Crop Science 42:2201-2206 (2002)
© 2002 Crop Science Society of America
NOTES
Application of fluorescence microscopy and image analysis for quantifying dynamics of maize pollen shed
Agustin E. Fonsecaa,
Mark E. Westgate*,a and
Robert T. Doyleb
a Dep. of Agronomy, Iowa State Univ., 1563 Agronomy Hall, Ames, IA 50011-1010
b Dep. of Molecular Biology, Iowa State Univ., 122 Molecular Biology Building, Ames, IA 50011-3260
* Corresponding author (westgate{at}iastate.edu)
 |
ABSTRACT
|
|---|
The objective of this study was to develop a simple, rapid and accurate technique to quantify maize (Zea mays L.) pollen shed under field conditions, capitalizing on the capacity of pollen to fluoresce and recent improvements in microscopic methods for acquiring, processing, and analyzing fluorescence signals from biological systems. Pollen shed naturally by the tassels was captured daily on passive pollen traps placed at apical ear level. Fluorescence microscopy was used to generate digital images of the trapped pollen, and pollen density per unit area was counted using commercial imaging software. Visual confirmation of fluorescing pollen grains indicated high measurement accuracy (r2 = 0.99) for the entire range of pollen shed densities typically encountered in the field. Pollen was randomly distributed across the surface of the pollen trap, and six to eight images per trap provided greater than 95% confidence for the mean trap value. The entire process of sample preparation, image capture, and image counting required less than 6 min per trap. The accuracy and ease of use of this technique make it ideal for characterizing the pattern of maize pollen production and dispersal under field conditions.
|
INTRODUCTION
|
|---|
IN MAIZE, pollination can occur only if airborne pollen shed by the staminate flowers on the tassel is captured by the stigmas (silks) of pistillate flowers borne on the ear. Because the durations of pollen shed and silk receptivity are limited, close synchrony between pollen shed and silk emergence is required for high kernel set in the field (Bassetti and Westgate, 1994; Carcova et al., 2000). Environmental conditions can affect pollen availability by modifying the synchrony between pollen shedding and silk emergence, or by changing the amount of pollen produced per tassel (Hall et al., 1982; Bolaños and Edmeades, 1993). Distinguishing between these possibilities has important implications for germplasm selection as well as ensuring genetic purity. To do so, however, requires a simple, rapid, and reliable technique for quantifying the dynamics of maize pollen shed under field conditions.
Studies attempting to describe the quantitative patterns of pollen release from maize tassels are limited, generally because collecting and counting pollen is laborious. Flottum et al. (1984) determined the intensity of pollen shed by collecting samples on spinning rods and counting adhering pollen visually. Sadras et al. (1985a) collected pollen shed daily within tassel bags, and counted the pollen with a light microscope. More recently, Bassetti and Westgate (1994) monitored pollen shed intensity with passive pollen traps, and counted the pollen grains by computer-aided video imaging. They found that accuracy of the counting procedure depended almost entirely on the quality of the video image, which was difficult to standardize with bright-field optics. None of these techniques readily distinguished pollen from similar-sized debris in the measurement field.
Fluorescence microscopy has great potential for qualitative and quantitative studies on the structure and function of plant cells, since fluorescence from a single cell can be detected both as an image and as a photometric signal (Wang and Taylor, 1989; Fricker et al., 1997). Recent advances in instrumentation for acquiring, processing and analyzing fluorescence signals have made it possible to capitalize on natural autofluorescence of pollen grains and tubes to quantify pollen germination in Nicotiana tabacum L. (Tirlapur and Cresti, 1992; Keller and Hamilton, 1998), Tradescantia paludosa ES Anderson & Woods (Keller and Hamilton, 1998), and Pennisetum ciliare (L.) Link. (Shafer et al., 2000). To our knowledge, pollen fluorescence has not been used to characterize any aspect of maize pollen development or shedding dynamics.
The objective of this study was to develop a simple, rapid, and accurate technique to quantify the dynamics of maize pollen shed under field conditions, capitalizing on the capacity of pollen to fluoresce. Such a technique would have immediate application for relating pollen production to staminate flowering characteristics, assessing patterns of pollen dispersal outside of shedding source fields, and providing true quantitative measures of pollen production by male inbreds parents in seed production fields.
|
Materials and Methods
|
|---|
Plant Culture
Maize hybrids, Pioneer Brand 3394 and 3489, were planted in the field (Aquic hapludoll) near Ames, IA, on 10 May 2000 in 76-cm rows in eight-row plots (6.10 by 9.15 m) at a density of 80 000 plants ha-1. Three replicated plots were randomly distributed within a larger experiment occupying approximately 0.47 ha. Plots received 168 kg N ha-1 as anhydrous ammonia before planting. Rainfall before anthesis was adequate to ensure a normal pattern of pollen shed.
Pollen Sampling
Pollen shed was monitored daily by means of passive pollen traps placed horizontally in the center of the plant canopy at the apical ear level, about 120 cm above the ground. The trap was supported on a clear plastic base (10.6 by 10.6 cm) mounted on a plastic-coated metal stake. The pollen traps were constructed on a base of white opaque high impact polystyrene (HIPSP) sheeting (approximately 8 by 9 cm). Two bands of 1.9-cm-wide smooth, black tape (Super 88-3M Scotch Brand, St. Paul, MN) were placed across the white base to produce a high-contrast background for imaging (area = 34.2 cm2). The black tape was covered with double coated tape (666-3M Scotch Brand). The double-sided tape was protected by a white liner, which was removed to expose the sticky surface when the trap was positioned in the field.
Pollen traps were placed in the center of each plot at about 1600 h each day throughout the pollen-shed period, and remained in place for 24 h. Upon collection, anthers and other debris were removed by hand, and the traps were covered with an acetate sheet (Highland Brand, 3M, Austin, TX) to prevent contamination with additional pollen or other debris. Pollen traps were stored in the lab under ambient conditions, about 20C and 50% RH, until counted by image analysis. There was no evidence of sample deterioration in term of pollen fluorescence after 7 mo of storage under these conditions.
Image Analysis
Fluorescence and bright field images of pollen adhering to the traps were collected with a Nikon Eclipse 200 EPI-Fluorescence microscope equipped with a Nikon Plan Fluor 4X/0.13 NA objective (Fryer Company, Huntley, IL). For fluorescing images, preliminary results showed that excitation illumination at 488 nm and monitoring emission at 510 nm provided the strongest signal. Digital images of the pollen were captured with a Hamamatsu CCD C4742-95 camera (Fryer Company, Huntley, IL) controlled with acquisition software from Prairie Technologies (Middleton, WI).
Typically, 10 0.25-cm2 images were collected from each trap. Camera exposure time was 1.07 ms, and gain (image sensitivity) was set at 100. The system was automated to collect an image every 4 s. The images were saved in tif file format. In most cases, all 10 images from different positions on a trap were captured and saved in less than 1 min.
To determine whether a portion of the pollen grains failed to fluoresce, images from identical 1-cm2 areas were collected for a wide range of pollen densities using both fluorescence and bright field microscopy. Images from both sources were counted by eye and compared.
Pollen counting was performed with the Metamorph Imaging System (Universal Imaging Corporation, West Chester, PA). Image contrast and threshold were adjusted to differentiate between fluorescence of the background and that of pollen grains to be counted. Typically, maize pollen grains were 60 to 90 pixels in size, so the software was set to ignore objects smaller than 50 pixels. At high pollen shed densities, the occurrence of touching pollen increased. The software identified these occurrences as large objects and reported them as multiples of single pollen grains. These were easily accounted for by visual inspection of the image.
|
Results and Discussion
|
|---|
Accuracy
The accuracy of the fluorescing image technique requires that all pollen grains autofluoresce when exposed to the actinic light. Therefore we examined a large number of pollen traps for grains that did not fluoresce. Figure 1
shows that nearly every pollen grain observed in a bright field image fluoresce when exposed to excitation illumination of 488 nm, even after being stored in the lab for up to 7 mo.
View larger version (15K):
[in this window]
[in a new window]
|
Fig. 1. Direct measure of fluorescing pollen grains. Images from identical 1-cm2 areas taken by fluorescence and bright field microscopy were counted by eye. The actual number of pollen grains per square centimeter is taken from the bright field images. The line represents the 1:1 relation.
|
|
Predicted results from the Metamorph software also were compared with actual image values (visually counted) for a wide range of pollen densities (Fig. 2)
. Predicted values for individual images were very accurate, especially at pollen grain concentrations less than 200 grains cm-2. At greater densities, the prediction tended to underestimate actual values slightly. This was due to two or more grains touching, which was increasingly common at high pollen densities. This counting error was easily corrected by a quick visual check on the computer screen for grains not counted (the software uses color to identify objects that are counted).
View larger version (19K):
[in this window]
[in a new window]
|
Fig. 2. Correlation between actual and predicted pollen density. Actual pollen density on the digital image was counted by eye, and predicted pollen density was provided by the Metamorph imaging software. The thick line represents the 1:1 relation, and the thin line is the linear regression. Each point is from a single image.
|
|
To determine if pollen deposition on the surface of the pollen traps occurred at random, pollen traps with contrasting amounts of pollen were analyzed at forty 0.25-cm2 positions across each trap surface. As expected, some variability in pollen density across the pollen trap was observed (Fig. 3)
. But there was no obvious bias in pollen deposition on the trap surface at any pollen shed density. Therefore, a practical way to obtain a representative value for each pollen trap was simply to divide the surface into uniform sections, collect an image from each section, and average the values from these images.
View larger version (87K):
[in this window]
[in a new window]
|
Fig. 3. Distribution of pollen counts on pollen traps collected at three pollen shed densities. Each value is the number of pollen grains per square centimeter. Average, standard deviation and coefficient of variation are shown for each density. Values more than one standard deviation greater than the mean (> mean +1 S.D.) are shown in black. Values more than one standard deviation less than the mean (< mean -1 S.D.) are shown in white.
|
|
A running average approach was used to determine the minimum number of images necessary to achieve a reliable estimate of the pollen density on traps collected at various times during pollen shed. Figure 4
shows that averaging counts from about eight images typically was sufficient to achieve a value within the 95% confidence interval of the overall trap average (derived from 10 or more images). But the numerical differences in pollen density between traps were evident even among single images. Thus, the number of images taken from each trap depends largely on the accuracy required and amount of time and effort available for data analysis.
View larger version (16K):
[in this window]
[in a new window]
|
Fig. 4. Pollen density estimated using an increasing number of images taken from pollen traps collected on different days during pollen shed. The average value for each trap was calculated from 10 images and is represented by the horizontal line. Triangles indicate values significantly different from the average value for the pollen trap (P < 0.05).
|
|
The fluorescence image presents pollen grains as bright spots on a dark field (Fig. 5)
. Accuracy of counting these bright spots depends almost entirely on the quality of the image. Consequently, particular attention must be paid during sample preparation and illumination to optimize image contrast and uniformity. Cleanliness of pollen trap will, of course, improve the image quality. On windy days, foreign debris often contaminates the traps. Large pieces, such as anthers or insects, are readily removed upon trap collection. Smaller particles also can be eliminated from the counting routine by means of the size-range option in the software program. Any dubious cases can be checked rapidly on the monitor and subtracted or added to the count as appropriate. In the 130 pollen grains image of Fig. 5, for example, the two long-narrow objects were counted as "pollen grains" and subsequently subtracted manually from the image count.
View larger version (117K):
[in this window]
[in a new window]
|
Fig. 5. Digital images of fluorescing pollen taken from pollen traps collected at varying pollen shed densities. Contrast and threshold intensity were adjusted to distinguish pollen from background fluorescence and foreign objects prior to counting. Numbers in the left top corner indicate the number of pollen grains in each image. Each image is 0.25 cm2.
|
|
Irregularities on the pollen trap surface, such as small bubbles, tape edges, or even a slight curvature in the pollen trap itself make it more difficult to achieve the uniform image background required for accurate pollen counting. Curvature of the pollen trap, for example, caused pollen grains to move in and out of the focal plane as the trap was repositioned for each successive image. Because pollen scatters the fluorescent light to some extent, pollen grains that are out of focus appear slightly larger in the digital image. This variation in pollen size is evident among the images in Fig. 5. In large part, focusing and background problems related to trap curvature are overcome by keeping the pollen traps perfectly flat on the microscope stage. Image problems associated with bubbles and roughness of the pollen trap surface were avoided only by careful attention to trap construction.
Ease of Use
With experience, the time required to obtain six to eight images from a pollen trap can be reduced to less than 1 min. If the original image is correctly focused, the auto-threshold function in the Metamorph software readily differentiates the pollen grains from background without operator adjustments. Small particles, such as dust, are eliminated easily from the count by a size limit option. And large objects, such as pieces of anthers or debris that were not previously removed, are easily detected and avoided. Once an operator is familiar with the imaging software, two to three images can be counted per minute. So the whole procedure to capture, process, and count each image requires only about 40 s. Relative to techniques that rely on light microscopy (Bassetti and Westgate, 1994) or scanners to generate digital pollen images, the fluorescence technique is more accurate, more rapid, and considerably easier to use.
Finally, another positive feature of our approach is that pollen traps can be stored for extended periods without affecting the counting results. We have observed that pollen stored on traps autofluoresce effectively even after 2 yr of storage in the laboratory. In this study, fluorescent images were prepared 7 mo after pollen traps were collected from the field. This represents a significant advantage over techniques such as coulter counting (Carre and Tasei, 1997), which requires collecting and storing pollen in an isotonic solution that has a limited shelf life.
Field Application
The accuracy observed across a broad range of pollen shed densities (Fig. 1 and 2) indicates this technique would be well suited for documenting variations in pollen production due to tassel size or male sterility, and characterizing pollen production in seed production fields. A practical application of this pollen measurement technique is shown in Fig. 6
. We quantified pollen shed in the field for two hybrids, grown at typical commercial plant densities. Both hybrids exhibited a similar pattern of pollen shed, which lasted about 2 wk and peaked about 2 d after anthesis (50% of the plants at maximum shed). Differences between hybrids for the maximum rate of pollen shed and total amount of pollen produced per square meter are readily apparent. Taking the area under the pollen shed curve as an estimate of the total amount of pollen shed by these hybrids, P3394 produced approx. 26 x 106 pollen grains m-2, while P3489 produced about 18 x 106 pollen grains m-2. These values correspond to 3.25 x 106 and 2.25 x 106 grains per plant, for P3394 and P3489, respectively. Total pollen production for these hybrids is considerably less than the 69 x 106 grains m-2 reported by Hall et al. (1982) and the 40 to 96 x 106 grains m-2 reported by Sadras et al. (1985b), who collected pollen in bags from selected plants and counted pollen by eye under a microscope. The lesser values for P3394 and P3489 in our study likely reflect the smaller tassel size of today's maize hybrids (Galinat, 1992; Duvick, 1997), some loss of pollen carried by the wind out of the plot area (Westgate et al., 2000), and the greater inherent accuracy of our pollen counting technique. In any case, our fluorescence-based measurements of daily and seasonal pollen shed density obtained with passive pollen traps represent pollen production by the local plant population that actually reaches the plane of exposed silks.
View larger version (16K):
[in this window]
[in a new window]
|
Fig. 6. Seasonal pattern of pollen shed relative to days after anthesis (50% of plants at maximum pollen shed) for 2 commercial maize hybrids (P3394 and P3489). The data are the average ± standard error for three replicate plots. Note that pollen shed lasts about 14 d and peaks 2 to 3 d after anthesis for both hybrids.
|
|
Because this method of pollen quantification is simple, accurate, and convenient, it has immediate application for relating staminate flower development to pollen release, documenting production conditions where pollen amount is limiting, and quantifying pollen dispersal downwind of corn fields. Future modifications of this technique focus on the possibility of using fluorescent images to distinguish pollen from different genetic sources (Abdul-Baki, 1992; Aronne et al., 2001).
Received for publication November 12, 2001.
|
REFERENCES
|
|---|
- Abdul-Baki, A.A. 1992. Determination of pollen viability in tomatoes. J. Am. Soc. Hortic. Sci. 117 (3):473–476.[Abstract/Free Full Text]
- Aronne, G., D. Cavuoto, and P. Eduardo. 2001. Classification and counting of fluorescent pollen using an image analysis system. Biotech. Histochem. 76:35–40.[ISI][Medline]
- Bassetti, P., and M.E. Westgate. 1994. Floral asynchrony and kernel set in maize quantified by image analysis. Agron. J. 86:699–703.[Abstract/Free Full Text]
- Bolaños, J., and G.O. Edmeades. 1993. Eight cycles of selection for drought tolerance in lowland tropical maize. II. Response in reproductive behavior. Field Crops Res. 31:253–268.
- Carcova, J., M. Uribelarrea, L. Borras, M.E. Otegui, and M.E. Westgate. 2000. Synchronous pollination within and between ears improves kernel set in maize. Crop Sci. 40:1056–1061.[Abstract/Free Full Text]
- Carre, S., and J.N. Tasei. 1997. Use of a flow-cytometer for pollen counting—An application to Vicia faba L. Acta Hortic. 437:369–372.
- Duvick, D.N. 1997. What is yield? In G.O. Edmeades et al. (ed.) Developing drought and low N-tolerant maize. Proc. of a symposium, El Batan. 25–29 March 1996 CIMMYT, El Batan, Mexico D.F.
- Flottum, P.K., D.C. Robacker, and E.H. Erickson. 1984. A quantitative sampling method for airborne sweet corn pollen under field conditions. Crop Sci. 24:375–377.[Abstract/Free Full Text]
- Fricker, M.D., C.M. Chow, R.J. Errington, M. May, J. Mellor, A.J. Meyer, M. Tlalka, D.J. Vaux, J. Wood, and N.S. White. 1997. Quantitative imaging of intact cells and tissues by multi-dimensional confocal fluorescence microscopy. Exp. Biol. Online 2: 19.
- Galinat, W.C. 1992. Evolution of corn. Adv. Agron. 47:203–231.
- Hall, A.J., F. Villela, N. Trapani, and C. Chimenti. 1982. The effect of water stress and genotype on the dynamics of pollen-shedding and silking in maize. Field Crops Res. 5:349–363.
- Keller, N.L., and D.A. Hamilton. 1998. Transient expression of the green fluorescent protein in pollen. Sex. Plant Reprod. 11:163–165.
- Sadras, V.O., A.J. Hall, and T.M. Schlichter. 1985a. Kernel set of the uppermost ear in maize: I. Quantification of same aspects of floral biology. Maydica 30:37–47.
- Sadras, V.O., A.J. Hall, and T.M. Schlichter. 1985b. Kernel set of the uppermost ear in maize: II. A simulation model of effects of water stress. Maydica 30:49–66.
- Shafer, G.S., B.L. Burson, and M.A. Hussey. 2000. Stigma receptivity and seed set in protogynous buffelgrass. Crop Sci. 40:391–397.[Abstract/Free Full Text]
- Tirlapur, U.K., and M. Cresti. 1992. Computer-assisted video image analysis of spatial variations in membrane-associated Ca2+ and calmodulin during pollen hydration, germination and tip growth in Nicotiana tabacum L. Ann. Bot. (London ) 69:503–508.[Abstract/Free Full Text]
- Wang, Y.-L., and L.D. Taylor. 1989. Fluorescence microscopy of living cells in culture p. 125–135. In Methods in cell biology, vol. 29, part A. Academic Press, London.
- Westgate, M.E, D.S. Ireland, and C. Daniel. 2000. Predicting maize pollen travel using particulate dispersion models. p. 146. In Agronomy abstracts. ASA, Madison, WI.
This article has been cited by other articles:
|
|
|
|
 
E. J. Rosi-Marshall, J. L. Tank, T. V. Royer, M. R. Whiles, M. Evans-White, C. Chambers, N. A. Griffiths, J. Pokelsek, and M. L. Stephen
Toxins in transgenic crop byproducts may affect headwater stream ecosystems
PNAS,
October 9, 2007;
104(41):
16204 - 16208.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
J. I. Lizaso, A. E. Fonseca, and M. E. Westgate
Simulating Source-Limited and Sink-Limited Kernel Set with CERES-Maize
Crop Sci.,
September 1, 2007;
47(5):
2078 - 2088.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
A. E. Fonseca, J. I. Lizaso, M. E. Westgate, L. Grass, and D. L. Dornbos Jr.
Simulating Potential Kernel Production in Maize Hybrid Seed Fields
Crop Sci.,
September 1, 2004;
44(5):
1696 - 1709.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
J. I. Lizaso, M. E. Westgate, W. D. Batchelor, and A. Fonseca
Predicting Potential Kernel Set in Maize from Simple Flowering Characteristics
Crop Sci.,
May 1, 2003;
43(3):
892 - 903.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
M. E. Westgate, J. Lizaso, and W. Batchelor
Quantitative Relationships between Pollen Shed Density and Grain Yield in Maize
Crop Sci.,
May 1, 2003;
43(3):
934 - 942.
[Abstract]
[Full Text]
[PDF]
|
|
|
TOP
ABSTRACT
INTRODUCTION
