The dark side of green fluorescent protein

Authors


Author for correspondence: Timothy C. Hall Tel: +1 979 8457750 Fax: +1 979 8624098 Email: tim@idmb.tamu.edu

Summary

  • • Here, severe interference of chlorophyll with green fluorescent protein (GFP) fluorescence is described for medicago (Medicago truncatula), rice (Oryza sativa) and arabidopsis (Arabidopsis thaliana). This interference disrupts the proportional relationship between GFP content and fluorescence that is intrinsic to its use as a quantitative reporter.
  • • The involvement of chlorophyll in the loss of GFP fluorescence with leaf age was shown in vivo, by the removal of chlorophyll through etiolation or by ethanol extraction, and in vitro, by titration of a GFP solution with chlorophyll solutions of various concentrations.
  • • A substantial decrease in fluorescence in early development of medicago and rice leaves correlated with chlorophyll accumulation. In all three species tested, removal of chlorophyll yielded up to a 10-fold increase in fluorescence. Loss of GFP fluorescence in vitro was 4-fold greater for chlorophyll b than for chlorophyll a.
  • • Differences exist between plant species for the discrepancy between apparent GFP fluorescence and its actual level in green tissues. Substantial errors in estimating promoter activity from GFP fluorescence can occur if pigment interference is not considered.

Introduction

The ability to visualize, track and quantify gene products in living tissues is essential for understanding the function of genes and the biological processes in which they are involved. The advent of fluorescent proteins, such as green fluorescent protein (GFP) from the jellyfish Aequorea victoria (Chalfie et al., 1994) and its variants, has endowed researchers with powerful tools to monitor gene expression (Harper et al., 1999; Sunilkumar et al., 2002) and gene silencing (Ruiz et al., 1998), and to define spatial (Johnson et al., 2005), developmental (Marton et al., 2005) and quantitative (Tang & Newton, 2004) properties of promoters. The lack of requirement for exogenous substrate, cofactors or histochemical fixation makes GFP a particularly valuable reporter for plant cells, which are not as permeable as animal cells (Hanson & Köhler, 2001). Several modifications of the native GFP have improved its thermotolerance and fluorescence, especially in plants (Siemering et al., 1996; Haseloff et al., 1997). For these reasons, GFP and its variants are widely used as attractive alternatives to β-glucuronidase (GUS) (Jefferson et al., 1987) for assessment of promoter function and in optimizing plant transformation (Stewart, 2001).

The success of GFP for macroscopically monitoring the existence and level of transgene expression in intact tissues or organs in promoter or coding region fusions relies upon the existence of a close correlation between the detected intensity of GFP fluorescence and GFP protein content in the plant tissue. Therefore, in studies characterizing transgene expression in the model legume medicago (Medicago truncatula), a surprising discovery was that a dramatic decrease in fluorescence (using either a 500-nm long-pass filter or a 525-nm band-pass filter) from GFP occurred shortly after leaf emergence. In contrast, no striking change in fluorescence with age was detected for nonphotosynthetic organs such as roots and petals (Zhou et al., 2004). Similar results were obtained using either 500 nm long-pass or 525 nm band-pass filters. Another surprise was the presence of similar levels of GFP transcript for young and old leaves, making it unlikely that gene silencing was the primary cause.

A decrease in observed GFP fluorescence with leaf maturation from a similar 35S-mGFP5er construct has been reported in transgenic tobacco (Harper & Stewart, 2000) and oilseed rape (Halfhill et al., 2003). Whereas the latter article attributed the decrease in GFP fluorescence to changes in soluble protein content, we now provide several lines of evidence suggesting that chlorophyll is a major culprit in the loss of detectable GFP fluorescence as leaves mature. In addition, we show that a decrease in the proportion of GFP in the total protein plays a role. The novel finding of the interference of chlorophyll with GFP fluorescence reveals a major complication in the application of GFP to plant research that deserves careful evaluation.

Materials and Methods

Plant material

Transformation of 35S-mGFP5er into medicago (Medicago truncatula) using the binary vector pCB302-phas-GUS in Agrobacterium tumefaciens strain EHA105 was performed as described previously (Zhou et al., 2004). Arabidopsis (Arabidopsis thaliana) plants also transformed (Bechtold & Pelletier, 1998) with 35S-mGFP5er were kindly provided by Tao Wang (Texas A&M University, TX, USA) and rice (Oryza sativa) plants transformed (Dong et al., 1996) with 35S-mGFP5er were generously donated by Yiming Jiang (Texas A&M University, TX, USA).

Fluorescence imaging and quantification

For qualitative detection of whole plants, an LT-9700 Little Luma excitation light (Lightools, Encinitas, CA, USA) was used together with a 470/40-nm filter and 500-nm (long-pass) emission filter, and images were taken with an Olympus Camedia C-3040ZOOM digital camera (Olympus, Melville, NY, USA).

Green fluorescence in leaves of plants transgenic for 35S-mGFP5er was recorded with a Zeiss SV11 stereomicroscope (Carl Zeiss, Oberkochen, Germany) coupled to an AxioCam HRc digital camera (Carl Zeiss), using 470-nm excitation and 500- or 525-nm emission filters. Quantification of fluorescence, as the average signal intensity in a selected area, was performed on digitized pictures captured in the presence of a 525-nm emission filter using Zeiss AxioVision software (Carl Zeiss).

Leaf segments were vacuum-infiltrated with water and mounted on a glass slide under a coverslip. Hand-cut cross-sections were mounted in the same manner. Specimens were viewed using a Zeiss Axiophot (Carl Zeiss) microscope equipped with a 63x/1.4 Plan Apochromat oil immersion objective and a GFP fluorescence filter set (excitation 470/20 nm; emission 525/25 nm). Image stacks were acquired at a 0.5-µm step between optical sections, using a Coolsnap cf CCD camera (Photometrics, Tucson, AZ, USA). The contrast and brightness of the stacks were then adjusted with ImageJ (http://rsb.info.nih.gov/ij) and projected (brightest point method) for extended depth of field. Differential interference contrast images of the same field of view were also captured. Assembly and annotation of figures were performed using Canvas 9 (ACD Systems, Saanichton, BC, Canada).

Protein extraction

Approximately 300 mg of leaf tissue was ground into fine powder in liquid nitrogen using a mortar and pestle, and transferred into an Eppendorf tube. Total leaf protein was then extracted with TE buffer [10 mm Tris and 1 mm ethylenediaminetetraacetic acid (EDTA), pH 8.0] for 20 min and centrifuged at 1.2 × 104 g for 20 min. The supernatant was collected and the total protein concentration was determined (Bradford, 1976).

GFP fluorometry

The fluorescence of plant samples (100 µg protein) was determined (Bio-Rad VersaFluor fluorometer; Bio-Rad, Hercules, CA, USA) relative to a standard curve for a GFP (Clontech, Mountain View, CA, USA) concentration range of 100 ng ml−1 to 1 µg ml−1 using 480/20-nm excitation and 510/10-nm emission filters.

To determine the effect of chlorophyll on GFP fluorescence, 1 mg of high-performance liquid chromatography (HPLC)-purified chlorophyll a or b (Sigma, St Louis, MO, USA) was dissolved in 2 ml methanol containing 20 mg ml−1 sodium (Na) cholate (Griffiths, 1978), dried in a Speed Vac Concentrator (Savant, Milford, MA, USA), and resolublized in 2 ml TE buffer. Various amounts of chlorophyll solution were then mixed with 1 µg of standard enhanced green fluorescent protein (EGFP) (Clontech) in a total volume of 1 ml TE buffer containing 20 mg ml−1 Na cholate. The relative fluorescence of the samples was determined with a Bio-Rad VersaFluor fluorometer calibrated to 0 RFU (relative fluorescence units) with a blank of TE buffer containing 20 mg ml−1 Na cholate and 1000 RFU with 1 µg of EGFP in TE buffer (1 ml) containing 20 mg ml−1 Na cholate.

Western blot analysis

Medicago leaf extract (5 µg protein) or rice leaf extract (20 µg protein), from various leaf positions, was separated on a sodium dodecyl sulphate (SDS)-polyacrylamide (5% stacking and 10% resolving) gel, then electrophoretically transferred to Immun-Blot PVDF membrane (Bio-Rad) using a Mini Trans-Blot Electrophoretic Transfer Cell (Bio-Rad). The membrane was incubated with a primary rabbit anti-GFP polyclonal antibody (Clontech) at 1 : 2000 dilution for 5 h at room temperature, and then incubated with a secondary goat-antirabbit antibody conjugated with alkaline phosphatase (Novagen, Madison, WI, USA) at 1 : 10 000 dilution for 1 h at room temperature. Hybridization signals were visualized by exposing the membrane to 1 ml CDP-Star Substrate (Novagen) and then recorded with a UVP BioImaging system (UVP Inc., Upland, CA, USA).

Chlorophyll content determination

Relative chlorophyll concentrations in leaves were determined using a SPAD-502 Chlorophyll Meter (Minolta, Mahwah, NJ, USA).

Results

GFP fluorescence driven by the CaMV 35S promoter diminishes in maturing transgenic medicago leaves

Medicago plants transgenic for the mGFP5er reporter driven by the putatively constitutive Cauliflower mosaic virus (CaMV) 35S promoter display an intriguing temporal–spatial pattern of fluorescence (Zhou et al., 2004). Using an excitation wavelength of 470 nm together with 500- or 525-nm emission filters, which reduce or eliminate red fluorescence from chlorophyll, strong GFP fluorescence can be observed in young seedlings, the root system, petals, meristematic tissue and emerging leaves. However, as young medicago leaves develop, a progressive loss of GFP fluorescence in aerial parts takes place, eventually resulting in the apparent absence of green fluorescence in mature leaves. Only red fluorescence from chlorophyll or a blank background is observed when 500- or 525-nm emission filters are used, respectively (Fig. 1a). To follow changes in fluorescence with leaf maturation, medicago leaves were numbered according to their relative positions on the shoot, with the emerging leaf closest to the shoot tip being labeled as number 1. To quantify the intensity of green fluorescence in leaves, digital images were taken (525-nm emission filter) using a Zeiss SV11 microscope fitted with an AxioCam Hrc digital camera. The average intensity of green light in a selected area was measured using AxioVision software (Carl Zeiss). A dramatic decrease in green fluorescence per unit area was evident during leaf maturation, the most rapid loss being during early development (from leaf 1 to leaf 3); leaf 5 and older leaves had only approximately one-eighth (12.9/100) of the fluorescence level of the first leaf (Fig. 1b and c).

Figure 1.

Loss of fluorescence in maturing medicago leaves reflects increased chlorophyll and decreased green fluorescent protein (GFP) concentrations. (a) Comparison of GFP fluorescence from nontransgenic (wt) and young to mature leaves (leaf positions 1–5) of medicago transgenic for 35S-mGFP5er. Images were recorded using 500- or 525-nm emission filters (top and bottom panels, respectively). The wild-type leaf is equivalent to a position 1 leaf. (b) Fluorescence (open circle) decreased and chlorophyll content (closed circle) increased with leaf maturity. Error bars indicate standard deviation. (c) Comparison of relative GFP fluorescence for intact leaves and leaf extracts with leaf age (position). (d) Western blot analysis of GFP in medicago transformed with 35S-mGFP5er. Extracts of leaves from positions 1–5 were loaded on a basis of equal total protein (5 µg: lanes 5–9) or fluorescence [50 RFU (relative fluorescence units): lanes 10–14]. Enhanced green fluorescent protein (EGFP) standards are shown in lanes 2–4 and a wild-type leaf extract (5 µg protein) was used as control (lane 1). (e) Leaves covered for 1, 4, 5, and 6 d were recorded using either a 500-nm emission filter (top panels) or a 525-nm emission filter (bottom panels). Wild-type leaves covered for 6 d were included as controls (wt). (f) The relative fluorescence (open circle) was plotted against chlorophyll content for leaves shown in (e). All pictures were exposed for 5 s. RU, relative units.

Decrease in GFP fluorescence during leaf maturation greatly exceeds the loss of GFP as a proportion of total medicago leaf protein

Leaves of various stages of maturity from medicago plants transformed with 35S-mGFP5er were ground in liquid nitrogen and extracted with 1 ml of TE buffer (see Materials and Methods). Total protein content was determined (Bradford, 1976) and the fluorescence of 100 µg protein determined relative to a standard curve for GFP (Clontech) using 480-nm excitation and 510-nm emission filters. A substantial decrease in relative GFP fluorescence was evident commensurate with leaf age (position), as shown in Fig. 1c.

The decrease in GFP content with leaf maturation was further substantiated by immunohybridization analysis using antibodies to GFP (Fig. 1d). As shown in Fig. 1d, using equal amounts (5 µg) of medicago leaf protein, the youngest leaf (leaf 1) yielded a stronger signal than older leaves, whereas loading samples on the basis of equal RFUs yielded similar western signal intensities. These data indicate that, in medicago, the relative concentration of GFP in total protein decreases with leaf maturity, contributing to the loss of fluorescence with leaf age. However, the ratio for loss of GFP fluorescence in intact leaves (100/12.9 = 7.8) was almost 3-fold greater than that in leaf extracts (100/36.9 = 2.7), unveiling the presence of a major interfering factor (or factors) in medicago leaves that are not present in leaf extracts (Fig. 1c).

Inverse correlation between chlorophyll level and fluorescence intensity observed in vivo implies the involvement of chlorophyll in the disappearance of GFP fluorescence

Because the apparent loss of GFP fluorescence is most evident in leaves, a likely candidate is interference by chlorophyll. The fact that chlorophyll absorbs strongly at the excitation wavelength of GFP protein, and the absence of chlorophyll in aqueous leaf extracts (for which there is little loss of fluorescence with leaf age, adjusted for GFP content; Fig. 1c and d) provide clues that chlorophyll is a likely culprit in the disappearance of GFP fluorescence.

A SPAD 502 chlorophyll meter (Minolta) was used to measure the relative chlorophyll levels in medicago leaves of various developmental stages (positions). As leaves aged, GFP fluorescence decreased with increasing chlorophyll content (Fig. 1b), the changes being most dramatic at early stages of leaf development (between leaves 1 and 3).

Removal of chlorophyll recovers strong fluorescence in both young and mature medicago leaves

Another approach used to study the effect of pigment concentration on fluorescence was to etiolate mature medicago leaflets by wrapping them in aluminum foil while still attached to the stem. A dramatic increase in fluorescence was obtained by 6 d of etiolation (Fig. 1e). Measurement of relative fluorescence and chlorophyll content showed a proportional increase in fluorescence with decreased chlorophyll content (Fig. 1f). The etiolated wild-type leaf control showed only background levels of fluorescence, demonstrating that the etiolation itself does not cause the leaf to fluoresce.

Further evidence that chlorophyll is a major factor in diminishing GFP fluorescence was obtained when medicago leaflets were extracted with 95% ethanol (Fig. 2a). Although GFP fluorescence diminished with duration of exposure to ethanol (data not shown), leaves extracted for up to 8 h with 95% ethanol showed a dramatic restoration of fluorescence compared with untreated transgenic control leaves or wild-type ethanol-treated control leaves (compare center panels of Fig. 2a with the left and right panels, respectively). The autofluorescence detected in the control leaves using a 500-nm emission filter (upper panels) was essentially eliminated when a 525-nm emission filter was used (lower panels).

Figure 2.

Recovery of green fluorescent protein (GFP) fluorescence in transgenic medicago by ethanol extraction. (a) Transgenic medicago leaves extracted with 95% ethanol (center panel, both rows) showed much stronger GFP fluorescence than untreated leaves using either a 500-nm emission filter (top panels) or a 525-nm emission filter (bottom panels). Wild-type leaves extracted with ethanol were used as controls (right). All images shown in (a) were obtained by a 5-s exposure of second trifoliate leaves. (b) An intact, untreated medicago plant (left) showed bright green GFP fluorescence in roots and emerging leaves, but not in mature leaves, whereas extraction of pigments with 95% ethanol for 8 h yielded green fluorescence in most mature trifoliate leaves (right).

A comparison between ethanol-treated and untreated whole medicago plants is shown in Fig. 2b. Whereas strong green fluorescence was evident for roots under both conditions, shoots of the untreated plant (left) showed bright red autofluorescence (except for the nodal meristem tissue). In contrast, most of the trifoliate leaves on the ethanol-treated plant (right) fluoresced green.

Titration of GFP with chlorophyll strongly decreases fluorescence

The effect of chlorophyll on GFP fluorescence was directly investigated by titrating a solution of EGFP (1 µg ml−1) in TE buffer against solutions of HPLC-purified chlorophylls a and b (Sigma) solubilized in 20 mg ml−1 Na cholate (Griffiths, 1978). The Na cholate was used to permit the mixing of chlorophyll, which is not soluble in water, with the aqueous solution of GFP. Fluorescence was measured as RFUs using a Bio-Rad VersaFluor fluorometer. Na cholate by itself decreases fluorescence by ∼20%, but was present at the same concentration for all readings. At low concentrations (10 µg ml−1), fluorescence loss was 4-fold greater for chlorophyll b than for chlorophyll a (Fig. 3), probably indicating their relative interference with GFP excitation and emission. At a concentration of 60 µg ml−1, chlorophyll a alone caused a decrease of more than 50% in GFP fluorescence; at 20 µg ml−1 chlorophyll b cut off GFP fluorescence completely (Fig. 3).

Figure 3.

Titration of EGFP with chlorophylls. Enhanced green fluorescent protein (EGFP) (1 µg) was mixed with increasing amounts of chlorophyll a or b in 1 ml TE buffer containing 20 mg ml−1 sodium cholate. Relative fluorescence units (RFU) were measured for both chlorophyll a (open squares) and b (closed squares). Error bars indicate standard deviation.

GFP fluorescence is clearly visible with microscopic detection methods

The presence of GFP fluorescence in medicago leaves was investigated using fluorescence microscopy with oil-immersion optics. In intact leaves viewed from either the top or the bottom, the endoplasmic reticulum (ER)-targeted GFP was readily detected in the epidermis (in which only the guard cells of the stomata have chloroplasts), but GFP signal intensity quickly decreased when focusing deeper into the tissue (not shown). However, in leaf cross-sections, the typical reticulate pattern of ER-localized GFP fluorescence was observed in all cell layers, including the palisade and spongy mesophyll cells (Fig. 4a). This suggested that the loss of GFP signal when focusing deeper into the tissue was caused by interference from chlorophyll in the chloroplast-rich mesophyll cells, as well as spherical aberrations resulting from the refraction index mismatch between the immersion medium and the aqueous sample. No discernable fluorescence was detected in the wild-type leaves under the same imaging conditions (Fig. 4b).

Figure 4.

Green fluorescent protein (GFP) fluorescence and differential interference contrast (DIC) microscopy of medicago leaf cross-sections. (a) Transgenic plant expressing 35S-mGFP5er. GFP fluorescence is shown in the upper panel, while the corresponding DIC image is shown in the lower panel. Bar, 10 µm. (b) Wild-type plant.

Medicago and rice, but not arabidopsis, leaves show early loss of GFP fluorescence

To explore the generality of our observations, use was made of available rice lines transgenic for the same 35S-mGFP5er construct as that used for medicago. As was found for medicago (Fig. 1b), a decrease in fluorescence with leaf age (position) was evident for rice, and was correlated with an increase in chlorophyll content, as shown in Fig. 5a. The relative GFP fluorescence of aqueous extracts was also determined. For line 2, which showed the highest fluorescence for the first leaf, the relative fluorescence of maturing leaves (at positions 2, 3 and 4) was followed. The in vitro data closely parallel the in planta data (Fig. 5b) in that relative fluorescence dropped most dramatically from leaf 1 to leaf 2. Western blot analysis (Fig. 5c) revealed a substantial (∼3-fold) decrease in GFP protein from the first to the second leaf (compare lane 2-1 with lane 2-2). Except for the nodes (Fig. 5d), GFP fluorescence in light-grown leaves of transgenic rice seedlings (7 d old) was hardly visible, paralleling the situation for maturing medicago leaves. Rice and medicago were also similar in that etiolation resulted in strong green fluorescence throughout the leaf (Fig. 5e). Thus, the diminution of GFP fluorescence with leaf maturity resulted from both chlorophyll interference and decreased GFP levels.

Figure 5.

Analysis of rice plants transgenic for 35S-mGFP5er. (a) Chlorophyll content (closed circles) and relative fluorescence (open circles) for leaves of line 2 at leaf positions 1–4. (b) Relative fluorescence of leaf extracts (100 µg total protein) from the first leaves of five independent transgenic lines (1-1, 2-1, 3-1, 4-1 and 5-1) and leaves 2 through 4 of line 5 (5-2, 5-3 and 5-4), selected on the basis of the relatively high fluorescence of its first leaf. (c) Western blot analysis of the leaf extracts in (b). (d) Leaves of transgenic (left) and wild-type (right) rice seedlings recorded using a 500-nm emission filter and a 525-nm emission filter (upper and lower panels, respectively). (e) Leaves of dark-grown transgenic (left) and wild-type (right) rice seedlings recorded using a 500-nm emission filter and a 525-nm emission filter (upper and lower panels, respectively). Pictures in (d) and (e) were exposed for 5 s. RU, relative units.

Interestingly, arabidopsis, the third species in which we examined the relationship between GFP fluorescence and leaf age, differed from rice and medicago. Six lines of arabidopsis transgenic for the same 35S-mGFP5er construct used in medicago and rice were chosen. In contrast to the dramatic decrease in fluorescence found between leaves 1 and 2 for medicago and rice, in arabidopsis, little difference in fluorescence was evident from leaf 1 to leaf 4 at 525-nm emission wavelength. This was the case for lines of both high and low intrinsic fluorescence (Fig. 6a). A similar situation was found for leaf extracts (Fig. 6b). Intriguingly, in arabidopsis, the chlorophyll content for leaves at different developmental stages (positions) showed only a minor decrease from leaf 1 to leaf 4 (Fig. 6c). Remarkably, upon extraction of chlorophyll with ethanol, a substantial increase in GFP fluorescence was attained for arabidopsis leaves of all developmental stages, especially young leaves (Fig. 6d, left panels).

Figure 6.

Green fluorescent protein (GFP) fluorescence in arabidopsis. (a) Relative GFP fluorescence in intact leaves at leaf positions 1–4 for six arabidopsis lines. (b) Relative fluorescence of leaf extracts (100 µg total protein) from leaves 1–4 of two arabidopsis lines. (c) Relative chlorophyll content in arabidopsis leaves of leaf positions 1–4. (d) GFP fluorescence in an intact transgenic arabidopsis plant (upper left panel) and the same plant extracted with 95% ethanol (lower left panel). Untreated and ethanol-extracted wild-type plants are shown in the right panels (upper and lower, respectively). All pictures in (d) were obtained by exposure for 5 s using a 500-nm emission filter. RU, relative units.

Discussion

Chlorophyll significantly interferes with GFP fluorescence

The involvement of chlorophyll in the diminution of GFP fluorescence was convincingly demonstrated both in vivo and in vitro. Two in vivo approaches were employed. In the first, chlorophyll in medicago leaves was reduced by etiolation (Fig. 1e). A caveat to this approach is the assumption that etiolation does not substantially enhance expression from the CaMV 35S promoter. We can find no evidence that this occurs, although light/dark transitions have been shown to alter the transcriptome profile for arabidopsis leaves (Ma et al., 2001). In the second approach chlorophyll was extracted with 95% ethanol. For both treatments, a dramatic recovery of GFP fluorescence was observed (Figs 1e and 2a). Ethanol per se did not cause the increase in GFP fluorescence; indeed, prolonged exposure to ethanol results in decreased GFP fluorescence. Further, no appreciable fluorescence was detected for ethanol-treated wild-type medicago leaves (Fig. 2a, right panels).

As an in vitro approach, a solution of GFP was mixed with a solution containing purified chlorophyll a or chlorophyll b. Chlorophyll b exhibited a strong negative effect on GFP fluorescence, while a milder effect was caused by chlorophyll a. As chlorophylls, especially chlorophyll b, absorb at the excitation wavelength (488 nm) of GFP, this makes them competitors of GFP for the excitation light. Considering the abundance of chlorophylls in light-grown leaves, it is quite conceivable that the fluorescence from GFP could be lowered to levels that are barely detectable macroscopically. It is certainly possible that pigments other than chlorophyll may interfere with the perception of GFP fluorescence.

Decreased GFP concentration and increased chlorophyll content combine to diminish fluorescence in developing leaves of medicago and rice

In the experiments reported here, GFP was driven by the CaMV 35S promoter which is widely considered to be constitutive (Benfey et al., 1989; Battraw & Hall, 1990). However, several reports provide evidence for variability in spatial and developmental expression from this promoter. For example, Mitsuhara et al. (1996) reported that, in tobacco, the highest (youngest) leaf had almost 8-fold greater CaMV 35S activity than the lowest (oldest) leaf, and Nagata et al. (1987) showed that the CaMV 35S expression level varies during the cell cycle. Additionally, Sunilkumar et al. (2002) described variation in CaMV 35S-driven expression levels with development and tissue type. Harper & Stewart (2000) also found that CaMV 35S-driven GFP fluorescence was higher in leaf primordia than in the lower (older) fully expanded leaves. However, the findings in the two latter papers relied on a direct relationship between GFP fluorescence and promoter activity and may need reconsideration in light of our finding that chlorophyll can substantially disrupt this relationship.

The above observations may partially explain the 2.7-fold difference in relative fluorescence of extracts from leaves 1 and 5 shown in Fig. 1c. However, the substantial effect of increased chlorophyll concentration with leaf maturity and relative fluorescence (Figs 1 and 3) must contribute to the significant difference between the respective 7.8-fold and 2.7-fold decreases in fluorescence found for intact leaves and leaf extracts of medicago, respectively (Fig. 1c).

A dramatic decrease of both GFP fluorescence and relative GFP protein content was also observed in rice, accompanied by an increase in chlorophyll level (Fig. 5a, b and c). Removal of chlorophyll by etiolation substantially restored GFP fluorescence in transgenic rice seedlings while causing little change in wild-type seedlings (Fig. 5d and e). Whereas no green autofluorescence was observed for mature wild-type medicago or arabidopsis leaves, a low but detectable level was evident in rice (not shown).

How arabidopsis hides its dark side

The results shown in Fig. 6a and b reveal that arabidopsis did not exhibit the remarkable decrease in GFP fluorescence level with leaf age found for medicago (Fig. 1b and c) and rice (Fig. 5a and b). However, because extraction of chlorophyll from mature leaves with ethanol resulted in a 5- to 10-fold increase in GFP fluorescence, there can be little doubt that chlorophyll-induced diminution (the ‘dark side’) of fluorescence actually occurred in arabidopsis as in medicago and rice. Figure 6c shows that a major difference between arabidopsis and the other species studied was that, rather than increasing with age (Figs 1b and 5a), leaf chlorophyll content in arabidopsis slightly decreased. Further, the data of Fig. 6b indicate that the proportion of GFP in the total leaf protein also remained constant, rather than decreasing as in the other species studied. These findings reveal that the chlorophyll content and the proportion of GFP in total protein contributed to loss of fluorescence in maturing leaves and that their relative contribution differed from species to species. In addition, the relative concentrations of chlorophyll a and b, and possibly other pigments, affected the relationship between GFP content and fluorescence.

Leaf optical properties may contribute to diminution of GFP fluorescence

Absorption spectra for green and white segments of variegated Coleus hybridus leaves (Stahlberg et al., 2000) reveal substantial absorption in green regions at the excitation (488 nm) and emission (509 nm) wavelengths of GFP. This implies that observation of GFP fluorescence in leaves can be hindered by both absorption of the excitation light (thus diminishing the excitation intensity) and reabsorption of the emitted green signal. In contrast, light absorption in nongreen leaf segments is significantly lower across the spectrum. Thus, the detectable fluorescence of a unit amount of GFP may be several fold lower in green leaves than in nongreen leaves. Because absorption spectra have been shown to vary with the physiological state of the leaf (Gitelson et al., 2001), apparent transgene expression levels based on GFP fluorescence may be affected even if the actual level is unchanged.

Whereas detection of GFP in green tissues using macroscopic methods can lead to substantial underestimation of expression levels, as exemplified by Fig. 6d, GFP fluorescence in thin sections used for microscopic analysis may be less affected by the presence of chlorophyll (Fig. 4a). We speculate that this may reflect the fact that tissues expressing GFP are less likely to be obscured by chlorophyll-rich cell layers in thin sample sections. Similar considerations lead to the possibility that differences in leaf architecture, such as the number of cell layers, cell and organelle shape and cell wall thickness or composition, may also contribute to the observed species differences in apparent diminution of GFP fluorescence.

Our data reveal that considerable deviation may exist between apparent fluorescence from GFP and transgene transcript levels in green tissues, especially leaves, and that the discrepancy can vary significantly between plant tissues and species. Thus, while GFP and other fluorescent proteins clearly have important and increasing roles to play in analysis of gene function, caution should be exercised when relating observed fluorescence to quantitative values.

Acknowledgements

We thank Yiming Jiang and Tao Wang for providing the transgenic rice and arabidopsis used in this study and Danny Ng and Guojun Yang for helpful discussions. We appreciate our involvement in stimulating medicago group meetings sponsored by NSF grant DBI-0110206 (D. Cook, PI).

Ancillary