Strategies for quantitative transient co-expression of independent markers
We evaluated three strategies to achieve stoichiometric co-expression of two fluorescent proteins. Constructs were expressed using the high-efficiency Agrobacterium-mediated transfection of tobacco leaf abaxial epidermal cells (3,31), which has been used widely in plant membrane trafficking studies with fluorescent protein reporters (3,6–9,11,12). Furthermore, we have previously used this method to establish quantitative imaging protocols for assaying membrane traffic as reported by a secGFP marker (3). In the first approach, leaves were co-inoculated with separate GFP-HDEL and yellow-fluorescent-protein-(YFP)-HDEL strains (Figure 1A and B), which target each fluorescent protein to the ER. However, individual epidermal cells expressed each marker independently, such that the fluorescence intensity of one marker in an individual cell did not predict the expression efficiency of the other (Figure 2A and B). This was confirmed by the low correlation observed in scatter plots of pixel intensities for the GFP and YFP channels (Figure 2C). Thus co-inoculation is not a generally applicable strategy for ratiometry.
Figure 1. Schematic representation of constructs. (A–S) The components of each construct described in this paper are named and illustrated according to the following scheme: G or GFP, green fluorescent protein (mGFP5); Y or YFP, yellow fluorescent protein; R or RFP, red fluorescent protein (mRFP1); sp, signal peptide; sec, secreted form of a fluorescent protein; H or HDEL, presence of an ER retrieval signal in addition to signal peptide; nls, nuclear localization signal; ST, Golgi-targeting signal from rat sialyl transferase; N, engineered N-glycosylation site (6); Rab, Arabidopsis Rab GTPase AtRAB-D2a either wild type or mutant forms; m or myc, c-myc epitope tag; nomyc, c-myc epitope tag is missing; f or flag, FLAG epitope tag; 2A, FMDV 2A peptide; 2A*, mutant form of the 2A peptide in which the two terminal residues [G and P, underlined in (T)] were each converted to alanine. (T) Amino acid and nucleotide sequence of the 2A peptide and upstream flanking sequence used in this study, taken from the constructs described by Halpin et al. (22); the arrow indicates the site at which 2A activity disrupts the polypeptide backbone; the underlined residues were converted to alanine in the 2A* sequence.
Figure 2. (opposite) Strategies for stoichiometric transient expression of fluorescent markers. Low (A, D, G) and higher (B, E, H) magnification confocal images of tobacco leaf epidermal cells transiently expressing a mixture of GFP-HDEL and YFP-HDEL (A and B), GH–YH (D and E) or Ym-2A-GH (G and H) with the YFP channel shown in red and the GFP channel in green in a single merged image. (C, F, I) Scatter plots of pixel intensities in the GFP channel (x axis) and YFP channel (y axis) for images similar to those shown in A, D and G, respectively; colours indicate the frequency with which particular intensity combinations are observed with blue being the lowest and red the highest. (J–L) Projections in z of 22 1.5-μm confocal optical sections through the cortical cytoplasm of cells expressing Ym-2A-GH with varying efficiency; YFP (red), GFP (green) and merged channel images are shown as indicated. Scale bars 100 μm (A, D, G), 20 μm (B, E, H and J–L). Agrobacterium strains carrying each constructs were infiltrated at OD600 0.03. (M) Coefficients of variance (V) for absolute (open bars) and ratiometric (shaded bars) measurements of GFP fluorescence in nine low-magnification confocal images, such as those in (D) and (G), taken from leaves expressing GH–YH or Ym-2A-GH; bars show the mean value of V and the standard error of four experiments. (N) Absolute (open diamonds) and ratiometric (shaded squares) measurements of GFP fluorescence in epidermal cells of tobacco leaves infiltrated with Agrobacterium strains carrying Ym-2A-GH at various titres; data are the means of two experiments. (O) Average coefficients of variance for each absolute (open bars) and ratiometric (shaded bars) measurement in (N).
In the second approach, we constructed a transfer DNA (T-DNA) in which GFP-HDEL and YFP-HDEL were each transcribed divergently from the same CaMV 35S enhancer elements (GH–YH, Figure 1C). This improved the overlap visible between signals from individual cells (Figure 2D and E) and increased the degree of correlation on a pixel-by-pixel basis in the scatter plot (Figure 2F). We used a quantification procedure (3) to calculate the mean GFP and YFP pixel intensities in nine random low-magnification images such as the one in Figure 2D and asked whether the YFP signal could be used to normalize for image-to-image variation in the GFP signal. The mean coefficient of variance (standard deviation divided by the mean, expressed as a percentage) for the GFP signal in four independent experiments was approximately halved by normalization to YFP, representing a significant improvement in data quality (Figure 2M, GH–YH). Nevertheless, it was clear that considerable variation still existed in the relative expression levels of each marker in individual cells, seriously limiting the predictive value of the single T-DNA approach for analysis of single cells.
In the third approach, we took advantage of the of the ‘self-cleaving’ FMDV 2A peptide (21,22,27) to link YFP and ER-targeted GFP-HDEL in a single construct, Ym-2A-GH (Figure 1G; the legend to Figure 1 explains the nomenclature used for this and other 2A constructs). Following transfection with this construct, there was a close correlation between the YFP and GFP signals for individual cells over a wide range of intensities (Figure 2G and H). The YFP accumulated in the cytoplasm and nucleoplasm, while GFP was confined to the ER and nuclear envelope (Figure 2H; see also Figure 3A–C), consistent with cleavage between the two fluorescent proteins and localization according to the ER targeting and retention sequences on the GFP moiety. The improved correlation between GFP and YFP intensities was confirmed by the scatter plot of individual pixel intensities in Figure 2I and by z-projections of higher magnification images of the cortical cytoplasm (Figure 2J–L). GFP and YFP signals substantially co-localized as the ER and cytoplasm are not readily resolved even at this level of magnification. The mean coefficient of variance was more than halved when the ratio was calculated (Figure 2M, Ym-2A-GH) and the improvement in the quality of the ratiometric data was manifest over a 30-fold range of Agrobacterium titres (from OD600 0.013 to 0.40, Figure 2N). Thus, although the absolute accumulation of GFP varied approximately 10-fold, the ratio of GFP:YFP varied only 1.4-fold. The ratio was unchanged above OD600 0.05 and was linear up to this point, suggesting that an OD600 close to 0.05 is optimal for use of constructs like Ym-2A-GH. The coefficient of variance for the ratio data was threefold lower than that of the absolute fluorescence data for bacterial titres of 0.05 or less (Figure 2O). We concluded that, in principle, 2A constructs offered an effective strategy to produce stoichiometric levels of reference and target markers suitable for ratiometric analysis of GFP expression levels in individual cells and in cell populations, with YFP accumulation reliably predicting GFP expression efficiency in each cell.
Figure 3. Targeting of translation products of 2A peptide fusions to different compartments. Confocal images of tobacco leaf epidermal cells transiently expressing (A–C) Ym-2A-GH, (D) Ym-2A-secGf, (E–G) nlsRm-2A-GH, (H) nlsRm-2A-secGf, (I–K) STN-Rm-2A-GH, (L) STN-Rm-2A-secGf, (M–O) secN-Rm-2A-GH, (P) secN-Rm-2A-secGf. YFP and RFP channels are shown in red and the GFP channel in green. Merged channel images are indicated. (M–P) Chlorophyll fluorescence is indicated in blue. Shown are single sections (A–C and E–H) and projections (D and I–P). Scale bar is 5 μm.
Constructs for ratiometric assays of membrane traffic
Images of cells expressing Ym-2A-GH indicated that the GFP moiety could be targeted to the ER while YFP remained cytosolic (Figures 2H and 3A–C). To obtain a 2A-based ratiometric fluorescent marker of biosynthetic membrane traffic, we replaced the ER-retrieval signal (HDEL) on the GFP moiety of Ym-2A-GH with a sequence encoding the FLAG epitope tag (Figure 1F). It was expected that the GFP moiety of the new construct, Ym-2A-secGf, would be trafficked to the apoplast, perhaps with a fraction being sorted to the vacuole as reported previously for secGFP in tobacco epidermis (3). In either location, GFP accumulation and fluorescence is known to be poor (3,6,7,11). Consistent with this prediction, when expressed in plant cells, GFP accumulation from Ym-2A-secGf was almost undetectable under conditions that revealed clear GFP fluorescence in the ER of cells expressing comparable amounts of Ym-2A-GH as judged by YFP accumulation (compare Figure 3C and D). As described in later sections, the difference between these two extremes of GFP accumulation, normalized to the level of expression of the reference (YFP) marker, provides a quantitative ratiometric assay of GFP trafficking.
While the Ym-2A-secGf construct successfully combined a trafficked GFP marker with a cytoplasmic YFP reference in stoichiometric quantities, we sought to extend the utility of the ratiometric approach and to test the versatility of the 2A peptide for targeting proteins to distinct endomembrane compartments. Therefore, we assembled a series of constructs in which the intended location of the reference marker was altered and YFP was replaced by red fluorescent protein (mRFP1) (32) to simplify spectral separation of the two fluorescence signals.
Targeting the reference to the nucleus provides a more convenient object for reliable semi-automated quantitative measurements. As shown in Figure 3E–G, nlsRm-2A-GH (Figure 1L) successfully targeted mRFP1 to the nucleus via a nuclear localization signal at its amino-terminus, while GFP was targeted to the ER as before. When the ER-retrieval signal on the GFP-HDEL moiety of nlsRm-2A-GH was replaced by the FLAG epitope tag to generate nlsRm-2A-secGf (Figure 1K), RFP fluorescence was still detected in the nucleus, but little GFP accumulated in the endomembrane system (compare Figure 3G and H), consistent with the expected export of the secGFPf moiety from the ER.
Targeting the reference marker to the ER may provide a better quantitative (intensity) and qualitative (morphology) reference for perturbation of the early secretory pathway as it would co-localize precisely with the accumulated trafficking marker. A 2A-based polyprotein that generated an ER-localized reference marker would require both the amino-terminal (RFP) and carboxy-terminal (GFP) moieties to be translocated across the ER membrane. In such a configuration, the timing of the proposed 2A cleavage is expected to be critical. If chain separation occurs before the signal peptide of the second fluorescent protein encounters the translocation channel, this signal peptide could be inserted independently into the translocon, carboxy-terminus first, allowing translocation as normal into the ER lumen. In contrast, if the nascent GFP-HDEL or secGf moiety emerged from the ribosome before 2A-mediated separation had occurred, the signal peptide would enter the translocation channel amino-terminus first and would be expected to act as a stop-transfer signal owing to its hydrophobic nature (33–35). This would cause the GFP moiety to accumulate on the cytoplasmic face of the ER membrane where the ER-retrieval signal would not encounter the luminal retrieval receptor.
To test whether the 2A peptide was able to generate stoichiometric quantities of ER-localized reference and trafficked markers, we constructed a secreted-RFP/ER-GFP dual reporter (secN-Rm-2A-GH) and a Golgi-RFP/ER-GFP dual reporter (STN-Rm-2A-GH) (Figure 1R and O) using either a cleavable signal peptide or the Golgi-targeting signal of rat sialyl transferase (ST), as described previously (6). We also constructed plasmids in which the HDEL signal of the GFP moiety was replaced by the FLAG epitope tag to test whether ER residency of the reference GFP marker was dependent on this retrieval signal as expected. The RFP moieties of these constructs incorporated N-glycosylation sites to allow glycan processing to be assessed in trafficking assays (6).
When expressed in tobacco leaf epidermis, STN-Rm-2A-GH efficiently targeted RFP to mobile punctate structures typical of the Golgi apparatus, while GFP fluorescence appeared to be confined exclusively to the ER (Figure 3I–K). Red fluorescent protein was not observed in the ER except at the early stages of transient expression, when faint ER labelling could sometimes be detected most probably due to the transport of new protein through the ER to the Golgi (6). When the HDEL signal on the GFP moiety was replaced with the FLAG epitope tag to generate STN-Rm-2A-secGf (Figure 1N), GFP accumulation was dramatically reduced relative to STN-Rm-2A-GH (compare Figure 3K and L), consistent with secretion of secGFPf. Thus, it appears that the GFP moieties were indeed translocated into the lumen of the ER where GFP-HDEL was retained by virtue of its carboxy-terminal HDEL signal while secGFPf was exported and were not simply associated with the cytoplasmic face of the ER membrane via an uncleaved signal peptide.
When secN-Rm-2A-GH and secN-Rm-2A-secGf were expressed in tobacco leaf epidermis, GFP accumulated in the ER in an HDEL-dependent fashion as expected, but RFP was unexpectedly observed to accumulate in the vacuole rather than the apoplast (Figure 3M–P). Indeed, the apoplast was evident as a dark line between adjacent cells (Figure 3M and P). It has been shown that mRFP1 is stable and fluorescent in the apoplast of tobacco epidermal cells [(3); see also Figure 4R below] so the absence of signal suggests that the RFP moiety of secN-Rm-2A constructs has not been transported there to any significant extent. In cells that expressed STN-Rm-2A-GH and STN-Rm-2A-secGf to comparatively high levels, RFP fluorescence was also observed in the vacuole (Figure 3I). This suggested that the RFP-2A moieties of both constructs were unexpectedly sorted to the vacuole with high efficiency after leaving the Golgi.
Figure 4. (on previous page) 2A acts as a vacuolar-sorting determinant in tobacco leaf epidermis. Confocal images of tobacco leaf epidermal cells transiently expressing various fluorescent proteins as indicated. (A–D′) secN-Rm-2A-GH infiltrated at OD600 0.25 and examined 48 h later. RFP (red) is visible in the vacuole (A and C′) and associates with the ER network marked by GFP (green) and is also clearly visible in distinct punctate structures that excluded the ER marker (D, arrows). B′–D′ show single sections through the vacuole of cells whose cortical cytoplasm is shown in projection (B–D). Chlorophyll fluorescence is indicated in blue. (E–G) secN-Rm-2A-secGf similarly imaged to reveal RFP and GFP in transit through the endomembrane system; each protein labels the ER network and putative PVC (arrows) and Golgi stacks (arrowheads). (H) A projection of cells expressing secN-Rm-2A-secGf, indicating that GFP but not RFP is detectable in the cell wall (arrowheads). (I–P) secN-Rm-2A-GH infiltrated either alone (I–L) or with a strain expressing YFP-RAB-F2b[S24N] (M–P). In the absence of the YFP-RAB-F2b[S24N], RFP signal is absent from the cell wall and confined to the vacuole (I and asterisked cells in M); however, in cells co-expressing the Rab GTPase mutant (numbered 1–5 in M–P), strong RFP fluorescence is detected in the apoplast in addition to the vacuole (arrowheads). The arrow indicates RFP that has apparently diffused from cells 1, 4 and 5 into the adjoining wall between two non-infected cells. RFP is in red, GFP in green and YFP in blue. (Q) Schematic representation of secRFP and secRFP-2A constructs. (R and S) secRFP accumulates exclusively in the apoplast (R) while secRFP-2A is almost exclusively vacuolar (S). (A, B′–D′) and (E–G) are single sections; (B–D, H, I–P) and (R and S) are projections. Scale bars 20 μm (I–P and R and S), 5 μm (H) and 2 μm (E–G).
2A acts as a sorting determinant for the Rab-F2 PVC in plant cells
To establish whether the RFP moiety of secN-Rm-2A constructs passes through the endomembrane system en route to the vacuole, we sought to visualize it in transit at steady state. Therefore, leaves were infiltrated with the secN-Rm-2A-GH strain at OD600 0.25, five times the usual titre and were examined by confocal microscopy 18 h after the onset of transient expression (48 h post-infiltration) when steady-state levels of RFP in transit through the endomembrane system would be at their highest (Figure 4A–D′). Under these conditions, RFP was visible in the vacuole (Figure 4A, C′), but much of the RFP moiety was associated with the ER network marked by GFP (Figure 4A, B and D). Red fluorescent protein was also clearly visible in distinct punctate structures that excluded the ER marker and probably represented intermediates on a vacuolar pathway (Figure 4D, arrows). When cells expressing secN-Rm-2A-secGf under similar conditions were imaged to reveal both RFP and GFP in transit through the endomembrane system, each labelled the ER network (Figure 4E–H), while GFP, but not RFP, was detectable in the cell wall at these high-expression levels (Figure 4H, arrowheads). Both markers, however, labelled small bright punctate structures (arrows in Figure 4E–G). These probably represent the same punctate RFP-labelled intermediates observed with secN-Rm-2A-GH (Figure 4D, arrows). Similar structures were previously observed with secreted and vacuolar GFP markers and identified as a PVC (3,12), suggesting that RFP may follow the same route to the vacuole. To test this possibility, we asked whether transport of the RFP moiety of secN-Rm-2A-GH to the vacuole was dependent on Rab GTPases of the Rab-F2 subclass. It has been shown that a YFP-tagged member of this subclass, AtRAB-F2b (ARA7) of Arabidopsis, localizes predominantly to this PVC in tobacco leaf epidermis and that the S24N mutant of this protein (YFP-RAB-F2b[S24N]) causes the vacuolar marker aleuGFP to be mis-sorted to the vacuole (12). The secN-Rm-2A-GH was infiltrated either alone (Figure 4I–L) or with a strain expressing YFP-RAB-F2b[S24N] and epidermal cells were imaged 54 h later (Figure 4M–P). In the absence of YFP-RAB-F2b[S24N], RFP signals were absent from the cell wall and confined to the vacuole (Figure 4I and asterisked cells in Figure 4M). However, in cells co-expressing YFP-RAB-F2b[S24N], strong RFP fluorescence was detected in the apoplast in addition to the vacuole.
It is notable too that RFP and GFP moieties of secN-Rm-2A-secGf were additionally observed in punctate structures that were larger and fainter than the PVC (arrowheads in Figure 4E–G). Similar structures have previously been observed with secGFP and identified as Golgi (3). Interestingly, the secN-RFP signal in these structures relative to the ER and PVC was significantly lower than that of secGFP, suggesting that its residence time in the Golgi may be lower. It also confirms that separation of RFP and GFP moieties occurs before the Golgi. We concluded that both the RFP and GFP moieties of the secN-RFPmyc-2A and STN-RFPmyc-2A fusions were translocated across the ER membrane, but that the RFP moieties were then sorted from the Golgi to the vacuole via the conventional PVC/Rab-F2-dependent pathway rather than following the default pathway to the apoplast.
We next investigated the location of the vacuolar-sorting signal within the RFP moiety of the 2A constructs. To test whether the signal resides within the 2A peptide itself, this sequence was fused to the carboxy-terminus of secRFP (Figure 4Q), a previously described secreted RFP molecule (3), and expressed in tobacco epidermal cells. While secRFP accumulated exclusively in the apoplast without appearing in the vacuole, as reported previously (3), secRFP-2A, which carried the 2A sequence was almost exclusively vacuolar (Figure 4R and S). The 2A sequence is, therefore, sufficient for vacuolar targeting of mRFP1 in the plant endomembrane system.
2A fusions behave similarly in tobacco epidermis and Arabidopsis seedlings
We were interested in knowing whether the 2A-based markers would exhibit similar behaviour in other plant species, particularly the genetic model organism Arabidopsis thaliana. Therefore, all of the constructs mentioned above were used to generate stable transformants in Arabidopsis and the roots and leaves of T3 homozygous plants were analysed by confocal microscopy. In every case, we observed efficient separation of YFP or RFP moieties from ER-resident or secGFP in the leaves. Notably, the STN-RFPm-2A and secN-RFPm-2A proteins accumulated clearly in the vacuoles of leaves as observed in tobacco leaf epidermis (Figure 5). Similar behaviour was observed in roots (data not shown) with the exception of STN-RFPm-2A-GH where RFP signals were clearly associated with the ER, particularly in meristematic and elongating cells. Thus, the 2A-based constructs behave similarly in the two most common dicot models, Arabidopsis and tobacco, and may, therefore, have general utility.
Figure 5. 2A fusions in Arabidopsis seedlings. Confocal microscopy of leaves of transgenic Arabidopsis plants expressing (A–C) Ym-2A-GH, (D) Ym-2A-secGf, (E–G) nlsRm-2A-GH, (H) nlsRm-2A-secGf, (I–K) STN-Rm-2A-GH, (L) STN-Rm-2A-secGf, (M–O) secN-Rm-2A-GH, (P) secN-Rm-2A-secGf. YFP and RFP channels are shown in red and the GFP channel in green. Merged channel images are indicated. (A–H) and (L) are single sections and while (I–K) and (M–P) are projections. Scale bars 50 μm (A–D) and 10 μm (E–P).
2A fusions involving YFP and RFP are not cleaved with equal efficiency
Confocal analysis suggested that the 2A fusions were cleaved with high efficiency as upstream and downstream moieties accumulated predominantly in distinct compartments. To investigate the efficiency of cleavage directly, protein extracts were prepared from transfected tobacco leaves or transgenic Arabidopsis seedlings and analysed on immunoblots using anti-GFP or anti-c-myc antisera. The Ym-2A, nlsRm-2A and STN-Rm-2A fusions each gave rise to bands of the expected molecular weight when probed with anti-c-myc antisera in tobacco (Figure 6A), although the YFPm-2A moiety accumulated as two bands in Arabidopsis (Figure 6C, lanes 2 and 3). Occasionally, a second higher band was seen, suggesting some heterogeneity of cleavage or some instability at one or other terminus of the primary cleavage product (data not shown). The same result was obtained with secN-Rm-2A-GH and secN-Rm-2A-secGf (data not shown). None of the fusions involving RFP gave rise to detectable uncleaved translation product. In contrast, a faint band corresponding to the uncleaved YFP fusions was consistently detectable with Ym-2A-GH and Ym-2A-secGf (Figure 6A and C, arrowheads).
Figure 6. Immunoblot analysis of 2A fusions in transfected tobacco leaves. Immunoblot analysis using anti-c-myc antiserum (A and C) or anti-GFP/YFP antiserum (B and D) of protein extracts prepared from tobacco leaves (A and B) or Arabidopsis transgenic seedlings (C and D) expressing Ym-2A-GH (lane 2), Ym-2A-secGf (lane 3), nlsRm-2A-GH (lane 4), nlsRm-2A-secGf (lane 5), STN-Rm-2A-GH (lane 6), STN-Rm-2A-secGf (lane 7) or non-transfected/non-transgenic plants used as a control (lane 1). Major bands of the expected mobility were detected in all cases, however, full-length, uncleaved, translation product was detected to varying extents with the YFP fusions (lower arrowheads) and mRFP1 fusions (upper arrowheads). (E) Immunoblot analysis using anti-GFP/YFP antiserum of Ym-2A-GH and nlsRm-2A-GH tested over a 10-fold range of Agrobacterium titres (OD600 as indicated). (F) Immunoblot analysis using anti-GFP/YFP antiserum of Ym-2A-GH over a 20-h period from 50 to 70 h post-infiltration. Lanes labelled U (E and F) show extracts of un-infiltrated leaf. Arrows indicate the positions of markers whose molecular weight is indicated in kDa.
We also probed the same protein extracts with anti-GFP antibodies. This revealed major bands of the expected molecular weight (29 kDa) for all constructs. Consistent with the fluorescence data, ER-resident GFP expressed from the -2A-GH constructs was more abundant than the secreted protein of the -2A-secGf constructs. Faint bands corresponding to the uncleaved product were detectable with the STN-Rm-2A fusions (Figure 6B and D, lanes 6 and 7, upper arrowheads), but significantly more uncleaved Ym-2A-GH and Ym-2A-secGf fusion proteins were detected in both tobacco and Arabidopsis (Figure 6B and D, lanes 2 and 3, lower arrowheads). Even allowing for the fact that Ym-2A fusions bear twice as many GFP epitopes as the RFP-based fusions, it is clear that a greater proportion of the YFP fusions accumulated in the uncleaved form. The ratio of full-length and cleaved translation product did not change when Ym-2A-GH or nlsRm-2A-GH were expressed using a 10-fold range of Agrobacterium titres (Figure 6E) or when Ym-2A-GH was monitored over a 20-h period starting approximately 20 h after the onset of transient expression (50–70 h post-infiltration; Figure 6F).
Thus, it appears that the 2A fusions with RFP at their amino-termini were cleaved with greater efficiency than those with YFP at that position. It is notable that the uncleaved YFP fusions were barely detected by the c-myc antibody so, despite the fact that c-myc occurs only once in the fusion protein whereas the anti-GFP epitopes occur twice, the c-myc epitope is apparently not recognized efficiently in the context of the full-length Ym-2A fusions.
Ratiometric analysis of membrane traffic in cell populations
We selected nlsR-2A-secGf and YFP-2A-secGf to test the effectiveness of ratiometric approaches for quantifying biosynthetic membrane traffic in tobacco epidermal cells. The YFP-2A-secGf is suitable for use with the quantification method used in Figure 2 as the cytosolic YFP and endomembrane-localized GFP signals are distributed similarly in low-magnification images of transfected cells. The incomplete cleavage of this fusion may not compromise its use in trafficking assays because the uncleaved product appeared to be transported efficiently through the endomembrane system as judged by the lack of intracellular GFP signal in confocal images (Figure 3D). We reasoned that the YFP domain on the cytoplasmic face of endomembrane compartments will experience a constant environment during transit, while the GFP moiety will encounter the same luminal/apoplastic environments as the cleaved secGf product. Thus, the YFP and GFP fluorescence intensities from the uncleaved transmembrane molecule were expected to respond like those of the separated YFP and GFP moieties when biosynthetic trafficking is inhibited.
Ym-2A-secGf or Ym-2A-GH was expressed either alone or with the dominant N121I mutant of the Arabidopsis Rab GTPase, AtRAB-D2a (ARA5; AtRab1b; At1g02130), a homologue of mammalian Rab1 (36) that is known to block ER to Golgi transport (3,6,19). The two extremes of the assay were set by the signal from the secGFP marker alone (Figure 7A) and the GFP-HDEL marker (Figure 7C). Consistent with previous observations [(3,6) and H. Betts and I. Moore, unpublished observations], absolute measurements of GFP fluorescence intensity indicated that the dominant-negative AtRAB-D2a[N121I] mutant had little or no effect on the accumulation of GFP-HDEL already trapped in the ER (Figure 7D and E) but increased the accumulation of secGFP (Figure 7B and E) to 50–75% of the GFP-HDEL value (Figure 7E). When the ratiometric approach, which normalizes for expression level and imaging efficiency using the co-expressed YFP signal, was used to analyse the same images, the results differed in three notable ways (Figure 7F). First, the statistical significance of the data was improved as illustrated by the lower coefficients of variance (Figure 7G). Second, AtRAB-D2a [N121I] caused secGFP to accumulate to levels comparable to those of ER-resident GFP-HDEL (expressed from Ym-2A-GH), whereas the absolute expression data in Figure 7E and previous studies suggested a figure of only 50–75% of GFP-HDEL. This can be explained most simply by a previously undetected reduction in transient expression efficiency in leaf areas co-infiltrated with the AtRAB-D2a [N121I] mutant strain. Third, the Rab mutant also caused a small but significant (P = 0.05) increase in the accumulation of GFP-HDEL owing most probably to escape of some GFP-HDEL molecules from the ER in cells expressing Ym-2A-GH alone.
Figure 7. Absolute and ratiometric measurements of GFP trafficking in tobacco leaf epidermal cells. (A–D′) Low-magnification images of Ym-2A-secGf expressed alone (A and A′) or co-expressed with AtRAB-D2a[N121I] (B and B′) and compared with Ym-2A-GH expressed alone (C and C′) or co-expressed with AtRAB-D2a[N121I] (D and D′). GFP channel (A–D) is green and YFP channel (A′–D′) is red. Scale bar is 100 μm. (E and F) absolute (E) and ratiometric (F) measurements of secGFPf and GFP-HDEL accumulation in cells expressing Ym-2A-secGf (Y2AsecG) or Ym-2A-GH (Y2AGH) with or without AtRAB-D2a[N121I] (D2aNI); means of three experiments normalized to the value for Ym-2A-GH; error bars are SE. (G) Average coefficient of variance for absolute (open bars) and ratiometric (shaded bars) measurements of GFP accumulation; n = 5 except for Y2AGH + D2aNI where n = 3. Agrobacterium strains were used at OD600 0.05 (Ym-2A-secGf and Ym-2A-GH) or 0.03 (AtRAB-D2a[N121I]), and images were acquired 70–74 h after infiltration.
Ratiometric analysis of membrane traffic in individual cells
The 2A-based ratiometric approach can also be applied to individual cells imaged at high magnification where it can establish whether the accumulation or mis-localization of a trafficking marker results from inhibition of biosynthetic trafficking or from high rates of marker synthesis. For example in Figure 8A, we infer that cell 2 is presumably co-expressing the co-infiltrated dominant inhibitory AtRAB-D2a[N121I] mutant as it accumulates more GFP than neighbouring cells or control cells (Figure 8B) with similar nlsRFP accumulation. This approach is potentially versatile, as it can be applied to transfected protoplasts or to specific cells of transgenic plants where the low-magnification method described above is not readily employed.
Figure 8. (on previous page) Use of 2A-based ratiometric markers for analysis of individual tobacco leaf epidermal cells. Single optical sections of cells from infiltrated leaves expressing (A) nlsRm-2A-secGf and AtRAB-D2a [N121I], (B) nlsRm-2A-secGf, only, or (C) nlsRm-2A-GH; in (C) transects used for image analysis are displayed as white lines. In (A), the expression level of the marker is similar in cells 1 and 2 and similar to the control cells in (B) as indicated by the intensity of the red nuclei, however, only cell 2 accumulates secGFP (green); note that the weak GFP signal below nucleus 1 is in the adjacent cell 3; chlorophyll fluorescence is indicated in blue. (D) GFP intensity in the nuclear envelope and adjacent cytoplasm plotted against nuclear RFP intensity in individual cells expressing either nlsRm-2A-secGFPf (shaded squares) or nlsRm-2A-GH (open squares); each datum represents a measurement from a single cell; the dotted line through the data for nlsRm-2A-GH shows the best fit (R 2 = 0.85) for nlsRFP values from 0 to 150 above which GFP-HDEL accumulation plateaus. (E–H) Projections of a series of optical sections from tobacco leaf epidermis expressing the ratiometric marker nlsRm-2A-secGf and C-terminally tagged AtRAB-D2a[N121I]-YFP. Cell 1 expresses both constructs to high level as indicated by the nlsRFP signal (red) and YFP signal (blue) and accumulates high levels of secGFPf (green); cell 2 expresses less of the marker and accumulates less secGFPf; cell 3 expresses only the marker and indicates the extent of secGFPf accumulation in control cells while cell 4 expresses only the YFP-tagged Rab GTPase and indicates that there is minimal bleed-through into the GFP channel under these conditions; (E–H) show the individual detection channels and merged image as indicated. (I) Quantitative analysis of secGFPf accumulation relative to the expression of YFP-tagged AtRAB-D2a wild type (shaded squares) and N121I mutant (open squares) in individual tobacco leaf epidermal cells; to assess the extent of secGFPf accumulation, GFP fluorescence is expressed as a percentage of the GFP fluorescence that would be expected if nlsRm-2A-GH was expressed at the same level as nlsRm-2A-secGf in each cell, based on its individual RFP signal and the slope of the line in (D) which was obtained from control cells expressing nlsRm-2A-GH only; GFP data is corrected for bleed-through of YFP fluorescence into the GFP channel using a function determined by analysis of cells that expressed only the YFP-tagged wild-type Rab GTPase. (J–M) Projections of optical sections through the cortical cytoplasm of cells from leaf areas co-transfected with the Ym-2A-secGf and nlsRm-2A-D2a[NI] constructs; (J–M) show the individual detection channels and merged image as indicated. (N) Ratiometric measurements of secGFPf accumulation in cells expressing Ym2AsecGf alone or together with nlsRm-2A-D2a[NI] or AtRAB-D2a[N121I] compared with Ym2AGH. Bars indicate means of three experiments normalized to the value for Ym-2A-secGf; error bars are SE. Scale bars in all images are 20 μm. Infiltrations were performed with nlsRm-2A-secGf and nlsRm-2A-GH strains at OD600 0.05 (A–C), 0.04 (D and I) or 0.02 (E–H); AtRAB-D2a -YFP and AtRAB-D2a[N121I]-YFP strains at OD600 0.02; and AtRAB-D2a[N121I] and nlsRm-2A-D2a[NI] at OD600 0.03.
Therefore, to quantify GFP accumulation ratios in individual cells, we developed an image analysis tool that can be used in conjunction with the nlsR-2A constructs. To measure signals unambiguously from specific cells, analysis was performed on the nucleus and perinuclear region of a series of confocal sections along the z-axis. Starting from a position in the nucleus, a series of transects was drawn across the cytoplasm into the surrounding vacuole (see Figure 8C). The maximum GFP signal along each transect was then extracted and averaged, while the nuclear RFP intensity was automatically extracted from the 3D data set from a user-defined seed. Figure 8D shows that in cells expressing nlsRm-2A-GH, under the imaging conditions used, there was a linear correlation (R 2 = 0.85) between the RFP and GFP fluorescence for RFP pixel intensities up to 150, after which GFP-HDEL signal plateaued at less than the maximum possible pixel intensity (255), suggesting saturation of the ER-retrieval mechanism at these expression levels. Similar analysis of cells expressing nlsRm-2A-secGf showed that little GFP accumulated in the nuclear envelope or ER over the entire range of RFP expression values (Figure 8D, shaded squares). The slopes of fitted lines indicated that GFP-HDEL accumulated in the ER 1000-fold more efficiently than secGFPf (GFP-HDEL = 1.2 × nlsRFP; secGFPf = 0.0009 × nlsRFP), providing a far greater dynamic range in this assay method in comparison to the low-magnification approach described hitherto (3) (Figures 2 and 7).
Ratiometric analysis of membrane traffic with quantification of effector levels
In the assays described so far, the presence or absence of the dominant inhibitory mutant in any cell could only be inferred from the value of the secGFPf ratio. Clearly if the protein can be visualized by tagging with a third fluorescent protein, its relative abundance in each cell could be determined directly and correlated with its effect on marker trafficking. Tagging AtRAB-D2a[N121I] with YFP at its amino-terminus renders it inactive and apparently unstable (data not shown), but fusions to the carboxy-terminus were stable and appeared to exhibit some residual inhibitory activity. We, therefore, asked whether this low activity could be quantified using the nlsRm-2A-secGf trafficking assay. As shown in Figure 8E–H, accumulation of the secGFP marker was clearly dependent on the abundance of both RFP and AtRAB-D2a[N121I]-YFP in the same transfected cell (compare cells 1 and 2 with cell 3). Using the image analysis tool described above, the YFP, RFP and GFP signal intensities in such images were quantified in cell populations co-infected with nlsRm-2A-secGf and either wild-type or mutant YFP-tagged AtRAB-D2a. For each cell, the YFP intensity was plotted against the GFP-to-RFP ratio. To provide a standard against which to compare the relative data generated by this analysis, the secGFP signal in each cell was expressed as a percentage of the GFP signal that would be expected for nlsRm-2A-GH expressed at the same level. Thus, the RFP intensity was used to read off the expected GFP-HDEL intensity from the slope of the dotted line in Figure 8D derived from cells expressing nlsRm-2A-GH in the same experiment. Wild-type AtRAB-D2a-YFP had no effect on secGFPf accumulation, but AtRAB-D2a[N121I]-YFP caused secGFPf to accumulate to approximately 10% of GFP-HDEL in the most highly expressing cells, confirming that it retained some residual inhibitory activity (Figure 8I). Note that this small increase would not be measurable using the previous low-magnification approach (Figure 7).
Use of 2A to monitor the expression level of cytoplasmic effectors
As direct tagging is likely to be problematic for many effector proteins, we next asked whether the 2A peptide could be used to provide a stoichiometric marker for the expression level of proteins such as AtRAB-D2a[N121I] which exhibit little or no activity when tagged at either terminus. Linking such a protein to a fluorescent marker via the 2A peptide is predicted to leave just a single additional proline residue at the amino-terminus of the tagged protein, which may, therefore, retain functionality. Thus, we fused the AtRAB-D2a[N121I] ORF downstream of the nlsRFPm-2A coding sequence to generate nlsRm-2A-D2a[NI]. nlsRFPm-2A was chosen because it appeared to be more efficiently cleaved than YFPm-2A and should give a readily measurable signal in the nucleus of each transfected cell. This construct was then used in conjunction with Ym-2A-secGf to assay GFP trafficking. When individual cells were imaged at high magnification, it was evident that high GFP-to-YFP ratios correlated with high-RFP fluorescence in the nuclei (Figure 8J–M).
In these and other membrane trafficking experiments, nlsRm-2A-D2a[NI] typically exhibited a weaker inhibitory influence on trafficking of secreted and Golgi markers than untagged AtRAB-D2a[N121I] (Figure 8N). Assays of secGFP accumulation using the low-magnification ratiometric method with Ym-2A-secGf confirmed that nlsRm-2A-D2a[NI] increased secGFPf accumulation by approximately half as much as native AtRAB-D2a[N121I]. This lower activity may reflect either lower expression efficiency, some sensitivity to the additional amino-terminal proline or inefficient separation of the Rab and RFP moieties leaving dysfunctional full-length protein. When immunoblots of extracts from cells expressing nlsRm-2A-D2a[NI] and control constructs were probed with anti-c-myc antibodies, we observed that approximately 50% of the protein accumulated in the unseparated form (Figure 9A). This result was surprising given the efficient separation of the RFP and GFP moieties in the 2A fusions described above. Therefore, we further investigated the mechanism responsible for cleavage of 2A-fusion proteins.
Figure 9. Analysis of fusion proteins with or without functional 2A sequences or signal peptides in tobacco epidermal cells. For each construct extracts were prepared from cells expressing two independently constructed plasmids. Arrows indicate the positions of markers and their molecular weights are given in kDa. (A) Immunoblot analysis of 2A fusions to AtRAB-D2a using anti-c-myc antiserum. Lanes: 1, uninfected tissue; 2 and 3, nlsRm-2A-D2a wt; 4 and 5, nlsRm-2A-D2a[NI]; 6 and 7, control nlsRm-2A-GH. (B) Immunoblot analysis using anti-GFP/YFP antiserum of constructs carrying wild-type or mutant 2A sequences. Lanes: 1, uninfected tissue; 2 and 3, Ym-2A*-GH; 4 and 5, Ym-2A-GH; 6 and 7, nlsRm-2A*-GH; 8 and 9, nlsRm-2A-GH. (C) Immunoblot analysis using anti-GFP/YFP antiserum of 2A fusions with or without a signal peptide. Lanes: 1, uninfected tissue; 2, Ym-2A-GH control; 3 and 4, Ym-2A-Gf; 5, nlsRm-2A-GH control; 6 and 7, nlsRm-2A-Gf; 8 and 9, Ym-2A-nlsRf. (D–L) Projections of series of confocal images of tobacco leaf epidermal cells expressing nlsRm-2A-GH (D–F), nlsRm-2A*-GH (G–I) and nlsRm-2A-Gf (J–L); RFP is in red, GFP in green, chlorophyll in blue, (F, I and L) merged channels. Scale bars 20 μm.
The FMDV 2A peptide exhibits little activity at the carboxy-terminus of YFP and RFP in tobacco leaf epidermis
The cleavage efficiency of the 2A protein is known to be influenced by sequences upstream but not by sequences downstream (21). Thus, the accumulation of uncleaved 2A fusion protein in cells expressing nlsRm-2A-D2a[NI] suggested that the apparently efficient cleavage of other nlsRFP-2A fusions may have resulted either from instability of the uncleaved fusion protein or from cleavage by a mechanism that is independent of 2A. One possibility for such a mechanism was cleavage by signal peptidase between the signal peptide and the GFP sequence of the various 2A-GH and 2A-secGf fusions.
To test these possibilities, we assembled a series of fusions that either lacked the signal peptide or carried a mutant form of the 2A sequence (here designated 2A*) in which the two terminal residues were each converted to alanine (Figure 1T, underlined). Either substitution is sufficient to prevent 2A-mediated polypeptide separation during translation in vitro or in mammalian cells (21,29,37). As shown in Figure 9B, Y m-2A*-GH and nlsRm-2A*-GH, which carried the mutant 2A sequence, gave rise to almost identical ratios of monomeric and dimeric translation products as the corresponding wild-type clones, although the mutant sequence sometimes resulted in lower quantities of expressed protein. Confocal analysis of transfected cells also indicated efficient separation of the GFP-HDEL moiety from the upstream YFP or RFP moiety of the 2A* fusions (Figure 9D–I and data not shown). Conversely, when the signal peptide was eliminated from Ym-2A-GH or nlsRm-2A-GH fusions, giving rise to Ym-2A-Gf and nlsRm-2A-Gf, or when nlsRFP was placed downstream of Ym-2A, the translation product accumulated predominantly as the uncleaved fusion (Figure 9C). Images of transfected cells confirmed that in the absence of a signal peptide, the GFP and RFP moieties of nlsRm-2A-Gf accumulated at the same ratio in nucleus and cytoplasm of infected cells (Figure 9J–L). Together, these results indicate that the presence of a signal peptide, but not a functional 2A sequence is critical for separation of the GFP moiety from the upstream RFP and YFP moieties in the Ym-2A-GH and nlsRm-2A-GH constructs. In the case of the nlsRm-2A-D2a proteins, as we do not have antibodies to detect the AtRAB-D2a moiety, it remains unclear whether the monomeric nlsRFPmyc species arose from partial 2A-mediated cleavage or instability of the Rab moiety in the fusion protein.
Given previous reports of successful separation of GFP-2A fusions in plant and animal cells and in vitro translation systems, we asked whether the presence of the c-myc epitope tag immediately upstream of the 2A sequence in all of our fusions influenced the results. Ynomyc-2A-Gf is identical to Ym-2A-Gf except for the absence of the myc sequence between the C-terminus of YFP and the 2A peptide (Figure 1D and E). As shown in Figure 10, this construct exhibited the same intracellular distribution as Ym-2A-Gf in transfected cells and accumulated predominantly as a dimer when analysed by immunoblot (Figure 10G). Thus, the behaviour of 2A in the constructs described here is not determined by the presence of the c-myc sequence at its amino-terminus.
Figure 10. Confocal and immunoblot analysis of Ynomyc-2A-Gf. Projections of series of confocal images of tobacco leaf epidermal cells expressing (A–C) Ynomyc-2A-Gf and (D–F) Ym-2A-Gf; YFP is in red and GFP is in green. (G) Immunoblot analysis of transfected tobacco leaf tissues using anti-GFP/YFP antiserum. Lane 1, uninfected tissue; lanes 2 and 3, independent Ynomyc-2A-Gf constructs; lanes 4 and 5, independent Ym-2A-Gf constructs. The minor bands detected (arrowhead) represent the cleaved translation products. Arrows indicate the positions of markers whose molecular weight is indicated in kDa.
The 2A sequence is required for correct assembly and targeting of STN-RFP-2A-GH translation products
The conclusion that nlsRFPmyc-2A is cleaved inefficiently in plant cells raised questions about the mechanism responsible for the correct insertion of RFP and GFP moieties into the ER during translation of STN-Rm-2A-GH and secN-Rm-2A-GH. As discussed earlier, because the upstream RFP moiety of these constructs is translocated through the ER, if cleavage did not occur the signal peptide of the GFP moiety would be expected to enter the ER translocation channel amino-terminus first where it would act as a stop-transfer signal leaving the GFP moiety, its signal-peptidase cleavage site, and its ER-retrieval signal on the cytoplasmic face of the ER membrane. In contrast, GFP appeared to be separated efficiently from the RFP moiety and to accumulate in the ER lumen (e.g. Figure 3I–P and 6B). To investigate whether 2A is indeed essential for the correct processing of STN-Rm-2A-GH, a derivative carrying the mutant 2A sequence was constructed. When this was expressed in tobacco leaf epidermal cells, the mutant 2A sequence caused a dramatic reduction in the accumulation of both the RFP and the GFP moieties (Figure 11A–F). When transfected cells were imaged under conditions that allowed the accumulated protein to be visualized, it was clear that some RFP was present in structures typical of the Golgi, but a significant proportion was associated with the ER which also accumulated a small quantity of GFP (Figure 11G and H). Immunoblot analysis confirmed that very little STN-Rm-2A*-GH translation product accumulated and that all the visibly accumulated protein was full length (Figure 11I). Thus, the presence of a wild-type 2A sequence is responsible for the efficient processing, sorting and stable accumulation of STN-Rm-2A-GH translation products. As discussed below, in light of the results with mutant 2A peptides in other proteins, it may be that correct processing of STN-Rm-2A-GH and secN-Rm-2A-GH constructs is dependent on a property of the 2A peptide that is unrelated to its self-cleaving activity.
Figure 11. The 2A sequence is important for proper assembly and targeting of STN-Rm-2A-GH. Projections of series of confocal images of tobacco leaf epidermal cells expressing STN-Rm-2A-GH (A–C and G) or STN-Rm-2A*-GH (D–F and H). RFP is in red and GFP in green. (G and H) merged channels from images acquired with enhanced sensitivity allowing the translation products of STN-RFPm-2A*-GH to be visualized (chlorophyll is in blue). Scale bars 20 μm (A–C and D–F) and 10 μm (G and H). Agrobacterium strains were inoculated at OD600 0.06. (I) Immunoblot analysis using anti-GFP/YFP antiserum. Lane 1, uninfected tissue; lane 2, nlsRm-2A-GH; lane 3, nlsRm-2A*-GH; lanes 4 and 5 two independently constructed STN-Rm-2A*-GH fusions; lane 6, STN-Rm-2A-GH. The weak bands in lanes 4 and 5 (arrowhead) represent the uncleaved translation product of STN-Rm-2A*-GH. Arrows indicate the positions of markers whose molecular weight is indicated in kDa.