITICAL STEP Using a negative binomial distribution to fit the mRNA copy-number histogram relies on assuming that the ‘on’ duration is much shorter than the mRNA lifetime. A more general expression for the mRNA copy-number distribution under the two-state model has been derived analytically^2,17. However, for the majority of cases we have studied, we find that a negative binomial distribution fits the data very well⁹.

TIMING

Steps 1–3, labeling FISH probes: overnight

Steps 4–10, ethanol precipitation: 1–1.5 d

Step 11, measuring labeling efficiency: 30 min

Steps 12–14, growing cell cultures: ~18–22 h

Steps 15–18, fixation and permeabilization: 2 h

Steps 19–23, hybridization: overnight

Steps 24–31, washing: 2 h

Steps 32–34, microscopy setup and recording: 2–4 h

Steps 35–41, data analysis: 2–3 d (depends on user's proficiency)

ANTICIPATED RESULTS

Single-molecule FISH (smFISH), followed by automated image analysis, allows the quantification of mRNA copy-numbers from a gene of interest in individual E. coli cells. Figure 1 shows the experimental workflow and representative data sets for the case of the endogenous lacZ gene in E. coli. Cell cultures of a strain lacking the lac operon (BW14894, negative control³⁶) and strains expressing lacZ at low and high levels (TK31035 grown with varying amounts of IPTG and cAMP) were fixed, permeabilized, hybridized with 6-TAMRA-labeled probes, washed and imaged according to our protocol described above. When the fluorescence images are inspected closely, it can be seen that each sample contains spots of distinct brightness and size (Fig. 3). The negative control sample (Fig. 3, left columns) shows dim “background” spots, corresponding to probes non-specifically bound in the cell. The low-expression control sample (Fig. 3, middle columns) shows brighter, well defined spots that correspond to single mRNAs, as well as background spots. The high-expression sample (Fig. 3, right columns) shows spots that are brighter and more variable than those visible in the low-expression control sample, reflecting multiple overlapping mRNA molecules. In the high-expression sample, background spots are typically not visible.

Figure 3. Different smFISH samples contain spots of distinct brightness and size.

Fluorescence channel images from three experimental samples (from left to right: negative control sample, low-expression control sample, high expression sample) are displayed at varying contrast levels (rows 2-4). The phase contrast and fluorescence images were taken at a single focal plane (z position). Scale bars, 2 μm. See “ANTICIPATED RESULTS” for a discussion of the observed features.

The properties of individual fluorescent spots are estimated using the Spätzcells program (Fig. 1b, right). We have found that the most robust feature for differentiating bona fide mRNA spots from false positives is their peak height (amplitude of the Gaussian fitted by Spätzcells). A low-expression control sample typically shows a peak-height distribution with two well-separated spot populations (Fig. 1c, red line). The values in the low-intensity population correspond closely to values found in the negative control sample (Fig. 1c, black line). A false positive threshold in spot peak height is then chosen so as to separate the two spot populations in the low-expression control sample. Thereafter in every sample examined, spots with peak heights lower than this threshold (Fig. 1c, shaded region) are discarded from the analysis.

After discarding false positive spots, the spot intensity (intensity integrated over the area of the Gaussian fitted by Spätzcells) corresponding to a single mRNA molecule is identified by examining the histogram of spot intensities in a low-expression control sample, where individual mRNAs are spatially separable (Fig. 1d). We have found that the spot intensity histogram for samples with a mean of < ~ 3 mRNA per cell shows a well-defined single mRNA peak, which can be successfully used for calibration purposes. In some cases, the peaks corresponding to two, three, etc. mRNAs can also be seen in the histogram (Fig. 1d).

The measured single mRNA intensity value is next used to convert the total intensity of all spots in a given cell to the number of target mRNA molecules in that cell. By measuring mRNA numbers in >1,000 cells, the population mean and variance are estimated (Fig. 1e). The copy-number histogram is fitted to a simple model of transcription kinetics^8,9. The parameters of the fit are used to calculate the frequency and size of transcription bursts (Fig. 1e). Using this analysis, the negative control sample typically shows a mean level and standard deviation of less than one mRNA. The estimated burst size and frequency for that sample (here, 0.4 mRNA and 0.1 min^–1 respectively) are indicative of the accuracy with which those parameters can be calculated in higher-expression samples. For the experiment depicted in Fig. 1e, the low-expression control sample has ~3 mRNA per cell (with burst size and frequency of 1.9±0.1 mRNA and 0.7±0.2 min^–1 respectively), while the same strain in high-expression conditions shows a mean level of ~50 mRNA per cell (with burst size and frequency of 18.8±1.3 mRNA and 1.3±0.2 min^–1 respectively).

GTGAATCCGTAATCATGGTC	TCACGACGTTGTAAAACGAC
ATTAAGTTGGGTAACGCCAG	TATTACGCCAGCTGGCGAAA
ATTCAGGCTGCGCAACTGTT	AAACCAGGCAAAGCGCCATT
AGTATCGGCCTCAGGAAGAT	AACCGTGCATCTGCCAGTTT
TAGGTCACGTTGGTGTAGAT	AATGTGAGCGAGTAACAACC
GTAGCCAGCTTTCATCAACA	AATAATTCGCGTCTGGCCTT
AGATGAAACGCCGAGTTAAC	AATTCAGACGGCAAACGACT
TTTCTCCGGCGCGTAAAAAT	ATCTTCCAGATAACTGCCGT
AACGAGACGTCACGGAAAAT	GCTGATTTGTGTAGTCGGTT
TTAAAGCGAGTGGCAACATG	AACTGTTACCCGTAGGTAGT
ATAATTTCACCGCCGAAAGG	TTTCGACGTTCAGACGTAGT
ATAGAGATTCGGGATTTCGG	TTCTGCTTCAATCAGCGTGC
ACCATTTTCAATCCGCACCT	TTAACGCCTCGAATCAGCAA
ATGCAGAGGATGATGCTCGT	TCTGCTCATCCATGACCTGA
TTCATCAGCAGGATATCCTG	CACGGCGTTAAAGTTGTTCT
TGGTTCGGATAATGCGAACA	TTCATCCACCACATACAGGC
TGCCGTGGGTTTCAATATTG	ATCGGTCAGACGATTCATTG
TGATCACACTCGGGTGATTA	ATACAGCGCGTCGTGATTAG
GATCGACAGATTTGATCCAG	AAATAATATCGGTGGCCGTG
TTTGATGGACCATTTCGGCA	TATTCGCAAAGGATCAGCGG
AAGACTGTTACCCATCGCGT	TGCCAGTATTTAGCGAAACC
AAACGGGGATACTGACGAAA	TAATCAGCGACTGATCCACC
GGGTTGCCGTTTTCATCATA	TCGGCGTATCGCCAAAATCA
TTCATACAGAACTGGCGATC	TGGTGTTTTGCTTCCGTCAG
ACGGAACTGGAAAAACTGCT	TATTCGCTGGTCACTTCGAT
GTTATCGCTATGACGGAACA	TTTACCTTGTGGAGCGACAT
GTTCAGGCAGTTCAATCAAC	TTGCACTACGCGTACTGTGA
AGCGTCACACTGAGGTTTTC	ATTTCGCTGGTGGTCAGATG
ACCCAGCTCGATGCAAAAAT	CGGTTAAATTGCCAACGCTT
CTGTGAAAGAAAGCCTGACT	GGCGTCAGCAGTTGTTTTTT
TACGCCAATGTCGTTATCCA	TAAGGTTTTCCCCTGATGCT
ATCAATCCGGTAGGTTTTCC	GTAATCGCCATTTGACCACT
AGTTTTCTTGCGGCCCTAAT	ATGTCTGACAATGGCAGATC
ATAATTCAATTCGCGCGTCC	TGATGTTGAACTGGAAGTCG
TCAGTTGCTGTTGACTGTAG	ATTCAGCCATGTGCCTTCTT
AATCCCCATATGGAAACCGT	AGACCAACTGGTAATGGTAG