text
stringlengths 1
6.27k
| id
int64 0
3.07M
| raw_id
stringlengths 2
9
| shard_id
int64 0
0
| num_shards
int64 16
16
|
|---|---|---|---|---|
pre-field labeling, cells were kept on ice for up to 5 h before use. For TRITC conjugation, to l mg of mouse IgE in l ml PBS was added 2 drops of 1 M carbonate buffer, pH 9. l0 #l of a 20 mg/ml solution of the isothiocyanate in dimcthylsulfoxide was added, the mixture homogenized by inversion and left on ice for 4 h. Free dye was removed by passage through a P6 column (Bio-Rad Laboratories, Budingame, CA) equilibrated with PBS and dialysis for 24 h (4°C) against PBS and NAN3. Using the absorbance at 280, 515, and 555 nm 05) the molar ratio of dye/protein was estimated to be ca. 2 and 5 for two different preparations. Subsequent gel filtration produced no change in these numbers, indicating that noncovalently bound dye was not bleeding off the conjugate. Texas Red was coupled with either rat or mouse IgE by mixing an equal volume of borate buffered saline (pH 8.0) containing 4-5 mg/ml of IgE with 0.25 M Na carbonate buffer, pH 9, and then adding powdered Texas Red (0.25 mg/mg IgE). The suspension of dye was inverted periodically during a l-h (rat) or 2-h (mouse) incubation on ice. Free dye was removed from the blue conjugate on a P6 column as above. From absorbance at 280 and 596 rim, the estimated dye/protein ratios were 2.5 for rat and 5.5 for mouse IgE (16). Before labeling cells, we routinely centrifuged the lgE for 30-60 min at 100,000 g (4°C) in a Beckman alrfuge (Beckman Instruments, Inc., Fullerton, | 3,068,800 | 7602999 | 0 | 16 |
CA), and then used only the upper 60-70% of the supemate. In some experiments we used a Texas Red-rat IgE (TRr-lgE) conjugate which had been purified ( Fig. 1) by chromatography on sequential columns of AcA 34 and AcA 22 (LKB Produkter, Bromma, Sweden) (17). The pooled monomer fractions were used directly (0.6 mg IgE/ml), but even after reconcentration to 4 mg/ml, analysis by nondenaturing PAGE (4-25%) showed no evidence of aggregated IgE. Since the D values obtained were equal to those for less rigorously purified TRr-IgE, we assume that aggregation of the latter was not a problem. Neither TRITC-nor Texas Red-mouse IgE was subjected to PAGE, but PAGE of the underivatized mouse IgE did reveal multimeric components. Because the mouse IgE was not purified by gel filtration after its reaction with TRITC or Texas Cell Labeling: For determination ofthe D value oflgE-bound receptors with pre-field labeling, RBL cells suspended in Tyrode's buffer (1-2 x 10~/ml) were incubated with 10-15 ug/ml of fluorescent IgE for 30 min on ice. Cells were diluted to 25 ml with either Tyrode's or CBA, centrifuged as above, and resuspended (1-2 x l&/ml) in the same buffer. As determined by spot photometry, binding of fluorescent rat IgE was inhibited ->96% by prior incubation in unlabeled rat or mouse IgE (100 ug/ml) for 15-30 min at 0*C. Electromigration chambers were filled with cell suspension (60 ul) and left for 10 min at room temperature to permit cells to adhere to the substrate before application of the field. In post-field labeling, the cells | 3,068,801 | 7602999 | 0 | 16 |
(1-2 x 106/ml) in CBA were added to chambers and allowed 10 min at room temperature for adherence. After electromigration (and back diffusion in the case of lgE-unoccupied receptors), slides were immediately transferred to a metal block (0*C) and three aliquots (50 ul each) of IgE (10-15 ug/ml) in CBA were passed through the chamber. After 15 min at 0*C the free IgE was rinsed away with three to five aliquots (50 ul each) of CBA; for back diffusion of bound receptors, slides were brought back to 24"C for various time periods. The asymmetric receptor distribution was then quenched by a 30-60-s fixation in cold (0*C) acetone, followed by replacement with CBA. By measuring the fluorescence intensity in a 4-#mdiam spot over the edge of cells (see below) before and after acetone fixation, the loss of lgE during fixation was apparently _<3%. By measuring the whole cell fluorescence with a larger photometer aperture the loss was apparently _<5% (n = 80). Apparatus a n d Procedure for field Application: Equipment was as previously described (5), with the following modifications. The electric field was applied to a cell chamber made from a glass microscope slide. Slides were washed in concentrated HNOa-H2SO4 (1:1, vol/vol) for 30-60 min and rinsed with distilled H20 before use to minimize fluorescence background due to adsorption of IgE conjugates; without this step the Texas Red conjugates gave uniquely high background fluorescence. Two parallel strips of doublecoated adhesive transfer tape (Y9469, 3M, St. Paul, MN) were used as the sidewalls (spacers) of the cell | 3,068,802 | 7602999 | 0 | 16 |
chamber, and a 24 x 40-mm coverglass served as the roof. This tape gave an average chamber thickness of 0.12 ram. Buffer wells at both chamber ends were dammed with a rim of rubber cement. To maintain isothermal conditions during post-field back diffusion and to compensate for heat production during field application, the cell chamber was placed on a constant temperature aluminum block. The block was machined to hold three slides, drilled for coolant circulation, and linked to a recirculating bath set to 24°C. With a maximum current applied to the chamber, namely, 6 mA (a field of 40 V/cm when using CBA; resistivity = 80 fl cm at 27"C), the slide temperature stabilized at 27"C. Post-field labeling of Fc~ receptors was performed on another metal block kept at exactly 0*C with a separate cooler. A small increase in this temperature (2-4"C) led to irreproducible and generally smaller starting asymmetries, possibly because of an abrupt increase in diffusion rates somewhere between 0-4"C. Tops of the slides were insulated by laying a styrofoam wafer over each. Using a tele-thermometer (YSI model 42SL) with a small-diameter probe (model 427, Yellow Springs Instrument Co., Yellow Springs, OH), we found that it took 45 s for the temperature of a slide to drop from 27* to 0*C, and an equal time for the reverse process. Only a few seconds were required to come within 75% of the steady value. Data Acquisition and Analysis: Receptor migration was measured with a microfluorimetric method as described before (5), except that a 4-umdiam aperture | 3,068,803 | 7602999 | 0 | 16 |
was used to collect photons from anode-and cathode-facing poles of the cell. An asymmetry index, A,(t) was calculated for each cell: where I_ and L are the fluorescence intensities at cathodal and anodal edges, after subtraction of the nearby cell-free background. The latter was determined separately for each cell. Photometer readings over the edge of unstained cells were only 4 -+ 2% above background on slides not exposed to IgE; cell "autofluorescence" was therefore neglected in both the pre-and post-field labeling experiments. For stained cells with no asymmetry, typical signal to noise ratios fell in the range of 6-8. A~ was determined for 40 cells on each slide (10 cells in each of four parallel and widely spaced scans), and an average A~ was calculated. Roughly half of the slides were scored single blind, with two of us exchanging experimental manipulations and A~ measurements. Because of photobleaching during the measurement of A~ on single cells, the true A~ is either smaller or larger than the measured A,, depending on which side of the cell is recorded first. To minimize bias we usually alternated the initial sampling side on sequential cells (population study). We did not try to analyze the contribution of any fluores-780 THE JOURNAL OF CELL BIOLOGY . VOLUME 99, 1984 cence quenching which may have occurred upon concentration of the fluorescent ligands on one side of the cell. A linear least squares program was used to fit the slope of the decay of In[AM)], for both the single cell and the population experiments. To | 3,068,804 | 7602999 | 0 | 16 |
the first order, on a single spherical cell, this slope equals -2D/r 2, where r is the cell radius (5). For a D value of 3.8 x 10 -~° cm2/s (the empirical value) the first-, second-, and third-order curves become superposable within 6 rain. Before 6' the maximum difference between first-and third-order slopes is 8%, the firstorder plot being steeper. In the population studies it is incorrect to equate, as we did, the decay of ln[A(t)] with -2Dt/~ 2, where D and t are population averages. The correct average is ln[A(t)]. However, presently available theory does not encompass the possibility that time zero asymmetries may be negative or zero, and decay to more negative values with time. Yet for the conditions of our population experiments, even the time zero distribution of A~ includes negative asymmetries, some of which are probably artifact (see Results) but many of which are real. Short of deriving more general equations, one compromises by forming the linear average and transforming it. This leads to an underestimate of the true average D value; applying both methods to the three cells in Fig. 4B (q.v.), the error is 33%. This is likely an upper limit to such error in the population studies, since ln(~,) and ln(x~) approach each other as the spacing between x~ decreases. Average cell diameters were measured with a reticle calibrated with a slide micrometer. RBL cells in CBA had an average diameter of 12.1 um (SD = 1.3 um; n = 330). This is close to the average value (12.3 | 3,068,805 | 7602999 | 0 | 16 |
~m) estimated from Coulter counter measurements of suspended RBL cells (18), indicating that not much flattening of the cells occurred in CBA, and that the assumption of spherical geometry is not bad. After acetone fixation and rehydration with CBA, the diameter had shrunk to 10.7 um (SD = 1.2 urn; n = 120). We made no attempt to correct for any difference between the measured A~ and the true A~ which might result from this shrinkage. Electron Microscopy and Photography: Cells for thin sectioning were processed as follows: after elution or treatment with enzymes cells were rinsed three times in 0.1 M Na cacodylate buffer, pH 7.2, and then incubated 1 h in the following buffer: 0.2 M cacodylate, pH 7.3, 0.5% ruthenium red, and 1.3% glutaraldehyde. After washing in cacodylate buffer the cells were incubated 3 h in a solution containing 0.2 M cacodylate, pH 7.3, 0.5% ruthenium red, and 1.3% osmium tetroxide. After dehydration the cells were embedded in Epon-Araldite, sectioned, and examined at 60 kV in a JEOL JEM 100C electron microscope (JEOL USA, Electron Optics Div., Peabody, MA). The sequence in Fig. 2 was taken with Kodak Recording film (No. 2475, Eastman Kodak Co., Rochester, NY) through a x40 oil immersion objective on a Zeiss epifluorescence microscope (Carl Zeiss, Inc., New York). Exposure times were 15, 25, 50, and 180 s for Fig. 2, a-d, respectively. The film was push processed by development in DK-50 for 8 min and printed on Ilford 3.1M paper. Electromigration of Vacant and IgE-occupied Fc Receptors In | 3,068,806 | 7602999 | 0 | 16 |
the first set of experiments, RBL cells were labeled with TRr-IgE and then plated in the cell chambers. Application of an electric field caused redistribution of the labeled Fc~ receptors on the cell surface from a uniform to a highly asymmetric distribution. Fig. 2, a-d depicts TRr-IgE-labeled RBL cells before and after application of an electric field of 20 V/cm for l0 min. Accumulation of the surface-bound IgE towards the cathode is obvious in Fig. 2b. This experiment was conducted with 30 mM NaNa and without glucose in the incubation medium, a condition known to block certain metabolic energy-dependent processes such as capping (19). On the majority of cells, essentially all of the fluorescence emanates from IgE exposed to the extracellular medium, because it is effectively quenched by treatment with dilute copper sulfate (Fig. 2, g and h). The paramagnetic Cu(II) ion is an effective quencher of fluorescence from rhodamine, and it enters living cells slowly enough to allow its use in distinguishing external from internal fluorophore (20). This is shown in Fig. 2, where e depicts a Swiss mouse 3T3 fibroblast containing rhodamine-labeled a2-macroglobulin (21 ), andfis 14 M NaCI, pH 4.5. Little quenching is detectable, and what is present is mostly due to photobleaching during first exposure; (g) four RBL cells in field of 20 V/cm. With field on, slide was scanned to find a "ringstained" cell showing only weak asymmetry. Such cells were of low frequency (<2%) with rat IgE (in a 20 V/cm field). (h) Same cells after treatment with CuSO4 as | 3,068,807 | 7602999 | 0 | 16 |
above. Note that only externally exposed IgE is redistributed by the field. Bar, 10/~m. x 723. the same cell after addition of 10 mM CuSO4 in 0.14 M NaC1. Although we plan to quantify field-induced internalization using radiolabeled IgE, the above qualitative observations are consistent with passive lateral migration of surface receptors under the influence of an electric field. When labeling of the surface receptors for IgE was carried out after field application (post-field labeling), an asymmetric distribution of Fc~ receptors was also observed, indicating that unoccupied receptors also migrate under the influence of the field. The graph in Fig. 3 shows the time evolution of asymmetry in the surface distribution of IgE receptors in pre-and post-field labeled cells, as determined microfluorimetrically (see Materials and Methods). Although both populations were exposed to a l0 V/cm field, the steady-state asymmetry of TRr-IgE-complexed receptors was much higher than that of the uncomplexed receptors (Fig. 3, a and b). Using a theoret-ical expression for m, the effective electrokinetic mobility (22), we estimate that at 10 V/cm the value of m for the complex is about four times greater than it is for the free receptor. 2 The simplest explanation for this observation is that bound IgE increases the electrokinetic mobility of its receptor. The difference was not due to back diffusion of free receptors 2 r is the cell radius, D the lateral diffusion coefficient, Eo the electric field strength, and As the steady-state asymmetry index. This is an approximate calculation, complicated by uncertainty in the true As for | 3,068,808 | 7602999 | 0 | 16 |
Eo > 10 V/cm; at these field strengths di begins to decrease after 15-20 min in the field, especially with post-field staining. during the post-field staining period: as shown in Fig. 5A (solid triangles) (q.v.), no appreciable back diffusion occurred if the cells were kept at 0*C. In theory, for equal m values, one would also expect a higher steady-state Ai for the IgEreceptor complex if its diffusion rate were much lower than that of the unoccupied receptor (Eq. 2). However, we know from independent measurements of D that this is not the case (see below). One referee suggested that long preincubation in the cold may have caused microtubule depolymerization and released constraints on the receptor; however, we observed marked enhancement by TRr-IgE even when cells were labeled at room temperature and never placed on ice (Ai[20'] = 0.65 at 10 V/cm). This referee also wondered whether noncovalently bound dye in the IgE preparation may have partitioned into the membrane during the preincubation, and in some way given an artificially high A~. This is clearly not so, because when cells were labeled at room temperature with nonfluorescent rat IgE, exposed to a l0 V/cm field for 20 min, quenched at 0*C, and then labeled with a mouse antibody against rat IgE (23) followed by fluorescein isothiocyanatelabeled rabbit anti-mouse IgG, the asymmetry was insignificantly different (P < 0.005) from that shown in the graph in Fig. 3. 782 THE JOURNAL Or CEtI_ BIOLOGY -VOLUME 99, 1984 FIGURE 3 IgE binding increases the electrokinetic mobility of Fce receptors. | 3,068,809 | 7602999 | 0 | 16 |
Ceils in micrographs were exposed for 20 min to a steady field of 10 V/cm at 24°C, quickly cooled to 0°C, and fixed with acetone. (a) TRr-lgE bound after field application (IgE was bound in the cold before fixation); (b) TRr-lgE bound before field application; (c) TRr-lgE bound before field but after treatment of cells for 30 min with neuraminidase (65 U/ml). Post-field incubation at 0°C was for 15 min, irrespective of whether IgE was bound before or during Rat IgE162 has an isoelectric point near 5.9 (24), and is thus negatively charged at the pH (7.2) of these measurements. Why then does it increase the tendency of its receptor to move toward the cathode, as if it were a cation? The most logical explanation that we can offer at this time is that drag on the receptor due to electro-osmotic flow along the membrane is the dominant force impelling it towards the cathode (25). The 185,000 Mr ligand may increase substantially the molecular area exposed to this flow. The marked diminution of free and bound rn values by prior removal of cell surface neuraminic acids (Fig. 3, c and graph) is consistent with this interpretation; i.e., the resulting lower charge density should reduce the electro-osmotic flow. Diffusive Recovery SINGLE CELLS (PRE-FIELD LABELING): Fig. 2, c and d show two stages in the diffusive recovery of TRr-IgE occupied Fc~ receptors on seven RBL cells after termination of the electric field. Asymmetry plots for two cells exposed to the same field are shown in Fig. 4, A | 3,068,810 | 7602999 | 0 | 16 |
and B. Although "ring staining" is often used as evidence against the internalization of externally applied fluorescent probes, as shown in Fig. 2, g and h, by itself it can be a rather poor criterion. The cell in the upper left of g has a crisp fluorescent ring, yet upon addition of a quenching agent (h), the ring remains. However, CuSO4 (10 mM) did quench the fluorescence of TRr-IgE on almost all cells during the period of back diffusion (Fig. 2, g and h). Moreover, the fluorescence of those few cells (<2%) showing weak asymmetry was only partially quenched by CuSO4, and the nonquenchable fraction (presumably internalized) was not formed into a crescent (Fig. 2 h). As shown in Fig. 4B, relaxation of asymmetry on single cells fits fairly well the exponential decay expected from the solution to the equation for diffusion on a spherical surface (5,22). Using TRr-IgE, recovery has been followed in the presence and absence of 30 mM NaN3 and 0.1% glucose without noticeably different outcomes. Together, these results are consistent with passive back-diffusion of IgE-FcE receptor complexes in the plane of the membrane. The electric field induced some cells to form one to three large vacuoles, both during field application and shortly thereafter. The vacuoles were fluorescent, and sometimes close enough to one edge or the other of the cell to contribute to the total fluorescence intensity measured there, thus yielding an artificially high or low Ai for that cell. This quantitative effect was observed directly during single cell recording as a | 3,068,811 | 7602999 | 0 | 16 |
large and sudden jump in the steady-state Ai at the time of vacuole formation. Although cells with obvious vacuoles were studiously avoided when scoring slides for the population studies, because of the inconvenience of changing focal planes on every cell, some were inadvertently included. Despite their usually low frequency, because of the large magnitude of (Ai -Ai), they probably contributed measurably to the standard deviation. Inexplicably, on a few slides vacuoles were very frequent. Vacuole induction may explain some of the unexpectedly large, negative asymmetries observed at short time points. From a series of 19 plots like those shown in Fig. 4 b we found an average D value of 4.6 x 10 -l° cm2/s (25"C) for TRr-IgE bound to Fc, receptors (range:l.5 x l0 -z° to 1.1 x l 0 -9 cm2/s). The average D value for bound mouse IgE (Texas Red-or TRITC-labeled) was 1.8 x l0 -~° cm2/s (25"C) for the 37 cells measured (range: 5 x 10 -u to 6 x l0 -l° cm2/s). However, the average coefficient of variation for the mouse IgE curves (0.56) was much lower than that for rat IgE (0.92), the "long-time" tails being much flatter. Furthermore, even without field application, after 1-2 h at room temperature the mouse IgE was much more heavily internalized than the TRr-IgE, as judged by its resistance to quenching by Cu(II) and its heterogeneous distribution. Without further work we cannot say whether the apparent difference between D values of murine and rat IgE is real or an artifact due to aggregation of | 3,068,812 | 7602999 | 0 | 16 |
the mouse IgE. This difference also showed up with the post-field labeling method, although the absolute D values estimated with the separate techniques were different. P O P U L A T I O N S T U D Y ( P O S T -F I E L D L A B E L I N G ) : Bylabeling the Fce receptors either immediately after field termination or after defined periods of PFR, we obtained diffusional recovery curves for IgE-bound and IgE-free receptors, respectively, as shown in Fig. 5. The initial asymmetry was created by application of a 40 V/cm field for 30 min at 27"C. For TRr-IgE the apparent D values of free and bound receptors were 4.0 and 3.1 x 10 -~° cm2/s. Identical experiments using mouse IgE yielded the plots shown in Fig. 5 B. The data is a combination of points for TRITC-and Texas Red-labeled mouse IgE, mostly the former; when used on separate days there was little difference between the D's obtained with the two fluorophores. The average D's for free and mouse IgEbound receptors were 4.5 and 2.3 x 10 -l° c m 2 / s (Table I). A Student's t test on all four slopes indicates they are truly different (P < 0.05 that they are not); but inasmuch as the distribution of Ai values on individual cells looks skewed, the validity of this t statistic is uncertain. Note that the mouse IgE not only yielded slower bound and faster free receptor D values, it also detected | 3,068,813 | 7602999 | 0 | 16 |
a lower initial asymmetry (0.31 vs. 0.44). N O N E F F E C T OF CELL COAT D E P L E T I O N : The native Fce receptor of RBL cells is trypsin resistant (26). On the other hand, treatment of several cell types with trypsin will release a large fraction of their cell surface proteoglycans (27,28). Recovery curves for bound receptor before and after treatment of RBL cells with 0.5% trypsin (in PBS-EDTA for 30 min at 37"C) were obtained by successively measuring different cells (every 30 s) on one slide during the period of back diffusion. Data from two to three slides were then averaged, the logarithms taken and fit to a straight line. D values for bound receptor before and after trypsinization were not significantly different (P < 0.05) (Table II) Representative electron micrographs of ruthenium redstained ceils that had or had not been trypsin-treated (as above) before staining are shown in Fig. 6. Ruthenium red is a hexavalent cationic complex routinely used for visualization of acidic mucopolysaccharides like those present in the glycocalyx (27). It reportedly also binds to sialoglycoproteins, provided their negative charge density is high enough (27). To be sure, substantial variation in the glycocalyx thickness was observed in control cells, and somewhat less in trypsintreated RBL cells, but the gestalt impression is faithfully depicted in Fig. 6. One sees a darker and ruthenium red on the control cells. The above is little more than suggestive evidence that the glycocalyx of RBL cells is not | 3,068,814 | 7602999 | 0 | 16 |
a major constraint to lateral diffusion of the Fc~ receptor, It is entirely possible that remnants of the glycocalyx that are left after trypsin retard diffusion of the receptor so much that no further encumbrance results from the trypsin-releasable component(s). DISCUSSION Our primary finding was that lgE does not cause a precipitous decline in the lateral diffusion rate of FCE receptors on RBL cells. This result is directly relevant to numerous studies of protein mobility that have employed high-molecular-weight fluorescent reagents, and to the possibility that these reagents hinder diffusion of the species being studied. The implicit assumption has usually been that diffusion of ligand-receptor complexes accurately reflects that of unligated receptors. Our work validates this assumption for Fc~ receptors on RBL cells. Whether the assumption is good for other receptors or other cells remains to be determined. A priori, there is some reason to expect that the glycocalyx of RBL cells should limit diffusion of IgE-bound Fc receptors, because in electron micrographs (Fig. 6) it often appears to extend well beyond the top-most projection of cell bound IgE (29). Wank et al. (30) also suggest that steric constraints imposed by glycocalyx components may account for the rate constant 30-fold slower for association of IgE with cell-bound Fc~ receptors than with Triton X-100-solubilized Fc~ receptors. We see no obvious reason why the glycocalyx could not limit the reaction rate without affecting lateral diffusion of the receptor, but the existence of either constraint is not yet proven. One would fike to extend the trypsin and neuraminidase results, | 3,068,815 | 7602999 | 0 | 16 |
and define more precisely what the contributions of different glycocalyx components are, both to the diffusional and the electrokinetic mobilities. Because of several complications, which arose in our application of PFR to RBL cells (see Materials and Methods and Results), we emphasize the comparison between free and bound Fc~ receptors; the values of the separate D's are no doubt inexact, though perhaps no more so than those derived from FRAP. 3 One complication is the limited field-induced internalization that we observed--even under conditions known to block internalization of other ligands on other cell types. This may be behind the previously noted biphasic time dependence of Ai at higher field strengths. Tank (8) found much more extensive field-induced internalization of lowdensity lipoproteins by human fibroblasts. In our case internalization was sometimes evident from the emergence of a fluorescent membrane delimiting one to three large intracellular "vacuoles." Very similar vacuoles also appear in rat peritoneal mast cells exposed to steady electric fields, and resemble the intraceUular cavities seen during mast cell degranulation (33,34). The considerable variation in Ai that one found among different cells on the same slide was a bit exasperating. We traced some sources of this variation to the following. First, human error in placement of the photometer aperture is appreciable with current methodology. Second, there is substantial variability in the A~ values between cells before field application, due not only to the above, but probably also to morphological distinctions, e.g., a higher concentration of surface projections on one side of the cell than the other. | 3,068,816 | 7602999 | 0 | 16 |
Third, cells were infrequently rotated during or after the acetone fixation so that their crescents became randomly oriented. Fourth, the diffusion coefficients differ substantially from cell to cell (Fig. 4B), and possibly the electrokinetic mobilities differ as well. A wide variation in single-ceU D values has been observed in previous FRAP studies of other membrane proteins (35,36). Whether the apparent difference between the D's of mouse and rat IgE is due to a species difference per se remains to be seen. From the relative degrees of cellular internalization and analytical PAGE we surmise that our fluorescent mouse IgE may have been more highly aggregated than the fluorescent rat IgE. Cross-linkage oflgE bound to Fc~ receptors is a signal for receptor internalization as well as cellular degranulation (37,38). Recent evidence also indicates that as with surface Ig on B lymphocytes (39) and C3b receptors on polymorphonuclear leukocytes (40), cross-linkage oflgE receptors on RBL cells renders them insoluble in nonionic detergents, probably due to association with the insoluble cytoskeletal matrix (41). It is intriguing to think that the slower D value we found for mouse IgE was due to signaling of cytoplasmic attachments 3 It is noteworthy that both FRAP and PFR give about the same D value for IgE-bound FCE receptors. Along with other evidence p~'esented here, this belies a suggestion (31,32) that the divergent FRAP and PFR results obtained with some other cell types differ because PFR and FRAP measure fundamentally different processes. The enhanced cathodal electromigration of Fc~ receptors upon binding of IgE was a | 3,068,817 | 7602999 | 0 | 16 |
surprise. Equally tenable explanations may exist, but our current hypothesis is that the occupied receptor has an increased exposure to the electro-osmotic flow. This might result from a less streamlined shape of the IgE-Fc~ receptor complex, from an increased cross-sectional area without shape changes, or from vertical extension into a plane where the electro-osmotic flow is greater than it is at the height of vacant receptors. There is an interesting contrast between the action of IgE and that of Con A, binding of which completely prevents electromigration of Con A receptors in Xenopus muscle cells (22). Unoccupied Fc~ receptors in RBL cells have an apparent electrokinetic mobility (at 10 V/cm) about 1/27 that of unoccupied Con A receptors in Xenopus muscle cells (5), whereas the ratio of PFR-measured diffusion coefficients is about 1/13. After binding of IgE, m is still lower by a factor of about 7, and D by a factor of about 17. It is tempting to try to relate these differences in m to what is known regarding the molecular structures of the receptors and cell surfaces, but there are too many unknowns at present. Future studies of electromigration using reconstituted lipid-protein systems may help isolate the parameters most important in determining m. | 3,068,818 | 7602999 | 0 | 16 |
Intra-particle plasmonic coupling of tip and cavity resonance modes in metallic apertured nanocavities Based on numerical studies of apertured metallic nanocavity structures, we describe a new intra-particle plasmonic interaction pathway that couples the plasmon resonance modes of the aperture edge and the cavity. In contrast to the inter-particle coupling schemes that require precisely arrayed nanoparticles, this intra-particle coupling scheme achieves the tunability in plasmonic resonance wavelength using a single standalone nanostructure. In addition, when the aperture edge is made sharp, it functions dually as a tip that amplifies its near-field producing the local filed enhancement effect. We investigate the details of the coupling mechanism and identify the dominant role of the tip mode in determining the coupling efficiency numerically. The numerical model results in good agreement with recent experimental results. This intra-particle coupling mechanism will help the monolithic integration of plasmonic functionalities and its application for the nanoscale spectroscopy of biological structures in vivo. ©2005 Optical Society of America OCIS Code: (240.6680) Surface plasmons; (000.4430) Numerical approximation and analysis Introduction The local field enhancement (LFE) associated with the plasmon resonance in metallic nanostructures has attracted intense research interest for its role in a number of useful nanophotonic phenomena such as surface-enhanced Raman scattering (SERS).The improvement of the LFE factor has been the focus of research.Especially for biological SERS applications, the ability to tune the plasmon resonance wavelength toward the "biological window" (700~1100 nm) in near-infrared (NIR) regime has been emphasized due to the low optical absorption and scattering within the spectral regime.The intrinsic resonance wavelengths of | 3,068,819 | 29598846 | 0 | 16 |
plasmonically active noble metals, however, mostly fall in UV-visible regimes.The inter-particle plasmonic coupling scheme, while being successful in lowering the resonance energy precisely as designed [1], requires the structure to be fixed on a substrate.For many in vivo and in vitro applications, mobile, standalone structures are needed. In order to modify the resonance characteristics of standalone nanostructures, the intraparticle plasmonic coupling scheme has often been adopted.In the scheme, multiple plasmon resonance modes hosted within a single nanostructure are coupled to generate a new mode.The nanostructure must be configured to host multiple, localized surface plasmon-polariton resonances (SPPR) modes within interaction range.The resonance modes residing on the vertices of polygonal metallic nanostructures are good examples [2].In curved nanostructures such as nanoshells [3] or nanorings [4], the coupling between the spherical or cylindrical SPPRs is utilized. In this Letter, we investigate the intra-particle plasmonic coupling phenomenon of metallic apertured nanocavities (ANCs) with an emphasis on LFE and resonance wavelength tuning.The ANC structure is schematically shown in Fig. 1(a).Such a geometry is already demonstrated based on masked deposition technique [5].3-D ANCs exhibiting geometries of the 2-D ANC axiosymmetrically are also demonstrated [6][7][8][9].Owing to its complex geometry, the ANC hosts a wider variety of SPPR resonance modes than its concentric counterpart and hence induces different intra-particle plasmonic coupling effects.While a number of ANCs have been experimented [5][6][7][8][9] and studied theoretically [9] and numerically [7,8,10], an analysis in terms of plasmonic coupling has rarely been reported.Our numerical studies focus on sharp-edged nanocavities with radii ranging from 1/4~1/3 of the excitation wavelength which | 3,068,820 | 29598846 | 0 | 16 |
facilitates intra-particle coupling.The sharp edge shown in Fig. 1(a) is naturally formed in deposition-based fabrication processes.We reveal that in such ANCs, the intra-particle coupling between the edge and nanocavity SPPR modes takes place and a hybrid mode with modified resonance characteristics is formed.Due to the sharpness, the edge plays a dominant role in the coupling process and causes high LFE.The edge modes are also sensitive to the shape and environment of the edge and the wavelength and polarization of the excitation.From an electromagnetic point of view, the ANC is an amplifying antenna feeding into a resonator. Simulation Model To investigate the LFE in ANCs, we compute their local fields by solving the Maxwell's equations.We adopt the finite element method (FEM) mainly for its compatibility with curved geometry and adaptive meshing capability.The latter becomes critical when dealing with a structure consisting of features differing by several orders in dimensions, i.e., the apex radius of the tip (0.5~4 nm) and the whole ANC (~400 nm).We use the complex permittivity of bulk gold [11].All simulations are transverse magnetic and the retardation is fully considered.The ambience is set as vacuum in accordance with the experimental setup. Supported 2-D Plasmonic Modes In 2-D, the localized SPPR modes supported by the ANC structure becomes more identifiable, as illustrated in Fig. 1(a).They reside at the tips, on the inner surface of the cavity, and the outer-surface.In addition, the surfaces [3,4,12] and the paired tips [13] are known to form coupled modes.Brief surveys of individual modes will be informative for the discussion of their | 3,068,821 | 29598846 | 0 | 16 |
interactions.Figure 1(b) shows the relation between the cylindrical cavity radius R and the excitation wavelength obtained from the characteristic equation where J l and H l are Bessel and Hankel functions, , and the prime denotes the differentiation with respect to its argument.ε i,o are the dielectric constants of the cavity and its surroundings [14].All parameters are consistent with the simulation.It is clear that with R = 150 nm, the nanocavity supports only the lowest TM (l=1) mode within the bandwidth of interest.As indicated in Fig. 1(b), the TM (l=1) mode attracts and repels charges to and from the cavity diametric points and its vicinities.Since there is no closed form description for the tip near-fields, we characterize those using numerically obtained field patterns shown in Fig.(a) 1(c).Beginning from the apex, the near-field repeatedly converges and diverges vertically along the tip and exhibits multiple nodes N n .As the excitation wavelength increases, the nodes move away from the tip apex and the convergence angle varies.Note that the amplification of the local fields also occurs near the tip apex due to the lightning rod effect (LRE).The interaction between two nanoscale tips becomes stronger at shorter gap width and/or longer excitation wavelength and is able to generate huge LFE [13].The near-field of the tip-to-tip interaction mode resembles the field pattern of an electric dipole [15]. Considering the dimensions of the present ANC, we exclude the outer-surface SPPR and its interaction with the cavity mode from further considerations. Plasmonic Coupling Analysis Based on the survey, we hypothesize that a strong | 3,068,822 | 29598846 | 0 | 16 |
intra-particle plasmonic between the tip and the cavity SPPR modes takes place within the ANC structures and that a strong LFE at a modified wavelength originates from it.The proximity of the tip and the cavity warrants the existence of such coupling effects.The phase retardation-induced SPPR coupling between two closely positioned nanostructures and the resultant LFE have already been predicted numerically [16].Such LFE can be attributed to the near-field interaction that increases the surface charge, and consequentially the field density, of the SPPR.In the case of sharp-edged ANCs, an additional improvement in LFE occurs owing to the LRE-induced amplifier action of the tip.Rigorous analysis of the coupling between the two SPPR modes is not straightforward.Due to the dissimilar and asymmetric geometries of tip-cavity, the elegant analytical approaches of Ref. 3 are not applicable.The tip-cavity coupling also depends on the phase retardation, which makes analytical approaches more complicated.Since the coupling occurs through the overlapping of the mode near-fields, we investigate the coupling mechanism and its optimum condition by closely examining the electric field patterns at wavelengths of LFE maxima and minima.The FEM simulation results for identifying the LFE maxima and minima are shown in Figs.2(a)-(c).The LFE factor is defined as the amplitude ratio between the excitation wave and the maximum value of the ANC near-field which occurs at the tip apex for all cases in our study.To emphasize the role of the sharp tip and its SPPR mode, the tip apex radius is set to 0.5 nm.The ANCs with tips of practical thickness will be discussed in the | 3,068,823 | 29598846 | 0 | 16 |
following section.The computation domain has dimensions of 1.6×3 µm and employs low-reflection boundary conditions.The width is made longer along the direction of wave propagation to ensure the generation of quasi-plane waves.The mesh consists of 46,588 adaptive triangular elements.Near the tip apices, the size of elements is kept under 0.1 nm.The sensitivity of the LFE spectra on the incidence angle of the excitation wave is evident from Fig. 2(b) and (c).This incidence angle dependence, however, is far lower than the extreme polarization sensitivity of LRE from an isolated tip. Figure 2(d) shows simulation results of two additional structures; a spoiled cavity with sharp tips that emphasizes mainly the tip-induced effects and a dull-tip ANC that mainly emphasizes the cavity-based effects.The fact that the computed LFE factor of a full ANC, ~10 3 at maximum, far exceeds those of the partial structures corroborates that the LFE in an ANC depends on the plasmonic coupling mechanism. Figures 3(a)-(c) show the field patterns within the ANC of Fig. 2(a) at wavelengths of the LFE peaks and valley.By examining the superimposed surface plots of LFE factors, we can examine the plasmonic coupling conditions for higher LFE.At LFE peaks, the TM (l=1) mode supported by the cavity is at least partially excited along with the tip mode.At the LFE valley, however, the field pattern within the cavity takes similarity to TM (l=2) mode which is not supported by the cavity.The inefficient excitation of a cavity mode leads to a weakened coupling between tip-cavity and lower LFE.Among the LFE peaks, the P 3 | 3,068,824 | 29598846 | 0 | 16 |
that exhibits minimally perturbed cavity mode results in higher LFE.The field patterns in the dotted boxes also indicate that it is the near-field of the tip mode at the nodes N n that determines the excitation efficiency of a certain cavity mode.This selective excitation is illustrated in Figs. 3(d)-(f). At P 1 , the field pattern at N 2 matches that of the TM (l=1) cavity mode at the position approximately and the mode is partially excited.The robustness of the cavity SPPR against the perturbation due to an aperture makes this partial excitation possible.Coyle et al reported that cavity SPPR modes persist until a spherical cavity becomes a hemispherical shell [9].At P 3 , the field lines are matched at N 1 which is closer to the tip apex than N 2 .This results in an almost full excitation of a TM (l=1), as evidenced by ~10 dB LFE over the whole cavity.In contrast, the field pattern of the tip mode at V does not match the TM (l=1) field pattern at any node and consequentially a cavity mode is not excited.This dominance of the tip mode can be attributed to its intensity much higher than that of cavity modes.The role of the tip SPPR makes the aperture width the most important geometric parameter in tuning the LFE peak wavelength.According to our model, a blue-shift of the LFE peak is expected with increasing aperture width since the shortened tip length enables the field patterns match near N 1 at shorter wavelengths.The LFE spectra as a function | 3,068,825 | 29598846 | 0 | 16 |
of the aperture width are plotted in Fig. 4 and show a good agreement with the prediction.The increase in LFE with increasing aperture width (shorter resonance wavelength) can be attributed to the influence of the intrinsic plasmon resonance which occurs near 500 nm. Discussion This model successfully explains the LFE characteristic between 600 and 900 nm.The increase in excitation wavelength required to match the tip and cavity field patterns is the origin of the LFE red-shifted into the NIR regime in ANCs.For < 600 nm, multipolar SPPR becomes dominant and the role of the tip mode diminishes.For > 900 nm, the tip-to-tip interaction mode becomes dominant and the cavity mode cease to be excited.The study of P 2 and P 4 due to 90° incidence angle is under way.We are aware that LFE from similar nanoscale apertured cavities have been reported, either directly or indirectly.Lopez-Rios et al reported strong field intensity near the upper corners of deep grooves of metallic gratings [17]. It is also numerically demonstrated that 2-D ANCs with multi-valued cross-section exhibit LFE near the aperture [10].While the LFE effects are well described, neither their intraparticle origin nor the role of the geometry has been investigated explicitly. We set the tip thickness to a practical value of 4 nm in accordance with the TEM image in the inset.We notice that as the sharpness of the tip is decreased, the tip-to-tip interaction mode begins overshadow the SPPR of the isolated tips at wavelengths shorter than the sharptip ANC case.The LFE peak near 700 nm is | 3,068,826 | 29598846 | 0 | 16 |
caused by the coupling between the tip-to-tip and nanocavity modes.The fact that the dipole-like near-field of the tip-to-tip mode does not overlap well with the cavity mode, in addition to the diminished LRE, explains the LFE factor lower than that of the sharp-edged ANCs.The simulated LFE is ~17 dB near 700 nm which is lower than experimental observation by ~7.7 dB [6].This discrepancy indicates the possible existence of a sharper edge and/or the involvement of additional Raman enhancement mechanisms such as the chemisorption effect.The ~600 nm peak is still from the tipnanocavity coupling. Conclusion In conclusion, we have investigated a novel intra-particle plasmonic coupling mechanism in metallic nanocavities with sharp-edged apertures.In such structures, our numerical simulations reveal that the sharp edge functions as an antenna that feeds the excitation wave into the and also as an amplifier that enhances the local filed utilizing its large curvature.The preliminary 3-D simulation results indicate that it is possible to adopt the 2-D approximation in which the aperture edge-nanocavity interaction is transformed into the coupling between two well-known SPPR modes: the tip mode and the cylindrical cavity mode.The interaction takes place through their near-fields and the requirement for matching the near-field patterns at the modal interface governs the wavelength and efficiency of the coupled mode.By examining the simulated electric field patterns, we show that the intense tip near-field plays a dominant role in this matching process.Accordingly, our simulations with varying geometric parameters reveal that changing the proximity of the tip to the cavity strongly affects the resonance wavelength of the | 3,068,827 | 29598846 | 0 | 16 |
coupled mode.It is important to note that the tip near-field under the tip-cavity coupled resonance exhibits higher LFE than the maximum achievable with paired tips without cavity support.This additional enhancement may be attributed to the synchronized transport of the free electrons due to the retarded modal interactions [16] but needs to be studied in more detail as a future work.This intra-particle coupling study provides criteria useful for the analysis and synthesis of monolithically integrated plasmonic functionalities. Fig. 1 .TM, l= 2 TM Fig. 1.(a) Schematic diagram of SPPR modes supported by a standalone ANC (b) Dispersion of the two lowest TM modes of an unperturbed cylindrical cavity in gold.(c) Electric field lines around a gold tip (Scale bar: 10 nm).The wave is incident from the left and polarized along the tip.(N n : n-th node for field convergence or divergence).Labels P n and V indicate wavelengths important to this study and will be explained in Section 4 and Fig.2. Fig. 3 . Fig. 3. (a)-(c) Simulation results at P 1 , V, and P 3 of Fig. 2(a).The electric field patterns are superimposed over surface plots of the LFE factors in log scale.The dotted boxes indicate the location of modal overlap.(d)-(f) Schematic representations of tip-cavity plasmonic coupling corresponding to (a)-(c), respectively. Fig. 4 . Fig. 4. Computed LFE factors at the tip as a function of the aperture width w (cavity radius: 150 nm, TM input at 0° incidence angle). | 3,068,828 | 29598846 | 0 | 16 |
Functional Analysis of Norcoclaurine Synthase in Coptis japonica* (S)-Norcoclaurine is the entry compound in benzylisoquinoline alkaloid biosynthesis and is produced by the condensation of dopamine and 4-hydroxyphenylacetaldehyde (4-HPAA) by norcoclaurine synthase (NCS) (EC 4.2.1.78). Although cDNA of the pathogenesis-related (PR) 10 family, the translation product of which catalyzes NCS reaction, has been isolated from Thalictrum flavum, its detailed enzymological properties have not yet been characterized. We report here that a distinct cDNA isolated from Coptis japonica (CjNCS1) also catalyzed NCS reaction as well as a PR10 homologue of C. japonica (CjPR10A). Both recombinant proteins stereo-specifically produced (S)-norcoclaurine by the condensation of dopamine and 4-HPAA. Because a CjNCS1 cDNA that encoded 352 amino acids showed sequence similarity to 2-oxoglutarate-dependent dioxygenases of plant origin, we characterized the properties of the native enzyme. Sequence analysis indicated that CjNCS1 only contained a Fe2+-binding site and lacked the 2-oxoglutarate-binding domain. In fact, NCS reaction of native NCS isolated from cultured C. japonica cells did not depend on 2-oxoglutarate or oxygen, but did require ferrous ion. On the other hand, CjPR10A showed no specific motif. The addition of o-phenanthroline inhibited NCS reaction of both native NCS and recombinant CjNCS1, but not that of CjPR10A. In addition, native NCS and recombinant CjNCS1 accepted phenylacetaldehyde and 3,4-dihydroxyphenylacetaldehyde, as well as 4-HPAA, for condensation with dopamine, whereas recombinant CjPR10A could use 4-hydroxyphenylpyruvate and pyruvate in addition to the above aldehydes. These results suggested that CjNCS1 is the major NCS in C. japonica, whereas native NCS extracted from cultured C. japonica cells was more active and | 3,068,829 | 21542493 | 0 | 16 |
formed a larger complex compared with recombinant CjNCS1. Higher plants produce divergent chemicals such as alkaloids, terpenoids, and phenolic compounds in secondary metabolism. Among these chemicals, alkaloids are very important in medicine because of their high biological activities. Alkaloids are low molecular weight, nitrogen-containing compounds that are found in ϳ20% of plant species. Most alkaloids are derived from amines produced by the decarboxylation of amino acids such as histidine, lysine, ornithine, tryptophan, and tyrosine. Although the coupling of amines to other products is the first important step in producing diverse alkaloids, this entry reaction has been poorly characterized. (S)-Strictosidine is a central intermediate for indole alkaloids. Strictosidine synthase, which catalyzes the formation of (S)-strictosidine from tryptamine and secologanin, is a rare exception in that its cDNA has been cloned from Catharanthus roseus and Rauvolfia serpentine (1,2). Benzylisoquinoline alkaloids are a large and diverse group of pharmaceutical alkaloids with ϳ2,500 defined structures. Norcoclaurine produced from tyrosine is a key entry compound from which various benzylisoquinoline alkaloids such as analgesic morphine, colchicines, antibacterial berberine, palmatine, and sanguinarine are produced through a multistep process. (S)-Norcoclaurine is produced by norcoclaurine synthase (NCS) 3 (EC 4.2.1.78), which condenses dopamine and 4-hydroxyphenylacetaldehyde (4-HPAA) ( Fig. 1) (3,4). Because this condensation reaction can occur relatively easily by a chemical reaction, it has been difficult to characterize the biochemical and molecular biological properties of NCS. Only recently has NCS been isolated and characterized from cell suspension cultures of meadow rue (Thalictrum flavum ssp. glaucum) (5,6), but the mechanism of the NCS reaction, including the | 3,068,830 | 21542493 | 0 | 16 |
stereospecificity of product, has not yet been determined. Cultured Coptis japonica cells are a unique system for producing large amounts of benzylisoquinoline alkaloid. Using this system, many enzymes in berberine biosynthesis, including norcoclaurine 6-O-methyltransferase, coclaurine N-methyltransferase, N-methylcoclaurine 3Ј-hydroxylase, 3Ј-hydroxy-N-methylcoclaurine 4Ј-O-methyltransferase, have been isolated and characterized (7)(8)(9). In this report, we describe the characterization of NCS in C. japonica cells. Because selected cultured C. japonica 156-1 cells expressed the genes for enzymes in berberine biosynthesis in greater amounts than non-selected cells 4 (9 -11), we screened expressed sequence tag (EST) clones with higher expression in a selected line and examined its NCS activity. Our isolated candidate cDNA (CjNCS1) had a dioxygenase-like protein family gene domain, and we examined its roles in NCS activity using both recombinant CjNCS1 produced in Escherichia coli and native NCS isolated from cultured C. japonica cells in comparison with recombinant Cj pathogenesis-related (PR) 10-like protein (PR10A) isolated based on sequence homology with Thalictrum NCS. All of our data suggested that the CjNCS1 gene product corresponds to native NCS in C. japonica cells. We discuss the significance of the co-existence of both CjNCS1 and CjPR10A proteins with NCS activity in C. japonica and the molecular evolution of NCS. EXPERIMENTAL PROCEDURES Cultured Cells and RNA Preparation-Four C. japonica cell lines, CjY, Cj8, 156-1, and 156-1S, were maintained in suspension culture as described previously (12). CjY was an unselected cell line, whereas 156-1 and Cj8 were selected cell lines for high alkaloid production (12), but Cj8 had lost its high productivity after inadequate subculture. The cell | 3,068,831 | 21542493 | 0 | 16 |
line 156-1S was transgenic 156-1 cells that had been transformed with scoulerine 9-Omethyltransferase cDNA expression vector driven by cauliflower mosaic virus 35S promoter (13). Total RNA was extracted from 1-, 2-, and 3-week-old cells with TRIzol and used for RNA blot analysis (10,14). Construction and Sequencing of a cDNA Library of C. japonica-Poly(A) ϩ RNA was isolated and a cDNA library was constructed as described elsewhere (8). The cDNA fragments were ligated into pDR196 vector (15). Sequencing of the cDNA library was performed using a MegaBACE 1000 DNA Sequencing System (Amersham Biosciences) in accordance with the manufacturer's instructions. RNA Gel Blot Analysis-10 g of total RNA isolated from each cultured cell line was electrophoresed on formaldehyde gels and then blotted onto BioDyne A TM (Pall) membranes. The EST clone sequences were amplified by PCR with primers (5Ј-GAA AGA AAA AAA ATA TAC CCC AGC-3Ј and 5Ј-TTT CGT AAA TTT CTG GCA AGG TAG AC-3Ј), separated on agarose gel, and recovered using GenElute TM (Sigma-Aldrich) to prepare probes. PCR product was labeled with 32 P-radioisotope, and RNA hybridization and signal detection were carried out according to standard protocols as reported previously (16). Construction of Expression Vectors for Recombinant CjNCS1 and CjPR10A-Expression vectors were constructed for fulllength CjNCS1 and CjPR10A cDNAs without the fused peptide derived from the vector sequence in pET-41a vector (Novagen, Madison, WI). CjNCS1 and CjPR10A cDNAs were amplified by PCR with 5Ј-GGG AAT TCC ATA TGA GCA AGA ATC TTA CTG GTG-3Ј (NdeI restriction site is underlined) and 5Ј-CCG CTC GAG TTA TAG TTT | 3,068,832 | 21542493 | 0 | 16 |
CAT CTG ATC CAA TAA GCT CTT ACC-3Ј (XhoI adaptor is underlined) primers for CjNCS1 and 5Ј-GGC ATA TGA GGA TGG AAG TTG TTC TAG-3Ј (NdeI restriction site is underlined) and 5Ј-GGC TCG AGT TAC TCT GAA CTC TTG TGT TTG-3Ј (XhoI adaptor is underlined) primers for CjPR10A. The PCR products were digested with NdeI and XhoI and ligated into the pET-41a vector digested with NdeI and XhoI using a DNA ligation kit (Takara Shuzo Co.). The resulting plasmid/vector was sequenced to confirm the correct construction. Isolation of Arabidopsis Homologue (AtSRG1) to CjNCS1 and Construction of Expression Vector-Total RNA was extracted from 5-week-old callus cultures of Arabidopsis thaliana maintained on Linsmaier and Skoog medium supplemented with 3% sucrose, 10 M 1-naphthaleneacetic acid, and 1 M benzyladenine using an RNeasy kit (Qiagen). Total RNA (2.5 g) was converted to cDNA using Superscript III according to the protocol of the manufacturer, and the AtSRG1 (GenBank TM accession number S44261), which is highly homologous to CjNCS1, was amplified by PCR using 5Ј-GGG AAT TCC ATA TGG AAG CAA AAG GGG CAG CA-3Ј (NdeI restriction site is underlined) and 5Ј-CCG CTC GAG TTA GAT TCT CAA AGC ATC TAG-3Ј (XhoI adaptor is underlined) as primers. PCR product was digested with NdeI and XhoI and ligated into pET-41a as described above. Heterologous Expression of Recombinant CjNCS1, CjPR10A, and AtSRG1 in E. coli-The expression vectors for full-length cDNAs of CjNCS1, CjPR10A, and AtSRG1 were introduced into E. coli BL21 (DE3). E. coli cells containing each plasmid were grown at 37°C in Luria | 3,068,833 | 21542493 | 0 | 16 |
Bertani medium. After induction with 1 mM isopropylthiogalactoside, E. coli cells containing CjNCS1 plasmid were incubated at 16°C for 48 h and E. coli cells containing CjPR10A and AtSRG1 plasmids were incubated at 37°C for 4 h. The resulting recombinant CjNCS1 protein was extracted from a 1-liter culture of E. coli grown in Luria Bertani medium with the extraction Tris buffer (0.1 M Tris-HCl, pH 8.0, containing 10% glycerol, 5 mM EDTA, and 5 mM 2-mercaptoethanol) and then purified to homogeneity. Crude extract was centrifuged at 10,000 ϫ g for 10 min, and the supernatant was subjected to fractionation with (NH 4 ) 2 SO 4 . The desired recombinant protein was precipitated between 30 and 60% (NH 4 ) 2 SO 4 saturation and made soluble in the extraction buffer again. The protein solution was dialyzed in the same buffer and applied to a DEAE-Sepharose CL-4B column (Amersham Biosciences) (2.5 ϫ 16 cm) that had been equilibrated with the extraction buffer. After the column was washed with 2 volumes of the same buffer, recombinant protein was eluted with 250 ml of a linear gradient of extraction buffer with 0 -0.5 M NaCl solution. The recombinant protein fractions were detected with SDS-PAGE, and these fractions were pooled and then applied to a hydroxyapatite column (Amersham Biosciences) (2.5 ϫ 9 cm) that had been pre-equilibrated with 10 mM potassium phosphate buffer (KPB) (pH 8.0). After the column was washed with 10 mM KPB, CjNCS1 was eluted with a linear gradient (0.01-0.5 M) of KPB solution (100 ml). | 3,068,834 | 21542493 | 0 | 16 |
The active fractions were pooled and then dialyzed in 50 mM KPB (pH 8.0). All operations were performed at 4°C, and all of the buffers contained 5 mM 2-mercaptoethanol and 10% glycerol unless stated otherwise. To examine the effect of ferrous ion in the NCS reaction, the enzyme solution was preincubated with 5 mM EDTA or 0.5 mM o-phenanthroline for 5 min at 4°C. After preincubation with chelators, the substrates were added to the reaction mixture, and the NCS activity was assayed as described above. Immunoprecipitation of NCS Activity with Antiserum against CjNCS1-Mouse polyclonal antibodies for C. japonica NCS were prepared against purified recombinant CjNCS1 at Japan Clinical Laboratories, Inc. (Kyoto, Japan). These antibodies were specific to CjNCS1, and no signal was detected by SDS-PAGE immunoblotting except for CjNCS1 in crude C. japonica cell extract. For immunoprecipitation experiments, the same volume of polyclonal antiserum for CjNCS1 was added to cell lysates. After 2 h of incubation at 4°C, 20 l of Protein G Plus/Protein A-agarose suspension (Calbiochem) was added to an antiserum/enzyme mixture, and the suspension was then mixed gently for another 12 h at 4°C. The supernatant was recovered by centrifugation, and NCS activity was determined as described above. RESULTS NCS Activity in C. japonica Cells-(S)-Norcoclaurine, the central precursor of pharmaceutically important benzylisoquinoline alkaloids, is produced by NCS, which condenses dopamine and 4-HPAA (Fig. 1). NCS activity is usually detected by the formation of total norcoclaurine (21), whereas racemic norcoclaurine is also formed by a chemical reaction during the enzyme reaction. Therefore, we established a | 3,068,835 | 21542493 | 0 | 16 |
system to detect the stereo-specific formation of (S)-norcoclaurine to estimate the native NCS reaction. Crude extract from cultured C. japonica cells clearly showed that the native enzymic reaction was stereo-specific, that is, it gave the (S) form, whereas the product of the chemical reaction was a mixture of (R) and (S) forms (Fig. 2). Molecular Cloning of a C. japonica NCS cDNA Candidate-Because it was rather difficult to assay NCS due to the background chemical reaction until the recent development of chiral column chromatographic analysis, we tried to isolate NCS cDNA using an EST library prepared from high berberineproducing cultured C. japonica cells. In fact, our EST library of C. japonica 156-1 cells has the advantage that transcripts of berberine biosynthetic enzyme have been highly enriched. 4 When the sequences of our EST library were determined, we noticed that it contained many oxygenase-like proteins that might be related to berberine biosynthesis. Additional RNA gel blot analysis showed that such oxygenase-like protein cDNA could be related to berberine biosynthesis, because the expression pattern of oxygenase-like protein was highly correlated with the expression of other biosynthetic genes in several Coptis cell lines with different berberine productivities (supplemental data). Because cDNAs for early steps in norcoclaurine biosynthesis from tyrosine have not yet been characterized, we examined the enzyme activity of this oxygenase-like protein. To examine the enzyme activity of oxygenase-like protein, we produced recombinant protein in E. coli. Although no enzyme activity was detected for tyrosine, 3,4-dihydrophenylalanine (DOPA), or tyramine, this protein produced a product with dopamine and 4-HPAA. | 3,068,836 | 21542493 | 0 | 16 |
The reaction product was identified to be (S)-norcoclaurine by LC-MS, although the enzyme activity was considerably lower than that of native enzyme. The other recombinant protein (putative 2-nitropropane dioxygenase) and pET 41-a vector, as the control, did not show NCS activity or any other enzyme activity for the early reaction from tyrosine to norcoclaurine. Thus, this clone was named CjNCS1 in this report. We determined the full-length sequence of CjNCS1 and found that this EST had 1172 nucleotides with an open reading frame that encoded 352 amino acids (supplemental data). The deduced amino acid sequence of CjNCS1 was similar to 2-oxoglutarate-dependent dioxygenases of plant origin (Fig. 3). CjNCS1 had no putative signal peptide and was predicted to be cytosolic protein. CjNCS1 was purified to electrophoretic homogeneity by ammonium sulfate fractionation with DEAE-Sepharose CL-4B and hydroxyapatite column chromatography. The molecular mass of the enzyme was determined to be 40 kDa by Superdex 200 HR 10/30 gel filtration (data not shown). SDS-PAGE of CjNCS1 gave a single band corresponding to a molecular mass of 40 kDa, indicating that the recombinant CjNCS1 was a monomer (data not shown). During our characterization of CjNCS1, NCS was isolated from a T. flavum cell culture and the amino acid sequences predicted from its cDNA showed 28 -38% identity with the Betv1 allergen and belonged to the PR10 protein family, whereas no homology with CjNCS1 was found (6). Thus, a gene homologous to T. flavum NCS (C. japonica PR10A) was isolated from a C. japonica EST library and compared with native NCS in | 3,068,837 | 21542493 | 0 | 16 |
C. japonica cells. Recombinant CjPR10A was produced in E. coli, and the protein was used as crude extract without further purification. Both CjNCS1 and CjPR10A catalyzed the condensation of dopamine and 4-HPAA and produced (S)-norcoclaurine, although the elution time of the products differed between CjNCS1 and CjPR10A in LC-MS analysis (Fig. 4). Confirmation of the structure of the product of the NCS reaction with CjNCS1 or CjPR10A by LC-NMR also showed that the products of both CjNCS1 and CjPR10A were norcoclaurine (data not shown). Characterization of C. japonica NCS Based on the Sequence Information of CjNCS1-To characterize which gene product corresponds to native NCS in C. japonica, native NCS extracted from cultured C. japonica cells was immunoprecipitated with antiserum against recombinant CjNCS1. The activity of native NCS after immunoprecipitation with anti-CjNCS1 antibodies was markedly reduced, whereas antibodies against Cj3Ј-hydroxy-N-methylcoclaurine 4Ј-O-methyltransferase, another biosynthetic enzyme in the berberine pathway, did not affect the NCS activity (Fig. 5A). Because a measurable amount of NCS activity remained even after immunoprecipitation with anti-CjNCS1, we examined whether CjPR10A might be involved in the NCS reaction in C. japonica cells as shown below. Because the sequence of CjNCS1 suggested that ferrous ion may be required (Fig. 3), we examined the effect of chelators on the NCS activity of native enzymes and both recombinant proteins. o-Phenanthroline, a chelator for ferrous ion, inhibited NCS activity in C. japonica cells and recombinant CjNCS1 (Fig. 5B), whereas EDTA did not (data not shown). o-Phenanthro- line as well as EDTA did not inhibit the NCS activity of CjPR10A. | 3,068,838 | 21542493 | 0 | 16 |
This result again showed CjNCS1 was involved in native NCS in C. japonica. However, it was possible that CjPR10A was still involved in native NCS, because residual NCS activity was visible. When we examined the substrate specificity of the NCS reaction, we found that the substrate specificity of CjPR10A was different from that of native NCS in C. japonica and CjNCS1. CjNCS1 and the crude extract from cultured C. japonica cells only accept dopamine and a phenylacetaldehyde group (phenylacetaldehyde, 3,4-dihydroxyphenylacetaldehyde, and 4-HPAA), whereas CjPR10A used pyruvic acid and 4-hydroxyphenylpyruvate for the condensation with dopamine (data not shown). This result clearly indicated that the native NCS reaction in C. japonica involved CjNCS1, not CjPR10A. The activity of recombinant CjNCS1 was lower than that of the native enzyme extracted from cultured C. japonica cells on a protein basis. Therefore, the compositions of recombinant CjNCS1 and native NCS were compared. The fractions by gel filtration for purified recombinant CjNCS1 and crude extract from C. japonica cells were analyzed by Western blotting. As a result, NCS from C. japonica was a complex, whereas recombinant CjNCS1 was a monomer (data not shown). 2-Oxoglutarate was not needed in the NCS reaction, as expected from the sequence of CjNCS1. Isopenicillin N synthase has structural characteristics similar to those of CjNCS1: requirement for ferrous ion, but not 2-oxoglutarate (25). Because isopenicillin N synthase requires oxygen for its enzyme activity, we examined the requirement for oxygen in the NCS reaction. The depletion of dissolved oxygen in the reaction mixture by the glucose oxidase-catalase system (26) | 3,068,839 | 21542493 | 0 | 16 |
did not reduce NCS activity. Thus, we concluded that NCS of C. japonica was a novel 2-oxoglutarate-independent dioxygenase-like protein that catalyzed the cyclase reaction without oxygen and was different from isopenicillin N synthase. Sequence Comparison of CjNCS1 and Other Plant Dioxygenases-The sequence of CjNCS1 was compared with those of other representative 2-oxoglutarate-dependent dioxygenases in plants to estimate the molecular basis of reaction (Fig. 3). The sequence of CjNCS1 showed a typical 2oxoglutarate-dependent dioxygenase-like protein sequence as in other plant genes. CjNCS1 had two completely conserved histidine and aspartate residues (His-228 and Asp-230) that wouldactasferrousligands,asreportedforother2-oxoglutaratedependent dioxygenases (27), and the motif His X Asp (53-57 amino acids) X His (28) (indicated by asterisks). However, CjNCS1 did not have the highly conserved consensus sequence for binding of the co-substrate, 2-oxoglutarate, Asn-Tyr-Tyr-Pro-Pro-Cys-Pro-Gln-Pro (Fig. 3). This corresponded to the result that 2-oxoglutarate was not required in the reaction of NCS. Isopenicillin N synthase also does not have this sequence and does not require 2-oxoglutarate (25). A homology search revealed that A. thaliana has AtSRG1, which is highly homologous to CjNCS1 (Fig. 6). AtSRG1 is a 2-oxoglutarate-dependent dioxygenase-like protein but has a conserved 2-oxoglutarate-binding domain. Because AtSRG1 retained a binding domain for 2-oxoglutarate, we examined the NCS activity of AtSRG1 expressed in E. coli. Although recom-binant AtSRG1 was successfully expressed in E. coli in soluble form, crude extract did not show any NCS activity with dopamine and 4-HPAA (data not shown). DISCUSSION Norcoclaurine formation is the key reaction in benzylisoquinoline alkaloid biosynthesis. NCS activity has been detected in most plant species that | 3,068,840 | 21542493 | 0 | 16 |
have been shown to produce benzylisoquinoline alkaloids (29). However, it has been difficult to characterize NCS due to the background activity of a chemical reaction. Although a gene in the PR10 family has been isolated from T. flavum and opium poppy (6, 29) during our characterization of CjNCS1, their physiological role was not yet been identified. Although NCS produce (S)-norcoclaurine stereospecifically, it has not been investigated empirically. In this study, we established an LC-MS system with chiral column to determine that NCS from C. japonica produced (S)-norcoclaurine stereo-specifically (Fig. 2). Because 6-O-methyltransferase and coclaurine N-methyltransferase in late steps of benzyliso- Functional Analysis of NCS in C. japonica quinoline alkaloid biosynthesis are not stereo-specific (7,8,30), NCS plays an important role in producing optical isomers in the biosynthesis. Although we also found that two gene products (CjNCS1 and CjPR10A) isolated from C. japonica cells could catalyze the stereo-specific condensation of dopamine and 4-HPAA, further studies showed that these enzymes had different enzymological properties and substrate specificities. CjNCS1 had a dioxygenase family domain and a ferrous ionbinding site, but CjPR10A did not. Furthermore, whereas CjPR10A accepted dopamine and 4-hydroxyphenylpyruvate as substrates for norcoclaurine formation, native NCS and CjNCS1 did not. Native NCS enzymes from Eschscholzia californica, Nandina domestica, and Corydalis pallida var. tenuis cells also did not accept phenylpyruvate as a substrate as did Coptis enzyme. 5 Whereas the formation of norcoclaurine by the condensation of dopamine and 4-HPAA is now clearly established as the central pathway in isoquinoline alkaloid biosynthesis (4,24), the formation of norcoclaurine via the condensation | 3,068,841 | 21542493 | 0 | 16 |
of dopamine and 4-HPAA or via the condensation of dopamine and 4-hydroxyphenylpyruvate was also intensively investigated. The formation of norlaudanosoline through norlaudanosoline 1-carboxylic acid by the condensation of dopamine and 3,4-dihydroxyphenylpyruvate was once seriously considered to be a possible pathway using in vitro and in vivo tracer experiments (23,31). The formation of norcoclaurine from dopamine and 4-hydroxyphenylpyruvate by recombinant CjPR10A suggested that a PR10-like protein might catalyze such an artificial reaction. Our current result provides the biochemical basis for solving the long-standing mystery of artificial activity through norlaudanosoline 1-carboxylic acid in isoquinoline alkaloid biosynthesis. CjPR10A and NCS from T. flavum had the signal peptide N-terminal (6) and were predicted to be localized in a vesicular compartment. On the other hand, major biosynthetic enzymes such as tyrosine/DOPA decarboxylase, 6-O-methyltransferase, coclaurine N-methyltransferase, and 3Ј-hydroxy-N-methylcoclaurine 4Ј-O-methyltransferase in the early steps in the isoquinoline alkaloid biosynthesis are cytosolic (32). Although dopamine synthesized in cytosol from L-DOPA by DOPA decarboxylase would be transported and accumulated within a vacuole at the concentration of 1 mg/ml in the latex of Papaver somniferum and Papaver bracteatum (33,34), dopamine produced in cytosol would be more efficient substrate for CjNCS1 and the following biosynthetic reactions. Immunoprecipitation with antiserum against CjNCS1 clearly indicated that native NCS in C. japonica cells consisted of CjNCS1 and not CjPR10A. Our preliminary characterization of the function of CjNCS1 using the transient RNA interference in C. japonica protoplasts (35) showed that accumulation of berberine clearly decreased with double-stranded RNA of CjNCS1, supporting the involvement of CjNCS1 in berberine biosynthesis. 6 Although recombinant | 3,068,842 | 21542493 | 0 | 16 |
CjNCS1 expressed in E. coli showed NCS activity, this activity was lower than that of native enzyme on a protein basis as estimated by immunoblot analysis. This low activity might be due to the instability of NCS, as reported in crude extracts of P. somniferum and E. californica (21). Extreme instability has also been reported for 1-aminocyclopropane-1-carboxylate oxidases and anthocyanidine synthase, which are 2-oxoglutarate-dependent oxygenases (36,37). Partially distorted folding or modification in E. coli also might decrease the activity. Our chromatographic analysis suggested that the native NCS extracted from cultured C. japonica cells would form a complex, whereas recombinant CjNCS1 was a monomer. Thus, the low activity of recombinant CjNCS1 might be caused by the difference in complex formation. Recently, Kristensen et al. (38) reported that the coordinated expression of two cytochrome P450 genes with a glucosyltransferase from sorghum enhanced the production of dhurrin in the heterologous host plant Arabidopsis through metabolon formation. RNA interference of codeninone reductase was also suggested to interrupt the metabolic channel from reticuline to codeine in P. somniferum (39). Although the NCS reaction in isoquinoline alkaloid biosynthesis may also form a similar complex to produce norcoclaurine from tyrosine, we need a more detailed characterization of the enzyme complex in isoquinoline alkaloid biosynthesis. Although we determined that CjNCS1 and CjPR10A produced (S)-norcoclaurine in LC-MS analysis, the different elution time of the products may suggest the different mechanism of NCS reaction between CjNCS1 and CjPR10A. The sequence similarity of CjNCS1 with the plant dioxygenase family would provide insight into the reaction mechanism of | 3,068,843 | 21542493 | 0 | 16 |
NCS, because genes in the dioxygenase family have been reported to be involved in divergent biosynthetic pathways as hydroxylase, hydroxylase/epoxidase, cyclase, desaturase, and ring expandase (40,41). One unique feature of CjNCS1 is that CjNCS1 lacks the 2-oxoglutarate-binding sequence, whereas it is highly homologous to plant 2-oxoglutarate-dependent dioxygenase. While the significance of this lack of a 2-oxoglutarate domain should be clarified in future studies, we expect that this modification would be an important step in the development of NCS activity. The alignment between NCS and homologous genes indicated that CjNCS1 was similar to isopenicillin N synthase, another cyclase in the dioxygenase family. Isopenicillin N synthase has a degree of homology with deacetoxycephalosporin C synthetase and deacetylcephalosporin C synthase, two 2-oxoglutarate-dependent dioxygenases involved in the later steps of cephalosporin biosynthesis (28). Interestingly, isopenicillin N synthase also has no 2-oxoglutarate binding region. Crystallographic characterization of the three-dimensional structure of CjNCS1 would provide important cues for the understanding of the molecular mechanism of NCS reaction. The molecular evolution of new enzymological activity has been discussed for 4,5-extradiol dioxygenase in betalain biosynthesis (42). Alignment of this gene family revealed a conserved motif that is present in all organisms except plants that synthesize betalain. In the alignment of CjNCS1 and other plant 2-oxoglutarate-dependent dioxygenases, the N-terminal domains are poorly conserved. The poorly conserved N-terminal domain may also represent binding sites for specific alkaloid substrates. 2-Oxoglutarate-dependent dioxygenases have also been reported to be involved in alkaloid biosynthesis in plants. Although there have been previous reports on hyoscyamine 5 H. Minami and F. Sato, | 3,068,844 | 21542493 | 0 | 16 |
unpublished data. 6 E. Fukusaki, C.-I. An, A. Kobayashi, and F. Sato, data not shown. 6-hydroxylase in scopolamine biosynthesis (43) and desacetoxyvindoline 4-hydroxylase in vindoline biosynthesis (44), this is the first report of the cloning of cyclase-type dioxygenase-like protein in alkaloid biosynthesis in plants. We expect that more dioxygenase-like proteins may play a role in alkaloid biosynthesis. It has been suggested that an early pathway in isoquinoline alkaloid biosynthesis is universal in the plant kingdom. During evolution, the catalytic properties of the key enzymes were created by amino acid substitution, and as such the enzymes became part of a biosynthetic pathway. In C. japonica, the universal pathway involving NCS plays a role in berberine biosynthesis. Recently Liscombe et al. (29) reported the monophyletic origin of benzylisoquinoline alkaloid biosynthesis prior to the emergence of the eudicots. The molecular characterization of NCS as key entry gene in comparison with cytochrome P450, such as CYP80 and CYP719, which are specific for isoquinoline alkaloid biosynthesis, would be useful for better understanding of the evolutionary pathway of this unique secondary metabolism. It is interesting that A. thaliana and rice (e.g. AK072706) have a CjNCS1-like protein, although we could not characterize the rice CjNCS-like protein in detail. As reported above, Arabidopsis SRG1 protein had a 2-oxoglutarate-binding domain and no NCS activity. SRG1 expression in Arabidopsis has been reported to be regulated by ethylene-induced senescence. Although the true function of SRG1 has not yet been clarified, it might be involved in the detoxification of amines or aldehydes as senescence proceeds, because both amine | 3,068,845 | 21542493 | 0 | 16 |
and aldehyde are highly toxic to physiological functions. Although the origin of alkaloid biosynthesis is not clear, the key entry enzyme genes are gradually being isolated. For example, strictosidine synthase couples tryptamine and secologanin as the first committed step in monoterpenoid indole alkaloid biosynthesis (2). Another example is homospermidine synthase, which catalyzes the first specific step in pyrrolizidine alkaloids from putrescine and spermidine (45,46). Homospermidine synthase has been shown to be recruited from deoxyhypusine synthase by independent gene duplication in several different angiosperm lineages during evolution (47). The amino acid sequences of both genes are considerably different from that of NCS. CjNCS1 could offer a new perspective for studies on the diversity and evolution of alkaloid biosynthesis. | 3,068,846 | 21542493 | 0 | 16 |
Wheat Ears Counting in Field Conditions Based on Multi-Feature Optimization and TWSVM The number of wheat ears in the field is very important data for predicting crop growth and estimating crop yield and as such is receiving ever-increasing research attention. To obtain such data, we propose a novel algorithm that uses computer vision to accurately recognize wheat ears in a digital image. First, red-green-blue images acquired by a manned ground vehicle are selected based on light intensity to ensure that this method is robust with respect to light intensity. Next, the selected images are cut to ensure that the target can be identified in the remaining parts. The simple linear iterative clustering method, which is based on superpixel theory, is then used to generate a patch from the selected images. After manually labeling each patch, they are divided into two categories: wheat ears and background. The color feature “Color Coherence Vectors,” the texture feature “Gray Level Co-Occurrence Matrix,” and a special image feature “Edge Histogram Descriptor” are then exacted from these patches to generate a high-dimensional matrix called the “feature matrix.” Because each feature plays a different role in the classification process, a feature-weighting fusion based on kernel principal component analysis is used to redistribute the feature weights. Finally, a twin-support-vector-machine segmentation (TWSVM-Seg) model is trained to understand the differences between the two types of patches through the features, and the TWSVM-Seg model finally achieves the correct classification of each pixel from the testing sample and outputs the results in the form of binary image. This | 3,068,847 | 49666754 | 0 | 16 |
process thus segments the image. Next, we use a statistical function in Matlab to get the exact a precise number of ears. To verify these statistical numerical results, we compare them with field measurements of the wheat plots. The result of applying the proposed algorithm to ground-shooting image data sets correlates strongly (with a precision of 0.79–0.82) with the data obtained by manual counting. An average running time of 0.1 s is required to successfully extract the correct number of ears from the background, which shows that the proposed algorithm is computationally efficient. These results indicate that the proposed method provides accurate phenotypic data on wheat seedlings. INTRODUCTION Wheat is an important primary food for a large proportion of the world's population, so methods to estimate its yield have received significant research attention (Bognár et al., 2017). The number of ears per unit area, the number of grains per ear, and 1000 grain weight are known as the three elements of wheat yield (Plovdiv, 2013). Of these, the number of ears per unit area is mainly obtained in the field. The wheat ear is an important agronomic component (Jin et al., 2017) not only is closely associated with yield but also plays an important role in disease detection, nutrition examination, and growthperiod determination. Thus, an accurate determination of the number of ears is vital for estimating wheat yield and is a key step in field phenotyping (Zhang et al., 2007). At present, two main statistical methods exist to obtain the number of ears per unit area: | 3,068,848 | 49666754 | 0 | 16 |
manual field investigation and image-based crop recognition (Nerson, 1980). Manual field investigation, which is the traditional method, is inefficient and costly, resulting in more and more interest in image-based crop recognition. However, because of the complexity of the field environment (e.g., illumination intensity, soil reflectance, and weeds, which alters the colors, textures, and shapes in wheat-ear images), accurate wheat ear segmentation and recognition remains a significant challenge (Mussavi and M. Sc. of Agronomy Ramin Agricultural and Natural Resources, 2011). In the field of image segmentation, a number of meaningful research results have emerged in recent years. These methods mostly focus on two approaches, the first of which is based solely on color information (Naemura et al., 2000). For example, Chen et al. proposed a threshold-selection algorithm for image segmentation based on the Otsu rule (Chen et al., 2012). Subsequently, Khokher et al. introduced an efficient method for color-image segmentation that uses adaptive mean shift and normalized cuts (Khokher et al., 2013). Moreover, Liao et al. used an edge-region active contour model for simultaneous magnetic resonance image segmentation and denoising (Liao et al., 2017). Additionally, the color information for wheat changes over the reproductive stage. Thus, different methods usually apply to different stages of reproduction. Therefore, in addition to the disadvantages described above, an excessive dependence on color information will lead to incomplete extraction. The second approach involves machine learning. For example, Kandaswamy et al. used the meta-segmentation evaluation technique to deal with the problem of image segmentation (Kandaswamy et al., 2013). Linares et al. introduced an | 3,068,849 | 49666754 | 0 | 16 |
imagesegmentation algorithm based on the machine learning of features (Linares et al., 2017). Soh et al. proposed a method based on a linear classifier that reveals a new method of segmentation (Soh and Tsatsoulis, 1999). In addition, Lizarazo et al. used a support vector machine (SVM) classifier to segment remotesensing data (Lizarazo, 2008). Because of its high accuracy and robustness, target segmentation based on classifiers was widely used for target recognition in the field of complex environments (Lizarazo and Elsner, 2009). This method mainly includes two key steps: (i) extraction and combination of image features and (ii) selection of classifiers to be trained. The first step above forms the basis of image recognition . Choosing the appropriate features directly impacts the final segmentation and recognition accuracy (Ding et al., 2017). Hu et al. proposed an image-feature-extraction method based on shape characteristics (Hu et al., 2016), and Yang et al. introduced multi-structure feature fusion for face recognition based on multi-resolution exaction (Yang et al., 2011). Datta et al. applied kernel principal component analysis (KPCA) to classify object-based vegetation species to fuse color and texture features, which has good results (Datta et al., 2017). To summarize, compared with the single-feature method, using a variety of features to express the red-green-blue (RGB) images can be more comprehensive and effective for improving the descriptive ability. Next, another key step of the classifier-based segmentation method is to use a general classifier to classify the features. The representative image classifier to be trained mainly includes a rough set, a Bayesian, and a | 3,068,850 | 49666754 | 0 | 16 |
SVM. Banerjee et al. used rough set theory to solve the problem of multispectral image classification (Banerjee and Maji, 2015). Zhang et al. proposed a method for multiple categories based on Bayesian decisions (Zhang et al., 2014). Finally, Park et al. introduced an automatic imagesegmentation method that uses principal pixel analysis and SVM (Park et al., 2016). Upon comparing with the other two classifiers, the SVM proves simpler in structure and offers global optimality and good generalization, so it has been widely used in the fields of image recognition and classification. However, the speed with which SVM learns a model is a major challenge for multi-class classification problems. To overcome these problems, the present study proposes a segmentation algorithm based on multi-feature optimization and twin-support-vector-machine (TWSVM) (Jayadeva et al., 2007). First, the algorithm extracts the color feature, texture feature, and edge histogram descriptor of wheat-ear images. Second, we use the KPCA to obtain the corresponding weights for each feature to rationally construct the feature space. The feature space is composed of multiple features to more comprehensively describe the target images, through which the advantages of each feature for classifying the different classes are manifested. Finally, the training of the TWSVM model is completed and better performance is obtained. The remainder of this paper is organized as follows: The next section describes in detail both the study area and image preprocessing. Section Methods describes the methodology. Section Results describes the experimental results and demonstrates the robustness of the method. Finally, we finish the paper with concluding remarks | 3,068,851 | 49666754 | 0 | 16 |
and possible directions for future work. Study Area The field planted with wheat was located in the Xinxiang comprehensive experimental base of the Chinese Academy of Agricultural Sciences. (Xinxiang, China, 35 • 9 ′ 32 ′′ latitude North, 113 • 48 ′ 28 ′′ longitude East). The sowing date was October 16, 2015. The experiment was conducted from 10 a.m. to 2 p.m. on June 9, 2016. For this paper, we collected data in overcast weather conditions, which resulted in totally scattered skylight with no direct illumination. These conditions eliminate shadows. While obtaining image data in the field, we made manual ground measurements of the corresponding plot to obtain manualrecognition data at the same time. The manual investigation area is 4 m 2 in each plot and the total area covered is 1200 m 2 with 300 plots. (Figure 1). Image Acquisition For each observation of an individual wheat ear, we systematically varied the illumination factors. Figure 2 shows an example of an image collected during a single observation. We imaged wheat ears from the side at 45 • above the horizontal because color and texture are typically substantial from this perspective. The camera aperture was f/3.9 with an exposure time of 1/90 s. The focal length of the camera was 50 mm. (A) Selection of dataset samples In conditions of varying illumination intensity from morning to afternoon, 1000 images were obtained with the same shooting mode (2 m imaging distance, 1.5 m imaging height, imaging angle at 45 • above horizontal) and the same camera | 3,068,852 | 49666754 | 0 | 16 |
parameters as mentioned above. As a result of the limitation of the number of sample images, 700 images were used as the training sample and the rest 300 images were used as the testing sample. This procedure gave us images under differing light intensities. We next divided these images into the following three categories by visual analysis: (a) high light intensity, (b) medium light intensity, and (c) low light intensity. All the images were selected from each category as the source of training set. This processing guarantees the robustness of the light intensity of the training results. (B) Image cutting The images used in this work were all obtained from oblique photography. As a result of the perspective, the wheat ears far from the shooting position are not well rendered in the images. The limitations imposed by camera resolution and the position of the camera focus make this part of the image low quality, so we cut the image to remove these parts and ensure uniform data quality. After this cutting process, the image size was reduced to 3500 × 1800 (Figure 2). (c) Counting results validation The performance of the image processing system to automatically counting the ears appearing in an image was tested in the images. In order to validate the algorithm, the machine counting result was compared with the manual image-based ear counting on the same image. Machine counting result depicts the binary image where the connected pixels in white color are considered as a wheat ear automatically detected by the image processing | 3,068,853 | 49666754 | 0 | 16 |
system; each of these regions are added and the final result is referred to as the algorithm counting. Besides, the number of ears in a subset of images has been counted manually and is referred to as the manual counting result. To ensure the precision of the manual statistics, two people repeated the counting operation according to the field range of the cut images. Moreover, in order to judge the accuracy of the segmentation, the wheat ear area is manual labeled as red small block. Then the labeled images were used as the mask images to compare with the machine segmentation and recognition results in order to ensure the accuracy of the method. METHODS After image acquisition, the main flow diagram of the proposed method includes off-line training of on-line segmentation, as shown in Figure 3. This research framework consists of five consecutive steps: (i) generating patches, (ii) establishing training and test sample sets, (iii) optical combination of multi-features space, (iv) training a classifier, and (v) noise reduction. Below, we discuss each step in detail and refer in particular to the variables, image types, and preprocessing strategies that we studied in our experiments. The specific algorithm (workflow) is as follows: Step 1: Select N images as training samples and extract patches of a certain size (20 × 20) from these samples; Step 2: Extract the color feature, texture feature, and edge histogram descriptor feature from the samples; Step 3: Use KPCA to extract the principal component features and calculate the weight for each feature in each | 3,068,854 | 49666754 | 0 | 16 |
class of samples; Step 4: Train the TWSVM classification model with the weighted features updated in Step 3; Step 5: Perform a weighting to the feature in the test sample with feature weights in each class, use the TWSVM-Seg model obtained in Step 4 to classify, and determine the image segmentation ( Figure 3). Generation of Patches Based on Simple Linear Iterative Clustering Method and Training Set and Validation Set Building region proportional to the superpixel size. This reduces the complexity to be linear in the number N of pixels-and independent of the number k of superpixels. 2. A weighted distance measure combines color and spatial proximity while simultaneously controlling the size and compactness of the superpixels. However, these SLIC superpixel regions are irregularly shaped, so they cannot be used directly as TWSVM input. Therefore, a small window called a patch (20 × 20 pixels) and centered on the weighted center of the current SLIC superpixel region is given to the TWSVM. Note that the code to implement the SLIC operation is based on open source code provided available at https://ivrl.epfl.ch/research/superpixels. After the SLIC generates the irregular superpixel regions as discussed above, the center of the patch is determined by the weighted center of the region. Then a regular patch is built according to the position of the center point. The percentages in each patch represent the ratio between the wheat ear area to the corresponding areas of the SLIC superpixel region. The sample patch is labeled category zero (background) if the percentage of the current | 3,068,855 | 49666754 | 0 | 16 |
patch is zero; otherwise, it is labeled category one (foreground). The fundamental part of any classification operation involves specifying the output or action, as determined based on a given set of inputs or training data. The classification system is formulated as a two-class model: positive patches and negative patches. The positive class contains image patches manually labeled from wheat ears under different illumination intensities. The negative class contains background images manually segmented from soil, rocks, etc. The dataset for the positive training class contains 8647 foreground patches, and that for the negative training class contains 7412 background patches. Meanwhile, the testing set was also generated by the SLIC through the same steps above. Multi-Feature Exaction and Combination (1) Multi-feature exaction Visual features are fundamental for processing digital images to represent image content. A good set of features should contain sufficient discrimination power to discriminate image contents. The feature-extraction section uses color coherence vectors (CCV) as the color feature (Roy and Mukherjee, 2013), the gray level co-occurrence matrix (GLCM) as the texture feature (Varish and Pal, 2018), and introduces the edge histogram descriptor (EHD) feature (Agarwal et al., 2013). Among them, CCV is Frontiers in Plant Science | www.frontiersin.org sufficiently robust to handle background complications and invariants in size, orientation, and partial occlusion of the canopy image. The GLCM feature has good performance in extracting information from local and frequency domains and it can provide good direction selection and scale selection characteristics. The EHD feature can effectively distinguish the images with very high similarity for colors and | 3,068,856 | 49666754 | 0 | 16 |
has good robustness for the color and brightness changes which are the features with very strong stability. Each feature describes the image content from different angles, performing a reasonable optimization and integration to achieve a more comprehensive description of the image content. The final feature matrix contains of three elements: f C 1 , f G 2 , and f E 3 . (2) Multi-Feature Combination based on Kernel Principal Component Analysis Based on the above description, we obtain a matrix composed of multi-dimensional features. Differences clearly exist for the importance of each feature in the classification process, then reasonably constructing the feature space, so it is important to assign weights to the features according to the importance of features. To achieve this goal, we use KPCA to extract the principal component of features, combining different features to determine the feature weights for the importance of different image classes (Twining and Taylor, 2003). The following details the specific method of classifying feature weights. We first normalized the fused feature to unify the range of values. The importance of features is inversely proportional to the dispersion of the feature distribution; features with a higher dispersion have a lower importance, which means that a smaller standard deviation leads to a higher importance for features. We thus use I n to indicate the importance of features: Where k n represents the standard deviation in class j of the sample set. When the distribution of one-dimensional features is more concentrated, the standard deviation is smaller, the corresponding k n is | 3,068,857 | 49666754 | 0 | 16 |
smaller, and the importance of the feature is greater. The formula for calculating the weight of features in each dimension is Through the above operation, we can merge the multiple feature vectors into a new feature matrix so that it can be used for machine learning with our own model. Image Segmentation Method Based on Twin-Support-Vector-Machine Segmentation Model and Noise Reduction We introduce a twin-support-vector-machine segmentation (TWSVM-Seg) model, which is based on the traditional SVM model is better for segmentation of wheat-ear images (Peng et al., 2016). It is similar in form to a traditional SWM with all its advantages. Moreover, it deals better with large-scale data. In the data X ∈ R (m * n) to be classified, we take positive samples m 1 with the "1" class from the training set to obtain matrix A m 1 ·n We then take negative samples m 2 with the "0" class from the train set to obtain matrixB m 2 ·n . We obtain a classification plane for each of the two classes. The data that belong to each class are, to the extent possible, near the corresponding classification plane. The required hyperplane parameters can be obtained by solving the following optimization problem: where K is the kernel function, A refers to m 1 positive (wheat ear) samples and B refers to m 2 negative (background) samples., e 1 and e 2 indicate the unit vector of the corresponding dimension, c 1 and c 2 are penalty coefficients, w is the normal vector of the optimal | 3,068,858 | 49666754 | 0 | 16 |
hyperplane, and b is the offset of the optimal hyperplane. q represents the discriminant coefficient. Here, the kernel function K is used to populate the TWSVM. Analysis shows that different kernel functions have very large impact on performance of TWSVM, and kernel function is also one of the adjustable parameters in TWSVM. Kernel functions, nuclear parameters and high-dimensional mapping space have a oneto-one relationship, so only select the proper kernel functions, nuclear parameters and high-dimensional mapping space when solving classification problem, we can get the separator with excellent learning and generalization ability. In this paper, we use the radial basis function (RBF) kernel K because of its excellent learning ability given large samples and low dimensions. We optimize the parameters of the kernel function after selecting. The error penalty factor c and gramma in the RBF are critical factors that impact the performance of the TWSVM, so these parameters strongly influence the classification accuracy and generalization ability of TWSVM. Here, we use the grid-search method to optimize and select parameters to obtain the global optimum results. Thus, the linear non-separable problem can be solved. Each sample in the training set belongs only to one of the two classes. By solving Equations (3) and (4), we get the following two hyperplanes: The two hyperplanes correspond to two different classes. For a sample to be classified, the distance to these two hyperplanes must be calculated. For each sample, the distance to each hyperplane is compared and the sample is classified into the nearest class. Through the above operation, | 3,068,859 | 49666754 | 0 | 16 |
the pixels in the test samples could be divided into two classes (wheat ear = 1, background = 0), which generates a binary image to achieve image segment. After all these operations, we then use the median filter w to minimize the noise and remove the result of burrs and noise over the binary image (Igoe et al., 2018). For this, we slide a window size of three pixels over the entire image, pixel by pixel, and numerically sort the pixel values in the window and replace them with a median value of neighboring pixels. This process provides several separate and disconnected bright areas, each of which represents an unidentified wheat ear. Here, we use the regionprops function in Matlab R2017b (Mathworks Inc., Massachusetts, USA) to count the independent regions in the image, which corresponds to counting the number of wheat ears. In addition, we apply the ground truth function to each image, and manually label the wheat ears in the image so as to compare with the result of computer recognition. Criteria to Evaluation Algorithm To evaluate the quality of the segmentation, we use the six indicators Qseg, Sr, structural similarity index (SSIM), Precision, Recall, and the F-measure. The following is a detailed description of the meaning and range of each index (Xiong et al., 2017). Qseg, which is based on both plants and background regions, ranges from 0 to 1. The closer Qseg is to unity, the more accurate is the segmentation. Thus, Qseg reflects the consistency of all the image pixels, including foreground | 3,068,860 | 49666754 | 0 | 16 |
ear part and the background part. Sr represents the consistency of only the ear part of the image. From the perspective of an image, it reflects the completeness of the segmentation results. The SSIM describes the degree of similarity between the segmentation images and the ground truth images. The SSIM ranges from 0 to 1, with higher values indicating more similarity between images. Precision and Recall are the most basic indicators for revealing the final segmentation results. Precision illustrates the accuracy of the segmentation algorithm, and Recall represents the completeness of the segmented images. In practice, Precision and Recall interact with each other. When Precision is high, Recall is low. The F-measure is proposed to balance these two indicators. The higher the value of the F-measure, the better the segmentation results. Table 1 shows how to calculate these indicators. Evaluation criteria Calculation formula Precision (1) and (2) represents a reference set of manually segmented ear pixels (with ω = 255) or background pixels (with ω = 0). a and b give the row and column of the image and i, j give the pixel coordinate of the image. In Equations (9-11), TP, TN, FP, and FN are the number of true positives, true negatives, false positives, and false negatives, respectively. True positives (TP) means when the predicted results and the corresponding ground truth are both wheat ear pixels. True negatives (TN) are when the predicted results and the corresponding ground truth are both background pixels. False positives (FP) are the pixels that are classified as wheat | 3,068,861 | 49666754 | 0 | 16 |
ear pixels, but the ground truth of those pixels is background. False negatives (FN) are the pixels that belong to the ground truth but are not correctly discriminated. RESULTS The performance of the proposed machine learning method is evaluated based on comparing its results against manual measurements. The algorithms were developed in Matlab R2017b. Segmenting a 3500 × 1800 image takes only 0.1 s on average on a Windows 10 PC with 4-core Intel Core i5 processor (2.71 GHz) with 12 GB RAM. For this paper, we separated the image dataset of 300 plots into three categories of equal size with different illumination conditions and show their segmentation results and corresponding ground truths. Results of Several Image-Segmentation Methods We apply three traditional segmentation methods to compare their results with those of the proposed method. The unsupervised methods are the Otsu method, mean shift and normalized cuts (MSNC), and the edge-region active contour model. The Otsu method is a global thresholding method. The Otsu threshold is found by searching across the entire range of pixel values of an image until the intra-class variances are minimized. MSNC first applies the mean shift algorithm to obtain subgraphs and then applies the normalized cut. Currency denomination and detection is an application of image segmentation. The edgeregion active contour model consists of two main energy terms: an edge-region term and a regularization term. This model not only has the desirable property of processing inhomogeneous regions but also provides satisfactory convergence speed (Cheng et al., 2001) (Figure 4). Figure 4 shows the | 3,068,862 | 49666754 | 0 | 16 |
input and output images that there are mainly three cases where these method has not worked properly: (i) pixels labeled as ear actually corresponded to leaves; (ii) contrast between the ear and soil was not great enough and (iii) whereas the algorithm labeled the area as an ear, those pixels are noise. Furthermore, the linear regression between the manual counting and the algorithm counting was calculated for 300 plots with different illumination (Figure 5, Table 2). We see from Figure 5 and Table 2 that the results of the proposed method correlate strongly (R 2 = 0.99) with the manual measurements for all selected images. Moreover, the standard deviation (SD) between the test set is smallest which means that the proposed method is the most stable. But the simple use of the correlation index cannot accurately evaluate the recognition accuracy, so we introduce below more evaluation criteria to verify the performance of these methods. We can draw a conclusion from the Figure 4 that there are mainly three kinds of regions in the image indicating examples where the algorithm has not worked properly: (a) Region 1 shows the case where two ears overlap together and are considered as one; (b) In Region 2, false negatives resulted in wheat ears that were not detected by the algorithm because the contrast between the wheat ear and soil was not great enough and the segmentation algorithm discarded that region; (c) In Region 3, whereas the algorithm labeled the area as a wheat ear, those targets are noise being a | 3,068,863 | 49666754 | 0 | 16 |
result of background brightness caused by a foreign object. Comparing the manual counting results with the statistical results obtained by different segmentation methods gives satisfactory results. To evaluate the segmentation results more comprehensively, six indices were introduced to judge the effect of the segmentation (Figure 6). Figure 6 shows that, compared with other three common methods mentioned in this paper, the proposed method gives the maximum mean value of the six indicators. The mean values of Qseg, Sr, SSIM, Precision, Recall and F-measure (%) are 0. 62, 0.72, 0.82, 0.82, 0.73, and 0.73, respectively. Moreover, Figure 6 shows that the proposed method gives the minimum standard deviation for each evaluation index, which means that it gives the most stable performance with images under different illumination conditions. Results of Segmentation Accuracy With Different Classifiers Differences in selecting the classifier can lead to quite different segmentation precision. To verify the proposed algorithm (TWSVM), we compare it against three well-established algorithms: rough set, Bayesian, SVM (Figures 7, 8). Figure 8 shows that the TWSVM provides better segmentation, and the wheat-ear integrity is well maintained. Except for the TWSVM, the SD of the other three algorithms is relatively large, which reflects their weak adaptability to different field testing images. In addition, the average of Q seg for the proposed algorithm is about 0.626, which is significantly greater than for the other three algorithms. Thus, the proposed algorithm is more consistent for both the panicle foreground part and the background part. In addition, the mean value of the SSIM for the | 3,068,864 | 49666754 | 0 | 16 |
proposed algorithm is greater than that of the other three contrast algorithms. Moreover, the F-measure is a comprehensive indicator and accounts for Precision and Recall; it is as high as 0.738 using our proposed algorithm compared with 0.398, 0.452, and 0.578 for the other algorithms, respectively. These results show that the proposed algorithm accurately FIGURE 6 | Results of evaluating segmentation with different methods. The color columns represent the means value and the black lines represent the standard deviations for the test images. In addition, the color differences between columns refers to the categories of segmentation methods. Blue is for the proposed method, green is for the Otsu method, yellow is for the MSNC, and purple is for the edge-region active contour model. segments the wheat ears and guarantees the integrity of segmentation. Results of Recognition Accuracy With Different Image Features The color feature, texture feature, and EHD feature are optimized to perform the segmentation of images, in Figure 8, the segmentation testing results by using different number of features are given (Figure 9). Results are found by Figure 9 that for each class of wheat ear image, the segmentation accuracy of the proposed algorithm is obviously better than that when using a single feature. And it can be seen that for the selected color feature, texture feature, and EHD feature, each feature has very different segmentation results, which also shows that there is a complementary FIGURE 8 | Comparison of segmentation with different classifiers. The color columns represent the mean values and the black lines | 3,068,865 | 49666754 | 0 | 16 |
represent the standard deviations for the testing images. Blue represent TWSWM, green is for rough set, yellow is for Bayesian, and purple is for SVM. relationship between each feature. After optimizing each feature, the proposed algorithm gives the weight to each feature, each sample constructs a reasonable feature space, and the average precision of the whole image is more than 82%, which is 12.4, 7.5, and 9.2% higher than the recognition accuracy of color feature, texture feature, and EHD feature, respectively. We use Q seg, and S r to judge the segmentation accuracy; the results are given in Table 3. As shown in Table 3, the use of multiple features is more robust against background noise and variations in illumination. As a result, we select the multi-feature method as the optimum technique and compare it with the state-of-the-art vegetation segmentation described herein. Moreover, robust hue histograms (RHH) (color) and Scaleinvariant feature transform (SIFT) (texture) were used as two typical features to participate in comparative experiments to verify the reliability of feature selection (van de Sande et al., 2010;Seeland et al., 2017). Here, CCV and GLCM were replaced with RHH and SIFT in order to test the variation of precision after different combination of features (Table 4). Table 4 could provide a conclusion that the combination of features given in this paper could get better segmentation effect. Although on some indices, for example, Q seg and S r , the proposed feature combination strategy was slightly lower than the match group (<5%). In general, it usually gives | 3,068,866 | 49666754 | 0 | 16 |
more accurate results especially in Precision and F-measure. DISCUSSION To be relevant for high-throughput phenotyping in field conditions, the segmentation algorithms must be sufficiently robust to handle dynamic illumination conditions and complex canopy architecture throughout the entire observation period. We find that the recognition accuracy of the classifiers differs substantially depending on the number of features and the illumination intensity. Here, we analyze how these factors affect the accuracy of the segmentation results, respectively. Effect of Illumination Intensity and Shadow on Recognition Accuracy Analyzing images acquired outdoors is a challenging task because the ambient illumination varies throughout the growing season. Unlike single plants grown in pots in greenhouse facilities, segmenting the vegetation from a field-grown plot is complex because of overlapping leaves and because portions of the canopy are shadowed or have high specular reflectance, each of which contribute to underestimating vegetation pixels in an image. To study the robustness of the method under different illumination conditions, we use the image brightness adjustment function of Photoshop CS6 (Adobe Systems Incorporated, California, USA) to adjust the luminance components. The original image brightness is called the "central value of brightness adjustment, " and the image results of five different luminance conditions are simulated by varying from dark to bright. The results are then associated with the artificial recognition results by using the proposed method to determine how the different illumination conditions affect this recognition method (Figure 10). Figure 10 shows that the segmentation accuracy reaches the highest value under conditions of lower brightness, which corresponds to overcast sky | 3,068,867 | 49666754 | 0 | 16 |
without overexposure or underexposure. We thus conclude that the illumination condition affects the recognition accuracy. Unlike the use of artificial light and enclosures, our flexible and fast image acquisition technique presents some major challenges related to image processing. Sharp shadows and bright surfaces may appear in the images as a product of the light conditions. As such, in order to provide robust results, the image processing algorithm pipeline must consider effects related with shadows. When the training set is set up, the images under different shading conditions are included. The recognition results in Figure 5 and Table 2 show that there is not much difference in the recognition results under different shading conditions. Analysis of Effect of Noise on Recognition Accuracy Noise may be generated through the entire process of the image processing and may be divided into two categories: system noise and environmental noise. System noise is usually caused by the imaging system itself and includes electronic noise and photoelectron noise. Environmental noise is caused by a poor image-acquisition environment and unreasonable image-acquisition methods. The proposed method relies on counting disconnected regions and fitting the obtained number to the manually counted amount of wheat ears via linear regression. So the excessive noise points will increase the error in statistical results. Here, Gauss noise, Rayleigh noise, exponential noise, and salt-and-pepper noise were introduced to test the noise robustness of the proposed method (Figure 11). Figure 11 shows that noise affects the accuracy of segmentation. Specifically, Rayleigh noise and salt-and-pepper noise reduce the accuracy by over 40% | 3,068,868 | 49666754 | 0 | 16 |
whereas the two other types of noise have little effect on the result (<20%). The first two types of noise are denser and larger and are easily mistaken for wheat ears. However, a Median filter or a Laplacian filter can effectively filter out these two types of noise and may be considered for denoising in actual production. Effect of Different Camera Angles and Field-of-View on Recognition Accuracy The performance of the algorithm was further tested through the different camera angles and fields of view. First, the images are taken at six different angles: 90, 75, 60, 45, 30, 15, and 0 • under same light intensity and camera parameters. Then, the center of each image is taken as the center of shooting, and the image is cut at 1/2, 1/4, and 1/8 long sides, respectively, then the imaging results of different fields of view are obtained. We use the same algorithm pipeline proposed for different camera angles and field-of-view images. As before, manual image-based counting is used as the validation data (Figure 12). The different camera angles show, with respect to the original images taken at 45 • , a decrease of over 20% in success rate while the shooting angle is close to 0 • . The interference of leaves and stems and the mutual occlusion between ears make it impossible to get an accurate number near the horizontal position. Meanwhile, the accuracy of image recognition from the vertical angle is also reduced by about 15%. The significant difference in wheat morphology between vertical and | 3,068,869 | 49666754 | 0 | 16 |
oblique observations may at the origin of this result. The different field-of-view results show an increase of 8% in success rate when the images are reduced to 75% of their original size. Success rates increased by a maximum of 13 and 15% for image size divided by 50 and 25% values, respectively. Near the edge of the image, distortion of the wheat ear shape reduces the recognition accuracy. At the same time, other interference factors affect the edge parts, such as noise, which will also affect the final result. In future work, the proper range of field of view should be studied. Analysis of Algorithm Efficiency We use the average running time of each segmentation method as a metric for the efficiency of the algorithm (Figure 13). We conclude from Figure 13 that the average running time of the proposed algorithm is 0.1 s for calculating the number of wheat ears in a single scene, which means that the proposed algorithm is an efficient method. Moreover, the running time increases as the number of wheat ears increases (compare Table 3 and Figure 13A). It seems that the increase in the number of target objects may lead to an increase in the time complexity of the algorithm. Thus, the proposed method may be used as a highthroughput post processing method to measure seeding statistics for large-scale breeding programs. We can also draw a conclusion from the Figure 13B that the most time-consuming step is patch classification by TWSVM. An intuitive improvement to further improve algorithm efficiency would | 3,068,870 | 49666754 | 0 | 16 |
be to parallelize all the procedures. However, such an improvement is likely to be hardware limited (due to the input-output speed of the memory and hard drive). To summarize, the proposed machine learning approach offers the advantage of versatility and can extract the number of green vegetation, such as wheat, maize, etc. Given an adequate training dataset, it could even detect disease or pest symptoms. As already mentioned, the performance of any supervised learning model strongly depends on the training datasets. Therefore, to have a good model, a substantial set of training data is important. Acquiring a training data is time consuming and can be subjective. Our aim is to expand this study by integrating a semi-adaptive approach to semi-automatically generate larger and more reliable training datasets. In addition, we must test the model on more varieties and different crops. CONCLUSION Accurately estimating wheat yield requires accurate statistics of the number of wheat ears per unit area. This is achieved in this study by using a method for automatic segmentation of target plant material in RGB images of wheat ears and by splitting these images into individual targets. The initial step in this proposed method requires minimal manual intervention to generate patches from the original images. This technique is partially verified by comparing its results with those of manual and automated measures of image segmentation. The good correlation between manual and automated measurements confirms the value of the proposed segmentation method. The segmentation performance is evaluated in this way because manual image segmentation is labor intensive | 3,068,871 | 49666754 | 0 | 16 |
and subject to observer bias. Manual inspection of segmented images indicates good quality segmentation in all images. Compared with other approaches, the proposed algorithm provides better segmentation and recognition accuracy. Moreover, this method can be expanded for use in different field environments and with different light intensities and soil reflectance. | 3,068,872 | 49666754 | 0 | 16 |
Insights and Opportunities Offered by a Rapid Ecosystem Service Assessment in Promoting a Conservation Agenda in an Urban Biodiversity Hotspot Regional and global scale ecosystem service assessments have demonstrated the socioeconomic value of protecting biodiversity and have been integrated into associated policy. Local government decision makers are still unsure of the applicability, return on investment, and usefulness of these assessments in aiding their decision making. Cape Town, a developing city in a globally recognized biodiversity hotspot, has numerous competing land uses. City managers, with a tightly constrained budget, requested an exploratory study on the links between ecosystem services and biodiversity conservation within this municipal area. We set out to develop and test a simple and rapid ecosystem service assessment method aimed at determining the contribution natural vegetation remnants make to ecosystem service provision. We took selected services, identified in conjunction with city managers, and assessed these in two ways. First we used an area weighted approach to attribute services to vegetation types and assessed how these had changed through time and into the future given development needs. Second, we did a regulatory and cultural service remnant distance analysis to better understand proximity effects and linkages. Provisioning services were found to have been most severely affected through vegetation transformation. Regulatory services have been similarly affected, and these losses are more significant because regulatory services can only function in situ and cannot be outsourced in the way provisioning services can. The most significant losses were in coastal zone protection and flood mitigation services, both of which will be | 3,068,873 | 53420791 | 0 | 16 |
placed under even greater pressure given the predicted changes in climatic regimes. The role of remnant vegetation in regulating and cultural services was shown to be a significant additional consideration in making the case for conservation in the city. Our rapid assessment approach does not allow for nuanced and individual understanding of the trade-offs presented by individual remnant patches, but is particularly strong in quickly identifying issues, key focus areas, and opportunities provided by this research direction, and thereby serving to facilitate and drive constructive engagement between ecosystem service experts and city planners. INTRODUCTION Ecosystem services are seen as a way of demonstrating the relevance and value of biodiversity to society, and the value of ecosystem service-based approaches is now, at a global scale, becoming well entrenched into national level policy (Seppelt et al. 2011).Ecosystem services assessments, however, have typically been carried out at regional scales and primarily in rural environments.Agriculture, water production, carbon sequestration services, and to a lesser extent cultural services, have formed the primary focus of these assessments (MEA 2005).These assessments have typically been large in scale, time, and people intensive, and thus expensive, as in the case of the Millennium Ecosystem Assessment (2005).Local level assessments of services remain under-explored and the urban context has received limited attention.The urban ecology underpinning this and true engagement with city-scale drivers and their effects, is still an emerging science (Pickett et al. 2001, Cadenasso andPickett 2008).Cities are key to securing long-term global sustainability, so interest in urban ecology is growing (Piracha and Marcotullio 2003).The loss of | 3,068,874 | 53420791 | 0 | 16 |
urban green space also impacts on conservation, particularly where cities and high biodiversity levels coincide.Where this happens an additional dimension is added, and the importance of existing urban nature, its functioning and connectivity, has to be more carefully considered in the planning process (Yli-Pelkonen and Niemelä 2005). The limited urban ecosystem services assessment work to date typically has had a northern hemisphere, developed world bias.Different cities face different issues, and developing cities within biodiversity rich areas face numerous, and specific challenges (Piracha and Marcotullio 2003).These relate to meeting local and global conservation expectations, local service delivery, and navigating the disputed territory between these as they play out around land-use allocation and associated trade-offs.What commonly emerges in developing cities is inefficient resource and land use, frequently with immediate negative environmental consequences and longer term realized negative social impacts (Piracha and Marcotullio 2003).The City of Cape Town in South Africa faces these http://www.ecologyandsociety.org/vol17/iss3/art27/types of difficulties, in particular motivating for biodiversity conservation at sites being promoted for development (Holmes et al. 2008).Although ecosystem services have been valued at the city scale (See De Wit et al. 2009), no spatial assessment of these services has been undertaken so clear links between the ecosystem services and the spatial representation of functioning ecosystems or natural vegetation remnants are lacking.Cadenasso and Pickett (2008) call for ecologists to get involved in city planning with some urgency.There is a pressing need for readily available information and spatial assessment tools for city planners and managers to use to guide, or at least inform, decision making around | 3,068,875 | 53420791 | 0 | 16 |
ecosystem service and biodiversity issues.Cities are typically complex decision making environments (Piracha and Marcotullio 2003) that require novel approaches to inform decision making forums.This paper introduces and applies a rapid assessment tool as a way of conceptualizing or scoping the spatial arrangement of, and temporal engagement with, ecosystem services in Cape Town.We use existing data, informed by the particular circumstances of Cape Town, and under the directive of the city's Biodiversity Management Branch, to determine the value of the approach.We used vegetation type and cover to characterize the ecosystem generating resources, both historically and currently, and then assessed how future land use change might affect services.In addition we examined remnant distance relationships for certain services.Both the findings and the value of the tool itself are considered. STUDY AREA The City of Cape Town is located on the southwestern tip of southern Africa and is situated in the Cape Floristic Region, a globally recognized biodiversity hotspot (Mittermeier et al. 2005) and conservation priority (Underwood et al. 2009).The city occupies about 2460 km², has a population of 3.7 million people, and includes 19 national terrestrial vegetation types, containing an estimated 3250 plant species, of which 190 are endemic (Rebelo et al. 2011, Holmes et al. 2012).The city has varied topography, with mountain ranges in the southwest (Table mountain range) and east (Hottentots Holland and Kogelberg ranges), a low lying central region referred to as the Cape Flats where urbanization is focused, coastal areas on the south and western edges, and agricultural areas in the northeast (Rebelo et al. | 3,068,876 | 53420791 | 0 | 16 |
2011).The urban spatial arrangement is typical of postapartheid cities in South Africa with racially defined spatial planning (Swilling 2010) still evident and aligned with significant wealth disparities.Transforming informal settlements into formal housing is one of the key development challenges (Swilling 2006, 2010, Holmes et al. 2008). Identifying ecosystem services The ecosystem service assessment team worked with the Biodiversity Management Branch of the City of Cape Town to identify the ecosystem services used in this investigation.City managers identified coastal security, water related issues, human well-being, and tourism as broad areas of concern.The managers and the project assessment team worked together to assemble available spatial data relating to ecosystem services, and looked for the potential linkages between the data and the services or issues of interest.The resources available for the analysis and assessment on this project amounted to a total of a one person week (senior scientist level). A simple assessment method was applied to relate ecosystem service values to vegetation types, distances, and land uses.The approach was two-pronged.First we assessed a suite of provisioning and regulatory services in relation to land cover change within the City of Cape Town by contrasting three different land transformation scenarios: completely natural (no transformation), current land use (current levels of transformation), and future possible land use (projected transformation) with four ecosystem services themes, namely agricultural provision, water run-off regulation, groundwater, and coastal zone protection.Under each of these themes one or more ecosystem service indicators or surrogates were examined.This was done to provide an indication of the current state of specific ecosystem | 3,068,877 | 53420791 | 0 | 16 |
services within the city in contrast to the potential maximum level, as well as to highlight how future scenarios of land cover change could be assessed and why this is important.We drew on the approach proposed by Deal and Pallathucheril (2009) who suggest the characterization of the resource and associated services in ways that facilitate modeling the future impacts of land use change on service delivery.Non-natural remnants, although recognized as providing a suite of ecosystem services which may complement or augment those from natural remnants, were combined with formal housing in this analysis.This potentially undervalues the ecosystem service contributions from these land types and is a recognized limitation of the study. Second, because much of the available data related to specific point localities, for example, where natural areas were in relation to potential users or beneficiaries and was not suitable for use in the land cover assessment, we developed an approach to incorporate these data.We demonstrate this by assessing a further suite of services, cultural and regulatory, in relation to their proximity to, and association with remaining natural vegetation remnants.Here we examined tourism sites, cultural sites, schools, and cultivated land associations with natural vegetation remnants. Ecosystem service assessment associated with land cover change Land cover mapping: potential, actual, and future Potential vegetation of the city was mapped using the South African National Vegetation map (Mucina and Rutherford 2006) combined with finer scale city vegetation subtype mapping (A.Stipinovich and P. Holmes, unpublished report).This layer was used as a representation of land-cover precolonial settlement, when ecological states and | 3,068,878 | 53420791 | 0 | 16 |
associated services were assumed to be at their highest potential level. We combined this indigenous remnant vegetation layer with the National Land-Cover 2000 (Van den Berg et al. 2008) to generate what we have called an actual land use or land cover layer for the city (Fig. 1).There were some discrepancies in these layers between what was natural and what was not.This related mainly to very small fragments that the National Land-Cover identified as natural remnants, and these were labeled as 'unknown'.The actual land-cover layer was then combined with each of the ecosystem service layers (listed below) to assign land-use-class status to service areas and service-level weights across land use classes.No ground truthing was performed in this study because the original National Land-Cover 2000 had been ground validated by Van den Berg et al. (2008). We generated an additional hypothetical land-cover layer, based on a single future scenario, both to demonstrate the potential value of an approach that introduces spatially defined future scenarios, and to determine likely changes in ecosystem service levels in the future (Fig. 1).Here we reclassified all natural vegetation remnants in the actual land cover that were not in protected areas or formally managed areas as urban or built-up (formal housing). Measuring change in services Although there are a variety of ways to map ecosystem services and examine changes in these over time, our emphasis on developing a simple and rapid assessment approach led us to focus on vegetation types.This is supported by Pickett and Cadenasso's (2008) argument that the functioning ecology of | 3,068,879 | 53420791 | 0 | 16 |
a city is hinged primarily on the presence of plants and plant communities.Yapp et al. (2010) also used a structural vegetation classification in assessing ecosystem services in Australia. For each service we identified appropriate datasets that enabled us to quantify services.We used an area weighted approach or proportional representation, to assign service values to the vegetation types in our potential vegetation map.This allowed us to determine how service levels had changed over time according to the change in area of the vegetation type associated with actual and future land cover maps.Although we made every attempt to use data with an appropriate resolution, including the city natural vegetation ; and potential future land cover, if all natural remnants, not formally protected, were converted to formal housing (c) for the City of Cape Town.http://www.ecologyandsociety.org/vol17/iss3/art27/remnants layer that was ground truthed for habitat condition, some datasets developed at the national scale had to be utilized, and may only be considered to provide a broad (generalized) overview when used at this scale, a further limitation of our selected approach. Agricultural provisioning services Although provisioning services were excluded from the initial discussions, we have included two of these services in our analysis, to demonstrate historical changes.Furthermore, these services may become increasingly important under conditions of climate change, in which food security may become a more localized concern. Land capability: We used the Land Capability data set (Agricultural Research Council 2002) as an index of agricultural potential within the city.This data set divides the national land surface into eight classes, ranging from 1 | 3,068,880 | 53420791 | 0 | 16 |
= very high potential, to 8 = very low potential. Grazing potential: The relative value of the land for livestock production was estimated by calculating the grazing potential of each vegetation type.This potential was derived from Scholes' (1998) national estimates of sustainable mean domestic livestock production.As Scholes (1998) provided categorical ranges of values we used the midpoint value of each class in assigning these values. Water run-off regulation Soil retention: Natural vegetation is important for stabilizing wind and water erosion prone soils (Gordon et al. 2008).Some forms of land cover also provide a measure of this service such as formal medium-density urban areas, however other land cover classes do not provide this services as effectively, for example, cultivated land when poorly managed (Van Noordwijk et al. 2004).We used the soil erodibility factor defined and mapped by Schulze and Horan (2007).They assigned each soil type (bare soils) within South Africa with an erodibility factor (K), ranging from 0.1 (soils with a low erodibility) to 0.7 (soils with a high erodibility).We developed a soil erodibility data layer for all vegetation types within the city. Critical infiltration zone mapping: Infiltration is a critical factor in rain water capture by the soil and a key factor in reducing overland run-off and flood peaks as well as recharging catchment water storage and sustaining river flows in the dry season (Gordon et al. 2008).There are synergies between this service and the flood mitigation service.We developed a critical infiltration zone data layer using a national rainfall coverage (Schulze 2007) and selecting areas of | 3,068,881 | 53420791 | 0 | 16 |
high and most intense rainfall (800 mm and above) as priority areas for ensuring maximum possible infiltration.All vegetation types that fell within these areas were assumed to have the same weighting. Flood mitigation zone mapping: We developed a flood mitigation zone layer for the city using a number of different spatial datasets.We took the national 1:500,000 rivers (large river systems) of the country and the 1:500,000 wetlands (see Nel et al. 2011), and selected those that fell within the city boundary.We created a 50 m buffer for these rivers.A 1:50,000 rivers dataset (smaller river systems) for the city was buffered by 32 m (DLA-CDSM 2007).The two river layers were merged with the flood prone areas data layer provided by the city, also buffered with a 50 m buffer, to create a single layer, which we call 'the flood mitigation zone.'This is an area that should remain undeveloped so as to allow for flood water spreading, infiltration, and calming.This area also is important for its potential to act as a buffer for improving the quality of water run-off from nonpoint sources of pollution. Coastal zone protection Natural vegetation and dune systems, particularly the foredune, play key roles in buffering the coastline against the impacts of periodic storms (Barbier et al. 2011).We developed a coastal zone protection layer based on the Integrated Coastal Management Act (Republic of South Africa 2008) recommendation of a 1000 m "no development zone" along coastlines, by buffering the city's coastline.We then calculated the areas for all land use classes for each of the | 3,068,882 | 53420791 | 0 | 16 |
three layers to obtain the predevelopment total area by vegetation type in this zone and the degree to which this has changed. Groundwater recharge, yield, and quality Groundwater recharge: Groundwater recharge is critical for sustained river flows in the dry season, for certain vegetation communities, and potential or actual human use for a range of purposes (Colvin et al. 2007, Scanlon et al. 2007).We used the national groundwater recharge data layer from a groundwater resource assessment (DWAF 2005). Groundwater yield: We used city scale groundwater borehole yield data in liters per second as service values (City of Cape Town 2002).The borehole yield is a direct measure of the potential for groundwater abstraction and use. Groundwater quality: We used the groundwater conductivity values (mS/m) to define classes (City of Cape Town 2002).Water quality determines the amount and cost of treatment required to purify it for human use so high quality water is of great value. Integrated service analysis To determine the level of service delivery generated by each vegetation type, and current and future land use or status, an expert group consisting of five senior scientists, all experts in fields associated with the selected services, scored cards of land use types in association with ecosystem services (between 0 -10).Scores were then averaged, and are presented in Table 1.The land use types included here were: natural vegetation in high (good), medium, and poor condition, cultivation, Table 1.Ecosystem services scored for different land use classes based on expert opinion.Scores were rated on a scale from 0 to 10 where | 3,068,883 | 53420791 | 0 | 16 |
0 represented no service and 10 the maximum potential service.forestry plantations, urban industrial, urban formal housing, urban informal housing, urban small holdings, mining and quarries, and unknown.The unknown category was scored as either being cultivated or natural in poor condition, based on firsthand knowledge of a sample of these fragments.The product of the vegetation type ecosystem service value and the land use score was then calculated for each of the nine ecosystem services.This process, in essence, combines the contribution of the original vegetation type with the associated land use and ecosystem service levels, and land use changes and captures their associated ecosystem service effects, for the past, present, and future scenarios.Values were then proportionally weighted between 0 and 1, and mapped according to the current land cover of the city. Pollination service potential We took the natural remnant data layer supplied by the city and created a 250 m gridded surface layer.We then calculated the distance from each grid cell classified as natural to the nearest land class area of cultivated land.We then selected natural remnants that were within 0 -0.5, 0.5 -5, and 5 -10 km based on known solitary and honey bee foraging distances (Beekman andRatnieks 2001, Gathmann andTschamtke 2002).This approach excludes bees that are moved in hives for commercial gain. Cultural features We took the created 250 m gridded surface layer of remnants and combined it with a cadaster layer, of all the heritage sites supplied by the city's heritage department.We calculated the distance from each grid cell classified as natural to the | 3,068,884 | 53420791 | 0 | 16 |
nearest land class area designated as a heritage site.All natural remnants areas within 1 km of a heritage site were selected, assuming that visitors to heritage sites may be interested in natural remnant areas if they were within close proximity or walking distance. Tourism services We acquired a tourism routes data layer from the city's tourism branch that consisted of three known routes with stop-off points within the city.This data layer was combined with the gridded natural remnant data layer and used to estimate the distance from each natural remnant to stop-off points.Using our knowledge we added additional points to one of the routes. We selected all natural vegetation remnants that are within 500 m of tourism route stop-off points.This distance was decided on by the project team and requires testing. Education service potential We took the natural remnant data layer, retaining only areas protected or managed as conservation areas and containing portions of river systems in natural condition and created a 250 m gridded surface layer.A schools database supplied by the city was used to calculate distances between remnants and schools.Remnants that were within a walking distance of 1 km of schools were plotted based on a small sample of known activities of this nature.The factors guiding exclusions may need to be reconsidered, but we excluded areas in a poor ecological state because these are less attractive to schools owing to lower biodiversity values and perceived pupil and staff safety issues. Ecosystem service assessment associated with land cover change The potential or original vegetation cover | 3,068,885 | 53420791 | 0 | 16 |
for the City of Cape Town (Fig. 1a) is considered to provide optimal potential ecosystem service delivery for the suite of selected services.http://www.ecologyandsociety.org/vol17/iss3/art27/ The current land use for the City of Cape Town (Fig. 1b) represents current ecosystem service delivery from natural remnants.The marked topography in parts of the city has focused development and transformation in the lowland areas. The mountainous areas have remained largely intact (Fig. 1b). The future scenario of conversion of unprotected natural vegetation remnants to formal housing further entrenches this pattern because unprotected areas that could be transformed occur mainly in low lying areas (Fig. 1b). There are marked changes in ecosystem service delivery between the original and current potential of the area, as well as projected service delivery under the future scenario described above (Fig. 2).The greatest reduction is in the grazing potential (49%), followed by land capability (32%), and flood mitigation (32%).Further reductions (additional 20%) are anticipated in grazing potential and land capability in the future.The future degree of change in the flood mitigation service may seem relatively small, but it is primarily the lower lying and flatter areas that are more at risk of flooding.The coastal protection zone has also decreased 25% because of land cover transformation, so too groundwater quality (27%), yield (20%), and recharge (20%), and further decreases in these services are anticipated under the future scenario.Soil retention has changed relatively little because most of the area with > 800 mm of rainfall is on slopes too steep to develop, on soils with little agricultural value, or | 3,068,886 | 53420791 | 0 | 16 |
in protected areas.The mapped ecosystem services highlight the degree to which services differ in their nature, but are generally affected by urban and agricultural transformation.An assessment of natural remnants indicates that much of the best agricultural and grazing land within the city has been converted to urban areas, or is under cultivation (Fig. 3a,b).This broad overviewhighlights the inherently low stocking rates of the Cape Flats, mainly because the soils are infertile sands with fynbos that provide low quality grazing.The high capacity in the mountain areas and the southern western peninsula is probably due to the errors in the underlying data. Natural vegetation generally provides the highest level of soil retention, preventing it from eroding and filling storm water systems and rivers with sediments (Fig. 3c).The scarring produced by soil erosion is generally highly visible and detracts from the aesthetic value.The critical infiltration areas within the city (Fig. 3d) play an important role in absorbing large volumes of rain water.This diminishes peak flood flows and is released during the dry season to sustain flows in the river systems.We have highlighted the areas with > 800 mm per year but lower rainfall areas also play a role.The city's flood mitigation zone (Fig. 3e) highlights the rivers systems and the extensive coastal wetlands that once characterized Cape Town (Fig. 1a).Much of this has been filled in and built up such as the coastline of Table Bay.The quality of this service is directly related to land use, with urban industrial areas, informal housing, and cultivation having the most severe effects. | 3,068,887 | 53420791 | 0 | 16 |
The natural vegetation in the coastal protection zone (Fig. 3f) provides an important buffer that absorbs wave induced storm damage.This highlights those coastal areas that have been developed predominantly for housing and industry. The highest groundwater yielding areas are the coarse sands of dunes in the northwestern areas and the southern areas (Fig. 3h).Unfortunately much of the groundwater in the southern region has become polluted, which reduces its value as a water source.The groundwater recharge potential for the city is strongly linked to the rainfall and also to the permeability of the rock formations and the associated soils (Fig. 3g).Sandstone and granite-derived soils have higher recharge potential than shale soils.The low potential on the highly permeable sand of the Cape Flats is largely due to the low rainfall in these areas.The areas with the best groundwater quality coincide with areas of sandstone fynbos on the Cape Peninsula and in the Kogelberg, and on adjoining granite fynbos areas (Fig. 3i).The Atlantis Sand Fynbos in the north also has particularly high water quality and coincides with high yields, indicating the importance of this area for both current and future extraction.The poorest quality groundwater is found on shale associated vegetation types in the east and Cape Flats Sand Fynbos in the west.The urban small holding area in the south stands out as polluted, possibly due to high nitrogen and phosphorous levels from vegetable farming.Informal settlements also appear to cause a decrease in water quality in this area. Remnant location service assessment Although only a small proportion of the remaining | 3,068,888 | 53420791 | 0 | 16 |
natural area is included in solitary bee foraging, the majority of natural remnants are captured by the long distance pollination class or honey bees of up to 10 km from a remnant (Fig. 4a).Natural remnants that are within 1 km of cultural heritage features are highlighted in Figure 4b.A number of large natural remnants have been classified by city officials as cultural sites and this accounts for the extensive nature of this category/class.The relationship between natural remnants and stop-off points on designated tourism routes (Fig. 4c) shows that a number of natural remnants are located relatively close to tourist stopoff points.The spatial extent is constrained by the distance tourists are likely to walk, which we set at 500 m.The relationship between natural remnants and city schools shows key areas situated within the City Bowl and the southwestern Ecology and Society 17(3): 27 http://www.ecologyandsociety.org/vol17/iss3/art27/Cape Peninsula coastal region (Fig. 4d).There are few easily accessible remnants in the low lying Cape Flats and the more extensive and better protected and conserved remnants are far from those schools.The access problem is further exacerbated by the fact that this is a low income area and the schools cannot afford transport.The walking threshold distances proposed here of 1 km to remnants could be adjusted to include other means of transport and their associated costs. DISCUSSION This rapid assessment must be viewed in the context of Cape Town as a recognized biodiversity hotspot (Myers et al. 2000).As such, the assessment set out to establish not whether biodiversity supplies ecosystem services, but rather whether | 3,068,889 | 53420791 | 0 | 16 |
the natural remnants within the city supply provisioning, regulating, and cultural services that could justify their continued existence as well as meeting critical conservation goals.The method adopts a pragmatic, rather than a purely conservation-driven approach by explicitly including the flow of services from various land uses and not just conservation areas as suggested by Piracha and Marcotullio (2003) and Palmer et al. (2004).For example, cultivated lands contribute to infiltration, though to a lesser degree than natural vegetation and this is acknowledged in the weighting of scores.Therefore, although some of these transformed spaces may not contribute directly to the important repository of species to be conserved, they do form part of a larger multifunctional matrix of space in the city, and most certainly contribute to ecosystem service delivery and ecological functioning.In this assessment natural vegetation always scored the maximum amount, but this is unlikely to be the case in all areas, for all services.In some instances transformed areas are likely to surpass natural vegetation, as is the case for crop production. Provisioning outsourced The temporal assessment shows that provisioning services have been critically compromised with the loss of grazing and land capability, largely because of the geographical position of these services on the low-lying lands that are also most readily transformed to housing.With much of the provisioning services now outsourced beyond the city boundary the loss of these services is largely unimportant.For example, water is predominantly sourced from outside the city (Quick 1995, Gasson 2002, as cited in Swilling 2006), and, because urban expansion has reduced available | 3,068,890 | 53420791 | 0 | 16 |
land for agriculture (del Mar Lopez et al. 2001, Anderson andO'Farrell 2012), produce is primarily brought into the city from outlying agricultural areas and further abroad (Gasson 2002, as cited in Swilling 2006).These trends, whereby the provisioning services that support a city are sourced beyond that city boundary, has been noted globally (Folke et al. 1997, Gutman 2007, Grimm et al. 2008).The fact that the areas that supply these services are also not adequately compensated for them is at the heart of many debates about payments for ecosystem services (Gutman 2007).The temporal view taken in this assessment, which shows both potential service delivery and actual service delivery, demonstrates this shift from an emphasis on provisioning to one on regulatory services in the city, in keeping with other historical narratives (Anderson and O'Farrell 2012). Regulatory services in situ Far more critical than the loss of provisioning services, which can be sourced elsewhere, is the substantial erosion of regulating services.Regulatory services are critical to city sustainability, can only be delivered in situ, and are generally of a scale that cannot be readily substituted with engineered infrastructure.The most significant loss has been in coastal zone protection and flood mitigation services.The consequence of the loss of coastal buffering will be felt in the future with predicted climate-related change, where sea-level rise combined with increased storm strengths can increase the risk of extreme wave conditions and consequent damage (Cartwright 2008, Theron et al. 2010, Barbier et al. 2011).Although this is a qualitative model, it suggests that sanctioning any further reduction in | 3,068,891 | 53420791 | 0 | 16 |
reserve land in this coastal protection zone would be unwise.The same can be said for flood plain infiltration and associated flood mitigation.The consequences of the reduction in this service are already felt by many in the low lying areas of the city near rivers.These remaining natural remnants should come under the strictest of land cover change controls to ensure the continued delivery of this mitigatory regulating service.This is particularly important given that the beneficiaries of this service are some of the most vulnerable and economically marginal in the city (Govender et al. 2011, Musungu et al. 2012).Although little information is available on actual pollination services or their importance within the city, our analysis highlights likely areas of importance for this service, according to two different pollinator groups. Cultural dimension, distance, and accessibility The examination of the distance relationships between remnants and broader society shows good potential for building a case based on cultural heritage and educational service potential.The inclusion of cultural services in the form of education and heritage in this study is based on the recognized importance of cultural service provision in the city, where remnant green spaces provide opportunities for urban dwellers to have contact with nature, which has numerous benefits.These include education about the environment and natural or cultural heritage, enhanced mental-health in response to space for recreation and relaxation, improved aesthetics, space to pursue religious or cultural rituals, cushioning of noise and air pollution, and the potential for inclusion in a tourism industry with economic benefits (Yli-Pelkonen and Niemelä 2005). Reflections on | 3,068,892 | 53420791 | 0 | 16 |
the loss of natural land, most pronounced on the lowlands or Cape Flats, immediately highlights the potential value of such remnant patches to neighborhoods and http://www.ecologyandsociety.org/vol17/iss3/art27/their schools.Research has found that urban fragments are of value to inner city schools and residents (Britton and Jackelman 1995, Le Maitre et al. 1997, Ashwell 2010) and education is a viable 'hook' on which to build a case for conservation.This study demonstrates diminishing opportunities for the numerous schools on the lowland areas to access natural land and a growing scarcity of opportunities for the provision of associated cultural services.There are opportunities for forging links between schools and remnant patches with a view to securing the future of biodiversity in these remnants and with positive benefits for the schools (Britton andJackelman 1995, Manuel 2006).The inhabitants hold the key to securing the sustainability of the city and the inclusion of this social dimension in this rapid assessment has proved valuable and warrants greater attention in future. Although, at a broad scale the natural environment most certainly contributes to Cape Town's tourism appeal, on a remnant scale this was not a significant driver.This assessment demonstrates that mass tourism, in its current form, does not present a viable 'hook' or case on which to motivate for the conservation of all remnant green spaces.Some of the remaining remnants are impressive from a scenic and biodiversity point of view, but are off current tourism routes.Strategic marketing campaigns that possibly linked these remnants with township tours may be a plausible strategy.It is also possible that a more | 3,068,893 | 53420791 | 0 | 16 |
specialized and focused form of tourism, targeting biologists, could be viable. Ecosystem services arguments for urban conservation This assessment indicates some very clear opportunities for invoking ecosystem services in motivating natural remnant conservation.Absorption of storm waves, for example, is clearly a critical regulatory service and a strong case can be made for coastal dune protection on this basis.The multiple service approach highlights the potential for building a case for using ecosystem service 'bundles' for the retention of natural remnants and green space, and the value of including a variety of services (Bennett and Balvanera 2007).The use of multiple layers demonstrates that a case can be made for a great number of critical remnants based on some ecosystem service.For example one area may serve as seasonal flood protection and also hold potential education value by its proximity to schools in an otherwise highly built up area.Or an area might not hold any tourism potential, but can be shown to be close to other remnants that warrant attention for the preservation of pollination services.The addition of a cultural service, and associated accessibility layer, supports the use of multiple services to demonstrate the societal value of remnants.It is worth noting that in some instances the ecosystem services argument will not gain sufficient purchase and here conservation must be fought on a different level, speaking to national responsibility and global biodiversity concerns.This method can readily demonstrate where those singular cases might be, and in turn different arguments can be mobilized around these remnants. The spatial arrangement of ecosystem service delivery | 3,068,894 | 53420791 | 0 | 16 |
is clearly presented through this method.The evident erosion of services through time demonstrates a history of urban planning taking place without cognizance of the importance of urban nature in ensuring city sustainability and resilience.Both current and future land use and environmental changes highlight the vital importance of regulatory services.If the scenario of developing all unconserved natural land in the city becomes reality, it could result in significant further loss if we fail to achieve densification and continue with urban sprawl.There is a critical need for spatial planning to engage with the various layers produced in an assessment such as this.Although any urban planning exercise and associated development initiative will require inevitable trade-offs such as ecosystem services in favor of housing, it is important that these trade-offs be informed.This is particularly pertinent to Cape Town, where the natural environment has a high biodiversity irreplaceability value and as a result the city presents a low choice environment where spatial planning decisions must be made under constrained circumstances.The option of matching ecosystem services and conservation targets, used for example by Egoh et al. (2011), cannot be exercised.In this minimal choice environment, prioritizing on the basis of ecosystem services and selecting only those sites of highest values, is potentially problematic from a biodiversity perspective.Although defining biodiversity hotspots may be a valuable approach in conservation prioritization, it is not necessarily a good approach for ecosystem services in the urban setting.In a hotspot such as Cape Town, biodiversity must lead over ecosystem services in building the conservation case. The value of a | 3,068,895 | 53420791 | 0 | 16 |
rapid assessment tool What emerges from this rapid assessment is not a full or complete ecosystem service assessment or inventory.In the same way as a scoping study would precede an environmental impact assessment, we argue that a rapid assessment such as this should precede a full ecosystem service assessment.This rapid assessment technique should be seen as a scoping tool highlighting areas where ecosystem services have been particularly adversely affected.Although this analysis may be limited by available biological and social data, such as tourist walking distance preferences, it is fairly simple to apply.Furthermore its use and implementation do not require dedicated training for users as in the case of other more sophisticated approaches and tools such as INVEST (Daily et al. 2009). The value of a scoping tool is that it can quickly derive the status quo, and simultaneously generate a platform for the formation of a common language and understanding to guide future discussions.In turn, areas of contention or contradictory http://www.ecologyandsociety.org/vol17/iss3/art27/circumstances, common when making ecological and social decisions, are issues quickly brought to light and can be readily addressed (Haila 1995).The development of this initial macro spatial and temporal scale understanding could in turn be used to direct more detailed assessment such as those proposed by Cowling et al. (2008) or foregrounding of various agendas.For example, it would be easy to run a scenario based on a housing development agenda (Turok and Watson 2001), or one to explore the specifics of climate change related impacts (Grimm et al. 2008), direct restoration efforts, or even to guide | 3,068,896 | 53420791 | 0 | 16 |
experimental design (Felson and Pickett 2005). In the same way this tool can highlight issues and provide opportunities for focusing future work, it can also serve to demonstrate who the players are, or should be.The tool requires expert input but finds its relevance only in the hands of practitioners; as a result its success is hinged on both successful discipline-driven interdisciplinary and issue-driven interdisciplinary engagement (Max-Neef 2005, Robinson 2008).Of particular significance to its success would be engagement among the various departments within the city.It is hoped that a rapid assessment such as the one presented here could serve as a vehicle to forward both this type of engagement, and in turn, in driving discussions toward innovative new solutions deemed critical to sustainability and resilience (Parnell et al. 2009). It must be acknowledged that South Africa has well-developed spatial datasets.This is not the case for most developing countries, and in these countries assessments of this nature would have to rely on globally produced datasets.These are often at coarse scales with poor resolution resulting in services being either grossly under or overestimated for these parameters being measured.Furthermore, rapid assessments as described here could in fact become lengthy processes when datasets need to be secured, or worse still developed, and in these instances alternative methods should be sought for assessing ecosystem services. CONCLUSION This study highlights the current and future potential service extent and changes within Cape Town based on natural vegetation and transformation.It demonstrates a useful rapid ecosystem services assessment method for understanding ecosystem services in an urban | 3,068,897 | 53420791 | 0 | 16 |
context at the city scale.It highlights both the value of ecosystem services to the city, in particular regulatory services, and shows where services are being eroded.The method shows potential to generate scenarios and assess specific ecosystem services, for example with further loss of specific vegetation types.The remnant distance analysis shows the value of considering multiple ecosystem services in building conservation arguments.Our scale of assessment does not allow for the nuanced and individual understanding of associated trade-offs presented by individual patches, and although it could be adapted to do so this would typically require more time and resources.Last, this method has the potential to facilitate and drive constructive engagement between ecosystem service experts and city planners, a crucial condition for urban sustainability. Fig. 1 . Fig. 1.Past, potential vegetation cover (a); present, current land-cover (b); and potential future land cover, if all natural remnants, not formally protected, were converted to formal housing (c) for the City of Cape Town. Fig. 2 . Fig. 2. Changes (present) and potential changes (future) in ecosystem service supply shown as a percentage of the potential service produced. Fig. 3 .Fig. 4 . Fig. 3. Maps of current ecosystem services based on individual service values associated with vegetation types and land transformation.Services values are weighted between 0 and 1 and plotted using Jenks natural breaks.Service include a) Land capability, b) Grazing provision, c) Soil retention, d) Critical infiltration, e) Flood mitigation, f) Coastal protection, g) Groundwater recharge, h) Groundwater yield, i) Groundwater quality. | 3,068,898 | 53420791 | 0 | 16 |
Validity of Washburn’s equation in sericin treated polyester fabric Wickability of fabrics has become an important test as it discloses information on comfort, dyeability and usefulness as a sportswear. A number of papers on the wickability of yarns and fabrics have been published and reviews have appeared [1]. The role of water in transporting moisture has been appreciated for a very long time. A considerable amount of work has been done on the application of sericin to polyester and cotton fabrics with a view to conferring antimicrobial property to them [2–4]. From the papers published it is found that wickability test, although was performed on the fabrics, has not been studied in detail. Wicking is the spontaneous transport of a liquid driven into a porous system by a capillary force [5]. Wicking height is proportional to root of time. Lucas-Washburn equation, which is a very popular one, includes properties such as surface tension, radius of the capillary, contact angle and viscosity of the liquid which has been used to study wickability. It is reported that the weft density pore size and the arrangement of void spaces in fabric have a significant effect on the wicking performance [6]. It is also reported that the motion of liquid in the void spaces between fibers in a yarn impacts the mechanism of fabric wicking critically [7]. It is found that the rate of movement of liquid is governed by the fibre arrangement in yarn which control the capillary size and continuity [8]. Validity of Washburn’s equation can be checked | 3,068,899 | 201239749 | 0 | 16 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.