Plant Materials and Field Trials
Plant material used for this study consisted of nine S. halepense genotypes that were collected from several locations in Texas, USA, one genotype collected in North Carolina, USA, and one genotype from the USDA National Plant Germplasm System  (Table 1). These plants were propagated vegetatively during the spring of 2010 to have sufficient plant material for field planting in autumn of 2010.
Winter survival trials were conducted at Commerce (33°11′ N, 95°55′ W) and College Station (30°32′ N, 96°26′ W), Texas, USA in 2010 and 2011. At both locations, genotypes were planted in a randomized complete block design with three replications. The plots at Commerce were located on the Texas A&M University–Commerce Research Farm, and the plants were grown in columns of soil inside Polyvinyl chloride (PVC) tubes 1 m in length and 15.2 cm in diameter (see below). The tubes were placed into 92-cm deep holes that were dug with a power take-off (PTO)-driven auger attached to a tractor. The distance between soil columns at Commerce was 1.8 m to allow for the movement of the tractor between plants.
Two experimental plantings were established at the Texas A&M University Research Farm near College Station. One was similar to the plot at Commerce except that, at College Station, the tubes were submerged into the soil to a depth of only 46 cm and the distance between rows and columns within a row was 1 m. The soil level of the columns at Commerce was even with the soil level of the field, but the soil level in the columns at College Station was 46 cm above the field soil level. This placement was intended to expose the upper half of the soil columns at College Station to more extreme temperatures. For the second College Station planting, the plants were transplanted directly into the soil at a distance of 2 m between plants within a row and 2 m between rows. Because one College Station experiment was conducted in soil columns and the second was planted directly into soil in the field, they were considered as different experimental environments.
The soil columns were made by longitudinally cutting 3 m lengths of 15.2 cm diameter PVC pipe into two equal halves. The halves were cut into 1-m sections. At the lower end of each half-tube, three evenly spaced 0.3-cm holes were drilled and threaded with galvanized wire, which crossed at the center to create a web-like design. The two pipe halves were then placed together and secured with duct tape at locations along the length of the tube. A piece of burlap cloth was folded several times and placed on top of the wire web at the bottom of each column for the purpose of retaining the soil while allowing water percolation.
Tubes used at both locations were filled with Westwood silt loam (fine-silty, mixed, superactive, thermic Udifluventic Haplustepts) that was taken to a depth of 15.2 cm from the Texas A&M University Research Farm. The tubes were filled with the soil, tamped on the ground, and then watered overnight using a soaker hose to insure similar soil compaction across all tubes. Following watering, soil compaction caused a drop in the soil level in each tube, and soil was added to bring the soil level to ∼7.6 cm below the top of each tube resulting in a 92-cm column of soil in each tube.
Planting and Sampling
For all studies, plants were vegetatively propagated via rhizome cuttings in March of 2010. Six cuttings from each genotype (three replicates with two plants each; one for autumn sampling and one for spring sampling) were used per experiment. After a few months of growth, plants were trimmed approximately 15 cm above the soil surface and transplanted into soil columns (on 16 August) or the field environment (on 8 July). At the time of planting, tiller number varied by genotype. Plants were fertilized at 84 kg N ha−1 as ammonium sulfate and watered. Columns designated for the autumn and spring sampling at Commerce were placed in the ground on 17 September. At College Station, the tubes designated for autumn were sampled directly without placement in the ground, while the tubes designated for spring sampling were placed into the ground on 12 November.
For sampling, the tubes were removed from the ground and the tape binding the two halves of the tubes together was cut. The tubes were then opened, leaving the soil column intact. Each soil column was cut laterally into two equal halves 46 cm in length. Each half of the soil column was placed in a tub of water and the soil was washed away so rhizomes could be collected. The rhizomes were collected separately from the upper (0–46 cm) and lower (46–92 cm) halves of each column. However, only the samples taken from the upper half of each column were used for analyses because only a limited number of columns had rhizomes below 46 cm. Rhizome diameters were measured at the center of the first complete internode from the cut site. Samples were then placed into plastic bags and stored in a −80 °C freezer. Autumn sampling at Commerce took place on 22 October prior to the first freezing temperature, which occurred on 25 November. College Station columns were sampled on 10 November prior to the first freezing temperature on 27 November.
Autumn sampling of the College Station field plots (planted directly in the soil rather than in tubes) took place on 17 and 19 November. The sampling procedure for the plants in the field differed from those in tubes. Above-ground vegetation of each plant was removed before the rhizomes were collected. Using a shovel, the whole plant was excavated. After the plant crown and rhizome mass were removed, the depth of the hole was increased to 46 cm and rhizomes were collected from the material removed from the hole. Rhizomes were not found below 46 cm; therefore, samples were not collected in the 46–92 cm depth range. The lack of rhizomes below 46 cm was likely due to a hard pan that was present approximately 41 cm below the soil surface. Rhizomes were washed and placed into plastic bags and stored on ice until they were transferred into a −80 oC freezer. Since the sampling method was destructive, the remaining plants of each genotype were left in the field to be sampled the following spring. Just prior to spring rhizome sampling, several traits associated with overwintering ability and spring regrowth (the presence of new leaf material in the spring) were measured for both the field-planted and column-grown plants at both locations. For the College Station field site (not in columns) a 1-m2 quadrat centered on the primary tillers of each plant was used to count the number of green shoots and the number of leaves. These data were collected at College Station on 2 March 2011. Spring regrowth measurements were made on the plants growing in the soil column on the day of sampling, which was 1 April at College Station and 7 April at Commerce.
Spring sampling at both locations was conducted in the same manner as the previous autumn sampling, with a few exceptions. Only live rhizome tissue was collected in the spring because some rhizomes had freeze damage. Rhizomes considered “live” had a healthy/white appearance and were rigid.
Rhizome Composition Analysis
To prepare the rhizome samples for analysis, the rhizomes were placed into mesh bags and dried in a lyophilizer (Labconco, Kansas City, MO, USA) for at least 3 days at −40 °C. All fibrous root material was removed from the rhizomes. Samples were then ground to a 1-mm particle size using first a coarse grinder and then a UDY Cyclone mill (Udy Corp, Fort Collins, CO, USA). The ground tissue was placed into glass vials and stored in cold storage.
Tissue samples were analyzed by Dairy One (Utica, NY, USA) using wet chemistry analysis for crude protein (CP), crude fat (fat), starch, water soluble carbohydrates (WSCs), and ethanol-soluble carbohydrates (ESCs). WSC consists primarily of glucose, sucrose, fructose, maltose, lactose, and fructans, whereas ESC is made up of primarily glucose, sucrose, fructose, maltose, and lactose. An estimate of fructan concentration was therefore obtained by determining the concentration difference between WSC and ESC for each sample.
Assimilate Analysis via NIR
Prior to wet chemistry analysis, the samples were scanned with a Thermo Scientific Antaris II FT-NIR Analyzer (Thermo Scientific Inc., Waltham, MA, USA) with a setting of 64 coaveraged scans per sample and a resolution of 4 cm−1 between wavelengths 4,000 and 9,999 cm−1 using the OMNIC 8 software suite (Thermo Scientific Inc., Waltham, MA, USA). To insure the tissue was packed uniformly into the vials, they were placed side by side in a small box and tamped on a lab bench 15 times prior to scanning. NIRS reflectance spectral data was converted to absorbance and combined with the wet chemistry analysis data in TQ Analyst software (Thermo Scientific Inc., Waltham, MA, USA) to create prediction equations. Approximately one third of the data were used as a validation set and the other two thirds were used as a calibration set. Many different standard mathematical treatments were applied over different spectral ranges before selecting the treatment with the highest performance index. The performance index validates the NIRS curve by measuring how accurately a calibrated method can quantify independent validation samples. A value of 0 reflects no prediction while a maximal value of 100 reflects perfect prediction.
Data were analyzed for genotype, environment, and genotype by environment (GxE) effects using a PROC MIXED statement with environment, genotype, and GxE as random variables in SAS version 9.1 . The variation for each effect was then divided by the total variance to get a percentage of total variance. The two sampling dates (autumn or spring) were not comparable and analyzed separately but with otherwise identical PROC MIXED analyses including all three environments. Correlations between CP, fat, starch, WSC, ESC, and overwintering capability were determined using a PROC CORR statement. These correlation analyses were carried out on all experiments and sampling dates individually. In addition, the two soil column environments were also analyzed jointly because of their similar design, using a PROC CORR statement. Logistic regression models were created for the prediction of spring regrowth (overwintering ability) based on autumn assimilate composition data using JMP® Software, Version 9 .