Xylem vulnerability to cavitation can be accurately characterised in species with long vessels using a centrifuge method

Study site and plant material

Chaparral shrub plant material was collected from a site in the San Gabriel Mountains of southern California (1040 m a.s.l.; 34^°21.613 N, 118^°26.125 W). The site has a mediterranean-type climate with hot, dry summers and cool, wet winters and a mean annual precipitation of 685 mm from 1996 to 2010 (Western Regional Climate Center 2011). Branches of long-vesselled V. vinifera L. cv. Glenora were collected from a mature, irrigated plant growing in Bakersfield, CA, USA. For V. vinifera, all branches were collected from the same individual to control for interplant variability and to reduce variability in the number of open vessels not due to differences in sample length. Branches of long-vesselled Q. robur L. cv. fastigiata were collected from plants grown in a greenhouse in Bakersfield with daily irrigation for 8 months, following their purchase as ~1-m plants from Lawyer Nursery Inc., MT, USA.

Two evergreen chaparral shrubs were chosen for the main experiments because of their widely divergent vessel lengths (Table 1): the long-vesselled Q. wislizeni var. frutescens Engelm. (Fagaceae) and the shorter-vesselled Heteromeles arbutifolia (Lindl.) M. Roem. (Rosaceae). Two additional long-vesselled species (mean vessel length >0.10 m), Rhus ovata S. Watson (Anacardiaceae) and Fraxinus dipetala Hook. and Arn. (Oleaceae), and two shorter-vesselled species (mean vessel length <0.06 m), Ceanothus leucodermis Greene (Rhamnaceae) and Ribes malvaceum Sm. (Grossulariaceae), were sampled for some experiments. Of these, F. dipetala and R. malvaceum were deciduous and the rest were evergreen.

We collected large branches (1.8–3.0 m in length) of Q. wislizeni and H. arbutifolia for laboratory dehydration vulnerability curves from January to June of 2011, before new leaves had flushed. Stem cutting was done under water, branches were enclosed in doubled and humidified plastic bags to reduce transpiration, and the cut end was kept under water during transport to the laboratory. Smaller branches for centrifuge vulnerability curves and a repeated centrifugation experiment (see below) were collected similarly during the same time period, except for V. vinifera and Q. robur, which were sampled in autumn 2011.

Stem hydraulic conductivity, loss of conductivity and centrifugation

We determined K_h from measurements of the net increase in fluid flow rate through a stem after applying a pressure head (Sperry et al. 1988a). We selected stem segments that were ca. 6–8 mm in diameter and prepared them for measurements by cutting segments from sampled stems under water and trimming them to the desired length with a razor blade. To make flow rate measurements, stems were inserted into a tubing apparatus that connected the basal end to an elevated fluid reservoir and the other end to a collection reservoir on an analytical balance (CP124S, Sartorius, Goettingen, Germany) connected to a laptop computer (Inspiron 5000e, Dell Computer Corp., Round Rock, TX, USA) that logged net flow over 10-s intervals. The pressure head was maintained below 2 kPa for stems of V. vinifera and below 4 kPa for all other stems to avoid refilling open vessels of large diameters. Stems were kept submerged in water during and between measurements. The perfusing solution used in the tubing apparatus was 20 mm KCl filtered to 0.1 μm (inline filter; GE Water and Process Technologies, Trevose, PA, USA) and vacuum-degassed for at least 20 min (Melcher et al. 2012). Before and after each measurement of pressure-driven flow, we measured background flow with no pressure head. We calculated K_h as the pressure-driven flow corrected for background flows divided by the pressure gradient (pressure head/stem length). Maximum K_h was measured on stems flushed with the same solution used for K_h measurements at 100 kPa for 1 h to refill embolised vessels. To calculate stem area specific conductivity (K_s), K_h was divided by stem sapwood cross-sectional area, which was calculated from caliper measurements of xylem and pith diameter.

We used centrifugation at a stage in each experiment to induce negative xylem pressure in stem segments (Holbrook et al. 1995). We mounted stems in a custom rotor (Alder et al. 1997) that enabled them to be spun in a centrifuge (Sorvall RC-5C; Thermo Fisher Scientific, Waltham, MA, USA) with stem ends submerged in solution contained in L-shaped reservoirs. The solution level in the reservoirs was kept even at both ends of the stem, so no flow was induced during centrifugation. During some experiments, we added foam pads to the vertical section of reservoirs to keep the solution in contact with stem ends when the rotor was not spinning (see below, Repeated centrifugation tests).

Laboratory dehydration vulnerability curves

Laboratory dehydration vulnerability curves were completed for 11 large branches of Q. wislizeni and three of H. arbutifolia. Both species had substantial native embolism (PLC >50% for many stems) that did not reverse during overnight rehydration of branches. The high native PLC made it impossible to fully characterise the vulnerability to cavitation of stems. Therefore, we refilled native embolism by re-cutting branches under water in the laboratory and attaching them to a tubing manifold pressurised to ~15 kPa for about 18 h with the same solution used for K_h measurements. We confirmed removal of native embolism by measuring the PLC of at least two stems per branch.

After refilling, 1–3 stem psychrometers (Stem Hygrometer; PWS Instruments Inc., Guelph, OT, Canada) connected to a data logger (CR7; Campbell Scientific, Logan, UT, USA) were installed on each branch to track branch xylem pressure. The bark was sanded away to expose a small patch of xylem where psychrometers were attached, and silicone grease was used to create an airtight seal. An ultrasonic acoustic emissions transducer (Model I15I; Physical Acoustics Corp., Princeton, NJ, USA) was attached to one of the branches to monitor cavitation as the branch dried. Bark was sanded off the branch where it was attached, and the xylem was covered with silicone grease to facilitate contact with the transducer. A spring-loaded attachment held the transducer against the stem with 30 N of force (Jackson & Grace 1996). The transducer was connected to hardware that conditioned the transducer output so that individual acoustic events were sampled only once, and that amplified the signal by 75 dB. A data logger (21×; Campbell Scientific) sampled this output every second and recorded 5-min sums of acoustic emissions.

Branches were equilibrated at increasingly negative xylem pressure through repeated cycles of drying and equilibration. Branches were dried by uncovering them for a short period, then equilibrated by sealing them in double plastic bags, which were externally covered with moist cotton sheets and more plastic bags, and allowing them to equilibrate for 16–23 h. Following equilibration, 1–3 stems were removed from each large branch for PLC measurements, with cuts made under water. At the same time, we collected up to five branchlets for water potential measurements with a pressure chamber (PMS Instrument Co., Albany, OR, USA) to compare with stem psychrometer values. We repeated these steps until the stem loss of hydraulic conductivity was near 100%.

Comparison of dehydration and centrifuge methods

We constructed centrifuge vulnerability curves using stems sampled during laboratory dehydration vulnerability curves for both Q. wislizeni and H. arbutifolia to test if negative xylem pressure induced by either air-drying large branches or centrifugation of stem segments resulted in different curves. After the PLC of stems sampled from large branches during a laboratory dehydration vulnerability curve was determined, we centrifuged the flushed stems for 5 min to a xylem pressure similar to that measured for the large branch with psychrometers. The centrifugation-induced PLC of these stems was then measured, which resulted in a matched pair of centrifuge and dehydration PLC values at similar water potentials for each stem segment.

Comparison of centrifuge vulnerability curves of two stem lengths

We constructed centrifuge vulnerability curves for stems of 140 and 271-mm length for Q. wislizeni, H. arbutifolia, Q. robur and V. vinifera to test the effect of changing the proportion of open vessels. After an initial flush, we measured the K_h of stems and then centrifuged them for 5 min (n = 10 for 140-mm stems and n = 12 for 271-mm stems of V. vinifera, n = 6 for all other species). Stem K_h measurement and centrifugation was repeated, inducing a more negative xylem pressure with each centrifugation, and continued until stem loss of hydraulic conductivity was near 100%.

Distribution of centrifugation-induced embolism

We measured the distribution of embolism along the length of stems to compare it with the theoretical negative pressure profile generated during centrifugation. We centrifuged 271-mm long flushed stems (n = 3) of the long-vesselled R. ovata at an angular velocity that induced -2 MPa of pressure at the rotational centre of the stem and reduced entire stem conductivity by 73–81%. We excised the basal, central and distal 40-mm segments of these stems and measured their PLC.

Repeated centrifugation tests

We repeatedly centrifuged stems of all six study species at the same angular velocity to test for increases in PLC unrelated to inducing more negative xylem pressure. After an initial flush and K_h measurement, stems were either centrifuged at −0.5 MPa for 5 min four times, with K_h measured after each centrifugation, or stems were centrifuged once at −0.5 MPa for 20 min, with K_h measured afterwards. All stems were then flushed, followed by a final K_h measurement. There were three stems per treatment, except R. ovata, for which there were six.

As a follow up, we tested if several treatments reduced the increase in PLC observed with repeated centrifugation at the same angular velocity. Because the centrifuge reservoir design relies on centrifugal force to maintain solution in contact with stem ends, they are exposed to air when the rotor is not spinning. We added foam pads (CoverGirl Make-Up Masters Sponge Wedges; Procter & Gamble, Hunt Valley, MD, USA) thick enough to be in contact with stem ends to the vertical section of reservoirs. The pads were saturated with solution and thereby kept the solution in contact with stem ends when the rotor was stationary. We tested the effect of foam pads by comparing the PLC increase with and without pads for repeatedly centrifuged stems of R. ovata (n = 12). We also tested the addition of 10 min at 0.1 MPa to the end of centrifugation as a ‘relaxation’ period on Q. wislizeni stems (n = 3). Finally, recent concerns about solution gas content (Espino & Schenk 2011) led us to test if additional degassing of the perfusing solution reduced increases in PLC with repeated centrifugation for R. ovata stems (n = 3). After our standard vacuum degassing of 10 l of continuously stirred solution for at least 20 min at an absolute pressure of ~10 kPa generated with a vacuum pump (model 5KH33DN16X; Millipore, GE, Billerica, MA, USA), we further vacuum degassed the solution using a membrane contactor (Liqui-Cel mini-module 1.7 × 5.5; Membrana, Charlotte, NC, USA).

Native PLC

We measured native PLC in Q. wislizeni at the end of the summer dry period in 2009, and again during the rainy season in early spring of 2010. For summer measurements, we collected large branches (n = 6) by cutting them in air near midday, and brought them to the laboratory sealed in doubled and humidified plastic bags. After an equilibration period, we determined branch water potential from branchlets using the pressure chamber and measured PLC of stems cut under water from the large branches at a distance from the original cut end that exceeded maximum vessel length. Spring measurements of native PLC were done in the field on stems collected pre-dawn, with cuts made under water. Pre-dawn water potential for each sampled shrub (n = 6) was determined from branchlets collected with stems. Laboratory and field-portable conductivity systems differed primarily in the method used to quantify flow, which was done in the field by timing the rate at which solution volume increased in a 0.1 ml pipette.

Vessel length distributions

Vessel length distribution was measured on six stem segments per species for the six chaparral study species using a silicon injection method (Sperry et al. 2005; Wheeler et al. 2011). We used stem segments that were 6–8 mm in diameter and longer than 250 mm. Prior to injection, stems were trimmed and flushed as described above for K_h measurements. The basal end of the segments were then injected with a 10:1 silicone hardener mixture (Rhodorsil RTV 141; Bluestar Silicones, distributed by Skycon, Toronto, OT, Canada) that was mixed with a soluble fluorescent dye (1% w/w; Uvitex OB, Ciba Specialty Chemicals, Tarrytown, NY, USA dissolved in chloroform) at 50 kPa for 24 h. Stem segments were cured at room temperature for at least 72 h before sectioning. Stems were sectioned at several distances from the injection surface, and cross-sections were analysed to determine the percentage of vessels filled. The vessel length distribution was calculated based on the equations described in Wheeler et al. (2011), and this distribution was used to estimate the percentage of vessels open all of the way through segments and percentage of vessels open to the middle of segments from either cut end.

Statistical analysis

Plant material was carefully matched within each experiment, and vulnerability curves were compared within each experiment. We tested if dehydration and matched centrifuge vulnerability curves differed by comparing models that included both curves using nested models constrained to fit a single curve (Crawley 2007). Models were fitted to mean PLC of stems (n = 1–3) at each branch equilibrium or centrifuge xylem pressure. A reparameterised Weibull model (Ogle et al. 2009) was fitted to Q. wislizeni values using generalised nonlinear least squares, and incorporated a variance model to account for non-constant variance (Pinheiro & Bates 2000). Models were fitted with maximum likelihood, and comparison of nested models resulted in a likelihood ratio test statistic (LRT). The shape of the relationship between PLC and xylem pressure for H. arbutifolia was poorly fitted with the Weibull model, so a generalised additive model (GAM) was fitted (Wood 2006). This type of model does not assume a specific functional form for the shape of the relationship between response and explanatory variables.

We tested if centrifuge vulnerability curves constructed with 140 or 271-mm stems differed using two separate analyses. In both analyses, curves were corrected for cavitation fatigue (Hacke et al. 2001) by calculating the PLC using the K_h measured following the initial spin at 0.5 MPa in place of the maximum K_h. This minimised the effect of xylem vessels unlikely to be functional in intact plants. To compare P₅₀ between vulnerability curves, we employed a common approach of fitting separate curves (GAM for H. arbutifolia and Weibull for other species) to data for each stem and then used the P₅₀ estimates from fitted curves in an analysis of variance (anova). We included species, stem length (140 mm or 271 mm) and their interaction in the model and tested for differences by stem length within species using a priori contrasts. A second approach that tested if the entire curves differed by stem length consisted of comparing models fitted to all stems that included two curves or were constrained to fit a single curve. This analysis was analogous to the dehydration and centrifuge curve comparison, but random effects were incorporated to account for the repeated measurements made on individual stems (Pinheiro & Bates 2000; Wood 2006).

The PLC response to repeated centrifugation was analysed as a mixed-model anova. We included centrifugation (repeat number), vessel length (long or short) and centrifugation type (repeated or single centrifugation) as fixed effects, and species and stem as random effects. Initial and final flushed K_s were similarly analysed by substituting flushed (initial or final) for centrifugation, and examined if initial and final K_s differed using a priori contrasts. For the repeated centrifugation experiment that tested the effect of foam pads in reservoirs, the PLC response to repeated centrifugation was analysed as a mixed-model anova that included centrifugation (repeat number) and treatment (with foam or without foam) as fixed effects, and stem as a random effect. The first and the fourth centrifugation PLCs for each treatment were compared using a priori contrasts. Additional treatments tested for their effect on PLC induced by repeated centrifugation were modelled similarly. The analyses were performed using R version 2.8.1 (R Development Core Team 2011) including packages nlme (Pinheiro et al. 2011) and mgcv (Wood 2006).

Comparison of dehydration and centrifuge methods

Dehydration and centrifuge methods for inducing negative xylem pressure in stems were compared by constructing matched vulnerability curves for stems that were initially dehydrated and subsequently flushed of emboli and centrifuged. For both long-vesselled Q. wislizeni and short-vesselled H. arbutifolia, the shape of the centrifuge vulnerability curve was similar to the dehydration vulnerability curve (Fig. 2). The P₅₀ for Q. wislizeni did not differ between dehydration and centrifuge curves (Table 2), but centrifugation produced modestly higher PLC than dehydration at a PLC >60% (Fig. 2a), resulting in differences between curves compared along their entire span (LRT₂ = 31.85, P < 0.001). Native PLC of Q. wislizeni stems sampled at the time of year when plants had high and low pre-dawn water potential closely matched the vulnerability curves (Fig. 2a). For H. arbutifolia, the dehydration and centrifuge vulnerability curves did not differ (Fig. 2b; F_{1.8, 52.1} = 2.68, P = 0.083), nor did P₅₀ values (Table 2). Thus, in contrast to results of Cochard et al. (2010) (Fig. 1), the cavitation response of stems did not diverge widely depending on whether negative xylem pressure was generated by either centrifugation or air-drying large branches or natural dehydration in the field.

As a second independent measure of cavitation, we measured ultrasonic acoustic emissions emitted by xylem tissue during dehydration vulnerability curves. A high rate of acoustic emissions was observed early in the dehydration of Q. wislizeni branches as xylem pressure began to decrease, but thereafter emissions declined to a low rate, even as xylem pressure continued to drop (Fig. 3). This pattern of acoustic emissions corroborates the shape of the Q. wislizeni vulnerability curves, which predict extensive cavitation at the xylem pressures where the rate of acoustic emissions was high, but less additional cavitation at more negative xylem pressures where the acoustic emissions rate was low.

While constructing dehydration vulnerability curves, we found that xylem pressure potential of branchlets from the equilibrated branches measured with the pressure chamber agreed with stem psychrometer water potential measurements for H. arbutifolia, but were more negative at water potentials below ~−4 MPa for Q. wislizeni (Figure S1).

Comparison of centrifuge vulnerability curves of two stem lengths

To test if the proportion of open vessels affected vulnerability curves measured with the centrifuge method, we compared curves for stems of 140 and 271-mm lengths. For H. arbutifolia, shortening the length of stem segments from 271 to 140 mm only substantially changed the percentage of vessels open to the middle, but both categories of open vessels changed substantially for Q. wislizeni (Table 1). Changes in percentage of open vessels between stem lengths for V. vinifera were similar to Q. wislizeni (e.g. entirely open changed from 1.5% to 9.4%; Jacobsen & Pratt 2012) and may also be similar for Q. robur, which has a similar maximum vessel length to Q. wislizeni. In spite of these changes, P₅₀ did not differ by segment length for any of the four species examined (Table 3; F_1,50 < 2.28, P > 0.137 for all segment length contrasts within species). Note that the P₅₀ values in Tables 2 and 3 cannot be compared directly, because Table 3 values were calculated from vulnerability curves corrected for cavitation fatigue (Hacke et al. 2001), a correction not possible for dehydration vulnerability curves. In a separate test for differences along the entire span of vulnerability curves, the curves for the two stem lengths did not differ for Q. wislizeni or H. arbutifolia (Fig. 4; LRT₂ = 3.36, P = 0.187 and LRT₃ = 0.02, P = 0.999, respectively), but did for Q. robur and V. vinifera (LRT₂ = 8.99, P = 0.011, LRT₃ = 7.48, P = 0.024, respectively). The magnitude of the difference between vulnerability curves of the two stem lengths for Q. robur and V. vinifera were small compared to differences in vulnerability curves related to stem length observed by Cochard et al. (2010) for Prunus persica and Q. robur (e.g. Q. robur, Fig. 1).

Distribution of centrifugation-induced embolism

We compared the distribution of embolism along stems with the theoretical negative pressure profile induced by centrifugation for a long-vesselled species. The PLC of stem segments excised from the centre of a centrifuged stem was higher than the PLC of segments excised from the ends (Fig. 5). This distribution of PLC matches the expected effect of the lower negative pressure generated near the axis of rotation at the centre of the stem during centrifugation (Alder et al. 1997).

Repeated centrifugation tests

Repeated centrifugation of stems at the same angular velocity led to increases in PLC with each additional centrifugation for six chaparral shrub species, but the magnitude of these increases was not related to having long or short vessels (Figure S2). For all species, flushing following repeated centrifugation fully recovered the initial flushed maximum K_h. Centrifugal force maintained solution in contact with stem ends during centrifugation, but stem ends were exposed to air when the rotor was not spinning. We substantially reduced the PLC increases during repeated centrifugation by adding foam pads to the reservoirs to maintain solution contact with stem ends at all times (Figure S3). We also tried adding 10 min at 0.1 MPa to the end of centrifugation as an equilibration period and additional degassing with a membrane contactor, but these treatments had no effect on PLC increases in Q. wislizeni and R. ovata, respectively (data not shown).

Centrifuge vulnerability curves agree with dehydration curves

We found no evidence for an artefact associated with open vessels for vulnerability curves constructed using a centrifuge technique. These findings contrast with those of Cochard et al. (2010) and Choat et al. (2010). The vulnerability curves resulting from laboratory dehydration or centrifugation were similar for both a long- and a short-vesselled species, and other studies have found similar agreement between these methods (Alder et al. 1997; Jacobsen et al. 2007a; Jacobsen & Pratt 2012).

Vulnerability curves with a so-called ‘r’ shape, like the Q. wislizeni curve, characterised by large increases in PLC with the initial drops in xylem pressure have been called into question by Cochard et al. (2010). The centrifuge vulnerability curve shape we report for Q. wislizeni was supported by a matching laboratory dehydration curve, the pattern of acoustic emissions of dehydrating branches and seasonal native embolism measurements. When employing the single-vessel air injection method (Melcher et al. 2003), which is a completely different methodology to that used here, Christman et al. (2012) found that another Quercus species had xylem with a similarly high proportion of highly vulnerable vessels. The ‘r’ shape has been previously observed for dehydration vulnerability curves of several other chaparral species (Jacobsen et al. 2007b), suggesting that multiple species have a high proportion of very vulnerable xylem vessels in this plant community. These highly vulnerable vessels may be functional and enhance carbon gain primarily during the wet season, or they may undergo daily cavitation and refilling cycles (Taneda & Sperry 2008; Christman et al. 2012), thereby maximising daily photosynthesis throughout the growing season. Alternatively, highly vulnerable vessels may be an efficient way to construct xylem having the required mechanical support.

The proportion of open vessels does not affect centrifuge vulnerability curves

We found no effect of changing the stem segment length and thereby the proportion of open vessels on the P₅₀ of centrifuge vulnerability curves. The lack of a substantial stem length effect for long-vesselled species, a finding consistent with Sperry et al. (2012) and Jacobsen & Pratt (2012), does not support the hypothesis that open vessels exhibit more vulnerability to cavitation during centrifugation. As described in Sperry et al. (2012), the increased vulnerability to cavitation of shorter stems for medium- to long-vesselled species reported by Cochard et al. (2010) may result from micro-bubbles being carried into the stem by the flow of solution induced as part of the unique Cavitron measurement technique. In contrast to Choat et al. (2010), we saw no evidence for an open-vessel artefact in centrifuge vulnerability curves of V. vinifera. A recent study suggests that V. vinifera branches used in dehydration vulnerability curves can develop vessel blockages, causing dehydration and centrifuge vulnerability curves to differ (Jacobsen & Pratt 2012).

Another result cited in Cochard et al. (2010) as demonstrating increased vulnerability of open vessels was the shift in Cavitron vulnerability curves following injection of air into the ends of stem segments. However, dramatically reducing hydraulic conductivity at the ends of stem segments with air injection (see Fig. 5 in Cochard et al. 2010) will shift vulnerability curves to appear more resistant to cavitation, regardless of which subset of vessels remain conductive. Centrifugation reduces hydraulic conductivity primarily near the middle of the segment length; therefore, a decrease in conductivity due to centrifugation results in a smaller proportional decrease in total stem conductivity when segment ends have been air-injected.

The pattern of cavitation is consistent with theoretical predictions

The differences between the distribution of embolism along the length of centrifuged stems measured in this study and that of Cochard et al. (2010) may partially explain why we do not observe a similar effect of stem length on centrifuge vulnerability curves. The distribution of PLC we measured for a long-vesselled species matches the theoretical expectation that embolism of a stem following centrifugation should be highest at the axis of rotation in the central part of the segment where the induced xylem pressure is most negative (Alder et al. 1997). Similar measurements conducted with the Cavitron found higher PLC at the basal stem end rather than at the centre (Fig. 5 in Cochard et al. 2010). Cochard et al. (2010) suggest this distribution could result from vessel draining, but the match between our results and the predicted pattern of PLC suggest that vessel draining is an unlikely explanation. Sperry et al. (2012) describe the high PLC at the basal stem end as likely arising from micro-bubbles or debris carried into the stem by the flow of solution during Cavitron measurements. The centrifuge vulnerability curve technique we used does not induce flow during centrifugation and therefore would not be vulnerable to this artefact.