Visualisation of PCNA Monoubiquitination In Vivo by Single Pass Spectral Imaging FRET Microscopy

Monoubiquitination of the DNA sliding clamp, PCNA, plays a central role in the control of damage bypass during replication. By combining a widely-spaced FRET donor/acceptor pair (CFP and mRFP) with spectral imaging, we have developed a simple method for the visualisation of PCNA monoubiquitination in both fixed and live cells with a single imaging pass. We validate the method with genetic controls in the avian cell line DT40 and use it to examine the intracellular dynamics of PCNA ubiquitination following subnuclear UV irradiation. This general approach is likely to be of utility for live imaging of post-translational modifications of a wide range of substrates in vivo.


Introduction
The monoubiquitination of the DNA sliding clamp, PCNA, plays a central role in the control of translesion synthesis in eukaryotes [1,2,3,4]. Conjugation of ubiquitin to lysine 164 of PCNA is induced by the arrest of replication by DNA damage. It is mediated by the E3 ubiquitin ligase RAD18 in conjunction with the E2 ubiquitin conjugating enzyme RAD6, although in higher eukaryotes there is evidence that other enzymes can be used as well [5,6]. The presence of ubiquitin on PCNA increases the affinity of the clamp for members of the Y-family of specialised DNA polymerases, which contain ubiquitin binding motifs, that are able to directly bypass DNA damage in a process known as translesion synthesis [7]. A single ubiquitin on PCNA can also act as a 'seed' for the formation of a non-canonical K63-linked polyubiquitin chain mediated by the E3 ligase RAD5 acting in concert with UBC13/MMS2 in yeast [2], or by the RAD5 homologues SHPRH/HLTF in vertebrates [8,9,10,11]. Formation of these chains is linked to a recombinational mode of bypass, sometimes termed template switching, that may also employ the helicase activity of RAD5 [12].
Ubiquitin can be removed from PCNA by the ubiquitin-specific protease USP1 [13], although blockade of this step does not appear to have a significant impact on the DNA damage response [14]. Further, it appears from Western blotting studies in human cells that ubiquitin-PCNA conjugates persist on chromatin for many hours after the removal of DNA damage, suggesting a model in which PCNA ubiquitination acts to promote bypass behind the replication fork [15]. In agreement with this, our own studies in the chicken cell line DT40 have shown that PCNA ubiquitination is required for post-replicative gap filling, but not for maintenance of fork progression on a damaged DNA template [16]. However, PCNA ubiquitination in DT40 appears to be rather more transient than in the human cell lines studied. DNA damage results in only a modest accumulation of conjugates as monitored by Western blotting unless USP1 is disrupted, in which case much higher steady state levels are observed [5,14].
Western blotting is currently the only method available for monitoring PCNA ubiquitination and is, by its very nature, rather crude as it only provides an indication of the average level of the modification in a large number of cells. Much has been learned about the subcellular dynamics of DNA repair proteins by studying their localisation within cells using fluorescence microscopy. However, such studies of PCNA ubiquitination have been hampered by the lack of an antibody specific for the ubiquitinated species of PCNA. A potential solution to this problem exists in the form of Förster Resonance Energy Transfer (FRET) [17]. FRET occurs when two fluorophores are brought into very close proximity (less than 10 nm). Under these conditions, emission energy from the donor can be directly, and non-radiatively, transferred to the acceptor fluorophore resulting in emission of photons by the acceptor. FRET is a widely used technique in biology that allows the determination of interactions between two proteins of interest within the range of 3-10 nm, thus allowing discrimination between simple co-localisation within the limits of the optical resolution of light microscopes and genuine interaction.
Three basic classes of method exist for detecting FRET [17]. 1. Fluorescence lifetime imaging monitors the reduction in the fluorescent lifetime of the donor fluorophore that occurs upon energy transfer. This technique requires specialist equipment, including expensive femtosecond lasers, and can be limited in the complex in vivo environment by uncertainty over the identity of the molecules to which energy is transferred. 2. Acceptor photobleaching monitors the emission intensity of the donor molecule before and after destruction of the acceptor fluorophore. This technique is quantitative and specific but necessarily destructive and is not well suited to long-term live cell imaging. 3. Sensitised emission monitors the indirect excitation of a donor fluorophore by energy transfer. Although conceptually straightforward and attractive in principle, the implementation of this technique for live cell imaging is complicated by the requirement for the acquisition and registration of multiple control images from different specimens.
Spectral imaging refers to the ability to acquire full spectral detail from an image rather than inferring colour from the more traditional use of filters to separate blocks of spectral information onto single detectors. Use of spectral imaging for monitoring FRET has a number of advantages as the intensities of the donor and acceptor can be simultaneously measured allowing ratiometric detection of FRET [18,19,20,21].
Here we report the detection of PCNA ubiquitination in both fixed and live cells through the use of spectral imaging to directly monitor FRET between fluorescently tagged ubiquitin and PCNA. We employ the rather under appreciated FRET pair cyan fluorescent protein (CFP) and monomeric red fluorescent protein (mRFP) [22], and confirm that energy transfer between them is comparable to the more commonly used pair, CFP and YFP. We demonstrate FRET between CFP-ubiquitin and mRFP-PCNA using genetic controls in the avian cell line DT40. By irradiating cells with UV light through a 3 mm microporous membrane to induce localised damage, we demonstrate a marked difference in the kinetics and extent of PCNA modification between cells irradiated in S phase and in G1 phase of the cell cycle.

CFP and mRFP Are a Widely Spaced, Yet Efficient FRET Pair
The most commonly employed fluorescent proteins in FRET studies are the spectrally shifted variants of green fluorescent protein, CFP (cyan) and YFP (yellow) [23]. However, use of the CFP/YFP pair requires extensive controls to correct for the spectral overlap of the CFP and YFP emission spectra. In general, more widely spaced fluorophores will not undergo efficient FRET since the efficiency of FRET is dependent on the overlap between the emission spectrum of the donor and acceptor spectrum of the acceptor (Jl) [24]. A further consideration is that many red fluorescent proteins e.g. DsRed exhibit strong tetramerisation, which limits their utility as protein tags. Monomeric RFP is a derivative of DsRed that exists as a monomer and exhibits very rapid maturation (,1 hour) [22]. The excitation spectrum of monomeric RFP (mRFP) has a significant shoulder at shorter wavelengths, which overlaps well with the emission spectrum of CFP ( Figure 1A, shaded green area). Equally, the overlap of the emission spectra of CFP and mRFP is minimal ( Figure 1A, black region). To measure the FRET efficiency between CFP and mRFP we constructed a fusion protein in which CFP was linked to mRFP via a short sequence containing a cleavage site for the Tobacco Etch Virus (TEV) protease ( Figure 1B). The emission spectrum of a solution of the CFP-mRFP fusion protein excited at 433 nm was determined before, and 2 hours after, the addition of TEV protease. The fusion protein was cleaved efficiently to yield CFP and mRFP ( Figure 1C). Cleavage resulted in an increase in fluorescence at the CFP peak of 38.6% with a corresponding diminution in fluorescence at 607 nm of 32%. This corresponds to a FRET efficiency of 0.28. This agrees closely with another recent determination of the FRET efficiency between CFP and mRFP of 0.25 [25]. Thus, CFP and mRFP are an efficient FRET pair that offer a number of potential advantages in their spectral properties compared with CFP and YFP.

Detection of FRET between CFP-Ubiquitin and mRFP-PCNA in Fixed Cells by Acceptor Photobleaching
In order to study PCNA ubiquitination by FRET we expressed ubiquitin tagged at its N-terminus with CFP and PCNA tagged at its N-terminus with mRFP in chicken DT40 cells. Previous work has demonstrated that both ubiquitin and PCNA tagged with fluorescent proteins are able to participate in normal cellular functions [26,27]. The genetically tractable DT40 cell line [28] provides a robust genetic platform to test the detection of FRET resulting from PCNA ubiquitination as mutants are available both in which PCNA cannot be ubiquitinated (pcnaK164R) [5,29] and in which the hydrolysis of ubiquitin from PCNA is inhibited (usp1) [5,14] (Figure 2A). For each of these mutants and for wild type cells we established clones expressing CFP-ubiquitin, mRFP-PCNA or both. In the case of the pcnaK164R mutant we expressed mRFP-pcnaK164R. Clones were matched for expression of each component using selection by cell sorting and this was confirmed by cytometry and by Western blotting ( Figure 2B & C). Relatively low levels of stable expression were selected to avoid a concentration-dependent FRET signal.
To obtain initial evidence of FRET between CFP-Ub and mRFP-PCNA in vivo we employed acceptor photobleaching to monitor the increase in the emission of the CFP donor fluorophore following destruction of the acceptor mRFP by exposure to high intensity excitation. This effect is clearly seen in the usp1 cell shown in Figures 3A. Following bleaching of acceptor fluorescence the intensity of the CFP signal increases ( Figure 3A-C). To compare the level of Ub-PCNA FRET in wild type, usp1 and pcnaK164R cells, we monitored CFP intensity following mRFP photobleaching in .30 individual cells for each line ( Figure 3D). A small increase in CFP signal was seen in wild type cells following photobleaching, while no change was seen in the pcnaK164R mutant. A more substantial increase in CFP signal was seen in usp1 cells. This is consistent with previous Western blot data in DT40, which reveals a low steady state of Ub PCNA in undamaged DT40 cells [5].

Direct Detection of PCNA-Ubiquitination by Spectral Imaging of Sensitised Emission FRET
Although acceptor photobleaching is a robust method for determining FRET, it is destructive and only suitable for imaging fixed cells and, in this respect, detection of FRET by sensitised emission is preferable. By allowing simultaneous microscopy and spectroscopy, spectral imaging can facilitate ratiometric FRET detection [19] and is now widely available on a number of microscopy platforms.
For each DT40 line (WT, pcnaK164R and usp1), clones expressing CFP-Ub and/or mRFP-PCNA (or mRFP-pcnaK164R as appropriate) were mixed on a slide, fixed and visualised with excitation at 407 nm or 515 nm over a range of wavelengths from 535 nm to 685 nm. Example images are shown in Figure 4A with regions of interest indicating a cell expressing CFP-Ub only (blue), mRFP-PCNA only (red) and both (green). Spectral data was collected from .50 whole cells of each expression pattern for each genotype, averaged and normalised to the CFP maximum. A FRET signal is expected at 607 nm, the emission maximum of mRFP [22]. However, this raw FRET signal [RawFRET l607 ] will also contain spectral bleedthrough from CFP [CFP Ex407/l607 ] and from mRFP directly excited by the 407 nm blue diode laser [mRFP Ex407/l607 ]. Thus, the actual FRET signal at pixel (x, y) [FRET l607 (x,y) ] is given by: FRET l607 x,y ð Þ~R awFRET l607 The CFP Ex407/l607 and mRFP Ex407/l607 components are clearly visible in the cells expressing only CFP-Ub or mRFP-PCNA ( Figure 4B, blue and red lines, respectively). In wild type cells an additional signal is present when both CFP-Ub and mRFP-PCNA are expressed ( Figure 4B, green line). This signal is lost in the pcnaK164R mutant and exaggerated in the usp1 cells, supporting it being RawFRET l607 . By examining cells expressing a range of levels of CFP-Ub and mRFP-PCNA, the contributions of [CFP Ex407/l607 ] and [mRFP Ex407/l607 ] to the raw FRET signal can be shown to be a linear function of the intensity of each fluorophore excited directly over the range of fluorophore concentrations present ( Figure 4C, blue and red points respectively. The FRET signal in usp1 cells is clearly distinguishable ( Figure 4C, light blue line) when compared with pcnaK164R cells ( Figure 4C, green line). Together, these data demonstrate that spectral imaging microscopy can directly detect PCNA ubiquitination by FRET between CFP-Ub and mRFP-PCNA at the level of whole cells.

Single-Pass Ratiometric Spectral Imaging FRET of PCNA Ubiquitination in Live Cells
The wide separation of CFP and mRFP coupled with spectral imaging therefore potentially allows the direct readout of FRET in living cells. However, there remains a contribution of donor bleed through [CFP Ex407/l607 ] and direct acceptor excitation [mRFP Ex407/l607 ] in the raw mRFP FRET signal although this is significantly less than with conventional FRET pairs, such as CFP and YFP.
To remove CFP bleedthrough we employed linear spectral unmixing. Linear spectral unmixing decomposes the spectral information in an image into defined constituent spectra, in this case reference spectra for CFP and mRFP excited at 407 nm and 515 nm. Reference spectra were all collected from control samples imaged under the same conditions as the experimental samples ( Figure 5A). We designated the main output channels of the unmixing algorithm SlCFP and SlmRFP ( Figure 5D, Panels [1][2]) corresponding to the extracted CFP and mRFP signals respectively. In order to remove any contribution of autofluorescence to the SlCFP and SlmRFP channels we also added a 'Junk' channel to the unmixing algorithm ( 407 SlmJunk, Figure 5A), which indicates signals produced by background and autofluorescence on excitation by the 407 nm laser. The implementation of the linear unmixing algorithm in the Nikon EZ-C1 software also generates a 'remainder' channel, which comprises the fraction of the total signal that was not assigned to one of the defined channels. This provides an indication of the accuracy of the unmixing process and, following unmixing into SlCFP, SlmRFP and SlJunk, accounted for less than 1% of the total signal.
Application of the unmixing algorithm to the spectral image effectively eliminated bleedthrough of CFP fluorescence into the mRFP channel ( Figure 5B), as previously described [19]. Following unmixing, we also removed any signal in the 407 SlmRFP channel that also appeared in the 'Junk' channel.  Figure 5D shows example unmixed images of a wild type DT40 cell with a localised region of FRET signal in its nucleus following UV irradiation through a microporous filter. The raw FRET ratio was calculated by dividing 407 SlmRFP by 407 SlCFP and the resulting ratiometric image is shown in Figure 5E. This raw FRET image still includes signal arising from direct excitation of mRFP by the 407 nm laser ( Figure 5C). We compared two methods that compensated for this signal with leaving the FRET ratio image uncorrected: 1. Pixel-by-pixel correction for the intensity of mRFP when directly excited; 2. the application of a threshold to the ratiometric image based on the genetic controls afforded by the pcnaK164R and usp1 cells. In Figure 5F, the direct excitation of RFP is corrected for on a pixel by pixel basis using a paired image of the cell acquired with excitation at 515 nm, the expected bleedthrough being computed using the relationship 407 SlmRFP = 0.025 ( 515 SlmRFP) +2.3, derived from the experiment shown in Figure 5C. Comparison of Figures 5E and F demonstrates that the contribution of the direct excitation of mRFP by the 407 nm laser is small. This is also demonstrated in Figure 5I, which shows a cell harbouring a focal accumulation of PCNA without any resulting FRET signal. Thus, we suggest that for the purposes of live cell imaging the direct excitation of RFP by the 407 nm laser can be ignored or, if more stringency is required, the threshold applied. Both approaches allow single pass imaging, thereby reducing phototoxicity and the risk of acceptor photobleaching.

The Kinetics of PCNA Ubiquitination Following Subnuclear UV Irradiation in DT40 Cells
In order to examine the localised ubiquitination of PCNA we UV irradiated cells through a 3 mm filter [30] prior to live cell imaging at 5 minute intervals using excitation with 407 nm light. Spots of FRET signal were seen throughout the 5 hour imaging window, an example sequence being shown in Figure 6A.
While PCNA ubiquitination has been principally described as an S phase phenomenon, mounting evidence suggests that it can also occur outside of S phase in both G2 and G1, at least in yeasts [31,32]. To establish whether the phase of cell cycle in which the cells were irradiated influenced the kinetics of PCNA ubiquitination, we synchronised wild type DT40 cells by centrifugal elutriation to produce populations in which .90% cells were in either G1 or S phase. The elutriated populations were UV irradiated through the 3 mm filter and the kinetics of PCNAubiquitin FRET signals monitored. Cells irradiated in S phase showed FRET signals at the earliest time point, 5 minutes post irradiation ( Figure 6B). In contrast, the proportion of cells irradiated during G1 exhibiting FRET spots was reduced ( Figure 6B). Further, significant levels of FRET signal was delayed by about 60 minutes. This is approximately the average length of G1 in DT40, which is 1.5-2 hours at 37uC. Thus, significant PCNA ubiquitination is not seen immediately in G1 and the appearance of the signal at later time points may reflect UV damage being passed through the p53dependent G1 checkpoint, which is defective in DT40, to S phase.

Discussion
Here we have described a straightforward method for monitoring the post-translational modification of a protein by monoubiquitination in vivo and validated it using genetic controls.
Previously, a method for detecting ubiquitination dynamically in living cells has been described using bioluminescence resonance energy transfer (BRET) [27]. However, BRET does not lend itself to cellular imaging and subcellular localisation. Detection of ubiquitination in vivo has also been achieved by the use of fluorescence lifetime imaging FRET and a non-fluorescent acceptor (REACh) [33]. However, this requires specialist lasers and detection equipment. A key feature of the FRET method presented here is its simplicity, utilisation of readily available equipment and a requirement for only a single imaging pass. This facilitates its use in live cell imaging while at the same time minimising phototoxicity and acceptor bleaching.
The success of this method derives from a combination of the use of a widely separated FRET pair with minimal bleedthrough coupled with spectral imaging microscopy. Although CFP and mRFP have not been widely employed in FRET studies, we show here that their performance in terms of FRET transfer efficiency is equal to, or exceeds that of, more conventional FRET pairs such as CFP and YFP.
The very wide spectral separation of the donor and acceptor fluorophores is a potential limitation of the method however as a large portion of the detectable spectrum is used, limiting the ability to combine this technique with other dyes or fluorescent proteins for colocalisation studies with ubiquitinated PCNA. However, this drawback is likely to be overcome by improvements in far-red and infra-red fluorophores [34].
We anticipate that further refinements to this method should make it applicable to the study monoubiquitination of other substrates as well as other protein post-translational modification such as SUMOylation across a range of eukaryotic cells.

Cell Culture & Synchronisation
DT40 cells were propagated and transfected as previously described [35]. Cell sorting was performed with a MoFlo automated cell sorter (Cytomation). Synchronisation by centrifugal elutriation was performed on a JE-5.0 Elutriation System (Beckman) with a 4 ml chamber as described [36] with elutriation being effected at constant flow with reducing rotor speed.

DNA Constructs
Expression vectors for CFP-ubiquitin and mRFP-PCNA were constructed using a previously described method in which a HindIII-SalI-NotI three-way ligation into pXPSN results in a SpeI fragment containing the expression cassette [37]. This SpeI fragment was transferred into the SpeI site of either pLoxBsr (in  the case of CFP-ubiquitin) or pLoxPuro (in the case of mRFP-PCNA) to provide a selectable marker [38]. To prepare the mRFP-TEV-CFP fusion construct, mRFP with a deleted stop codon was cloned into the SalI and HindIII sites of pGEX6P1 to create pGEX6-mRFP. CFP was amplified with the primers CFPTEVF [59-AAAAGCTTGAAACCCTATACTTCCA-AGGAATGGTGAGCAAGGGCGAGGAGC] and CFPTEVR [59-AAAAGCGGCCGCCTACTTGTACAGCTCGTCCA-TGCC] and the resulting products cloned into the HindIII-NotI sites of pGEX6-mRFP. The resulting plasmid pGEX6-mRFP-TEV-CFP drives the bacterial expression of mRFP fused to CFP with a short linker [E-N-L-Y-F-Q-G] incorporating the TEV protease cleavage site.

Protein Expression and Purification
For expression of the mRFP-TEV-CFP fusion protein, E. coli BL21DE3 (Invitrogen) were transformed by heat shock. Following selection with ampicillin, a single colony was grown for 8 hours at 37uC in 2 ml LB, after which 250 ml of this culture was used to inoculate 250 ml LB for overnight growth. 12.5 ml of this culture was inoculated into 500 ml 2XTY and grown to an O.D. of 0.7 at which point protein expression was induced with IPTG at a final concentration of 1 mM. The culture was then grown at 16uC overnight. The bacteria were pelleted and the pellet resuspended and incubated in 40 ml ice cold PBS with 0.1 ml/ml lysozyme, 10 mg/ml DNAse 1, 2 mM DTT and 1 tablet of 'Complete' EDTA-free protease inhibitor (Roche) for 10 minutes. Following disruption by sonication the lysate was centrifuged in at Ti45 rotor at 40 krpm for 45 minutes at 4uC. The supernatant was filtered through a 0.45 mm membrane and incubated with rotation with 1 ml glutathione sepharose beads per litre of initial culture, prewashed in PBS, at 4uC overnight. The purified protein was cleaved from the GST tag with 120 ml PreScission protease (GE Healthcare) for 4 hours at 4uC. The protein was dialysed in PBS (with protease inhibitor and 1 mM DTT), concentrated in a Vivaspin column (Vivascience) and flash frozen in aliquots. mRFP and CFP protein was detected in Western blots with anti-mRFP-HRP (Abcam ab34767) at 1:500 dilution and anti-GFP-HRP (Abcam ab6663) at 1:1000 dilution, with detection by ECL+ (GE Healthcare).

Spectrophotometry
Experiments were performed on an LS55 spectrophotometer (Perkin-Elmer) at 22uC with the following settings: excitation slit = 5, emission slit = 15, PMT voltage = 900 V, emission detection range = 450-650 nm, emission step = 1 nm, integration time = 2 s, stirrer speed = 2. The optimal excitation of CFP was determined to be 433 nm and was used for subsequent determinations of FRET. Experiments were performed in TEV cleavage buffer (1 M Tris-HCl pH 7.4, 10 mM EDTA, 10 mM DTT) using 2.7 mg/ml mRFP-TEV-CFP protein in a volume of 20 ml. 10 units of TEV protease (Invitrogen) was added and spectra taken every 30 minutes until the CFP emission peak at 478 nm plateaued.

Cell Fixation
Approximately 1610 5 DT40 cells in PBS were applied to poly-llysine (Sigma) treated 22 mm diameter coverslips for 5 minutes in a humidified box following which they were fixed in 4% paraformaldehyde (Alfa Aesar) in PBS for 5 minutes. Cells were mounted in Fluoromount G (Southern Biotech) and sealed under a coverslip.
Microfilter UV Irradiation 2.5610 5 cells were added to a LAB-TEK II Chamber (Nunc) and incubated for 30 minutes at 37uC in 10% CO 2 . For irradiation the medium was removed and replaced with 200 ml of PBS and a 3 mm pore size filter (Millipore) applied to the cell layer. The cells were irradiated with 100 J/m 2 265 nm light delivered by a calibrated UV lamp (UVP Inc.) housed in a specially designed shuttered enclosure. The PBS was removed and replaced with pre-warmed RPMI with the usual supplements but without phenol red. The cells were transferred directly to the heated CO 2 incubation chamber on the microscope.

Microscopy & Image Analysis
Microscopy was performed with a Nikon C1-si microscope equipped with a Tokai-HIT heated CO 2 incubation chamber. The objective used was the Plan-Apo Uv-corrected 60x N.A. 1.4. Images were taken in spectral mode with a 10 nm bin size. For illumination of fixed cells the 50 mW 407 nm blue diode laser was used at 10% and 515 nm line of the 40 mW Ar-ion laser (Melles-Griot) also at 10%. For live cell imaging the 407 nm laser was used at 1%. In both cases pixel dwell was set to 9.6 ms. Images were captured every 5 minutes. For photobleaching mRFP, the selected field was scanned with the 515 nm and 561 nm laser lines at full power for 2 minutes. Raw image files were low pass filtered and spectrally unmixed using the high accuracy unmix algorithm in the Nikon EZ-C1 software v3.7 with the reference spectra shown in Figure 5A. Subsequent processing was performed in ImageJ v1.42 [39].