Functional repair of embolized vessels in maize roots after temporal drought stress, as demonstrated by magnetic resonance imaging


  • Ilja Kaufmann,

    1. Lehrstuhl für Experimentelle Physik 5, Physikalisches Institut, Universität Würzburg, Am Hubland, D–97074 Würzburg, Germany;
    Search for more papers by this author
  • Thomas Schulze-Till,

    1. Lehrstuhl für Experimentelle Physik 5, Physikalisches Institut, Universität Würzburg, Am Hubland, D–97074 Würzburg, Germany;
    Search for more papers by this author
  • Heike U. Schneider,

    1. Lehrstuhl für Biotechnologie, Biozentrum, Universität Würzburg, Am Hubland, D–97074 Würzburg, Germany;
    Search for more papers by this author
  • Ulrich Zimmermann,

    1. Lehrstuhl für Biotechnologie, Biozentrum, Universität Würzburg, Am Hubland, D–97074 Würzburg, Germany;
    Search for more papers by this author
  • Peter Jakob,

    1. Lehrstuhl für Experimentelle Physik 5, Physikalisches Institut, Universität Würzburg, Am Hubland, D–97074 Würzburg, Germany;
    Search for more papers by this author
  • Lars H. Wegner

    1. Lehrstuhl für Biotechnologie, Biozentrum, Universität Würzburg, Am Hubland, D–97074 Würzburg, Germany;
    2. (current address) Plant Bioelectrics Group, Karlsruhe Institute of Technology, Campus North, Building 630, Hermann-v-Helmholtz Platz 1, D–76344 Eggenstein-Leopoldshafen, Germany
    Search for more papers by this author

Author for correspondence:
Lars H. Wegner
Tel:+49 7247 824302


  • • Xylem sap under high tension is in a metastable state and tends to cavitate, frequently leading to an interruption of the continuous water columns. Mechanisms of cavitation repair are controversially discussed.
  • • Magnetic resonance (MR) imaging provides a noninvasive, high spatial and temporal resolution approach to monitor xylem cavitation, refilling, and functionality.
  • • Spin density maps of drought-stressed maize taproots were recorded to localize cavitation events and to visualize the refilling processes; c. 2 h after release of the nutrient solution from the homemade MR imaging cuvette that received the root, late metaxylem vessels started to cavitate randomly as identified by a loss of signal intensity. After c. 6 h plants were rewatered, leading to a repair of water columns in five out of eight roots. Sap ascent during refilling, monitored with multislice MR imaging sequences, varied between 0.5 mm min−1 and 3.3 mm min−1. Flow imaging of apparently refilled vessels was performed to test for functional repair. Occasionally, a collapse of xylem vessels under tension was observed; this collapse was reversible upon rewatering.
  • • Refilling was an all-or-none process only observed under low-light conditions. Absence of flow in some of the apparently refilled vessels indicates that functionality was not restored in these particular vessels, despite a recovery of the spin density signal.


Severe drought stress leads to a gradual water loss of plant tissues and, in turn, to a loss of turgor pressure that can result in wilting and, eventually, in irreversible damage of tissue cells. Desiccation is retarded by stomatal closure. Previously, detailed studies have been performed on the effect of drought stress on stomatal conductance, turgor pressure and leaf water potential in various crop species including maize (Premachandra et al., 1992; Fromm & Fei, 1998; Tuberosa et al., 2002) and on the responses of these parameters to rewatering (Grams et al., 2007). Turgor pressure loss is accompanied by the development of more negative pressures in the xylem vessels (Schneider et al., 1997a,b; Wegner & Zimmermann, 1998). Xylem sap under negative pressure is in a metastable state; local input of energy, substantial lowering of the extremely high activation energy barrier for vapor bubble formation or air seeding result in rupture of the tensile water column. Xylem walls are partly hydrophobic (Laschimke, 1989; Zimmermann et al., 2004), a feature that compromises wettability of the surface and favors cavitation events. Cavitation frequency is also strongly correlated with xylem tension. When a vessel is embolized following a cavitation event, this will ultimately lead to a loss of function of that particular vessel as a pathway for long-distance water transport. Consistently, negative pressure values down to c. −0.6 MPa (corresponding to tensions of 0.7 MPa) could only occasionally be recorded in very few xylem vessels of herbaceous plants subjected to osmotic stress and drought (Zimmermann et al., 2004). Cavitation thresholds have been quantified by measuring acoustic emissions in intact plants, by measuring the sudden loss of hydraulic conductance of cut branches when subjected to air-dehydration and/or by measuring the loss of hydraulic conductance of cut branches as a function of air pressurization of the xylem (Tyree & Sperry, 1989; Tyree, 1997). However, these methods are based on a questionable physical background, as discussed in detail elsewhere (Zimmermann et al., 2004). A more elegant approach is that of Canny and others (Canny, 1997; McCully et al., 1998; McCully, 1999; Pate & Canny, 1999; Canny et al., 2001; Facette et al., 2001). These authors have investigated diurnal emptying of vessels and subsequent refilling, which has also puzzled many plant scientists, by rapidly freezing stems and roots of transpiring plants and subsequently imaging the air/ice distribution in the xylem. The drawbacks of this method are that it is destructive and that gas bubble formation may artificially occur owing to freezing of xylem water under tension (Cochard et al., 2000; Canny et al., 2001; Richter, 2001).

High-resolution 1H magnetic resonance (MR) imaging (for the basics and application to plants, see Köckenberger, 2001; Van As, 2007) is the only method that provides a nondestructive alternative for measuring the dynamics of the water content at the level of an individual xylem vessel of intact transpiring plants (Kuchenbrod et al., 1995; Schneider et al., 2000a, 2007; Holbrook et al., 2001; Zimmermann et al., 2002, 2004; Clearwater & Clark, 2003; Scheenen et al., 2007). Spin density (SD)-weighted and T1-weighted 1H MR images of well-watered and water-stressed plants have shown that the number of embolized vessels increases under water shortage conditions (Holbrook et al., 2001; Clearwater & Clark, 2003; Zimmermann et al., 2004; Scheenen et al., 2007; Schneider et al., 2007) and decreases again upon resupply of water (Holbrook et al., 2001; Scheenen et al., 2007). However, SD-weighted and T1-weighted 1H MR images do not allow conclusions about the functional state of the vessels. This is only possible if such measurements are combined with noninvasive bulk flow measurements in the vessels (Kuchenbrod et al., 1996, 1998; Rokitta et al., 1999a,b; Schneider et al., 2000a; Scheenen et al., 2002, 2007; Zimmermann et al., 2004). Furthermore, elucidation of the processes resulting in cavitation and leading to refilling of vessels upon resupply of water also requires spatial and temporal resolution of the water dynamics in the surrounding tissue.

In this communication we present high-resolution SD-weighted and flow-weighted 1H MR images acquired for the roots of intact 7- to 15-d-old maize (Zea mays) plants that underwent drought stress and subsequent rewatering.

Cavitation and refilling of individual metaxylem vessels was monitored with a high spatial and temporal resolution. Functional recovery of previously embolized vessels was verified by flow imaging. Moreover, a new method was designed, based on a multislice MR imaging sequence, to monitor the velocity of sap rise during refilling.

Materials and Methods

Magnetic resonance imaging – experimental setup

All experiments were performed using vertical-bore Bruker AMX or Avance 500 spectrometers (Bruker Bio-spin, Rheinstetten, Germany) (field strength 11.75 T) equipped with a 660 mT/m microscopy gradient system providing a usable inner diameter of 40 mm. Figure 1 shows a sketch, as well as photographs, of the home-built probehead used in this study. In addition to a 1H radiofrequency (RF)-channel, the probe comprises a climate chamber system (Kuchenbrod et al., 1996, 1998; Olt et al., 2000) with sensors for monitoring of temperature, relative humidity and illumination. Simultaneously, transpiration and CO2 assimilation rates of the whole plant were measured by means of a volumetric gas exchange apparatus with an infrared gas analyser (Zentraleinheit CMS400; Firma Walz, Effeltrich, Germany). The plants were illuminated during the experiments by means of a 50 W halogen reflector lamp (Decostar 35; Osram, Munich, Germany) in combination with an IR-filter (KG1, Schott, Wiesbaden, Germany) on top of the magnet. The light was guided into the probe through a paper tube, the inner surface of which was covered with aluminum foil. The tube had the same inner diameter as the chamber of the probe, resulting in a much more homogeneous light distribution than obtained by illumination by fiber optics. The light intensity could be increased up to 800 µmol photons m−2 s−1 using an iris. Temperature and humidity within the climate chamber were 19–22°C and 60–80%, respectively.

Figure 1.

Detail of the magnetic resonance (MR) imaging setup including probehead. (a) Diagram showing the probe (1) with inserted sample (glass cuvette (2) with maize plant and reference capillary (3)) and attached radiofrequency (RF)-coil (4), air supply tubes (5) with humidity sensor (6), temperature sensor (7), glass fiber (8) that is attached to a photo diode (at the lower end – not depicted). The shoot is encased by a climate chamber (9) comprising an RF-shield and closed by a glass lid (10) at the upper end. Light within the MR magnet is provided by a metalized tube (11). The horizontal arrow indicates the slice of imaging in the center of the RF-coil. (b) Photograph showing the probehead without climate chamber. (c) Photograph showing the same part of the setup after closing the climate chamber.

The plants were grown in 35-cm long glass MR cuvettes that were custom-designed for MR microscopy on plant roots. The lower 30-cm-long part of the cuvette had an outer diameter of 15 mm and an inner diameter of 12 mm. The outer and inner diameters of the upper part of the cuvettes were reduced to 5 mm and 3.5 mm, respectively, in order to fit the MR imaging tube into the Helmholtz coil (r = 5 mm) used for application of RF pulses and for signal reception. The cup-shaped region at the top (height 2.5 cm, outer diameter 2 cm) served as a reservoir for the growth medium.

Plant material and treatment

Seeds of Z. mays (L.) cv. Bangui were germinated in darkness on damp tissue paper in a covered platter at 22–26°C. After 3–5 d plants were transferred to the MR imaging cuvettes filled with constantly aerated Long Ashton growth medium (9 mosmol kg−1; Hewitt, 1966). Subsequently, the cuvettes with the plants were placed in a growth cabinet (15 h light : 9 h dark regime; light intensity was 800 µmol photons m−2 s−1 using a mercury-vapor lamp; relative humidity (RH) 25–35%; temperature 25–27°C). The culture medium was replenished on a daily basis. The roots continued to grow down into the MR imaging cuvettes through the area of reduced diameter. During cultivation roots were kept in the dark by wrapping the cuvettes in aluminum foil. For MR imaging experiments, 7- to 15-d-old plants with main root lengths of 20–35 cm and several lateral roots were used. The MR microscopy was performed on 12 plants 3–4 cm below the seeds. The nutrient solution was not aerated during the MR imaging experiments to avoid perturbation of the measurements by rising air bubbles and associated root movements. Plants 1–7 were grown during February and March 2004, plant 8 was grown in March 2005 and plants 9–12 were grown in November and early December 2005.

Magnetic resonance imaging methods

In this study, only spin echo sequences were used for MR microscopy. With these sequences, disturbances (artifacts) caused by flaws in magnetic susceptibility within the inhomogeneous root tissue could largely be avoided.

Spin-lattice relaxation time, T1, and SD maps were determined with the AMX using a magnetization-prepared MR microscopy sequence (Haase et al., 1993). To this end, a saturation recovery pre-experiment (Evelhoch & Ackerman, 1983; 12 different relaxation time steps varying from 8.8 ms to 4.0 s) was combined with a centrically reordered 8-echo train rapid acquisition with relaxation enhancement (RARE) sequence (Hennig et al., 1986; for details on imaging contrast see also Mulkern et al., 1990). This sequence was especially optimized and implemented for functional MR-microscopy on roots, providing a relatively short imaging duration of 14 min while reducing T2 weighting (effective echo time was 9.1 ms) and avoiding susceptibility artifacts. With a field of view (FOV) of 5 × 5 mm and 128 × 128 pixels the isotropic in-plane resolution was 39 µm. The slice thickness was adjusted to 0.5 mm. For each image SD values were normalized to the mean value measured in the reference capillary, to compensate for variations of the signal sensitivity during the course of the experiments.

Magnetic resonance flow imaging was performed using a stimulated echo for flow weighting (Bourgeois & Decorps, 1991) followed by a spin echo imaging sequence; detailed discussion of this flow imaging method can be found in Kuchenbrod et al. (1998) and Rokitta et al. (1999a). The echo time and the repetition time were 10.6 ms and 1.1 s, respectively. The acquisition time for one flow image was 19 min. The slice thickness was set to 4 mm for a sufficient signal-to-noise ratio. For calculation of the images the 128 × 32 complex data points were zero filled, yielding image arrays of 128 × 128 pixels to make them comparable to the T1 and SD maps. In addition, in some experiments flow velocity maps were acquired at the beginning and at the end of the entire experiment with 128 × 128 pixels yielding a real in-plane resolution of 39 × 39 µm, without zero filling. Since T2 only had a minor effect – if any – on SD maps and the temporal changes during drought stress (data not shown), T2 correction was generally omitted in favor of a higher time-resolution of SD imaging.

On the Avance instrument, a multislice spin echo sequence was used for additional studies with a higher time-resolution to resolve the dynamics of the rise of the sap level in embolized vessels during refilling. Three slices (thickness 0.5 mm, gap 2.0 mm) were acquired simultaneously with an in-plane resolution of 39 × 39 µm (FOV 5 × 5 mm). An echo time of 7.3 ms and a repetition time of 2.0 s resulted in mainly SD-weighted images. The total scan time for two averages was 8.9 min.

Measurement of the velocity of sap ascent during refilling of embolized vessels

From the multislice spin echo imaging experiment performed with the plants 9 and 12 the rise velocity (vrise) of the sap level within an embolized vessel during the refilling process could be calculated. The method introduced here is based on the assumption that the rise velocity is almost constant, at least while the sap level is passing the three virtual MR imaging slices. The variable for describing the sap rise time through one slice is trise, with trise = 0 when the sap reaches the lower edge of the slice. The percentage F(trise) of each xylem voxel that is filled with sap is a function of trise, which is defined in three sections:

image(Eqn 1)

where d = 0.5 mm is the slice thickness. Considering three slices (lower, middle, upper) at slice positions p = {−2.5 mm, 0 mm, +2.5 mm} (Fig. 2) that were filled consecutively we can specify a time translation between trise for each slice and the experimental time t:

Figure 2.

Mathematical analysis of the rise time in individual xylem vessels using multislice imaging. The signal intensity S(p,T) depends on the rise of the sap over three virtual imaging slices: In the sketched example the xylem sap (dark tinted area) has already passed the lower slice 1 at p1 = −2.5 mm (not shown), moves within the middle slice 2 at p2 = 0.0 mm and has not yet entered the upper slice 3 at p3 = +2.5 mm. The thickness of each slice was d = 0.5 mm; trise describes the rise time of the sap within a single slice, starting with trise = 0 when the sap reaches the lower edge of each slice.

image(Eqn 2)

where t0 defines the point in experimental time when the sap level reaches the center of the middle slice. With trise(p,t) the stated function F describes the filled voxel percentage for each slice depending on the ‘general’ time coordinate t. The MR signal intensity that is achieved within the xylem during refilling should be described for each slice position and each data acquisition. Thus, T and ΔT were introduced describing the acquisition start time and acquisition duration, respectively. Calculating the definite integral of F(p,t) for each imaging process and dividing the result by ΔT, the duration of the imaging process, results in

image(Eqn 3)

S(p,T) describes the total signal intensity acquired in a xylem pixel during one imaging process, depending on the slice position p and the acquisition start time T. The value of S(p,T) ranges from 0 for empty to 1 for voxels that were completely filled with sap during the whole acquisition of one image.

The weighting function G(t – T) has to be taken into account. It describes the variation of the sensitivity of the imaging sequence during the acquisition interval [0, ΔT]. This time dependency results from different contributions of the single phase encoding steps to the total signal intensity of a certain image feature. In the case of linear phase encoding (as used here), G is the Fourier transformation of the image feature in question. For a spot that is two pixels in diameter, such as the xylem vessels investigated, G corresponds to the center maximum of a sinc function.

A fit of the two-dimensional function S(p,T) vs the signal intensity measured for all three slices of one xylem vessel (Fig. 2), results in the rise velocity vrise and the refilling time t0 of the center slice.

Experimental procedure

A single cuvette with one plant was transferred from the growth cabinet to the MR probehead. Preparation and mounting of the sensors around the plant (see Fig. 1) took about 30 min. During this period the plant was exposed to laboratory light only. Once placed within the magnet the plant was allowed to adapt to the specific environmental conditions (see earlier) during system calibrations for c. 1 h.

Plants 1–8 The plants were exposed to drought stress for c. 6 h and were then rewatered at the low light level (< 1 µmol m−2 s−1). For plants 1–7, an alternating sequence of T1 and flow velocity maps was applied resulting in a time-resolution of 33 min. Sometimes, slightly longer time-intervals occurred that resulted from the necessity to adjust the MR sequence parameters after draining the nutrient solution. Additional acquisition of T2 maps (Kaufmann et al., 2008) reduced the time resolution for plant 8 to 47.5 min.

When the first set of MR maps was completed the nutrient solution was released from the cuvette. The usual day–night cycle of the light regime continued. Thus, after 2.7–5.1 h of desiccation, depending on the starting time of the experiment, the photon flux density was reduced from c. 800 to < 1 µmol m−2 s−1. This time shift between a light–dark cycle and drought treatment did not result in any detectable differences in the reactions of the plants during drought stress or recovery. The plants were rewatered after 5.7–6.9 h of drought stress by refilling the cuvettes with nutrient solution up to c. 3 cm below the site of measurement. Five to eight hours after rewatering the low-light period ended and a light regime of 800 µmol m−2 s−1 was re-established, with the exception of plants 6 and 7, which stayed in the dark throughout the experiment.

Plants 9–12 These plants were also exposed to 6 h of drought stress, but rewatering occurred during exposure to the elevated light intensity. The time resolution for the multislice imaging experiments was 8.9 min. The plants were illuminated at c. 800 µmol m−2 s−1 for 15.0 h. The light intensity was kept at this value when nutrient solution was resupplied and was reduced to < 1 µmol m−2 s−1 c. 9 h later. Additional flow velocity maps were acquired before the beginning and at the end of the complete experiment (in-plane resolution 39 × 39 µm, 128 × 128 pixels, no zero filling).


Interpretation of virtual root cross-sections obtained by SD maps

Virtual cross-sections of roots were obtained by preparing SD images. Figure 3(a) shows a SD map of roots in an MR imaging cuvette at the site where the RF coil was placed. The contours of the reference capillary filled with nutrient solution, of the thick tap root (diameter c. 1 mm) and of > 10 lateral roots and/or shoot-borne roots immersed in nutrient solution can be clearly identified. Anatomical features of the root are reflected by differences in free water content between different tissues resulting in differences in SD. Figure 3(b, right half) shows a SD image of a well-watered maize tap root (plant 7) at an enlarged scale under control conditions (i.e. before drought stress was imposed). On the left side, part of an image of a free-hand cross-section of a tap root obtained by light microscopy is depicted. The most conspicuous anatomical features in the root cross-sectional SD images are the late metaxylem vessels (diameter c. 50–70 µm) that stand out as bright spots in the root center. The signal intensity values of these spots are similar to those measured in the solution surrounding the root indicating that they are filled with a dilute medium. The inner ring of late metaxylem vessels is encircled by an outer ring of vascular tissue that can be identified in the MR image as a bright ring-shaped structure. Early metaxylem vessels and phloem tissue remain indiscernible in this area, and individual cavitation events cannot be detected. Peripheral to the stele, the root cortex is visible.

Figure 3.

Magnetic resonance (MR) microscopy of maize roots and identification of anatomical structures. (a) Virtual cross-section (MR spin density (SD) map) of the glass cuvette containing a maize seedling at the site where the radiofrequency (RF)-coil is positioned (compare Fig. 1a,b). The cuvette is filled with nutrient solution (1). Inside, the reference capillary (2), the main (primary) root of the seedling (3), a secondary root (4) and several lateral roots (5) can clearly be discerned. (b) Cross-section of a main tap root arranged from a light micrograph that was obtained from a freehand cross-section (left half) and from an MR micrograph of a well-watered root (SD map; right half). Arrows indicate the position of late metaxylem elements filled with xylem sap that can be identified as bright spots. Bars indicate stele (S) and cortex (C).

Drought-stress induced cavitation of individual xylem vessels and subsequent refilling as visualized by MR imaging

Maize seedlings were subjected to c. 6 h of drought stress by release of the nutrient solution from the MR imaging cuvette. Preliminary experiments had shown that plants had started to wilt after 6 h of drought stress. Seedlings were rewatered in the dark (n = 8) or in the light (n = 4).

The effect of temporal water deprivation on the transpiration rate is depicted in Fig. 4 for a representative seedling (plant 2). About 1 h after release of the nutrient solution from the MR imaging cuvette, the transpiration rate of the maize plant started to decrease from c. 2 µmol H2O s−1 to < 0.5 µmol H2O s−1. A further, rapid drop of transpiration to c. 0.1 µmol H2O s−1 was observed when irradiation was reduced from c. 800 to < 1 µmol m−2 s−1, conforming to the standard diurnal light regime. Resupply of nutrient solution had only a small effect on transpiration. The transpiration rate increased to c. 0.6 µmol H2O s−1 when the initial light regime was restored.

Figure 4.

Transpiration of a maize seedling as influenced by temporal drought stress (plant 2). The dotted lines mark the release of nutrient from the cuvette and subsequent refilling, respectively. The tinted area indicates exposure to the low light regime (< 1 µmol m−2 s−1). Before and afterwards, the photon flux density was adjusted to c. 800 µmol m−2 s−1. The spike at t = 2.5 h results from a baseline check of the gas exchange measuring device during the experiment.

In the absence of the nutrient solution the main root shrunk and started to shrivel. Upon rewatering, the root started to recover, ending up at 94–102% of its initial diameter. In some of the plants, the root regained its initial morphological appearance (n = 3 out of plants 1–8), whereas in others the root was apparently irreversibly damaged by the drought treatment, as revealed by visual inspection. In these roots, drought stress caused a (local) deformation of the root (and a nonuniform root diameter) that, together with a color change from beige to brown, remained several hours after rewatering. Development of shoot-borne adventitious roots occurred with these plants when culturing was continued for several days after the experiment was terminated.

In Fig. 5a, a sequence of images is shown that was recorded on plant 2 during a drought-stress experiment at intervals of 33 min (compare transpiration data for the same seedling in Fig. 4). The time indices denote the start of data acquisition for the particular image. In the first (control) image, acquired at well-watered conditions, six late metaxylem vessels can clearly be identified as bright spots. Subsequently, the nutrient solution was removed from the cuvette at t = 0 : 00 h. Within the first 2 h of water deprivation, SD maps revealed little change in root water status apart from some shrinkage of the root diameter. A local water film still remained between the surface of the root, the reference capillary and the glass cuvette owing to capillary force, as can be seen in the image taken at t = 0 : 28 h after the release of the nutrient solution. The film disappeared in the following scan. In the image taken at 2 : 17 h, three dark spots (corresponding to low signal intensity in about four pixels each) emerged in the center of the root (left half). From their size and position within the virtual root cross-section, these local losses in signal intensity can clearly be assigned to late metaxylem vessels (note that the signal intensity is affected by partial volume effects if part of the pixel area is outside the vessel). They apparently represent spontaneous cavitation of three of these vessels. Note that in the same image three late metaxylem vessels had ‘disappeared’ in the right half without any transition to low signal intensity indicating cavitation. These vessels apparently collapsed, perhaps because of increased tension. During subsequent root drying the number of embolized vessels remained constant. Upon resupply of nutrient solution at 5 : 49 h, the root diameter started to increase again, but the filling status of the late metaxylem vessels remained unchanged for several hours. The ‘disappeared’ vessels recovered and became detectable by MR imaging again at 8 : 01 h, i.e. 2 h and 12 min after the root was rewatered. Interestingly this was associated with a slight increase in the transpiration rate (see Fig. 4; c. 08 : 00 h). Only at t = 10 : 46 h (i.e. 4 h and 57 min after rewatering) one black spot representing an embolized vessel turned into an area of high signal intensity, indicating refilling of this vessel. After another 33 min, the two remaining embolized vessels were refilled; apparently all late metaxylem vessels contained xylem sap again from this time onwards.

Figure 5.

Cavitation and refilling of late metaxylem vessels induced by temporal drought stress: monitoring by spin density (SD) mapping and flow imaging. (a) Series of SD maps of a main root of a maize seedling (plant 2). The top left image shows the complete cross-section of the cuvette at the site of measurement; only the root (see frame) is shown in subsequent images. Nutrient solution was removed at time index 0 : 00 h and resupplied at 5 : 45 h. Three cavitations (as identified by a strong decrease in local SD in the root center) appeared at 2 : 17 h and were refilled between 10 : 13 h and 11 : 19 h. Note that the time indices denote the start of data acquisition for the particular image. The SD (given in relative units; r.u.) was normalized to the mean value measured within the reference capillary. (b) Flow velocity (vflow) maps superimposed on the SD maps. No flow was detected during drought treatment (not shown), but flow was restored when roots were supplied with nutrient solution again. Note that flow is also detected in previously cavitated vessels after SD maps indicate refilling. In two vessels (left side of the stele), flow ceases at the end of the experiment even though vessels still appear to be filled with liquid at the site of measurement, indicating the occurrence of cavitations outside the measuring plane.

The filling status of xylem vessels, as revealed by SD maps does not provide sufficient evidence that these vessels can serve for long-distance transport, since vessel embolism below or above the 0.5-mm thick slice of imaging may remain undetected (Zimmermann et al., 2004; Clearwater & Goldstein, 2005). Selected SD images with superimposed flow velocity maps are depicted in Fig. 5b. The leftmost image shows the initial situation under well-watered conditions. All stelar pixels display a flow of 0.7–0.9 mm s−1, indicating that all vessels were conducting under these conditions (including those that had not cavitated during the course of the experiment). The flow velocity was in the order of magnitude reported previously for transpiring maize plants (Kuchenbrod et al., 1996). Note that flow ceased completely 1 h after drought stress treatment had started. The next panel (t = 8 : 34 h) shows a flow velocity map recorded after termination of exposure to drought (i.e. 2 h and 45 min after rewatering). Flow detected by MR imaging was restricted to vessels that retained a high signal intensity in SD maps (located in the right half of the root image), providing evidence that these vessels still contributed to long-distance transport. Note that the average flow velocity remained below the initial value because the light intensity had been lowered from c. 800 µmol m−2 s−1 to a value below 1 µmol m−2 s−1 (it has to be kept in mind that plants were exposed to their usual light regime during the drought experiment). The following flow velocity map depicted here was taken 2 h later when one embolized vessel had apparently been refilled. Consistently, flow could also be detected in this refilled vessel that had apparently resumed full functionality. When high SD signal intensities had been restored in all vessels 70 min later, the conducting area also had increased again, although the initial extent of the conducting area was not re-established. Flow velocities started to increase again when the initial light intensity was applied (at t = 14 : 49 h), but the conducting area decreased again because flow stopped in two of the vessels even though these vessels remained filled with xylem sap at the site of the measurement. Interestingly, flow recovered in most seedlings (n = 4 out of 5 seedlings that showed cavitations and subsequent refilling), independent of the degree of damage to roots by drought stress assessed by visual inspection (see earlier).

Figure 6a shows a summary of the dynamics of vessel cavitation and refilling as obtained from SD maps in eight seedlings subjected to the same protocol as documented in Fig. 5 (seedlings 1–8; see also Table 1). Cavitation was observed in six of these seedlings. Note that nearly all cavitation events occurred in a narrow time-period c. 2 h after water had been released from the cuvette. In four cases (plants 1, 2, 4 and 5) it was observed that areas of high signal intensity identified as a late metaxylem vessel ‘disappeared’ during drought stress treatment without any indication for a cavitation (compare Fig. 5). These vessels recovered and became detectable by imaging again after the root had been rewatered. In all four cases this was associated with an increase in the transpiration rate. Interestingly, in seedlings 4 and 8 an additional cavitation event occurred with some delay about 2 h after resupply of nutrient solution (Fig. 6a). One of those vessels (plant 4) had collapsed temporarily during drought stress and cavitated during recovery.

Figure 6.

Dynamics of vessel cavitation and refilling in maize roots induced by temporal drought stress. Plots of the number of cavitations per plant root with time for plants 1–8 (a; rewatering at very low light irradiation of < 1 µmol m−2 s−1) and for plants 9–12 (b; rewatering at full light irradiation of c. 800 µmol m−2 s−1). Each plant can be identified by its characteristic line style (see legends at a and b). Note that a small vertical offset was introduced for each plant in order to facilitate identification of individual traces. Gas filling of vessels was persistent until the end of the experiment (dash–dotted lines) or cavitation repair occurred (dashed lines). The left part of each graph reflects the formation of cavitations after removal of nutrient solution from the cuvette (time index 0 h): The end of the traces indicates rewatering of the individual plants (after c. 6 h of drought stress). In the right part of graphs (a) and (b), the dynamics of cavitations after resupply of nutrient solution (time index 0 h) is summarized for the same roots. Note that in two of the plants shown in (a), a cavitation event occurred even after rewatering. The arrows in (a) and (b) indicate the point in time at which spin density maps for each plant with maximum number of cavitations in virtual root cross-sections were recorded as shown in (c) and (d), respectively (plant number given at each image).

Table 1.  Summary of some relevant properties of the plants used in this study and of drought stress-induced cavitation and recovery of xylem vessels
  1. Plants 1–8 were resupplied with nutrient solution at a low light intensity, whereas plants 9–12 were rewatered at a high light intensity. 1, The youngest leaf was not unfolded yet; nd, not determined. For more details, see text.

Plant number   1 2   3 4 5 6 7 8 9101112
Age (d)  1310  15101113 71111121112
Root length (cm)> 3528> 35253326142530322934
Shoot length (cm) (longest leaf)nd17  23191726 81618222323
Number of leavesnd 3   4 34141 2 34141 4 4
Number of late metaxylem vessels  6 6  5 6 7 7 5 5 7 7 6 7
Number of cavitations  1 3  0 2 2 4 0 4 3 0 0 1
Number of refilled vessels  1 3 0 0 0 4 3 1
Number of refilled vessels exhibiting flow  1 1 4 3 0

The number of cavitations detected in the measuring plane varied between zero and four; it did not correlate with any of the parameters measured in this study (including time-courses of SD and T1 not shown here) and, hence, was apparently random. Interestingly, for root cells in none of the experiments were all vessels cavitated in the root investigated. Figure 6c shows tap root SD maps of plants 1–8 taken at the time when the maximum number of vessels (up to four) in all plants was embolized. Note that the position of the xylem vessels that exhibited cavitations within one plant was apparently random. Refilling was again observed in a narrow time-period c. 6 h after resupply of nutrient solution in three of the six seedlings with embolized vessels. In these plants (nos. 1, 2 and 8) all embolized vessels were ‘repaired’. In the other three plants (nos. 4, 5 and 6), vessels remained embolized until the end of the experiment c. 11 h after drought stress was terminated.

In order to test whether refilling of embolized vessels was restricted to plants that were kept in the dark, the experiment was repeated, but the nutrient solution was resupplied while plants were exposed to light (plants 9–12). In two out of four plants treated in this way, one or more cavitation events occurred during desiccation. All embolized vessels were refilled 11 h and 10.7 h after rewatering (Table 1, Fig. 6b), but refilling was only observed upon reduction of the photon flux density from 800 to < 1 µmol m−2 s−1 (i.e. approx. 9 h after rewatering). Spin density maps were qualitatively similar to those recorded on plants that were rewatered during the dark. Again, the position of embolized vessels was apparently random, (i.e. cavitation events did not follow a certain spatial pattern; Fig. 6d). This indicates that vessels will not cavitate more readily because of their position (e.g. in the vicinity of one that is already cavitated).

These additional experiments also confirmed that refilling of embolized vessels was obviously an all-or-none phenomenon. In toto, cavitations occurred in eight out of 12 seedlings subjected to 6 h of drought stress during this study; in five of these, all embolized vessels were refilled following rewatering. Refilling did not necessarily occur simultaneously in all embolized vessels of one root (see Figs 5a, 6a,b), but it was completed within 1 h, at a maximum. In the case of the other three plants all vessels remained cavitated throughout the experimental period.

In two plants (experiment nos. 9 and 12), the time course and direction of refilling could be evaluated using a new method based on a multislice MR imaging sequence. Ascent of the sap in the embolized vessels was tracked by evaluation of three consecutive imaging slices (see the Materials and Methods section; Fig. 2). In plant 9, three embolized vessels were refilled (Fig. 7a). The rise velocities were determined to be 3.3, 1.8 and 1.0 mm min−1, respectively (Fig. 7b). The sap level of the embolized vessels of plant 9 passed the center of the slice of imaging 2 : 11 h, 2 : 12 h and 2 : 08 h after the low light intensity had been established (i.e. within a narrow time-interval). Flow imaging 6 : 15 h after refilling at the site of the measurement revealed that the vessels became functional again. In plant 12, the single vessel was refilled with a velocity of 0.5 mm min−1. Refilling of the virtual section occurred 1 : 41 h after the low light regime had been established (see also Table 2 for a summary of the refilling dynamics and the corresponding vessel sizes and root lengths). However, obviously no functional refilling occurred, since flow had not recovered in this vessel when a flow map was recorded 8 h after refilling.

Figure 7.

Calculation of the velocity of xylem refilling in individual vessels using the new method based on changes in signal intensity in parallel imaging slices with time (see Fig. 2). (a) Time series of images obtained on plant 9 (see times of the day along top of panel) in three different slices obtained during refilling. Note that in some vessels an ‘intermediate’ signal intensity is observed when the meniscus was just passing the measuring plane; an example is marked by arrows. (b) Fit of the function S(p,T) to the signal intensity at the central pixel of one late metaxylem vessel in all three slices (see arrows in a) during refilling of a cavitation (plant 9). The signal occurred at first at the lower slice 1, then at slice 2, and finally refilled the pixel of slice 3. The point in experimental time when the sap level reached the center of the middle slice, t0 = 17:12 h; rise velocity, vrise = 1.8 mm s−1.

Table 2.  Summary of the properties of the refilled vessels for which the velocity of sap ascent has been measured
Plant number99912
Vessel number1231
Velocity of sap ascent (mm min−1)
Estimated vessel diameter (µm)65716265
Time that elapsed between start of the dark phase and vessel refilling (h)2 : 112 : 122 : 081 : 41
Root length below measuring slice (cm)26262630


This study clearly demonstrates that MR imaging is a powerful tool to survey cavitation phenomena in roots of intact plants. Previously, this method had successfully been applied at different sites of the shoot (Holbrook et al., 2001; Clearwater & Clark, 2003; Zimmermann et al., 2004; Scheenen et al., 2007; Schneider et al., 2007). Spin density maps allow the identification of drought-induced cavitation events in individual xylem vessels and of subsequent refilling of embolized vessels upon rewatering. In contrast to other, destructive techniques such as cryo-scanning electron microscopy (McCully et al., 1998) the fate of individual xylem vessels can be monitored continuously using MR imaging. Cavitation is observed by a ‘sudden’ loss of signal intensity in individual xylem vessels, visualized by a color shift from white to dark gray. Identification of embolized vessels in this study was restricted to the large metaxylem vessels of the tap roots of the 7- to 15-d-old maize seedlings (diameter c. 50–70 µm). Cavitation was rarely observed in well-watered seedlings under control conditions but was induced in the majority of the seedlings (eight out of 12 plants) under drought conditions.

Successive cavitation events may be favored by the entry of water vapor (or air) from neighboring vessels that have already undergone cavitation, as suggested by the air seeding hypothesis (Tyree & Sperry, 1989). However, in maize roots the spatial distribution of the xylem vessels that exhibited cavitations within one root was random (Fig. 6c,d), indicating that immediate spreading of a cavitation event to neighboring vessels is apparently prevented, so at least some of the vessels remain functional for several hours of drought stress (Table 1). This is in accordance with data obtained previously on plant stems (Holbrook et al., 2001; Clearwater & Clark, 2003; Scheenen et al., 2007). A second type of response to drought-induced tension, vessel collapse, has also directly been shown here in the experiment on plant 2 (Fig. 5) and on three other plants. A collapse of xylem vessels under tension has been postulated before (Jacobson et al., 2005) but direct evidence has been scarce. Cochard et al. (2004) and Brodribb & Holbrook (2005) used cryo-scanning electron microscopy to investigate vessel collapse in water-stressed pine needles and in leaf veins of Podocarpus grayi, respectively. However, this approach is open to criticism because vessel collapse may be an artifact arising from a preceding cavitation that was induced by freezing of the sample (Cochard et al., 2000; Richter, 2001). A collapse of xylem vessels is likely to affect axial resistance of root segments and to have an impact on ‘vulnerability curves’, especially those obtained on root segments (Jackson et al., 2000). Most importantly, the collapsed vessels recovered after rewatering and became functional again. Vessel collapse may be part of the plant's strategy to cope with a shortage of water. The ‘advantage’ for the plant is that no gas phase has to be resorbed in order to restart mass flow in a vessel.

After resupply of nutrient solution, refilling of xylem vessels was observed in five out of eight plants with one or more embolized vessels. When more than one cavitation event during drought treatment had occurred in a root either none or all vessels were refilled with xylem sap when water was resupplied. After resupply of medium, some O2 depletion will occur around the apical part of the root being immersed in stagnant nutrient solution for several hours (Trought & Drew, 1982). We cannot exclude that cavitations in three plants persisted because the root suffered from hypoxia, although it has to be noted that the site of measurement always protruded from the nutrient solution (see the Materials and Methods section). It should also be noted that sap exudation from cut roots (as measured by MR imaging of sap flow in the xylem) proceeded for at least 15 h when the root was immersed in stagnant nutrient solution (Kaufmann, 2008).

Evidence for xylem refilling of individual xylem vessels has also been presented in a previous long-term MR imaging study on grapevine (Holbrook et al., 2001), but in that study no parallel monitoring of flow was performed. Thus, it could be argued that the gas phase shrunk, but was not resorbed completely, and prevailed outside the measuring plane. Indeed, Scheenen et al. (2007) presented additional flow mapping on the stem of a cucumber plant, which revealed no flow within cavitated and subsequently refilled xylem vessels, which demonstrated that signal recovery at the site of measurement is not sufficient to show functional repair of embolized xylem vessels. In the present study we were able to detect flow in the majority of the previously cavitated vessels after refilling (but not in all of them; see Table 1). Hence, we were able, for the first time, to identify individual xylem vessels that experienced cavitation when drought stress was imposed and that regained their functionality after refilling. Moreover, flow spontaneously ceased in two vessels even though the filling status had apparently not changed in these vessels (Fig. 5a,b), suggesting that cavitations had occurred outside the measuring plane. However, it should also be kept in mind that MR imaging can only be used to detect a gas phase that spread across the imaged lumen of a vessel, not the cavitation event per se. Finally, the presence of a gas phase in a vessel does not necessarily exclude that this particular vessel is conducting water (Zimmermann et al., 2004). Despite these limitations, the conclusions drawn by Holbrook et al. (2001) concerning cavitation repair after rewatering of drought-stressed plants at low light intensities appear, retrospectively, to be largely correct in the light of the present study.

Previously, repair of embolized vessels under high tensions in adjacent, functional vessels has been claimed for several species (Salleo et al., 1996; Zwieniecki & Holbrook, 1998; Melcher et al., 2001; Stiller et al., 2005), including maize (McCully et al., 1998), but could not be confirmed by this study. In plants that were irradiated continuously after rewatering (9–12), refilling of embolized vessels was observed only 1–2 h after the light regime had been lowered to the background level and c. 9 h after rewatering, whereas in plants rewatered under low-light conditions (1–8), vessels were repaired 5–6 h after the end of the drought-stress period. This is consistent with the results of Holbrook et al. (2001) for grape stems.

In a previous study on cavitation and refilling in maize, McCully (1999), making use of the cryo-scanning electron microscopy technique, argued that under stress conditions individual vessels might be subjected to a constant cycle of cavitation and refilling. By contrast, MR imaging showed that a variable number of vessels per root cavitated and remained cavitated during drought stress treatment, whereas other vessels were continuously water-filled.

We were also able to monitor the refilling of single xylem vessels by the ascent of the sap level and to measure its velocity. It was relatively low, ranging from 0.5 mm min−1 to 3.3 mm min−1 (Table 2). Measurements by Scheenen et al. (2007) using a different MR imaging approach revealed an even slower refilling of embolized vessels (5–14 h for the refilling of a 3-mm thick imaging slice) in the stem of a cucumber plant, after cavitations had been induced by cooling of the roots. In contrast to our approach, the method introduced by Scheenen et al. (2007) did not directly reflect water ascent, but the observed signal increase could also have been caused by a continuous secretion of water into the xylem along the whole vessel. It appears that the timescale and perhaps even the underlying mechanisms for refilling of embolized vessels differ between species, different parts of plants (roots, stem, etc.) or even between different stress situations (compare Schneider et al., 2000b).

The velocity of vessel refilling determined here was more than one order of magnitude lower than the values measured for root pressure-driven flow (Kaufmann, 2008). However, it has to be taken into account that those values were measured on excised roots, whereas the velocities of the ascending sap level determined here were obtained on intact plants. In case that some overpressure, acting as a counterforce, was built up during refilling the low rates measured by us are not at variance with root pressure-driven flow.

Several mechanisms for vessel repair after cavitation have been proposed previously (Salleo et al., 1996, 2004; Holbrook & Zwieniecki, 1999; Clearwater & Goldstein, 2005). More functional studies like the present one are required to decide between them. Recently, the putative role of tissue pressure in these processes has been discussed (Clearwater & Goldstein, 2005). It is unlikely that tissue pressure in the stele is involved when collapsed vessels are recovered, since hydrostatic pressure exerted by cells counteracts vessel expansion. It has to be noted, however, that recovery of cavitated vessels clearly preceded repair of embolized vessels (Fig. 5), so that tissue pressure could still be involved in embolism repair in maize (Bucci et al., 2003; Clearwater & Goldstein, 2005), which is essentially the replacement of a gas phase by a fluid phase in a rigid ‘tube’.


Technical assistance by Elsbeth Fekete, Würzburg, is gratefully acknowledged by the authors.