Guard cell-specific microarray data were analyzed side by side with mesophyll cell-specific microarray data 26 to search for strong guard cell promoter candidates with low expression levels in mesophyll cells. Additional guard cell and mesophyll cell microarray experiments were conducted covering 23,000 genes (ATH1 Affymetrix) (See Additional files 1, 2, 3, 4, 5, 6, 7 and 8). Furthermore, candidate genes were analyzed using Genevestigator to select genes with low expression levels in non-leaf tissues across more than 2000 microarray experiments 27. Guard cells and mesophyll cells exposed to ABA were also analyzed, as ABA synthesis is induced under several stress conditions. The following criteria were used for selection of strong guard cell promoter candidates. The raw signal in guard cells was set above 10000, the raw signal in mesophyll cells was set below 1000, and the reduction or induction fold by ABA was set to be less than two. Transcriptional profiles of several genes passed these criteria (Figure 1 and Additional file 9). The putative promoters (1–2 kb upstream of the annotated ATG start codon) (Figure 2) were amplified by PCR and cloned into a GUS reporter vector. GUS staining of the T1 transgenic plants showed guard cell specific staining for one particular promoter candidate (At1g22690), designated as pGC1. At1g22690 is among the most highly expressed genes in guard cells. It showed relatively high expression in guard cells and low expression in mesophyll cells. At1g22690 encodes a small cysteine rich protein (119 amino acids). It belongs to the GASA family (GA-stimulated transcript (GAST1) protein homolog). A study by Wigoda et al. 28 suggested that GIP2 (a GASA protein from Petunia hybrida) exhibited in planta antioxidant activity. T-DNA insertional line in At1g22690 did not yield any noticeable stomatal phenotypes under our typical laboratory conditions (unpublished data). Furthermore, our guard cell microarray data showed that two other GASA genes also showed high expression level in guard cells (GASA 1 (At1g75750) and GASA 4 (At5g15230)).

We analyzed GC1 (At1g22690) gene expression in response to different treatments in the microarray data compiled by Genevestigator 2729. Among 96 treatments, 8 treatments affected At1g22690 expression more than two fold. Salt and osmotic stress dramatically deceased At1g22690 gene expression (more than 10 fold) 30. Meanwhile, light, ABA, GA, cold or drought did not induce more than a two-fold change in gene expression of At1g22690. This suggests that GC1 (At1g22690) has a relatively constant expression under most common situations.

Interestingly, the pGC1::GUS not only delivered strong GUS expression in guard cells in leaves (Figure 3A,B), but also in guard cells in petioles and hypocotyls (Figure 3C,D,E). GUS staining from other candidate promoter-GUS fusions was either not very strong in guard cells and/or showed reporter expression in other tissues (data not shown). We therefore focused on pGC1 for the rest of this study. The GC1 promoter was also fused to a second reporter, a GFP-based calcium reporter, yellow cameleon 3.60 (YC3.60) 31. Most T1 transgenic plants (approximately 75%) transformed with pGC1::YC3.60 exhibited strong guard cell specific fluorescence, indicating a high degree of guard cell expression efficiency per transformant. Some plants also showed fluorescence in some leaf epidermal cells (data not shown). However, younger or immature guard cells showed no or much less GFP expression (Figure 3F,G). Furthermore, guard cells in sepals and hypocotyls also showed GFP expression (Figure 3H,I,J,K).

We further examined whether the GC1 promoter could drive guard cell specific reporter expression in a guard cell development mutant, too many mouths (tmm) 32. The tmm mutant was transformed with either the pGC1::GUS or the pGC1::YC3.60 construct. GUS staining showed reporter gene expression in clustered stomata (Figure 3L). Similarly, GFP expression was observed in clustered stomata in tmm plants transformed with pGC1::YC3.60 (Figure 3M).

To test if the GC1 promoter can drive guard cell specific reporter gene expression in plants besides Arabidopsis, we also transformed pGC1::YC3.60 into tobacco plants. Interestingly, strong guard cell GFP expression was observed in tobacco leaves (Figure 3N).

A promoter region may contain both enhancer and repressor elements. To probe which part of the original 1716 bp promoter (full length, FL, -1693 bp/+23 bp) is required for strong guard cell specific reporter expression, four 5' truncated versions of the GC1 promoter were generated as D1 (-1140 bp/+23 bp), D2 (-861 bp/+23 bp), D3 (-443 bp/+23 bp), and D4 (-224 bp/+23 bp) (Figure 4). These truncated promoters were fused to the GUS reporter to generate the following constructs: pGC1(D1)::GUS, pGC1(D2)::GUS, pGC1(D3)::GUS and pGC1(D4)::GUS. These GUS reporter constructs were transformed into Columbia wild type plants side-by-side with the original pGC1(FL)::GUS construct. T1 seedlings (n = 50–100) from each transformation event were pooled and stained. The truncated pGC1(D1) drove similar or stronger GUS expression in seedlings than the original full-length promoter (Figure 4A), suggesting that elements in the region from -1693 bp to -1140 bp might repress promoter activity in guard cells. Promoters pGC1(D2) and pGC1(D3) led to weaker reporter gene expression in guard cells than pGC1(FL), suggesting elements in the region from -1140 bp to -443 bp might enhance the promoter activity in guard cells. The shortest promoter, pGC1(D4), drove reporter gene expression in tissues other than guard cells, such as roots and seed coats, suggesting the region from -861 bp to -224 bp was required for guard cell specific activity. This region contains 8 (T/A)AAAG elements that have been shown to be required for guard cell specific activity of the KST1 promoter in potato 24. The truncated promoter, pGC1(D1), showed strong guard cell expression, suggesting that it contains elements for both guard cell specificity and promoter strength.

Many physiological stimuli in plant cells induce changes in the intracellular calcium concentration. Calcium acts as a secondary messenger in many signal transduction cascades 33. Cytosolic calcium concentrations can be monitored either by chemical reporters such as the ratiometric Ca²⁺-sensitive fluorescent dye fura-2 3435, the genetically encoded calcium sensitive luminescent protein aequorin 14 or the fluorescent ratiometric calcium reporter yellow cameleon 121536. Stomatal closing signals, such as ABA and CO₂, have been shown to induce calcium elevations in guard cells 161819373839404142. Spontaneous calcium transients in leaf epidermal samples have also been observed without any ABA treatment 154344. It is not clear whether spontaneous calcium transients occur in guard cells in intact plants as fura-2 injected Vicia faba guard cells did not show such transients 45. A new generation calcium indicator, yellow cameleon, YC3.60, shows an enhanced calcium-dependent change in the ratio of YFP/CFP by nearly 600% compared with yellow cameleon 2.1 31. By combining the GC1 promoter with YC3.60, pGC1::YC3.60, as described before, we could observe strong guard cell expression of the YC3.60 in intact leaves, hypocotyls, and sepals (Figure 3).

We first measured calcium transients in intact leaf epidermis from plants transformed with pGC1::YC3.60 by imposing calcium oscillations as described previously 1146. Robust calcium transients with ratiometric changes of up to a factor of 4 relative to the baseline ratio could be observed in guard cells (Figure 5B). Ratiometric changes of approximately 0.5 were observed using 35S::YC2.1 in response to imposed calcium transients 15434446. This further confirmed the robust ratiometric signal to noise efficiency of YC3.60. Next, we performed calcium imaging in intact Arabidopsis seedlings by mounting leaves to a microscope cover glass. Two different methods were tested: the first one was to submerge only the root with water and leave the shoot in air, and the second one was to submerge the entire plant in water. Spontaneous calcium transients were detected under both conditions (Table 1). A representative calcium transients/time course is shown in Figure 5D. Interestingly, the spontaneous calcium transients of two guard cells from the same stomate were often not synchronized (Figure 5C,D and additional files 10 and 11). These experiments clearly demonstrate that spontaneous calcium transients occurred in guard cells of intact plants and were not an artifact of imaging excised epidermis and illustrate the potential of the pGC1 promoter as a method for driving transgene and reporter expression in guard cells.

Manipulation of specific gene expression in guard cells, either by highly expressing the wild-type gene or a dominant mutant form, or reducing its expression in guard cells, would be very powerful to probe a specific gene function in guard cells. To further explore the application of the GC1 promoter, we took the antisense approach to analyze reduction of gene expression in guard cells. For this purpose, a 35S::GFP transgenic line with stable GFP expression in both guard cells and epidermal cells (Figure 6A,B) was transformed with a pGC1(D1)::anti-GFP construct (anti-GFP fused to the truncated GC1 promoter pGC1(D1)). 34 out of 40 T1 plants of 35S::GFP plants transformed with pGC1(D1)::anti-GFP showed greatly reduced GFP expression in guard cells while the GFP expression level in epidermal cells was unchanged (Figure 6C,D). These observations suggest a remarkable antisense repression efficiency using pGC1(D1). Interestingly, less suppression of GFP expression was observed in immature guard cells (yellow arrow in Figure 6D). This is consistent with the observation that pGC1 drove less reporter gene expression in immature guard cells (Figure 3G). This experiment strongly indicates that an antisense approach can be used to reduce expression of selected genes in guard cells without affecting its expression in other cell types.

As the central regulator of water transpiration and CO₂ uptake, guard cells have been developed as an integrative model system to investigate interplay among ion channel/transporter activities, light, plant hormones, secondary messengers, the cytoskeleton and membrane trafficking in regulating the physiological output: the stomatal aperture 2454748. Several guard cell promoters have been reported. The KAT1(At5g46240) promoter delivered specific reporter expression in guard cells even though it sometimes induced reporter expression in other cells and tissues such as roots and inflorescences 25. AtMYB60 (At1g08810) also showed specific expression in guard cells based on promoter::GUS and promoter::GFP study 49. AtMYB61(At1g09540) has also been shown to be mainly expressed in guard cells 50. Based on our guard cell-specific microarray data, we estimated the average transcription levels in Figure 1 and additional file 9. The AtMYB61 gene expression signal was the lowest among these genes. In the case of KAT1, its expression in guard cells was much higher than that in mesophyll cells. But its raw signal was approximately 5 to 10 fold lower than that of GC1. AtMYB60 also exhibited highly guard cell specific expression compared with its expression in mesophyll cells. However, the raw signal of AtMYB60 was only approximately one third of that of GC1. Furthermore, AtMYB60 is also highly expressed in seeds based on Genevestigator microarray analyses 272951525354. Similarly, RAB18 (At5g66400) is also highly expressed in seeds besides its strong expression in guard cells. pGC1 drove very strong and specific reporter gene expression in guard cells (expression is very low in non-leaf tissues/organs), although reporter gene expression was observed in epidermal cells in some plants transformed with the pGC1::YC3.60 (data not shown). In summary, the GC1 promoter is a very strong guard cell promoter among those analyzed.

Our current study with intact Arabidopsis plants using the genetically encoded calcium reporter YC3.60 driven by the GC1 promoter showed that spontaneous calcium transients occurred in guard cells in intact Arabidopsis plants. This is consistent with previous observations of spontaneous calcium transients in Arabidopsis guard cells 154344. However, the mechanisms causing spontaneous calcium transients are not yet characterized in depth. Several lines of evidence suggest a connection between hyperpolarization of the guard cell plasma membrane and spontaneous calcium transients in guard cells. In experiments where membrane potential and [Ca²⁺]_cyt were measured simultaneously, hyperpolarization caused ABA-induced [Ca²⁺]_cyt increases. Maintaining guard cells in a more hyperpolarized state produced spontaneous [Ca²⁺]_cyt oscillations in Vicia faba guard cells 38, in a sub-population of Commelina guard cells 39 and in Arabidopsis guard cells 43. Calcium imaging analyses in intact Arabidopsis plants using pGC1::YC3.60 show that spontaneous calcium transients also occur in intact plants. These spontaneous Ca²⁺ transients may also be the result of integrated signaling by multiple stimuli converging in guard cells, such as light conditions, CO₂ and water balance. In Vicia faba no spontaneous calcium transients were observed in guard cells in intact plants 45. In this case fura-2 (ca. 100 μM) was injected into guard cells. High concentrations of fura-2 may inhibit spontaneous calcium elevations, as loading the close fura-2 analogue, BAPTA, into Arabidopsis guard cells effectively inhibits these calcium transients 44. By contrast, the estimated yellow cameleon concentration in guard cells of pGC1::YC3.60 transgenic plants was approximately 1 μM (see Methods). The lower concentration of yellow cameleon should interfere less with guard cell calcium homeostasis and could monitor more faithfully calcium concentration dynamics. Note that low concentrations of injected fura-2 also allowed resolution of repetitive calcium transients in guard cells 3839. Note that BAPTA-derived fluorescent dyes such as fura-2 and indo-1 have certain complementary advantages to cameleon, as they can be loaded into cells that are not easily transformed 55 and these dyes can report rapid millisecond scale Ca²⁺ transients that occur in neurons56, but have presently not yet been reported in plants using fura-2 or indo-1.

Circadian calcium oscillations at the whole plant leaf level with a daily rhythm have been demonstrated by several groups using aequorin as the calcium reporter 575859. Most likely this circadian calcium oscillation results from synchronous changes in baseline cytosolic calcium in a cell population 60. As the circadian calcium oscillation is related to the baseline of intracellular calcium, the rapid spontaneous calcium transients in individual guard cells likely would be filtered from circadian calcium measurements 60. Repetitive calcium transients may reflect functions that include continuous calcium homeostasis between extracellular calcium, cytoplasmic calcium, and intracellular calcium stores. Spontaneous calcium transients in guard cells also correlate with the recent proposed calcium sensor priming hypothesis for calcium specificity in signaling, in which the stomatal closing signals ABA and CO₂ are proposed to prime (de-inactivate) calcium sensitive steps that mediate stomatal closing 4461.

(T/A)AAAG, a binding motif for Dof zinc finger transcription factors, has been suggested to play a critical role for guard-cell specific expression of KST1 promoter activity in potato based on block mutagenesis 24. However, the putative promoter regions (1800 bp before ATG start codon) for AtACT7 (At5g09810), KAT1 (At5g46240), RAB18 (At5g66400), AtMYB60 (At1g08810), AtMYB61 (At1g09540) and GC1 (At1g22690) all contain a similar number of Dof factor binding motifs, the (T/A)AAAG elements, even though some of them do not show guard cell expression preference (see additional file 12). AtMYB61, which showed low expression in guard cells (Figure 1), contains 29 (T/A)AAAG elements in its putative promoter region, while the AtACT7 promoter contains 23 (T/A)AAAG elements. Promoter truncation suggests that the region from -861bp to -224 bp in the GC1 promoter contains elements for guard cell specific promoter activity (Figure 4). This region contains 8 (T/A)AAAG elements. However, block mutagenesis of the central TAAAG motif on the sense strand (marked by a star in Figure 4B) in this region did not affect reporter expression in guard cells (data not shown). Thus the (T/A)AAAG element alone may not explain why GC1 and other guard cell-specific genes exhibited guard cell-specific expression.

Arabidopsis thaliana (Columbia ecotype) plants were used for transformation experiments unless otherwise specified. The 35S::GFP transgenic line was generated for a previous study 62. The guard cell development mutant, too many mouths, was a kind gift from Dr. Fred Sack at the University of British Columbia, Vancouver.

Plant growth, ABA treatment, guard cell protoplast isolation, and RNA extraction were performed as previously described 26. Affymetrix Arabidopsis ATH1 genome arrays (Santa Clara, CA) were used, representing approximately 23,000 genes. Transcripts were amplified, labeled, and hybridized at the University of California, San Diego Gene Chip Core facility. For each condition (with or without ABA treatment, guard cell or mesophyll cell), two independent hybridizations were performed. Transcriptional inhibitors (33 mg/L actinomycin D and 100 mg/L cordycepin) were added during protoplast isolation for RNA samples for four chip hybridizations as described 26. ATH1 microarray data were deposited at MIAMExpress 63 with an accession number E-MEXP-1443 and also on our laboratory's website for public downloading (see Additional files 1, 2, 3, 4, 5, 6, 7 and 8).

To amplify the GC1 (At1g22690) promoter from the Col genomic DNA by PCR, primers YZ27 (5'-CATG

CCATGG

GTCGAC

To clone the GC1 promoter into the pBI101 vector, pGEM-T-pGC1 was first cut by NcoI. The sticky end was then filled-in by T4 DNA polymerase (New England BioLabs) to create a blunt end. The pGC1 fragment was then released by SalI digestion. Meanwhile, the destination vector, pBI101, was cut sequentially by SmaI and SalI. The pGC1 fragment was then inserted upstream of the GUS reporter gene in the pBI101 vector to create pBI101-pGC1::GUS construct (simplified as pGC1::GUS).

To create the 5'-deletion series of the pGC1 promoter, primer YZ27 was used with primers YZ159 (5'-GC

GTCGAC

GTCGAC

To create pBI101-pGC1::YC3.60 construct, YC3.60 was first released from pcDNA3-YC3.60 31 by EcoRI/BamHI double digestion. Then the BamHI-5'-YC3.60-3'-EcoRI fragment was cloned into pSK vector (prepared by EcoRI and BamHI digestion) to create pSK-YC3.60 construct. The pSK-YC3.60 was then digested with NotI and NcoI to receive NotI-5'-pGC1-3'-NcoI fragment from pGEM-T-pGC1. This ligation resulted in the pSK-pGC1::YC3.60. The pGC1::YC3.60 fragment was released by SalI/SacI double digestion, meanwhile the pBI101 vector was digested with SalI/SacI to remove the GUS reporter gene. The pBI101(SalI/SacI) was ligated with SalI-5'-pGC1::YC3.60-3'-SacI to create pBI101-pGC1::YC3.60 construct.

To create pGreenII 0179-pGC1(D1)::anti-GFP binary vector with hygromycin selective marker in plant, the 35S terminator was amplified with YZ439 (5'-AAGAGATCTA

TCTAGA

GAGCTC

CCATGG

TCTAGA

The central TAAAG motif (-579-->-575) on the sense stand was changed to CGGGA by block mutagenesis using the QuickChange Site-Directed Mutagenesis Kit from Stratagene (La Jolla, California).

The binary constructs, pBI101-pGC1::YC3.60, pBI101-pGC1::GUS, pBI101-pGC1(D1)::GUS, pBI101-pGC1(D2)::GUS, pBI101-pGC1(D3)::GUS and pBI101-pGC1(D4)::GUS were transformed into the Agrobacterium tumefaciens strain GV3101 by electroporation. The transformants were selected on LB plates with both kanamycin (selective marker for the construct) and gentamycin (selective marker for the Agrobacterium). Arabidopsis plants were then transformed by Agrobacterium GV3101 hosting respective constructs following the dipping method as described by Clough and Bent 65. The T0 seeds were selected on 1/2 MS plates with 50 μg/ml kanamycin.

In the case of pGreenII 0179-pGC1(D1)::anti-GFP, the GV3101 with the helper plasmid pSOUP was used as the host strain, and the selection for Agrobacterium transformants was carried on LB plates with Kanamycin, gentamycin, and tetracyclin. This was used to transform 35S::GFP transgenic plants (kanamycin resistant). The T0 seeds were selected on 1/2 MS plates with 25 μg/ml hygromycin (Roche).

Seedlings were stained following a previously described protocol 62.

Transgenic Arabidopsis seedlings or sepals of pBI101-pGC1::YC3.60 were simply placed between a microscope slide and a cover glass. A Nikon digital camera was attached to the microscope. Exposure time for the bright image is 5 seconds and 15–25 seconds for fluorescence image (excitation wavelength is 440 nm). For 35S::GFP plants and 35S::GFP plants transformed with pGreenII 0179-pGC1(D1)::anti-GFP, intact leaf epidermis were used for epi-fluorescence image acquisition.

In vitro sterile shoot cultures of Nicotiana tabacum cv. SR1 were maintained on 1/2 MS agar medium containing 15 g/l sucrose. The pH was adjusted to 5.5 before autoclaving. The tobacco culture was grown at 25°C, with a light/dark cycle of 16/8 h (light intensity was approximately 70 μmol m^-2 s^-1). Stable transformation of Nicotiana tabacum SR1 with pBI101-pGC1-YC3.60 was performed as described previously 66. Transgenic regenerated tobacco shoots were selected by kanamycin (100 μg/ml) resistance and were then transferred on 1/2 MS agar medium containing 15 g/l sucrose supplemented with kanamycin (100 μg/ml) and cefotaxime (200 μg/ml). T1 regenerated plants, which were able to set up root organogenesis in presence of kanamycin, were then analyzed for cameleon expression.

The tobacco leaves of plant transformed with pBI101-pGC1-YC3.60 were observed with a Leica TCS SP2 laser confocal microscope (Leica Microsystems). For cameleon detection, excitation was at 514 nm and emission between 525 and 540 nm. The images acquired from the confocal microscope were processed using Image J 67.

All calcium imaging in this work was performed with a TE300 inverted microscope using a TE-FM Epi-Fluorescence attachment (Nikon Inc. Melville, NY). Excitation from a 75 W Xenon lamp (Osram, Germany) was always attenuated 97% by using both 4× and 8× neutral density filters (3% transmission) to reduce bleaching of reporters during time-resolved imaging. Wavelength specificity was obtained with a cameleon filter set (440/20 excitation, 485/40 emission1, 535/30 emission2, 455DCLP dichroic; filter set 71007a Chroma Technology, Rockingham, VT). Filter wheel, shutter and CoolSNAP CCD camera from Photomerics (Roper Scientific, Germany) were controlled with Metafluor software (MDS, Inc., Toronto, Canada).

Intact leaf epidermes of pGC1::YC3.60 transgenic plants were prepared for microscopy as described in Mori et al. (2006) 11. On the microscope, intact epidermis was perfused with depolarization buffer (10 mM MES-Tris buffer, pH 6.1 containing 25 mM dipotassium imminodiacetate, and 100 μM BAPTA) for 10 minutes to obtain a background. Subsequently hyperpolarizing buffer containing Ca²⁺ (10 mM MES-Tris buffer, pH 6.1, 1 mM dipotassium imminodiacetate, and 1 mM CaCl₂) was applied for 2 minutes intervals, followed by 5 minutes of depolarizing buffer.

Both intact leaves and intact plants were used in this study. Medical adhesive (Hollister Inc., Libertyville, IL) was used to attach leaves to microscope cover glasses. A paintbrush was used to gently press the leaf to the coverslip. In the case of intact plants two different methods were followed. The first method was to submerge only the root with water while the shoot was left in air. The second method was to completely submerge entire seedlings in water. Sometimes submerging only the root but not the shoot caused the leaf attached to the cover slip to show wilting in less than 10 minutes with subsequent closure of the stomata. Most of the intact plant imaging experiments were therefore carried out by submerging both the shoot (leaves) and the root in water. The submersion of the entire plant prevented the leaf from drying out and no stomatal closure was observed for more than 50 minutes. The imaging protocol was the same as in Mori et al., 2006 11.

Recombinant yellow cameleon protein was isolated after expression in E coli. Recombinant cameleon protein was then added at defined concentrations to a glass cover slip for fluorescence imaging. Then two additional cover slips were used to create a slanted gradient of cameleon solution thicknesses. This enabled analysis of various solution thicknesses in the range of stomatal guard cell thicknesses. Diluted yellow cameleon protein solutions at different concentrations were analyzed and the florescence intensity was measured for each concentration at various thicknesses. Calibration curves were generated for protein concentrations and florescent intensities at different thicknesses. This was utilized to estimate the yellow cameleon protein concentration in guard cells of pGC1::YC3.6 transgenic plants.