Self-Reporting Arabidopsis Expressing pH and [Ca2+] Indicators Unveil Ion Dynamics in the Cytoplasm and in the Apoplast under Abiotic Stress1[w]

For noninvasive in vivo measurements of intra- and extracellular ion concentrations, we produced transgenic Arabidopsis expressing pH and calcium indicators in the cytoplasm and in the apoplast. Ratiometric pH-sensitive derivatives of the green fluorescent protein (At-pHluorins) were used as pH indicators. For measurements of calcium ([Ca2+]), luminescent aequorin variants were expressed in fusion with pHluorins. An Arabidopsis chitinase signal sequence was used to deliver the indicator complex to the apoplast. Responses of pH and [Ca2+] in the apoplast and in the cytoplasm were studied under salt and “drought” (mannitol) stress. Results are discussed in the frame of ion flux, regulation, and signaling. They suggest that osmotic stress and salt stress are differently sensed, compiled, and processed in plant cells.

How plants sense and respond to various environmental stimuli, such as touch, wind, cold, light, oxidative stress, high salinity, and drought, has become an area of intense investigations in the past decades. Many cellular compounds such as Ca 2ϩ , lipids, H ϩ , cyclic nucleotides, and inositolphosphates are listed as messengers used by plants to forward and compile cellular signals (Sanders et al., 1999).
The cytoplasmic free Ca 2ϩ concentration ([Ca 2ϩ ] cyt ), in particular, has been found to play a central role in a wide range of cellular events (Trewavas and Malhó, 1998;Knight, 2000;Sanders et al., 2002), although its alleged universality has been critically questioned recently (Plieth, 2001;Scrase-Field and Knight, 2003).
The apoplast is the first plant compartment encountering environmental signals. It has been suggested that the apoplast is involved not only in the response but also in the perception and transduction of various environmental signals in cooperation with the plasma membrane (Hoson, 1998). pH and calcium concentrations in the apoplast have been measured in many plants by selective electrodes (Zieschang et al., 1993;Felle, 1998), by collecting apoplast fluid (Dannel et al., 1995;Cramer and Jones, 1996;Lohaus et al., 2001) or by impermeable fluorescent dyes (Hoffmann and Kosegarten, 1995;Taylor et al., 1996;Sakurai, 1998aSakurai, , 1998b. However, little in vivo information on the change of apoplastic pH and [Ca 2ϩ ] is available from whole and undisturbed (i.e. not infiltrated) plants under abiotic stress. Such knowledge can help to understand the role of the apoplast in more detail. To overcome drawbacks of other methods, we started to develop a novel approach: the design of transgenic plants that simultaneously express their own [Ca 2ϩ ] and pH indicators at the desired location. We expressed modified pHluorins (Miesenbö ck et al., 1998;Moseyko and Feldman, 2001;Plieth et al., 2001) and aequorin (AQ; Knight et al., 1991Knight et al., , 1997a as fusion proteins in the cytoplasm of Arabidopsis (Fig. 1, A, E, and F). For studying extracellular ion dynamics, we also targeted pHluorins and an AQ variant with low calcium affinity (Kendall et al., 1992) as fusion proteins to the apoplast by using an Arabidopsis chitinase signal sequence (Fig. 1, B-D). A series of experiments were performed to characterize the indicators  and to demonstrate that the produced self-reporting plants are able to monitor changes of pH and [Ca 2ϩ ] in the cytoplasm and in the apoplast . Salt (NaCl) and drought stress were studied in particular with this novel technique (Figs. 6-10) to discover new aspects of ion flux, regulation, and signaling.

Characterization of AQ Variants in Bacteria
For comparison of native AQ and LAAQ, both AQ variants were expressed in bacteria. Bacteria were reconstituted at room temperature for 30 min with 10 m coelenterazine (CTZ) in Luria-Bertani medium and then stimulated with 1 mm CaCl 2 solution. The kinetics (data not shown) were similar to what has been shown previously (Jones et al., 2002). Relative luminescence of initial response was lower in LAAQexpressing bacteria in comparison with native AQ. The measured relative luminescence indicates that less LAAQ than AQ is discharged during Ca 2ϩ entry into the cells. This verifies that LAAQ is, because of its lower Ca 2ϩ -binding affinity, the better choice for a higher [Ca 2ϩ ] range as is expected in the apoplast.

Characterization and in Vitro Calibration of Modified pHluorins
Modified At-pHluorins isolated from bacteria were characterized fluorometrically. Ratiometric At-pHluorin (ratioGFP) has pH-dependent spectra with isoexcitation point at 428 nm ( Fig. 2A). The excitation maxima of ratioGFP are at 395 and 475 nm, and the emission maximum is at 508 nm. Its k d is 6.73 Ϯ 0.03, and its optimal dynamic range is in the interval 5.6 Ͻ pH Ͻ7.8, which makes it suitable for both cytoplasmic and apoplastic pH measurements (Fig. 2B).
Ecliptic At-pHluorin (eclipticGFP) displayed a ratiometric behavior in the emission mode with an isoemission point at 490 nm (Fig. 3A). The excitation maxima of eclipticGFP are at 398 and 477 nm, and the emission maximum is at 510 nm. Its k d is 7.25 Ϯ 0.02, and its optimal dynamic range is in the interval 6.5 Ͻ pH Ͻ 8.0, which makes it suitable especially for cytoplasmic pH measurements (Fig. 3B).
The spectra (Figs. 2 and 3) show that all modifications made to produce At-pHluorins hardly altered spectral properties and/or pH dependency when compared with original pHluorins from Miesenbö ck et al. (1998). In addition, when fused to AQ, the emission spectra were negligibly broadened by 3 nm in half band width, and peaks were shifted by about 1 nm toward the blue. Also, the pH dependence of Figure 1. Schematic structures of the gene cassettes. A, Fusion construct from pCM2. B, Apoplast targeting construct: insertion of chitinase signal sequences (66 bp) upstream of GFP5 giving pchitGFP5:AQ. C, Fragment exchange (Cla-Xho) to yield pH sensitivity and (XhoI-PstI) for insertion of low-affinity AQ (LAAQ), leading to pchitratioGFP:LAAQ. D, Same as C but insertion of ecliptic pHluorin fragment giving pchite-clipGFP:LAAQ. E and F, Replacement of Cla-Xho fragment for introducing pH sensitivity and introduction of point mutation for increased solubility. This produced cytoplasmic expression of ratioGFP (E) and ecliptic green fluorescent pro-At-pHluorins did not significantly alter when fused with AQ and chitinase propeptide. Moreover, the NaCl concentration (0-200 mm NaCl) at constant pH had negligible influence on the fluorescence ratios (data not shown).

Luminescence. In Vivo Reconstitution of AQs
Plants expressing the GFP5:AQ fusion in the cytoplasm and in the apoplast were in vivo reconstituted with two different CTZ derivatives. Figure 4 shows absolute luminescence during the first hours of reconstitution. The time courses demonstrate that maximal basal luminescence in the cytoplasm (Fig. 4A) is reached after 4 h. The apparent binding constant of cp-AQ is pk Ca Ϸ 6.4 compared with pk Ca Ϸ 5.9 of native AQ (Plieth and Trewavas, 2002). As a consequence, basal luminescence produced by cp-AQ decays, even with the low resting level of free [Ca 2ϩ ] in the cytoplasm (Fig. 4A, circles), whereas basal luminescence from native AQ is stable here (Fig. 4A, squares). A much higher [Ca 2ϩ ] is present in the apoplast. Hence, even luminescence from native AQ decays and cp-AQ luminescence dissipates within 5 h in this compartment (Fig. 4B). We found that basal luminescence even from LAAQ is not stable in the apoplast when reconstituted with native CTZ and also decays below minimum detection level within 2 d (data not shown). Nevertheless, there is a time window of more than 12 h for [Ca 2ϩ ] measurements after in vivo reconstitution of targeted AQ with native CTZ.

[Ca 2؉ ] Response to Cold in the Cytoplasm and in the Apoplast
The in vivo response of luminescence from AQ expressed in the apoplast to cooling was compared with luminescence responses reported by AQ expressed in the cytoplasm (Fig. 5). These responses, too, show that AQ has successfully been targeted to a compartment different from the cytoplasm. At first glance, the data suggest that [Ca 2ϩ ] is slightly lowered in this compartment during cold period. However, AQ luminescence is temperature dependent (data not shown), and the decrease of luminescence Figure 2. A, Excitation spectra of ratiometric GFP at different pHs (emission wavelength ϭ 508 nm). Curves were normalized by F(428ex; 508em). B, Corresponding calibration curve for fluorescence excitation ratio R(395ex/475ex; 508em); dissociation constant, i.e. midpoint of the fitted curve ϭ pK d ϭ 6.9 Ϯ 0.03; optimal dynamic range for pH measurements is between 5.6 and 8.0. Figure 3. A, Emission spectra of ecliptic GFP at different pHs (excitation wavelength ϭ 400 nm). Curves were normalized by F(400ex; 490em). B, Corresponding calibration curve for fluorescence emission ratio R(400ex; 490em/510em); dissociation constant, i.e. midpoint of the fitted curve ϭ pK d ϭ 7.25 Ϯ 0.02; optimal dynamic range for pH measurements is between 6.5 and 8.0.
with lowered temperature in Figure 5C might well originate from this effect.

Fluorescence. Expression and Localization of the Indicator Complex
The expression level of indicator protein-as scrutinized by fluorescence-seemed to be higher in the apoplast expressing lines than in the cytoplasmicexpressing lines. This can be explained with the permanent export of indicator in apoplast-targeted lines and extracellular accumulation. Confocal laser scanning microscopy (CLSM) revealed staining of the cell wall (supplemental data).
The in vivo spectra taken from ratiometric pHluorins expressed in roots (Fig. 6) are slightly different compared with the in vitro spectra (Fig. 2) because of slight autofluorescence in the plant specimen and different emission bands (fluorescence emission wavelength F em ϭ 508 Ϯ 5 nm in Fig. 2 and F em ϭ 535 Ϯ 25 nm in Fig. 6). However, the spectra clearly indicate compartments of different acidity: pH Ϸ 6.3 in the apoplast and pH Ϸ 7.2 in the cytoplasm.
A change of external pH ( Fig. 7) induced no change in fluorescence ratio of cytoplasmic-expressed At-pHluorin. The pH indicator expressed in the apoplast, in contrast, reported a strong dependence of apoplastic pH on external pH (Fig. 7). This demonstrates that the indicator complex is located in a compartment that has direct access to the outer medium and confirms successful targeting of the protein to the extracellular space.

Free Calcium Concentration in the Apoplast ([Ca 2؉ ] apo ) and Cytoplasmic Free Calcium Concentration ([Ca 2؉ ] cyt ) Responses during "Drought" and NaCl Stress
Changes of [Ca 2ϩ ] apo and [Ca 2ϩ ] cyt were determined in response to repeated periods of NaCl stress Figure 4. Development of basal AQ luminescence during in vivo reconstitution of Arabidopsis roots with different coelenterazine (CTZ) derivatives. Squares, Reconstitution with native CTZ; circles, reconstitution with cp-CTZ. Addition of CTZ at t ϭ 0.6 h to 10 M final concentration. Note: Absolute luminescence (ϫ10,000) is given here, which is luminescence of each integration interval divided by total luminescence produced by the specimen. Relative luminescence (ϫ10,000) is given in all other figures, which is luminescence of each integration interval divided by luminescence still remaining in the specimen. A, Reconstitution of AQ expressed in the cytoplasm. B, Reconstitution of AQ targeted to the apoplast. (i.e. 100 mm NaCl) and "drought." The latter was mimicked by isosmotic mannitol.

Osmotic ("Drought") Stress
Repeated 30-min periods of mannitol treatment were applied to roots. The results (Fig. 8A) indicate that the [Ca 2ϩ ] apo is hardly affected by this treatment (Fig. 8A, gray curve). Apart from tiny responses during mannitol washout, there are no significant changes in the [Ca 2ϩ ] apo signal. The [Ca 2ϩ ] cyt , in contrast, is drastically affected, and pronounced [Ca 2ϩ ] cyt transients are observed (Fig. 8A, black curve). This experiment produces two different kinds of stimuli: first, a hyperosmotic stimulus when mannitol is applied and second a hypo-osmotic stimulus when mannitol is washed out. The very first hyperosmotic treatment is a weak stimulus giving a [Ca 2ϩ ] cyt transient with small amplitude (at t ϭ 0.5 h; Fig. 8, A and C). All following hyperosmotic treatments (t ϭ 1.5, 2.5, and 3.5 h; Fig. 8A) do not give spectacular [Ca 2ϩ ] cyt transients anymore, suggesting that some kind of adaptation must have happened. The hypo-osmotic-triggered [Ca 2ϩ ] cyt response, however, is much more pronounced indicating that this stimulus is more critical. In both cases (hyper-and hypo-osmotic shock), the [Ca 2ϩ ] cyt is brought back to base level by cellular Ca 2ϩ clearance mechanisms after the transient. [Ca 2ϩ ] cyt does not stay on elevated levels as is the case for NaCl treatment (described below).

NaCl Stress
Repeated periods of NaCl treatment give drastic transients and prolonged alterations of the [Ca 2ϩ ] level in both compartments (Fig. 8, B and D). The very first NaCl treatment leads to a permanent increase of the cytoplasmic [Ca 2ϩ ] ([Ca 2ϩ ] cyt ) after a short transient peak (Fig. 8D). The following washout, however, produces a further permanent increase ( Fig. 8B) that is not seen during mannitol treatment (Fig. 8A). These two increased [Ca 2ϩ ] cyt levels alternate during the rest of the experiment as NaCl solution is replaced by water and vice versa. The [Ca 2ϩ ] apo is markedly affected by NaCl as well (Fig.  8B, gray curve). During the very first NaCl treatment, a distinct increase in [Ca 2ϩ ] apo occurs. The following treatments bring also about an increase or at least an elevated steady state of [Ca 2ϩ ] apo during NaCl periods and a decrease during washout (water). To discriminate NaCl stress and osmotic response, we switched between mannitol and NaCl solutions with same osmotic pressure (Fig. 9). This brings about a uniform osmotic stress. The very first transition from mannitol to NaCl (t ϭ 1.5 h; Fig. 9) resulted in an elevation of [Ca 2ϩ ] cyt as was observed before (t ϭ 0.5 h; Fig. 8B), but with a negligible transient peak. The subsequent change from NaCl to mannitol (t ϭ 2 h; Fig. 9) unexpectedly produced a [Ca 2ϩ ] cyt transient, but no change of the reached steady-state level. All subsequent transitions from mannitol to NaCl and vice versa caused similar alterations between two [Ca 2ϩ ] cyt states-with intermediate transients-as has been observed with NaCl-water treatment (Fig. 8). The same treatment of the apoplastic expressing plants (Fig. 9, gray curve) confirmed that external Na ϩ caused a drastic increase in [Ca 2ϩ ] apo and gave a similar kinetic pattern as with NaCl-water treatment (Fig. 8B, gray curve).

Cytoplasmic pH (pH cyt ) and Apoplastic pH (pH apo ) Responses during NaCl and "Drought" Stress
The F 395 to F 475 fluorescence excitation ratio of the pH-sensitive GFP revealed that "drought" (mannitol) treatment ( Fig. 10A) does not influence pH cyt and pH apo . NaCl stress, however, resulted in significant alterations of pH in the apoplast and to a minor extent in the cytoplasm (Fig. 10B). The pH responses Figure 6. In vivo spectra of ratioGFP expressed in Arabidopsis roots (black curve, cytoplasmic expression; gray curve, apoplastic expression). Curves were normalized by F(428ex; 530em).
Figure 7. pH changes in the apoplast (gray curve) and the cytoplasm (black curve) of Arabidopsis roots in response to changes in external pH. External pH was adjusted with 10 mM HEPES (pH 7.6) and 10 mM MES (pH 6.0) in standard medium (KCl, CaCl 2 , and MgCl 2 : 1 mM each).
became more pronounced with the number of NaCl treatments in both compartments. This sort of "sensitization mechanism" (i.e. increasing response amplitudes with increasing number of treatments) is opposite to "adaptation" mechanisms (i.e. decreasing response amplitudes with increasing number of treatments) usually observed with other abiotic stimuli (e.g. cold; Fig. 5A).

DISCUSSION
For studying ion flux and regulation when the plants are under abiotic environmental stress, genetically encoded indicator proteins proved to be the most elegant way (Plieth, 2001). An important advantage is that they can be targeted to many different organelles, compartments, and tissues by fusion with specific promoters, signal sequences, or by trapped enhancers (Kiegle et al., 2000;Plieth, 2001). When using reporter plants expressing their own indicators, loading problems, as is the case for low-M r indicator dyes (Plieth and Hansen, 1996), are circumvented, and artifacts resulting from loading procedures are avoided. Measurements can be performed in unperturbed, whole, intact plants and under "physiological" conditions; i.e. conditions similar to what plants experience in the wild. Thus, any study which makes use of "self-reporting plants" relates more closely to conditions in the field and deserves to be called "biologically relevant."

Scrutiny of Transgenic Lines
The fact that none of our transgenic lines differ in phenotype from the wild type verifies that expression of the indicators do not interfere with signal transduction, growth, and development. It suggests that neither cellular ion regulation in general nor ion buffering in particular is affected by the alien Ca 2ϩand H ϩ -binding indicator proteins. Figures 4 to 7 show that the targeted indicator complex is expressed in a compartment with high [Ca 2ϩ ] and low pH compared with cytoplasmic conditions. Both details have been anticipated for the extracellular space, and there are more details confirming correct targeting of the indicator in the apoplast.
The overall continuing decay of the apoplastic signal (Fig. 8A, gray curve) indicates permanent washout of Ca 2ϩ ions by the perifusion medium which was not supplemented with extra Ca 2ϩ .
The in vivo spectra (Fig. 6) taken from dissected roots expressing ratioGFP in the cytoplasm (black curve) and in the apoplast (gray curve) are markedly different and their F 395 to F 475 ratios indicate pH cyt Ϸ 7.2 and pH apo Ϸ 6.3, which is close to what has been expected. The apoplastic pH has been reported many times from different species, and the majority of values varies between 5.3 (Kosegarten and Englisch, 1994) and 6.7 (Dannel et al., 1995).
However, the measured resting apoplastic pH (i.e. Ϸ6.3) was always slightly higher than expected (i.e. pH Ͻ 6.0). Therefore, we reexamined the transgenic lines by CLSM at high magnification (supplemental data). In the cytoplasmic expression lines, cell walls could be easily detected as dark lines between fluorescent protoplasts. The apoplastic expression lines, in contrast, also showed fluorescence at these locations. For a better resolution, the cell walls of these lines were separated from the protoplasts by pretreatment with 100 m cycloheximide and plasmol- Figure 8. A, Cytoplasmic and apoplastic [Ca 2ϩ ] in response to "drought" stress (200 mM mannitol versus water) measured by AQ luminescence from Arabidopsis roots. B, Cytoplasmic and apoplastic [Ca 2ϩ ] in response to NaCl stress (100 mM NaCl versus water). C and D, Close-ups of A, and B, respectively. ysis with 500 mm mannitol. This finally revealed a clear cell wall staining with GFP.
However, in these lines, we also found residual expression of indicator protein inside the cells located in so-called "fusiform organelles." These organelles are part of the endoplasmic reticulum (ER) network (Hawes et al., 2001). This finding is not surprising because the protein is translocated via the ER because of the preceding N-terminal chitinase signal propeptide (22 amino acids), carried along the secretory pathway, and then excreted to the apoplast because of the missing HDEL retention sequence. So, there is locally always some cellular expression with fluorescence adulterating the global extracellular fluorescence ratio. We found this cellular expression appearing mainly in undifferentiated, meristematic, and growing cells near the root tip and less in the fully differentiated, basal tissue. Therefore, for pH measurements, we chose the fully differentiated tissue of the root hair zone of lateral roots were side effects caused by residual expression in the ER and autofluorescence are negligible. The fact that the "bulk freight" of targeted indicator has reliably been excreted is demonstrated by the test experiment shown in Figure 7: A change of external pH does produce an immediate change of fluorescence ratio. This shows that the main part of indicator, sufficient for reliable pH measurements, is located in the apoplast. Furthermore, there was no At-pHluorin within vacuoles detected by CLSM. This is in contrast to Persson et al. (2002), who also found protein in vacuoles when targeted to ER with C-terminal HDEL sequence removed. For our constructs, this means that either no At-pHluorin has been trafficked to vacuoles or any vacuolar fluorescence has been quenched because of an extremely low pH vac . The latter, however, is unlikely because other studies show that GFP is stably fluorescent even in acidic vacuoles of Arabidopsis (Neuhaus, 2000

pH and [Ca 2؉ ] Respond Differently to "Drought" and NaCl Stress Situations
The amplitude of the [Ca 2ϩ ] cyt response to water loss (hyperosmotic stimulus) is much smaller compared with that seen during hypo-osmotic stimulus (Fig. 8A, black curve). This has been observed before (Pauly et al., 2001) and verifies that shrinking is less serious for the cells than swelling, which might lead to rupture if cell wall and/or membrane do not stand the rising pressure (Takahashi et al., 1997). The decrease in [Ca 2ϩ ] cyt amplitude with repeated mannitol treatment periods (Fig. 8A, black curve) reflects some sort of adaptation that also has been observed with other abiotic stimuli such as cold and gravity (Plieth et al., 1999a;Plieth and Trewavas, 2002). The apoplastic [Ca 2ϩ ] remains almost unaffected by mannitol treatment (Fig. 8A, gray curve). This is a surprise because we expected that a significant amount of Ca 2ϩ ions is shifted from the extracellular space during hypo-osmotic treatment into the cytoplasm. This should give a significant transient fall in the apoplastic luminescence signal. Because the latter fails to  appear, we conclude that the cellular [Ca 2ϩ ] cyt transients during mannitol washouts (Fig. 8A, black curve) are mainly produced by release of Ca 2ϩ ions from internal stores (e.g. vacuole) rather than influx from the outside.
NaCl, in contrast, produces [Ca 2ϩ ] and pH changes in both compartments (Figs. 8-10): The finding that [Ca 2ϩ ] cyt is increased during NaCl stress is in line with observations from Lynch et al. (1989), who observed a [Ca 2ϩ ] cyt increase in maize (Zea mays) root protoplasts during NaCl stress. The finding that [Ca 2ϩ ] cyt responses to NaCl is markedly different from "drought" has been anticipated by Shi et al. (2002), who showed that root growth deprivation is different under NaCl and drought stress.
There are two differences between NaCl and "drought" [Ca 2ϩ ] cyt responses: First, the whole [Ca 2ϩ ] cyt is permanently shifted by NaCl toward a higher level (Figs. 8B and 9, black curves). Second, the short-term response to NaCl is prolonged when compared with the [Ca 2ϩ ] cyt kinetic under "drought" treatment (Fig. 8,C and D). In particular, these shortterm differences are well in line with Knight et al. (1997b).
Prolonged [Ca 2ϩ ] cyt elevations have been seen with other studies where the cytoplasm has been challenged with an excess of other monovalent ions, namely H ϩ (Plieth et al., 1997(Plieth et al., , 1999b. The main conclusion drawn in these studies is that calcium acts as protecting and ameliorative agent. Our assumption is that this can hold under Na ϩ stress, too. The ameliorative action of calcium may result from various cellular interactions. First, an increase of [Na ϩ ] also does affect cellular calcium buffer systems like H ϩ ions do: Monovalent ions are able to displace stabilizing Ca 2ϩ ions from their binding sites (i.e. phospho-and carboxy head groups in membranes, Dawson and Hauser, 1970;Hanson, 1984;and uronic groups in the cell wall, DeMarty et al., 1984). Thus, it makes sense for a plant cell to respond to a strong and dangerous increase of toxic monovalent ions with an active release of Ca 2ϩ ions into both the cytoplasm (Figs. 8B and 9, black curves) and the apoplast (gray curves) to counteract loss of membrane and cell wall integrity even if enzyme regulation, signal transduction, and, thus, cellular activity are thereby negatively affected. This somewhat unspecific role could be designated as an "emergency brake function" of calcium. Because there is evidence for a calcium buffer capacity in the millimolar range in the cytoplasm (Plieth et al., 1997;Plieth and Hansen, 1998) and the apoplast (DeMarty et al., 1984;Smart and Trewavas, 1984), the increases of free [Ca 2ϩ ] reported here (Figs. 8 and 9) constitute only a tiny percentage of the total amount of activated Ca 2ϩ ions. Hence, the quantity of calcium mobilized by NaCl seems rather massive.
Second, one more specific effect of Ca 2ϩ in the apoplast is its action on voltage-independent ions channels (VICs). VICs are believed to be the major doorway for Na ϩ into the cell (Amtmann and Sanders, 1999). Because Ca 2ϩ inhibits VIC-mediated Na ϩ currents (White and Lemtiri-Chlieh, 1995;Maathius and Amtmann, 1999;White and Davenport, 2002); thus, it helps to establish a favorable intracellular K ϩ to Na ϩ ratio. Hence, calcium is an "ion transport modulator." This specific modulator function is not a matter of Ca 2ϩ to Na ϩ ratio in the extracellular matrix but rather a question of a critical external Ca 2ϩ concentration necessary to be maintained for effective Na ϩ influx inhibition (Davenport et al., 1997;Tyerman et al., 1997;Murata et al., 1998;Kinraide, 1999).
Third, Ca 2ϩ ions have signaling properties. A calcineurin-like protein SOS3 has been identified to play a significant role in NaCl stress response in Arabidopsis (Liu and Zhu, 1998). SOS3 defective mutants are NaCl hypersensitive. This phenotype is partly suppressed by increased levels of external Ca 2ϩ . Additional support for this Ca 2ϩ calcineurin signaling pathway has been given by Pardo et al. (1998). Hence, calcium has a "signaling function." The current proposed regulatory pathway for intracellular Na ϩ and K ϩ homeostasis and Na ϩ tolerance in plants assumes an increase in [Ca 2ϩ ] cyt under Na ϩ stress (Halfter et al., 2000;Zhu, 2001) that is needed to switch kinases and phosphatases related to NaCl tolerance (Liu and Zhu, 1998;Pardo et al., 1998;Xiong and Zhu, 2002) and phosphoinositol turnover (Hirayama et al., 1995), which in turn is related to [Ca 2ϩ ] signaling.
It has been shown that calcium activates cell wall phosphatases (DeMarty et al., 1984). Moreover, it has been reported several times that apoplastic calmodulin plays an important role in signal transduction (Sun et al., 1994(Sun et al., , 1995Tang et al., 1996;Ma and Sun, 1997;Ma et al., 1999). Thus, calcium seems to have a signaling function in the apoplast as well. However, despite all findings dedicating [Ca 2ϩ ] apo an important role in toxicity alleviation and NaCl stress response, no effect of external Ca 2ϩ on Na ϩ toxicity was found recently in Arabidopsis (Essah, 2000). Hence, no conclusive picture can be drawn yet of how Ca 2ϩ is actually working in the apoplast in Arabidopsis.
Taken together, extracellular and cellular calcium perform different tasks. The experiments shown here provide the first information, to our knowledge, of how calcium behaves extracellularly and intracellularly in Arabidopsis under NaCl stress. Massive permanent Ca 2ϩ shifts are broad-spectrum responses that may address unspecific protection mechanisms, ion transport, and specific signaling under NaCl stress circumstances.

Plasmid Constructs
Standard PCR and cloning techniques (Sambrook et al., 1989;Ausubel et al., 1999) were employed to engineer the plasmid constructs described below. For cytoplasmic expression, we started from a plasmid pCM2 (Moore, 2000), which is a pUC18 derivative equipped with a plant expression unit consisting of a short CaMV35S promoter and the corresponding terminator (528 and 193 bp, respectively;Pietrzak et al., 1986). pCM2 contains an in-frame fusion of the GFP gene (GFP5; Haseloff et al., 1997) downstream linked with AQ cDNA (Fig. 1A).
Because there is a ClaI site in GFP5 383 bp downstream of ATG, and all relevant mutations that render GFP pH sensitive are downstream of ClaI, we decided to exchange the second part of GFP5 between ClaI and XhoI with corresponding fragments produced by PCR amplification from ratiometric and ecliptic pHluorins (Miesenbö ck et al., 1998). In this way, pratioGFP:AQ and peclipGFP:AQ were constructed.
For cytoplasmic expression and better cellular distribution, we introduced the F99S mutation of smGFP (soluble modified GFP) into the gene constructs by substituting the parts between XbaI and ClaI sites with the corresponding portion of smGFP (CD3-326, Davis and Vierstra, 1998) obtained by PCR. Resulting plasmids were named psmratioGFP:AQ and ps-meclipGFP:AQ (Fig. 1, E and F).
For apoplastic targeting, we used the Arabidopsis chitinase signal sequence (22 amino acids) from plasmid pBINm-gfp5-ER, which is known to target to the ER (Haseloff et al., 1997). Omitting the sequence coding for HDEL at the C-terminal end of the protein interrupts the continuous retrieval of the protein from the cis-Golgi to the ER lumen, causing secretion of the gene product and, thus, accumulation of the protein in the apoplast. The first part of the gene cassette in pBIN35S-mgfp5-ER, which contains a 66-bp sequence coding for a chitinase signal propeptide (22 amino acids) together with the first 383 bp of GFP sequence flanked by XbaI and a ClaI recognition sites, were generated by PCR. The PCR product was used to replace the corresponding XbaI-ClaI fragment in pCM2, and pH-sensitive mutations of ratiometric and ecliptic pHluorins were introduced as has been done with the cytoplasmic expression constructs, resulting in two plasmids: pchitratioGFP:AQ and pchiteclipGFP:AQ. Finally, we replaced the AQ gene in these constructs with an engineered AQ with reduced Ca 2ϩ affinity (LAAQ;Kendall et al., 1992) using XhoI and PstI restriction sites. Thus, two more gene constructs with a 66-bp apoplast signal sequence fused to pH and [Ca 2ϩ ] indicator genes, namely pchitratioGFP:LAAQ and pchiteclipGFP: LAAQ (Fig. 1, C and D), were produced. All gene constructs were confirmed by sequencing.
All gene cassettes (Fig. 1) were finally transferred to the binary vector pBI121 (Jefferson et al., 1987) using XbaI and SacI. This vector confers kanamycin resistance in plants and enables stable insertion of the genes into the Arabidopsis genome via Agrobacterium tumefaciens-mediated transformation (described below).

Bacterial Expression of Indicator Protein
To verify that modifications did not alter indicator properties, all indicator and fusion cDNAs were also cloned into bacterial expression vectors (pRSET; Invitrogen GmbH, Karlsruhe, Germany). Indicator proteins were expressed in bacteria and assessed fluorometrically (Figs. 2 and 3) and luminometrically. Because bacterial expression of GFPs is sometimes problematic (Gonzalez and Ward, 2000), and pRSET is leaky expressing when not induced, we grew bacteria under strong selection (70 g mL Ϫ1 chloramphenicol and 200 g mL Ϫ1 ampicillin) on "Terrific Broth" medium (no. 243820, DIFCO Laboratories, Detroit) for 8 h at 37°C/150 rpm. Protein production was induced by 1 mm isopropylthio-␤-galactoside at 20°C/300 rpm overnight (i.e. 15 h). For protein isolation, bacteria were cracked in 200 mm phosphate buffer (pH 7.5) using either a French press or several freezethaw cycles. Debris was removed by centrifugation and filtering the supernatant through a 0.45-m nylon filter. Cleared bulk protein extract was used for in vitro fluorescence assays at room temperature (Figs. 2 and 3) and appropriately buffered pH.

Plant Transformation
The floral dip method for A. tumefaciens-mediated transformation was used (Clough and Bent 1998; http://plantpath.wisc.edu/ϳafb/protocol. html)to transform Arabidopsis of ecotype Columbia (Col0, Nottingham Arabidopsis Stock Centre no. N1092). Seeds from transformed plants were vapor phase sterilized in accordance with Clough and Bent (http:// plantpath.wisc.edu/ϳafb/vapster.html) and sown on sterile solid medium (1ϫ Murashige and Skoog, no. M0222, Duchefa, Haarlem, The Netherlands; and 0.75% [w/v] agar, no. A1296, Sigma, St. Louis) containing 50 g mL Ϫ1 kanamycin and supplemented with 1% (w/v) Suc. After 3 d at 4°C vernalization in the dark, plates were transferred to a growth chamber and cultivated at 70% relative humidity, 21°C, and 16-h photoperiod. Kanamycin-resistant F 1 transformants were transferred to soil in small pots, covered with ARACONs (BetaTech, Gent, Belgium), and further cultivated under same conditions. After selfing, plants of the F 2 generation were screened for fluorescence and luminescence. Lines with highest expression of the indicator fusion protein were selected, and individual F 2 plants were used for further proliferation into F 3 to find homozygous lines.

Plant Material
According to Figure 1, B to E, we obtained five transgenic lines expressing different indicator proteins: GFP5 and AQ in the apoplast (Fig. 1B), ratiometric pHluorin and LAAQ in the apoplast (Fig. 1C), ecliptic pHluorin and LAAQ in the apoplast (Fig. 1D), soluble modified ratiometric pHluorin and AQ in the cytoplasm (Fig. 1E), and soluble modified ecliptic pHluorin and AQ in the cytoplasm (Fig. 1F).
Plants were used for experiments when grown for 1 to 4 weeks on vertical 1.2% (w/v) plant agar (no. A1296, Sigma) supplemented with 0.5ϫ Murashige and Skoog medium (no. M0222, Duchefa) and no sugar.

In Vivo Measurements of pH and [Ca 2؉ ]
Because we used a double excitation fluorescence setup (described below), only Arabidopsis lines expressing ratiometric pHluorin were used in this study for pH measurements. For pH experiments, whole plants were placed in a plastic petri dish with a thin (0.13-mm) glass bottom. Only fully developed lateral branch roots were chosen and mechanically fixed to the glass bottom.

Solutions and Perifusion
NaCl stress was applied by 100 mm NaCl (Merck, Darmstadt, Germany). Drought stress was mimicked by application of isosmotic mannitol (Merck) solution (i.e. 200 mm). Osmolarities of both NaCl and mannitol solutions were checked by osmometry (OSMOMAT 030 Gonotec GmbH, Berlin). All solutions were unbuffered, if not stated otherwise. During all experiments, plant material was rinsed with freshly prepared and aerated solutions. For pH measurements, the petri dish was perifused with a flow rate of 5 mL min Ϫ1 . For [Ca 2ϩ ] measurements, the perfusion rate was 3 mL min Ϫ1 in the luminometer cuvette. Perifusion solutions of different composition were automatically switched by computer-controlled magnetic pinch solenoid valves (Sirai, Steinhö ring, Germany).

pH Measurements
The dish with fixed plant material was mounted on an inverse microscope (Diaphot, Nikon, Dü sseldorf, Germany) and perifused at all times with freshly made and aerated medium. The microscope was equipped with a 20ϫ objective (Fluor Ph3DL, Nikon) and an excitation ratio imaging device (TILL Photonics, Gräfelfing, Germany). GFP fluorescence was excited with light of 395, 428, and 475 nm from a monochromator (TILL), and emission was measured at 535 nm (emission filter, HQ535/50; dichroic mirror, 500dcxr; AHF, Tü bingen, Germany). Fluorescent images were taken every 12 s. The F 395 to F 475 ratio was used as a measure for pH (calibration; Fig.  2B), and the F 428 signal (i.e. isoexcitation point of ratio pHluorin; Fig. 2A) was taken as control to detect possible artifacts such as movement of the specimen or pH-sensitive autofluorescence. Macro programming of experiments, image acquisition, data collection, and evaluation were carried out using the software TillVision version 3.31 (TILL) extended with customtailored software for control of magnetic valves.

Reconstitution of AQ and Ca 2؉ Measurements
Reconstitution of AQ with CTZ was performed in planta essentially as described previously (Knight et al., 1997a). Synthetic native CTZ was from Prolume Ltd. (Pittsburgh), and cp-CTZ was from Molecular Probes Europe (Leiden, The Netherlands). In brief, CTZ was dissolved first in methanol to have a 1 mm stock solution and then added to pure water to produce a 10 m solution for reconstitution. Roots were dissected from plants grown on vertical agar and incubated in reconstitution medium in the dark. Roots with apoplastic expression lines were reconstituted for 3 to 4 h. Roots with cytoplasmic expression were reconstituted longer (i.e. overnight; 8-15 h) if not stated otherwise.
Reconstituted roots from about five plants were placed in a standard 4.5-mL acryl cuvette (no. 67.755, Sarstedt, Nü mbrecht, Germany) equipped with inlet and outlet for perifusion. The cuvette was fixed in a purpose-built, light-tight sample housing in front of a chemiluminometer (PMT 9829A, Electron Tubes Ltd., Ruislip, UK). Luminescence of the specimen was integrated every 12 s.

Distribution of Materials
Upon request, all novel materials described in this publication will be made available in a timely manner for noncommercial research purposes.