Expression of Multiple Transgenes from a Single Construct Using Viral 2A Peptides in Drosophila

Expression of multiple reporter or effector transgenes in the same cell from a single construct is increasingly necessary in various experimental paradigms. The discovery of short, virus-derived peptide sequences that mediate a ribosome-skipping event enables generation of multiple separate peptide products from one mRNA. Here we describe methods and vectors to facilitate easy production of polycistronic-like sequences utilizing these 2A peptides tailored for expression in Drosophila both in vitro and in vivo. We tested the separation efficiency of different viral 2A peptides in cultured Drosophila cells and in vivo and found that the 2A peptides from porcine teschovirus-1 (P2A) and Thosea asigna virus (T2A) worked best. To demonstrate the utility of this approach, we used the P2A peptide to co-express the red fluorescent protein tdTomato and the genetically-encoded calcium indicator GCaMP5G in larval motorneurons. This technique enabled ratiometric calcium imaging with motion correction allowing us to record synaptic activity at the neuromuscular junction in an intact larval preparation through the cuticle. The tools presented here should greatly facilitate the generation of 2A peptide-mediated expression of multiple transgenes in Drosophila.


Introduction
Co-regulated expression of genes is important in many biological contexts and also has great utility when introducing exogenous genes into model organisms. An obvious application is to co-express a reporter gene with an effector to identify or ''label'' the cells manipulated by the effector gene. Co-expression of several proteins may also be required to produce a desired biological effect. To achieve expression of multiple transgenes in cultured cells, separate plasmids, each with their own promoter, are co-transfected often in combination with a reporter plasmid at a lower concentration. Since the majority of cells that take up DNA will take up both plasmids, reporter-expressing cells are also likely to express the co-transfected genes. While co-transfection with multiple plasmids is possible for in vitro cell transfection and even in vivo through electroporation, it is far from ideal for most in vivo applications as it can still result in genetically manipulated cells that are not marked by a reporter.
Internal ribosome entry site (IRES) sequences have been used as an alternate method to separate two coding sequences under the control of a single promoter. In theory both proteins could be expressed in the same cells and at the same level, but in reality the level of the second protein is often significantly reduced ( [1]). In addition, the length of the IRES sequence can be prohibitive in experiments using viruses with small packaging capacity.
The discovery of viral 2A peptides and their use in generating multiple separate proteins from a single mRNA circumvents many of these previous difficulties. The first such peptide characterized is the 2A peptide from foot-and-mouth disease virus (FMDV; [2]), which separates the 2A and 2B protein products. Originally thought to encode a protease cleavage site, the ''cleavage'' of proteins separated by the 2A sequence surprisingly occurs cotranslationally, in an unconventional process where a peptide bond often fails to form between the glycine and terminal proline tRNA in the 2A peptide motif (D(V/I)EXNPGQP, where Q represents the skipped peptide bond). Despite this failure, translation can proceed and two distinct proteins are produced with equimolar stoichiometry. However, this bond skipping does not always occur; if a bond forms, a stable fusion protein is generated that will not subsequently cleave ( [3]). A short stretch of the 2A peptide sequence coding for approximately 20 amino acids is sufficient to cause the bond-skipping effect, and similar sequences have been found in several other viruses; the most commonly used versions are from equine rhinitis A virus (ERAV; E2A), FMDV (F2A), porcine teschovirus-1 (P2A), and Thosea asigna virus (TaV; T2A).
Previous work demonstrated that the various 2A peptides have different efficiencies of peptide bond skipping and that these efficiencies vary depending on the experimental context. For example, the 2A peptide from FMDV was reported to exhibit ,90% skipping efficiency ( [4]), compared with reports of nearly 100% ( [5]), and most recently only 40% ( [6]). This variability is almost certainly due in part to differences in experimental conditions, including use of different model organism or cell line. With this in mind, we tested which of the four most widely used 2A peptides work best in Drosophila, both in the S2 cell line and in vivo for the efficient expression of multiple polypeptides from a polycistronic mRNA transcribed from a single promoter. As a demonstration of the utility of this approach, we use the P2A peptide to co-express the red fluorescent protein tdTomato and the genetically-encoded calcium indicator GCaMP5G to enable ratiometric calcium imaging at the larval neuromuscular junction (NMJ). The tools presented here should greatly facilitate the generation of 2A peptide-mediated expression of multiple transgenes in Drosophila.

DNA Cloning
DNA sequences coding for codon-optimized 2A peptides were generated from the oligonucleotides in Table 1 MyrGFP was created by PCR using F-myrGFP BclI and R-GFPo StuI and pJFRC19 template ( [9]) which contains a fly codon-optimized myristylation sequence derived from the first 85 amino acids of the Drosophila Src homolog (Src64B) followed by a codon-optimized EGFP (F64L, S65T). This PCR product was cut with BclI and StuI and ligated with pC5-Kan 2A cut with BamHI/StuI.
DsRed-nls is also called RedStinger ( [10]), which is a DsRed variant containing a nuclear localization signal from the Drosophila transformer gene. It was amplified by PCR with F-dsRed BamHI and R-dsRed XbaI, and cloned into pCR8GW TOPO, then excised with BamHI/XbaI and cloned into pC5-Kan myrGFP-2A cut with BglII/AvrII to make pC5-Kan myrGFP-2A-RedStinger.

S2 Cell Culture
S2 cells (selected for adhesion; gift of Dr. Cedric Wesley, University of Wisconsin-Madison) were grown in DMEM (Invitrogen) supplemented with 10% FBS and 1% penicillinstreptomycin and transfected using Effectene (QIAgen) according to manufacturer's protocol.

Western Blotting
S2 cells were harvested, washed with PBS and resuspended in 2x SDS sample buffer, then boiled and centrifuged. Supernatant was then loaded on a 14% SDS-PAGE gel. 15 adult flies were anesthetized, transferred to tubes and homogenized in 2x SDS sample buffer, boiled and centrifuged. Approximately 1 fly was loaded per lane.
Protein was transferred to a PVDF membrane, blocked with Odyssey blocking buffer (Licor) and probed with rabbit anti-GFP (A6455, Invitrogen) at 1:1000, followed by donkey anti-rabbit 800 IRDye (Licor) at 1:10,000. Quantitative imaging was performed using an Odyssey infrared imaging system (Licor). Integrated intensity measurements were performed with Odyssey software. Separation efficiency was calculated as the lower band intensity divided by the sum of lower and upper band intensities after subtracting the background.

Immunohistochemistry and Imaging
S2 cells were harvested 48 hours post-transfection and resuspended in fresh media and then plated on a poly-lysine coated coverslip for 10 minutes. The solution was removed and replaced with PBS and then 4% parafomaldehyde in 0.1 M phosphate buffer, pH 7.20 for 5 minutes. Fixative was removed and the coverslip washed with PBS before placing on a droplet of Vectashield (Vector Laboratories) on a slide and sealed with nail polish.

Image Analysis
The 3D Object Counter plugin in ImageJ software (FIJI package,NIH) was used to determine the size and abundance of bright GFP/RFP-positive aggregates from confocal stacks of larval muscles.
Wide-field microscopy was performed on an Olympus BX51WI microscope fitted with a 606 (0.9 NA) water-immersion objective. A Lumen Dynamics (Mississauga, Ontario, Canada) X-Cite XLED1 LED excitation unit with GYX (540-600 nm) and BDX (450-495 nm) modules was used to supply excitation light which was then further filtered with a dual-band pass 470/556 nm filter. Images were captured with an Andor Technology (Belfast, Northern Ireland) EMCCD camera (DV887). A Cairn Instruments (Kent, UK) Optosplit II emission beam-splitter was placed before the camera for dual-emission wavelength imaging. Simultaneous imaging of tdTomato (ex./em. 556/640 nm, [12]) and GCaMP5 (ex./em 480/530 nm, [15]) was achieved using a primary 500/530 nm bandpass dichroic mirror in the microscope housing, a secondary 600 nm dichroic mirror mounted in the emission beam-splitter, and 512/25 nm and 660/20 nm emission filters in the beam-splitter. Filters and dichroic mirrors were obtained from Chroma Technology (Bellows Falls, VT, USA) or Semrock (Lake Forest, IL, USA). The imaging system was controlled through a Dell PC running Andor iQ software (version 2.7.1). All imaging was performed at room temperature (,24uC)

Preparation of Specimens
For ex vivo imaging, female wandering 3 rd instar Drosophila larvae were collected and fillet-dissected in chilled HL6 to expose the longitudinal muscles of the body wall and the segmental nerves that project from the ventral ganglion. The posterior ventral ganglion was cut from the cerebral ganglia to stop endogenous nerve activity ( [17]) and a loop of a segmental nerve was drawn into the lumen of a glass micropipette with the ventral ganglion still attached ( [14]). Nerves were stimulated as described previously ( [18]). For in vivo imaging, intact larvae were restrained between a glass slide and coverslip.

Statistical Analysis and Figure Preparation
SigmaPlot (version 12.3; Systat Software) was used for statistical analysis. One-way ANOVA (with a Bonferroni post hoc test) was used for comparisons. Differences were considered to be statistically significant at a-values of P#0.05.

A Versatile Shuttle Vector for Cloning 2A Peptides
As a first step to compare the properties of different 2A peptide sequences in Drosophila, we generated a vector that facilitates cloning of transgenes separated by the 2A sequence. This set of vectors contains a 2A peptide sequence in the middle of two multiple cloning sites and is based on the C5 series of vectors ( [8]). They have a large number of unique restriction enzyme cut sites that make it easy to transfer cloned DNA from small shuttle vectors into injectable expression vectors for in vivo transformation of Drosophila. A diagram of this vector and the restriction enzyme sites are shown in Figure 1. As an additional feature, we designed the vector so that once a gene is added 59 to the 2A sequence, it can be easily removed and added downstream of the 2A sequence in another vector ( Figure 1B and 1C). This modular design with gene-2A cassettes can improve cloning efficiency and enable rapid construction of chains of transgenes that can be expressed together from the same promoter. After the desired construct has been generated in the pC5-Kan 2A shuttle vector, it can be easily subcloned into the pUAS-C5 expression vector using flanking rare restriction enzymes ( Figure 1D). We produced vectors to test four of the most commonly used 2A peptide variants from equine rhinitis A virus (ERAV; E2A), human foot-and-mouth disease virus (FMDV; F2A), porcine teschovirus-1 (P2A), and Thosea asigna virus (TaV; T2A).

Comparison of 2A Peptides In vitro
Using the vectors described above, we tested the separation efficiencies with which each of the four most widely used 2A peptides. Previous studies have used a number of different cell-free and cell-based assays to determine separation efficiency and a range of values have been reported for each peptide ( [4], [5], [6], [19]). Our strategy was to express a membrane-targeted green fluorescent protein (mGFP) and a red fluorescent protein with a nuclear localization signal (dsRed-nls) separated by the 2A peptide (Figure 2A; see methods for details). Using this approach, we could image 2A-dependent polypeptide separation in fixed cells and also quantify the level of separation using SDS-PAGE followed by western blotting.
We transfected the Drosophila S2 cell line with each of the four versions of mGFP-2A-dsRed-nls, as well as one without any 2A peptide sequence, driven by the ubiquitous actin5c promoter. Figure 2B shows sample confocal micrographs of GFP and RFP fluorescence in fixed S2 cells after transfection. Since any fusion protein resulting from failure of the 2A peptide bond-skipping process will contain both a membrane localization signal (Nterminal myristyl group) and a nuclear localization signal, we expected to detect these full-length polypeptides in nuclear or perinuclear domains. In the absence of a 2A peptide, GFP and RFP fluorescence is concentrated in perinuclear puncta (figure 2 B5). We observe similar perinuclear puncta more frequently in cells transfected with E2A-and F2A-containing vectors than with the P2A and T2A vectors, indicating that E2A and F2A have lower separation efficiency in Drosophila S2 cells. More quantitatively, we measured separation efficiency by western blots using an antibody against GFP. As noted previously ( [2]), we detect the presence of slower-migrating bands of the size expected for a fulllength mGFP-2A-dsRed-nls protein product, and the relative abundance of this upper band varies depending on the particular 2A peptide sequence; the P2A and T2A peptides yield the best separation efficiency ( Figure 2C and D). E2A had an efficiency of 68+/23%, compared with 65+/24% for F2A, 96+/21% for P2A and 88+/23% for T2A (average +/2 SEM, n = 5, p,0.001 by one-way ANOVA between E2A and F2A compared with either P2A or T2A; P2A and T2A were not significantly different). The unconventional way in which the second polypeptide is generated using 2A peptides results in an N-terminal proline which could limit the protein's perdurance. Additionally, the ribosome could disassociate when encountering the conserved NGPG motif, leading to reduced translation of the second protein. To test whether the protein following the 2A sequence is expressed ratiometrically, we transfected S2 cells with a construct encoding GFP-P2A-GFP-dsRednls. In this way we can directly compare the GFP-2A* (separated) band with a higher GFP band whose apparent molecular weight has been shifted by fusion with dsRed. A western blot probed with anti-GFP shows both bands are present with similar intensities ( figure 2E). This is quantified in figure 2F, demonstrating that the proteins are produced at very similar levels (n = 2).

Comparison of 2A Peptides In vivo
To test how the 2A peptides performed in vivo, we generated transgenic fly lines where the mGFP-2A-dsRed-nls construct was placed under the control of the yeast upstream activation sequence (UAS) so that it could be expressed under the control of tissuespecific Gal4 lines ( Figure 3A). To control for position effects on transgene expression level, we integrated all four constructs into the same genomic locus using the phage phiC31 system ( [20], [9]). Separation of GFP and RFP was best demonstrated in large cells with good separation between membrane and nuclear compartments, such as larval salivary gland and somatic muscles. When we express these transgenes in the larval salivary gland, we can easily see differences among the different 2A peptides in accumulation of perinuclear GFP/RFP-positive puncta ( Figure 3B). Similar to the results with S2 cells, expression utilizing E2A and F2A peptides in salivary glands results in the most prevalent production of perinuclear GFP/RFP-positive puncta, which indicate failure of the peptide-bond skipping event. In contrast, we observe few of these puncta in cells expressing the constructs containing the P2A and T2A peptides, indicating a high efficiency of polypeptide separation in vivo (figure 3B). When DsRed-nls was expressed alone in the salivary gland cells we do not observe extra-nuclear spots of red fluorescence, suggesting that these accumulations result from mis-targeting due to the myristylation sequence of mGFP (data not shown). Differences in separation efficiency are recapitulated with expression in other cell types including larval body wall muscles (figure 3C) and larval multi-dendritic sensory neurons (data not shown). The analysis of GFP/RFP-positive aggregate number from muscle cells shown in figure 3D demonstrates that increased numbers of larger aggregates are found in cells expressing E2Aand F2A-containing transgenes.
We also analyzed separation efficiency of 2A-linked fluorescent proteins by western blot analysis of whole fly extracts. Surprisingly, the band representing full length myrGFP-2A-dsRed-nls was present at a much lower level than in S2 cell extracts ( figure 3E). This observation suggests that some type of clearance mechanism is operating to eliminate the myrGFP-2A-dsRed-nls fusion protein, perhaps because of the competing localization signals.

Using P2A to Facilitate Ratiometric Calcium Imaging
The invention of optogenetics and the evolution of fluorescent reporters and sensors in recent years have provided strong impetus to develop efficient new ways of applying these tools to biological problems. In many instances, significant experimental advantages are gained by expression of multiple reporters and sensors in the same cell, such as the paired expression of channelrhodopsin and halorhodopsin for dual color optical control of electrical activity ( [21]). Another example is tandem expression of a fluorescent indicator along with a control fluorescent protein, enabling measurement of baseline differences and comparison between cells and experiments.
Because of the very high polypeptide separation efficiency of the P2A peptide, we used it to develop a tool for in vivo calcium imaging. The genetically encoded calcium indicator GCaMP has undergone a series of refinements since its invention, with the most recent versions offering increased signal-to-noise ratios (SNR) by 2-3 fold ([15] [16]). Unfortunately, this increased SNR has been primarily achieved by significantly lowered basal fluorescence, which can make it difficult to focus on a region of interest before and even during calcium imaging, especially if there is any movement of the preparation during recording. We sought to circumvent these problems by co-expressing a bright and stable red fluorescent protein along with one of the newer GCaMP indicators. We therefore made a construct that expresses the red fluorescent protein tdTomato followed by the P2A peptide and then GCaMP5G, all under the control of the Gal4/UAS system.
We expressed UAS-tdTomato-P2A-GCaMP5G in motoneurons and recorded red and green fluorescence intensities with muscle contraction blocked in a dissected larval preparation. We were easily able to detect calcium flux into the nerve terminals upon electrical stimulation of the axons and the GCaMP fluorescence signal increased with increasing stimulation frequency ( Figure 4A and 4B). Because movement of the preparation was blocked, the tdTomato signal shows minimal fluctuation over the duration of the experiment, so correction of the GCaMP signal by normalizing to tdTomato has little effect ( Figure 4C). We could also easily quantify peak GCaMP fluorescence intensity at different stimulation frequencies using regions of interest demarcated by tdTomato fluorescence ( Figure 4D).
To test whether a strong red signal would allow us to measure changes in the cytoplamsic calcium concentration in a preparation where contraction of body wall muscles was not blocked and movement occurred, we recorded fluorescence intensity from motoneuron terminals through the cuticle of an intact larva ( Figure 4E and 4F). Raw GCaMP fluorescence signals recorded in this experimental setup are difficult to interpret alone, but when they are normalized to the tdTomato signal, we can distinguish real calcium-dependent signals from movement-derived artifacts ( Figure 4G). Thus, by use of the P2A peptide vector, we were able to achieve a significant refinement over existing imaging techniques demonstrating an example of the useful ways in which this technology can be applied.

Discussion
Expression of multiple polypeptides at similar levels from a single mRNA is made possible by inclusion of a short viral 2A peptide coding sequence between the polypeptide-encoding transgenes. Here we compare four of the most commonly used 2A peptides from different viruses and find that the P2A and T2A peptides exhibit the highest efficiency of polypeptide separation in Drosophila both in cultured cells and in vivo enabling separation of co-expressed, differentially targeted fluorescent proteins. We also demonstrate use of the P2A peptide to improve ratiometric calcium imaging with GCaMP5G. Furthermore, the series of vectors we have developed that incorporate 2A peptides permit easy cloning of 2A-linked transgenes for expression in Drosophila.

Previous Use of 2A Peptides in Drosophila
The use of 2A peptides in Drosophila has been described in four recent reports. Gonzalez et al., 2011([22]) utilized a 2A peptidecontaining vector to co-express a fluorescent marker (GFP or mCherry) and an antibiotic resistance gene (Neo R ) to facilitate the peptides. The vector used in these experiments only permits cloning a single gene for stable expression in S2 cells, whereas our vectors are more general and make it easy to clone multiple genes back-to-back. Inagaki et al., 2012 ([23]) utilized a 2A peptide that was not described to co-express two parts of an artificial signaling cascade and did not report any data on separation efficiency. Diao and White, 2012 ([24]) utilized a 2A peptide to effectively co-express Gal4 from the translational start site of a protein of interest. They compared the same four 2A peptides we studied here as well as the 2A peptide from Drosophila C virus (DCV) and concluded that T2A worked the best. However, their assay was only qualitative and based on examining nuclear versus membrane localization of Gal4 in cultured S2 cells (and the subsequent signal amplification by a UAS-reporter). Although we are in agreement that T2A is far better than either E2A or F2A, our quantitative evidence indicates that the P2A peptide is equally efficient as T2A at generating distinct proteins in Drosophila. These minor discrepancies may be due to use of slightly different sequences surrounding the 2A peptide or differences inthe particular assays used.
Finally, Masuyama et al., 2012 ([25]) utilized a reporter with two membrane protein-GFP fusions separated by an undisclosed 2A peptide to increase fluorescence intensity.
Only one of these previously published papers describes a useful vector for cloning, and that vector only enables use of T2A to express a single gene upstream of a resistance gene ( [22]). Significant improvements incorporated into our vector include a greater selection of cloning sites before and after the 2A sequence, the capability for easily generating multimeric gene-2A repeats, and the ease with which the finished construct can be shuttled into an expression vector (such as pUAS-C5 attB) for in vivo transformation. Using a small shuttle vector for the initial cloning is beneficial in many ways. First, more unique sites are present in the shuttle which increases the likelihood that a gene of interest can be inserted in frame with the 2A sequence. Second, the smaller vector is easier to amplify in bacteria and purify with higher yields which is helpful for sequencing inserts. Third, the inserts from the shuttle vector are generic and can be subcloned into expression vectors with different purposes, for example expression in cultured cells or for transformation of genetic model organisms.
Another method for inserting a 2A peptide would be to directly add the sequence during PCR using an oligonucleotide containing the 2A coding sequence and homology to the gene of interest, as at this exposure (asterisk is next to non-specific band also seen when myrGFP is expressed alone), despite prevalence of lower band representing separated GFP-2A* protein. doi:10.1371/journal.pone.0100637.g003 was done by Diao and White ( [24]). However, making gene1-2A-gene2 constructs such as those we present here would require multiple PCR steps and may be less efficient than stepwise addition into the pC5-Kan 2A vector. This would be especially true for longer chains of transgenes. We expect that the P2A and T2A versions would still work best if added using PCR.
Of the four vectors we present here, the E2A-and F2A-carrying versions may not be as useful in Drosophila, given their poor efficiency at polypeptide separation. However, the use of both P2A and T2A sequences in a longer construct instead of solely P2A may reduce DNA rearrangements in bacterial plasmids. More recent reports on different versions of F2A have found that longer sequences can perform better ( [19]), but we only tested the short 22 amino acid version. Despite significant differences in separation efficiency in S2 cell culture and in many in vivo tissues examined, we observe far less unseparated protein in vivo than we expected. For example, given the data in figure 2D, we would expect ,35% of E2A-containing protein to be unseparated. Although we find prominent GFP/RFP-positive aggregates in salivary gland and muscle using immunohistochemistry, they are limited, and western blots show very little protein of the expected size for unseparated myrGFP-2A-dsRed-nls. One explanation for this difference could be that membrane-targeted proteins with nuclear localization signals are selectively degraded by protein quality control pathways that are less efficient in S2 cells or become saturated at high expression levels. Another possibility is that separation efficiency could depend on the tissue or developmental stage, although we think this is unlikely given the qualitative similarities between the diverse cell types we examined. While the apparent dearth of unseparated protein on western blots would on the surface seem to indicate that all 2A peptides are efficient in vivo, we caution that this could be an artifact of the particular localization signals we used, and not a general finding. In cell types where we can easily detect aggregated GFP/RFP, the relative accumulation recapitulates the separation efficiency measured in cell culture. Further studies will be needed to investigate separation efficiency in vivo in a variety of Drosophila cells as well as at different developmental stages to confirm the generality of our results. To demonstrate the usefulness of the 2A peptide methodology, we have applied it to ratiometric Ca 2+ imaging at the larval Drosophila neuromuscular junction. Use of the P2A peptide greatly simplified co-expression of tdTomato with GCaMP5G compared with the crossing schemes that would otherwise be required to generate a fly capable of expressing both proteins. The 2A peptide technique also enables the co-expressed proteins to be produced at approximately equal levels, which is an important consideration in a number of applications. Additionally, 2A peptides allow production of multiple proteins using the same promoter sequence, which reduces the dilution of transcription factors that could result from the presence of too many binding sites, as can occur with weakly-expressing Gal4 lines driving multiple UAS-linked transgenes. Newer versions of GCaMPs have lower fluorescence in the unbound state, which makes expression of another fluorescent marker protein essential for identifying the particular region or cells selected for experimental analysis. As our results demonstrate, by use of a co-expressed marker protein we can also control for movement artifacts during imaging. Thus, we were able to monitor presynaptic activity at larval NMJ in a living, intact, contracting larval preparation, which is highly desirable in a number of experimental situations.
There are a large number of potential applications for controlled co-expression of multiple polypeptides from a single polycistronic mRNA transcribed from a transgenic construct. Increasingly, new techniques in Drosophila and other genetic model organisms rely on genetic lines harboring multiple transgenes which can be complicated to obtain using standard recombination and crossing schemes. The 2A peptide technology has the potential to reduce this complexity and simplify stock construction. The new vectors we present here greatly facilitate synthesis of 2A peptide-mediated polycistronic constructs enabling convenient and broad application of this technology.