Field Assessment of Alfalfa Populations Recurrently Selected for Stem Cell Wall Digestibility

Genetic variability for stem cell wall (CW) digestibility could be exploited to improve rumen-fermentable energy in alfalfa (Medicago sativa L.) forage. We evaluated in the field the response to recurrent selection for stem CW digestibility in alfalfa. Digestibility was assessed as the concentration of glucose released after enzymatic hydrolysis of fiber (enzyme-released glucose, ERG). Two initial cultivars, 54V54 and Orca, and populations obtained after successive cycles of divergent selection for stem CW digestibility (D−1, D−2, D+1, and D+2) were established at three field sites in north, central, and south of Québec. Field trials conducted over two growing seasons showed that populations obtained after two selection cycles (D+2) had significantly higher CW digestibility (+20.7  mg ERG g−1 CW) than initial cultivars (average of 13% improvement of digestibility). The D+2 populations did not differ from the initial cultivars with regard to biomass yield, winter survival, and stem water soluble-carbohydrate concentration. Increases of ERG concentrations were observed in response to each selection cycle, and broadsense heritability highlights a moderate control of genetic factors over environmental factors for CW digestibility. Recurrent selection for stem CW digestibility is a valuable approach to increase fermentable energy in alfalfa forage and improve N utilization by ruminants. A. Bertrand, A. Claessens, M.-N. Thivierge, S. Rocher, and Y. Castonguay, Québec Research and Development Centre, Agriculture and Agri-Food Canada, 2560 Hochelaga Blvd., Québec, QC G1V 2J3, Canada; J. Lajeunesse, Québec Research and Development Centre, Agriculture and Agri-Food Canada, 1468, Saint-Cyrille St., Normandin, QC G8M 4K3P, Canada; P. Seguin, Macdonald Campus, McGill Univ., 21111 Lakeshore Rd., Sainte-Anne-de-Bellevue, QC H9X 3V9, Canada. Received 20 Feb. 2018. Accepted 20 Apr. 2018. *Corresponding author (annick.bertrand@agr.gc.ca). Assigned to Associate Editor Ali Missaoui. Abbreviations: CW, cell wall; D+, high digestibility; D−, low digestibility; DM, dry matter; ERG, enzyme-released glucose; HPLC, high-performance liquid chromatography; IVTD, in vitro true digestibility; NDFD, neutral detergent fiber digestibility; NIRS, nearinfrared reflectance spectroscopy; WSC, water-soluble carbohydrate. Published in Crop Sci. 58:1632–1643 (2018). doi: 10.2135/cropsci2018.02.0119 © Crop Science Society of America | 5585 Guilford Rd., Madison, WI 53711 USA This is an open access article distributed under the CC BY license (https:// creativecommons.org/licenses/by/4.0/). Published June 14, 2018


2016
).As a result, producers face a tradeoff between yield and digestibility in their decision to harvest alfalfa.The development of alfalfa germplasm that maintain higher CW digestibility during stem maturation would provide growers with more management flexibility to produce high-yielding alfalfa with superior nutritive value.
Options to increase fiber digestibility in alfalfa forage include genetic engineering (Chen and Dixon, 2007) and conventional breeding (Buxton and Redfearn, 1997).According to Buxton et al. (1987), breeding using selection traits that specifically target the stems instead of total biomass would have a greater impact on alfalfa digestibility, since most alfalfa fiber is found in stems.When comparing nutritive value traits between whole plants, leaves, and stems, Milic et al. (2014) concluded that alfalfa breeding programs should focus on stems since larger values of narrow-sense heritability for feeding value traits were obtained with stems.
The concentration of glucose that is released after enzymatic hydrolysis of purified CW (enzyme-released glucose, ERG) provided a reliable assessment of CW digestibility of alfalfa stems (Duceppe et al., 2012) and of the in vitro true digestibility (IVTD) of timothy (Phleum pratense L.) roots (Bertrand et al., 2014).Near-infrared reflectance spectroscopy (NIRS) was validated as a highthroughput technique to predict ERG concentration in large collections of lignified alfalfa stem samples.Using this NIRS-based screening approach, alfalfa was found to naturally exhibit a large genetic diversity among cultivars and genotypes for stem CW digestibility with a strong negative correlation between that trait and total lignin concentration (Duceppe et al., 2012).These initial results indicated the potential to improve stem CW digestibility in alfalfa using ERG concentration as a selection criterion.At this point, the indirect impact of selection for stem CW digestibility on agronomic attributes such as yield, winter survival, and water-soluble carbohydrate (WSC) concentration and the stability of that trait under various agroclimatic conditions are unknown.Although yield and persistence remain determinant criteria in the choice of alfalfa cultivars and need to be maintained, gains in CW digestibility should not be made at the expense of WSC concentration, which has a positive effect on forage intake and milk production (Brito et al., 2008(Brito et al., , 2009)).
To assess the impact of recurrent selection for CW digestibility, assays were conducted under field conditions in the province of Quebec using two distinct alfalfa genetic backgrounds.In a first series of assays conducted at three sites in the province of Quebec over two growing seasons, we evaluated (i) the response of alfalfa to selection for stem CW digestibility based on ERG; (ii) the effects of this selection on important attributes including yield, stem WSC concentration, and winter survival; and (iii) the stability and heritability of this trait under various agroclimatic conditions.In a second assay, we evaluated the broad-sense heritability of stem CW digestibility by comparing three cycles of positive selection for that trait.

Divergent Recurrent Selection
Five hundred alfalfa genotypes from the forage-type cultivar Pioneer 54V54 (thin lodging-susceptible stems) and 500 genotypes from the biomass-type cultivar of Flemish origin Orca (large nonlodging stems) were seeded in a greenhouse and transplanted into a field in Lévis, QC, Canada (Kamouraska clay loam; 46°47¢22¢¢ N, 71°8¢8¢¢ W; altitude » 68 m) in June 2008.In September 2009, after two growing seasons, a visual selection based on plant vigor and yield was done, and the bottom parts (25 cm) of the stems of 136 selected 54V54 genotypes and 228 Orca genotypes were harvested at the seventh greenpod maturity stage (Kalu and Fick, 1981).We selected the bottom part of the stems, which are more lignified than the top, and we harvested at a late maturity stage to maximize contrasts between highly lignified stems.Our screening enzymatic test has been designed to discriminate between mature lignified stems (Duceppe et al., 2012).
For each genetic backgrounds (54V54 and Orca), 20 genotypes with high (D+) and low (D−) CW digestibility were identified using NIRS predictions of ERG concentration.The goodness of fit of the NIRS calibration model or the ratio of performance to deviation was constantly >2.6, indicating a fair to good prediction model.Their CW digestibility was subsequently validated by enzymatic hydrolysis (Duceppe et al., 2012).These D+ and D− genotypes were transplanted in 15-cm-diam.pots and transferred into a greenhouse.Plants within each group were intercrossed by hand with toothpicks to generate four Cycle 1 populations (54V54 D+1, 54V54 D−1, Orca D+1, and Orca D−1).An identical amount of seeds was harvested from each genotype.Seeds were pooled within each population.For each population, 130 plants were seeded in a greenhouse and transplanted into the field in Lévis in spring 2010.At the end of the growing season, the 110 most vigorous plants from each population were transferred into a greenhouse.The bottom parts (25 cm) of the stems of each individual plant were harvested at the seventh greenpod maturity stage (Kalu and Fick, 1981), and their ERG concentration was assessed by NIRS predictions and subsequently validated by enzymatic hydrolysis.For a second cycle of divergent phenotypic selection, the 20 genotypes with the higher CW digestibility in populations 54V54 D+1 and Orca D+1 and with the lower CW digestibility in populations 54V54 D−1 and Orca D−1 were intercrossed by hand with toothpicks to generate four new Cycle 2 populations (54V54 D+2, 54V54 D−2, Orca D+2, and Orca D−2) from which mature seeds were harvested.Progeny from these two selection cycles were used for Exp. 1.
Using a similar approach, a third cycle of positive recurrent selection was performed in 2012.For this purpose, 1000 plants from populations 54V54 D+2 and Orca D+2 were seeded in a greenhouse and transplanted into a field in Lévis (St-André sandy loam; 46°46¢17¢¢ N, 71°12¢05¢¢ W; altitude» 43 m) in the spring.At the end of the growing season (fall 2012), a visual selection based on plant vigor and yield was performed, and site (Fig. 2a and 2b).In 2014 to 2015, temperatures at the soil surface were measured every 2 h using data loggers (Hobo pro V2 external temperature data logger, Model U23, ONSET) placed at the soil surface (Fig. 2c).
The 10 populations were assigned at each site to a randomized complete block design with four replications.Rainfed plots consisted of single 5-m rows of 25 plants, each row being spaced 90 cm apart.Fertilization consisted of P, K, and B, based on soil analyses and local recommendations (CRAAQ, 2010).In Normandin, this resulted in 22 kg N, 90 kg P 2 O 5 , 90 kg K 2 O, and 0.9 kg B applied in the establishment year and 38 kg P 2 O 5 , 75 kg K 2 O, and 0.5 kg B applied annually after the last cut on the two production years (2014 and 2015).In Saint-Nicolas, fertilization consisted of 13 kg N, 50 kg P 2 O 5 , 50 kg K 2 O, and 0.5 kg B applied in the establishment year and 38 kg P 2 O 5 , 75 kg K 2 O, and 0.5 kg B applied after the last cut on the two production years (2014 and 2015).In Sainte-Anne-de-Bellevue, fertilization consisted of 59 kg N, 25 kg P 2 O 5 , and 49 kg K 2 O applied in the seeding year and 40 kg N, 25 kg P 2 O 5 , 30 kg K 2 O, and 1.0 kg B applied in spring of the two production years (2014 and 2015).At Saint-Nicolas and Normandin, weeds were controlled by the application of paraquat dichloride (1,1¢-dimethyl-4,4¢-bipyridinium dichloride) at a rate of 0.8 kg a.i.ha −1 , sprayed within 48 h after the first and second harvests.At Sainte-Anne-de-Bellevue, weeds were controlled by hand.
One harvest was taken in the establishment year (2013).In post-establishment years (2014 and 2015), three harvests per year were taken at Sainte-Anne-de-Bellevue and Saint-Nicolas, and two at Normandin.At each cut, plants were harvested at late flowering (Maturity Stage 6; Kalu and Fick, 1981).Plants were cut 5 cm aboveground, except for the last harvest of each year that was cut 10 cm aboveground for snow retention and insulation 532 plants of 54V54 D+2 and 490 plants of Orca D+2 were retained.The bottom parts (25 cm) of the stems of each individual plant were harvested at the seventh greenpod maturity stage (Kalu and Fick, 1981), and their ERG concentration was assessed by NIRS predictions and subsequently validated by enzymatic hydrolysis.The 50 genotypes with the highest CW digestibility in each population were intercrossed by hand with toothpicks to generate two additional Cycle 3 populations (54V54 D+3 and Orca D+3).Progenies from five selection cycles (D−2, D−1, D+1, D+2, and D+3) were used for Exp. 2.  S1).Air temperature and precipitations were recorded daily from May to October at each site during the 2013, 2014, and 2015 growing seasons (Fig. 1).During the overwintering period (from November to March in 2013-2014 and 2014-2015) standard air temperatures measurements (at 1.5 m in 2013-2014 and 2014-2015) were recorded at each against winter damage.The number of plants alive per row was noted at the first and last harvest of every year.Winter survival rate (%) was calculated by dividing the number of plants alive at the first harvest in the spring by the number of plants alive at the last harvest the preceding year (´100).Harvest dates and the accumulation of growing degree days (5°C base) for each harvest are presented in Table 1.For a given year, DM yield is the sum of the biomass harvested for all cuts during the growing season.
The last harvest of each year was used to measure stem ERG and WSC concentrations.For that purpose, five plants were harvested in each row, and a 500-g subsample of total biomass (leaves + stem) was dried at 55°C until constant weight for the determination of DM percentage of biomass.For the 20 remaining plants, the bottom 25 cm of stems was harvested separately from the rest of the plant above.Both parts (stems and the above part) were weighed for yield determination.A fresh 300-g subsample of stems was weighed and dried at 55°C until constant weight to determine DM percentage of stems.Dried stem subsamples from the bottom part were manually defoliated to remove any remaining leaves and ground a first time using a Wiley mill (Model Digital ED-5 midsized mill, Arthur H. Thomas Company) to pass a 2-mm screen, and a second time with a Cyclone laboratory sample mill (UDY Corporation) to pass a 1-mm screen.Defoliated stems were analyzed for ERG and WSC concentrations as described below.

Experiment 2: Field Evaluation of the Heritability of Stem Cell Wall Digestibility after Three Cycles of Selection
In a separate assay, 12 alfalfa populations (initial alfalfa cultivars 54V54 and Orca, and populations derived from both genetic backgrounds after one [D+1], two [D+2], and three [D+3] cycles of recurrent selection for higher CW digestibility, as well as after one [D−1] and two [D−2] cycles of recurrent selection for lower CW digestibility) were seeded in a greenhouse in spring 2013.After 7 wk of growth, seedlings were transplanted into a field in Lévis in June 2013 (Saint-Pacôme sandy loam; 46°46¢10¢¢ N, 71°12¢20¢¢ W; altitude » 41 m).Annual average temperature was 5.1°C, whereas the average temperature during the growing season (1 May-31 Oct.) was 15.2°C, and annual cumulated precipitation was 1164 mm (average of 2013 and 2014).Soil characteristics are presented in Supplemental Table S1.Each population was planted in rows of 20 plants, with 90-cm spacing between plants under a randomized complete block design with five replications (rows), resulting in 100 plants for each population.Model 2410 refractive index detector, controlled by Empower 2 software (Waters Corporation, 2008).The WSC were separated on an Aminex HPX-87P column (300 mm ´ 7.8 mm ´ 9 mm) preceded by a Carbo-P pre-column (Bio-Rad) and eluted isocratically at 80°C, at a flow rate of 0.5 mL min −1 , with deionized H 2 O. Carbohydrate concentrations were expressed on a 105°C DM basis determined using a thermogravimetric analyzer (Model TGA 701, Leco Corporation).

Cell Wall Preparation and Enzyme-Released Glucose Quantification
Starch was hydrolyzed by adding 3 mL of digestion buffer (200 mM sodium acetate, pH 4.5) with amyloglucosidase (15 U mL −1 , Sigma-Aldrich) in the tubes used for WSC extraction, containing the remaining supernatant and pellet.After homogenization, tubes were incubated for 60 min at 55°C and then centrifuged 10 min at 1500g.Supernatant was discarded, and pellets remaining after extraction of WSC and starch were washed three times with methanol at 60°C and air dried to obtain purified CW.Enzymatic hydrolysis was performed as described by Selig et al. (2008) and modified by Duceppe et al. (2012).Briefly, CWs (?115 mg each) were incubated for 48 h at 50°C while shaking in 10 mL of buffer containing 50 mM sodium citrate, 0.02% (w/v) sodium azide, and 25 mL of each of the following enzyme preparations: Accellerase 1500 (2638 U g −1 cellulase activity and 674 U g −1 b-glucosidase activity, Genencor), Accellerase XY (22,454 U g −1 xylanase activity, Genencor), Accellerase XC (2500-3800 U g −1 xylanase activity and 1207 U g −1 cellulase activity, Genencor), and Pectinex 3XL (3000 U g −1 pectinase activity, Sigma-Aldrich).Stuctural sugars (glucose, cellobiose, mannose, xylose, arabinose, and galactose) released from CW after enzyme hydrolysis were separated on an Aminex HPX-87P column preceded by a Deashing precolumn (Bio-Rad).Peak identity and quantity were determined by comparison with standards eluted isocratically with deionized One harvest was taken during the establishment year (2013), and three harvests were taken at the seventh greenpod maturity stage (Kalu and Fick, 1981) in 2014.The number of plants alive per row was noted at each harvest.Fertilization consisted of 38 kg P 2 O 5 , 75 kg K 2 O, and 0.5 kg B applied in the establishment year.Weeds were controlled by an application of paraquat dichloride at a rate of 0.8 kg a.i.ha −1 , sprayed within 48 h after alfalfa's first and second cuts in 2014.For the last cut in September 2014, the bottom 25 cm of stems on all plants of each row was harvested, and a fresh 300-g subsample of stems was weighted and dried at 55°C until constant weight to determine DM concentration.Dried stems were manually defoliated and ground as described in Exp. 1.The ERG concentration was predicted by NIRS after wet chemistry calibration as described below.

Near-Infrared Reflectance Spectroscopy
Ground alfalfa stems samples were scanned over 400 to 2498 nm, at 2-nm intervals, using a NIR-Systems 6500 monochromator (Foss), as described in Nie et al. (2009).Seventy-five samples were selected based on the NIR-Systems scores to form a calibration set (60 samples) and a validation set (15 samples).The 75 selected samples were analyzed for WSC and ERG concentrations by high-performance liquid chromatography (HPLC) as described below.New sets were calibrated each year.

Water-Soluble Carbohydrate Quantification
Approximately 200 mg of ground stem samples were incubated for 90 min in tubes with 7 mL of deionized H 2 O at 100°C to inhibit enzyme activity.Tubes were cooled and left overnight at 4°C for optimal extraction of soluble sugars.Tubes were then centrifuged for 10 min at 1500g, and 500 mL of the supernatant was collected for quantification of WSC (sum of raffinose, sucrose, glucose, pinitol, and fructose) by HPLC using a Model 515 pump, a Model 717Plus autosampler, and a Table 1.Harvest dates, the number of days between harvests, and accumulation of growing degree days (GDD, 5°C basis, calculated from 1 April on) for each cut during the growing season at three sites (Sainte-Anne-de-Bellevue, Saint-Nicolas, and Normandin) for Exp. 1. water at 80°C using the chromatographic system described for WSC quantification.The concentration of glucose released by the enzymatic hydrolysis of cellulose from purified CWs, and expressed on a 105°C dry CW basis, was used for the screening of the genotypes for their digestibility.

Stem and Fiber Digestibility
Stem IVTD of DM and neutral detergent fiber digestibility (NDFD) were measured in ground alfalfa stems collected on the last harvest of 2015 at the Saint-Nicolas and Normandin sites.The IVTD and NDFD were measured using a 48-h rumen fluid incubation followed by a neutral detergent fiber determination of the post-digestion residues (Goering and VanSoest 1970).The rumen incubation was performed using an ANKOM Daisy II incubator with rumen collected from a lactating dairy cow, following the procedures outlined by ANKOM Technology Corporation.The IVTD and NDFD were calculated with standard equations (Bertrand et al., 2014) and expressed as grams per kilogram DM.The Pearson correlation coefficient (r) for IVTD and ERG, as well as for NDFD and ERG, were calculated on raw data by the CORREL function of Microsoft Excel software.

Heritability Estimation
Variance components were calculated using the MIXED (Littell et al., 1996) and REML procedures of SAS (SAS Institute, 2006) according to the methods of Holland et al. (2003).Broad-sense heritability was calculated on a plot mean basis as described by Nyquist and Baker (1991)  is the variation due to the population ´ location interaction, s 2 PY is the variation due to the population ´ year interaction, s 2 PLY is the variation due to the population ´ location and year interaction, s 2 e is the residual variation, and l, y, and r are the number of locations, years, and replications, respectively.

Statistical Analyses
For the dependent variables yield, ERG, and WSC, data were subjected to an ANOVA using the MIXED procedure in SAS (SAS Institute, 2006).Replications were considered random effects, whereas sites, cultivars, selection cycles, and years were considered fixed effects.Since the same plots were harvested during multiple years, the variable "year" was considered as a repeated measure.Correlation coefficients between ERG and IVTD and between ERG and NDFD were established using raw data for each variable.
Winter survival was analyzed as binomial data (alive or dead plant) using the LOGISTIC procedure in SAS (SAS Institute, 2006), with the FIRTH option to correct for quasi-complete separation of data points.Random effects of the replications were not taken into account.Sites, cultivars, selection cycles, and years were considered as fixed effects.
Data normality was verified using the UNIVARIATE procedure, and the Shapiro-Wilk test (Shapiro and Wilk, 1965) was used to determine whether the residuals were normally distributed.The homogeneity of variance was verified visually with graphics of residuals.When model parameters had a significant F test (for yield, ERG, and WSC variables) or c 2 test (for winter survival), the LSD method was used to compare means, and mean separation output was converted to letter groupings using the Pdmix800.sasprogram (Saxton, 1998).The SEM is reported in the figures.Statistical significance was postulated at P £ 0.05.

Multisite Assay: Experiment 1
Environmental Conditions at Field Sites The three field sites used for Exp. 1 differed both in terms of air temperature and precipitation.The average temperature from 2013 to 2015 was 1.7°C for the most northern site of Normandin, whereas it was 4.4°C in Saint-Nicholas and 6.6°C for Sainte-Anne-de-Bellevue. Cumulated annual precipitation (2013)(2014)(2015) was the least at Sainte-Anne with 890 mm, the greatest at Saint-Nicolas (1257 mm), and intermediate in Normandin (975 mm) (data not shown).Weekly averages of air temperature from 1 May until 31 October show that temperature was constantly lower in Normandin, intermediate in Saint-Nicolas, and higher in Sainte-Anne (Fig. 1a-1c).Yearly distribution of the precipitations at each site during the growing season is presented in Fig. 1d to 1f.During winter 2013-2014 (from 14 November until 13 March), air temperature at the three sites dropped below 0°C before 21 November and remained constantly below 0°C throughout winter (Fig. 2a).During winter 2014-2015, the air temperature drop to below the freezing point was delayed for 1 wk in Sainte-Anne (21 November) as compared with the two other sites (14 November) (Fig. 2b).During both winters, air temperature in Normandin was constantly lower than at the two other sites.Temperature at the soil surface during winter 2014-2015 fluctuated between 1 and −11°C and was lower in Normandin between 14 November and 12 December, whereas it was lower in Sainte-Anne between 2 and 30 January (Fig. 2c).Cutting management varied between sites: two harvests were performed at Normandin, whereas milder climatic conditions allowed for three cuts to be taken at the two other sites.Number of days between harvests and accumulation of growing degree days (5°C basis) also differed among the three sites and are presented in Table 1.

Multisite Field Evaluation of Forage Attributes
We observed a significant positive response of stem CW digestibility to selection (P < 0.001) with no difference observed between the two cultivars (Fig. 3a, Supplemental Fig. S1).Averaged over the two cultivars, there was a progressive increase in CW digestibility from the initial populations to the first positive cycle of selection (+8 mg ERG g −1 CW), and from the first to the second positive cycle of selection (+13 mg ERG g −1 CW).Relative to initial populations, the average increase in ERG concentration in D+2 was of 13% (+20.7 mg ERG g −1 CW, Fig. 3a).There was no significant difference between populations D-1 and D-2, but both populations had a significantly lower ERG concentration than the initial populations (−10 mg ERG g −1 CW).Overall, CW digestibility did not significantly differ between cultivars Orca and 54V54, but there was an interaction between sites and years for that trait (P < 0.001, Fig. 3b, Supplemental Fig. S1a): the digestibility increased from 2014 to 2015 in Saint-Nicolas and Normandin (+21 and +38 mg ERG g −1 CW, respectively), whereas it decreased slightly in Sainte-Anne (−8 mg ERG g −1 CW, Fig. 3b).For a further validation of our digestibility assessment approach, we measured ERG, IVTD, and NDFD in a large collection of stem samples collected on the last harvest of 2015 at Saint-Nicolas and Normandin, and we established the correlation between values of ERG and IVTD and values of ERG and NDFD.We obtained a Pearson correlation coefficient (r) of 0.72 between ERG and IVTD and of 0.79 between ERG and NDFD (data not shown).
To assess if selection for CW digestibility had an indirect effect on other important traits, we assessed plant DM yield, stem WSC concentration, and winter survival of initial populations and populations selected for increased (D+1 and D+2) and decreased (D−1 and D−2) stem CW digestibility.Averaged over the two cultivars, we observed a significant difference in yield between populations (P = 0.018, Fig. 3c).Relative to initial populations, plant DM yield was significantly higher in D+1 and D−2.There was no significant difference between the yield of D+2 and initial populations.Some interactions were significant between years, sites, and cultivars (Fig. 3d).From 2014 to 2015, an increase in DM yield was observed in Saint-Nicolas, whereas a decrease was observed at the two other sites, mainly in Sainte-Anne.Cultivars behaved similarly over years in Saint-Nicolas and Normandin, but Orca had lower yields than 54V54 in Sainte-Anne; this difference was more striking in 2015 than in 2014 (Fig. 3d).
Averaged over the two cultivars, the effect of recurrent selection on stem WSC concentration was significant (P = 0.002, Fig. 3e).Relative to initial populations, WSC concentration was significantly lower in D−1 and D−2, whereas no difference was observed between D0 and D+1, as well as between D0 and D+2.A three-way interaction occurred between sites, cultivars, and years (Fig. 3f ).The highest WSC concentration was observed in Saint-Nicolas in 2014.At this site, a significant difference between the two cultivars was observed only in 2014, with a higher concentration of WSC in Orca than in 54V54.Globally, stem WSC concentration decreased from 2014 to 2015, and this decrease was greater at Saint-Nicolas than at the other sites.It is noteworthy that variations in stem WSC between years and sites were noticeably greater than those observed with stem CW digestibility.
Selection for stem CW digestibility did not affect winter survival (data not shown).The average survival rate for the two winters (2013-2014 and 2014-2015) was of 90% and was similar for all populations.

Heritability of Stem Cell Wall Digestibility in the Field
Broad-sense heritability was calculated in the first experiment conducted at multiple sites with the first two cycles of recurrent selection.In a second experiment conducted at one site, heritability estimates included a third cycle of positive recurrent phenotypic selection for CW digestibility (D+3).In the latter experiment, there was significant interaction between selection cycles and cultivars for stem CW digestibility (P = 0.014) as a result of a more progressive response to positive recurrent selection in 54V54 than in Orca (Fig. 4).Although the two genetic backgrounds did not initially differ, the average gain in ERG concentration with each positive cycle of selection (from D0 to D+3) was 8.4 mg ERG g −1 CW in 54V54, whereas it was of 4.6 mg ERG g −1 CW in Orca.Accordingly, broad-sense heritability estimates based on plot means indicated that the ERG concentration is under substantial genetic control in these populations and seems to be higher in 54V54 (H 2 = 0.29 and 0.73 for Exp. 1 and 2, respectively) than in Orca (H 2 = 0.21 and 0.44 for Exp. 1 and 2, respectively) (Table 2).Furthermore, the population ´ location variation (s 2 PL ) was, in both populations and both experiments, an order of magnitude smaller than the variance associated with population effects (s 2 P ).

DISCUSSION Field Evaluation of Cell Wall Digestibility in Recurrently Selected Populations
The imbalance between rapidly degradable protein and rumen-fermentable energy in alfalfa forage adversely affects ruminant performance and reproduction and contributes to environmental pollution (Berthiaume et al., 2010).Improvement of stem CW digestibility would increase energy available to the rumen microflora ( Jung and Lamb, 2006;Lamb et al., 2012).Casler (1999) estimated that breeding cultivars with 3 to 10% higher digestibility could take up to 20 yr.However, this process could be accelerated and gains increased by the development of high-throughput screening methodology specifically targeting stem CW digestibility.Measurements of ERG and NDFD provide an estimate of fiber digestibility based on a 48-h enzyme incubation period ( Theander et al., 1995) with either rumen fluid (NDFD) or a cocktail of commercial enzymes (ERG).In the current study, stem CW digestibility was determined as NIRS-predicted concentration of ERG.We validated that ERG was a good estimate of both IVTD (r = 0.72) and NDFD (r = 0.79).Duceppe et al. (2012) documented that high ERG is associated with low CW lignin underscoring the negative relationship (r = −0.83) between lignin and fiber digestibility previously documented in alfalfa ( Jung and Lamb, 2006;Tecle et al., 2008).The method we developed to assess CW digestibility based on ERG has various advantages: the 48-h hydrolysis is made using a standardized enzyme cocktail that is stable across time and among laboratories and does not depend on cyclic and variable rumen composition coming from different animals ( Jami and Mizrahi, 2012).Furthermore, whereas IVTD is determined by DM disappearance after a 48-h incubation, our method provides a precise quantification of glucose released from CW cellulose by HPLC.Finally, ERG measurements were easily calibrated with NIRS, and reliable prediction equations were developed.
Our results show that recurrent selection based on ERG had a significant impact on stem CW digestibility of two alfalfa cultivars of unrelated genetic backgrounds.We also documented that response to selection is stable under various agroclimatic conditions.Two cycles of recurrent selection for higher ERG concentration improved stem CW digestibility by 13% compared with the initial populations.This range of increase is within 10 to 15% in NDFD of a biotechnology-derived alfalfa germplasm with a suppressed lignin gene (Grev et al., 2017).Jung and Allen (1995) estimated a 5.5% increase in net return in response to a 10% increase in NDFD for a 100-cow and 100-ha dairy farm.The progressive increase in CW digestibility in response to recurrent selection is a typical feature of quantitative traits and is indicative that more gains could be achieved with additional cycles of selection.This is supported by the significant increase in ERG observed at one site after a third cycle of positive selection (Fig. 4).A more comprehensive assessment will be required to unequivocally estimate the further gains that can be achieved with additional cycles of selection.
We observed a significant site ´ year interaction for ERG indicating that the level of stem CW digestibility over years varied depending on the site where plants were grown.This was expected given the contrasted climate conditions encountered between sites and years, as well as the differences in harvest dates and frequency between sites.Previous studies highlighted the marked impact of the environment on alfalfa stem digestibility traits ( Jung and Lamb, 2006;Rock et al., 2009;Lamb et al., 2012).In that context, the systematic confirmation of higher stem CW digestibility in populations recurrently selected for higher ERG, regardless of the agroclimatic conditions, is noteworthy and underscores the stability of that trait.
Cultivars Orca and 54V54 responded similarly to selection cycles for stem CW digestibility at the three sites and during the 2 yr of Exp. 1.However, when a third cycle of selection was included, there was a significant interaction between cultivars and selection cycles, indicating that cultivars may differ in their response to recurrent selection.The same observation is true for negative selection for that trait showing a significant decrease in CW digestibility after two cycles of selection (D−2) only in Orca (Fig. 4).Divergent populations obtained from selection (D−2 and D+3) were shown to be highly contrasting in their CW digestibility.This unique genetic material is highly valuable for the search of molecular markers associated with CW digestibility.As mentioned before, a more detailed assessment of stem digestibility in D+3 populations will be required to confirm this indication of a differential response to additional cycles of selection between cultivars.Populations 54V54 D+3 and Orca D+3 obtained after three cycles of selection for higher ERG had 24 and 13% more stem CW digestibility than their respective backgrounds.Field assessment of an additional cycle of selection is currently underway to confirm if gains are sustained with additional cycles of selection.
An understanding of the inheritance of CW digestibility is important to ensure that the trait could be passed from the selected genotypes to their progeny and facilitate breeding (Wang et al., 2016).This can be estimated using broad-sense heritability (H 2 ), which indicates the proportion of variation among individuals that is the result of genetic factors over environmental factors (Holland et al., 2003).We calculated H 2 in a multisite trial (Exp. 1) and by comparing populations developed after three cycles of recurrent selection in a positive direction with their respective initial cultivars, Orca and 54V54 (Exp.2).The H 2 values obtained in Exp. 1 and 2, were 0.21 and 0.44 for Orca and 0.29 and 0.73 for 54V54, respectively, which highlights a moderate control of genetic factors over environmental factors for CW digestibility.This result is consistent with the significant phenotypic progress obtained by recurrent selection and suggests good prospects for genetic selection for that trait.However, it has to be taken into account that heritability estimates depend on experimental design, and estimates for 54V54 and Orca in the present study may depend on environmental conditions that could modify the proportion of variation arising from genetic and environmental factors (Vandemark et al., 2006).Dubé et al. (2013) identified several DNA polymorphisms associated with stem CW digestibility using pools of contrasted genotypes.This observation, associated with the quantitative nature of CW digestibility, suggests polygenic control.In that perspective, we are currently genotyping alfalfa populations recurrently selected for improved stem CW digestibility with high-throughput sequencing techniques to identify regions of the genomes affecting that trait.This information could pave the way to genome-wide selection approaches that could help rationalize plant material to be phenotypically screened and accelerate the development of germplasm with superior stem fiber digestibility.
Taken together, our results show that ERG appears as a good selection trait for genetic improvement of fiber digestibility and rumen fermentable energy in alfalfa forage.

Effects of the Selection on Other Traits
Yield and Winter Survival Yield and digestibility have often been reported to be negatively related, and it was thus important to document whether selection for CW digestibility would have an impact on alfalfa yield.In spite of yield differences between populations issued from selection cycles, there was no clear relationship between yield and ERG.For instance, we observed a greater DM yield after one cycle of positive selection and two cycles of negative selection, but there was no difference in DM yield between the initial populations and populations D+2 obtained after two cycles of selection for improved CW digestibility.This result indicates that increase in alfalfa stem CW digestibility was not achieved at the expense of DM yield.Although the early study by Kephart et al. (1989) concluded that alfalfa selected for low lignin in the total herbage had lower DM yield than high-lignin lines, our study shows that selection specifically targeting stem CW digestibility had no systematic impact on DM yield.The different outcome of the two breeding strategies could be related to the observation by Kephart et al. (1989) that breeding based on the digestibility of whole aboveground herbage caused the unintended selection of plants with higher leaf/stem ratio with lower stem yield.Furthermore, the increase in alfalfa yield that we observed in some of the selections could be due to our preselection of the most vigorous plants before CW digestibility analysis (Hopkins et al., 1993).
An improved CW digestibility in mature alfalfa stems could allow farmers to harvest at a later stage, therefore potentially obtaining greater DM yield.A less intensive cutting regime could also improve alfalfa stand life (Lamb et al., 2007).Considering the low amount of fossil fuel input required when growing a perennial legume, the utilization of alfalfa stems with higher CW digestibility along with relatively unaffected DM yield would result in a positive net energy balance for forage.
We observed interactions between sites and cultivars for yield.The lower DM yield observed in Normandin measured for both cultivars each year is likely explained by the colder growing season temperature at this site located at the northern limit of agriculture in the province of Quebec.Moreover, only two harvests per year were taken in Normandin, compared with three harvests per year in Sainte-Anne and Saint-Nicolas.The lower DM yield for Orca than for 54V54 in Sainte-Anne in 2014 to 2015 can be explained by the difference in cold tolerance between the two cultivars.Orca is an alfalfa cultivar of Flemish background known to be susceptible to winter kill (Barnes et al., 1977;Lamb et al., 1997).On the other hand, 54V54 (Pioneer Hi-Bred International) is a winterhardy-type cultivar previously selected for cold tolerance.During the winter between growing seasons 2014 and 2015, temperature recorded at the soil surface (under the snow) was lower in January at Sainte-Anne than at the other sites, likely due to insufficient snow protection (Fig. 2).Even if there was no difference in winter survival (% of plants alive in spring 2015), living plants of cultivar Orca may have suffered more damage from the cold, reducing plant vigor and yield in 2015.
This multisite study showed that selection for CW digestibility did not affect alfalfa winter survival during the 3 yr of study.Winter survival was high for all populations for both years (average of 90%).In a field study evaluating the winter hardiness of switchgrass (Panicum virgatum L.) populations obtained after several cycles of recurrent selection for increased in vitro DM digestibility, Buxton et al. (1995) observed a reduction of winter hardiness in the Cycle 3 switchgrass populations.One of the hypotheses for this result is that lower winter survival in more digestible, recurrently selected populations could be attributed to the diminution of lignin or lignin-like compounds that play a role in regulating cold tolerance.Winter survival of alfalfa populations undergoing additional cycles of selection is currently under evaluation to further characterize the effect of recurrent selection for CW digestibility on winter survival.

Stem Water-Soluble Carbohydrate Concentration
The selection for CW digestibility was accompanied with changes in stem WSC concentration, but populations D+1 and D+2 did not differ from the initial unselected populations, confirming that an improvement in stem CW digestibility is possible without a decrease in stem WSC concentration.The reduction of WSC in response to negative recurrent selection could indicate that the two traits may be partially linked.
The observed three-way interaction between sites, cultivars, and years for WSC concentration was not surprising, since plant WSC concentration is strongly affected by the environmental conditions that prevail during growth (Pelletier et al., 2010;Morin et al., 2011).The much larger spatiotemporal variability of WSC as compared with ERG underscores the greater stability of the latter trait and its value as a reliable selection criterion to improve rumen fermentable energy.

CONCLUSIONS
This study showed that recurrent phenotypic selection for higher stem ERG concentration in mature alfalfa can improve CW digestibility without affecting alfalfa DM yield, winter survival, and WSC concentration.Stem ERG concentration has a good potential as a selection trait for CW digestibility improvement with a high digestibility gain per selection cycle and a moderate heritability.The populations developed in this project are currently undergoing further cycles of recurrent selection, and the molecular characterization of this unique genetic material will accelerate the development of germplasm with superior stem fiber digestibility.

Fig. 3 .
Fig. 3. Effects of the divergent phenotypic selection on stem cell wall digestibility measured by (a) enzyme-released glucose (ERG) concentration average for the two cultivars Orca and 54V54, (c) plant dry matter (DM) annual biomass yield average for Orca and 54V54, and (e) stem water-soluble carbohydrate (WSC) concentration average for Orca and 54V54 for five populations (initial population and populations issued from selection cycles for cell wall digestibility [D+1, D+2, D−1, and D−2]).Panel b shows stem cell wall digestibility average for cultivars Orca and 54V54 grown at three sites (Sainte-Anne-de-Bellevue, Saint-Nicolas, and Normandin) in 2014 and 2015.Panel d shows dry matter yield of cultivars Orca and 54V54 grown at three sites (Sainte-Anne-de-Bellevue, Saint-Nicolas, and Normandin) in 2014 and 2015.Panel f shows stem soluble carbohydrate concentration of cultivars Orca and 54V54 grown at three sites (Sainte-Annede-Bellevue, Saint-Nicolas, and Normandin) in 2014 and 2015.Stem cell wall digestibility (a, b) and soluble carbohydrate concentrations (e, f) were measured on the last harvest of each growing season.The SEM is illustrated by vertical bars.Only significant main effects and interactions are indicated and illustrated.

Fig. 4 .
Fig. 4. Cell wall digestibility measured by enzyme-released glucose (ERG) concentration in six populations (initial population and populations issued from selection cycles for cell wall digestibility [D+1, D+2, D+3, D−1, and D−2]) of alfalfa cultivars Orca and 54V54 grown at Lévis in 2014.The SEM is illustrated by vertical bars.Only significant main effects and interactions are indicated.
using the equations

Table 2 .
Variance component estimates for genotype (s 2