Evaluation of sainfoin (Onobrychis viciifolia) for forage yield and persistence in sainfoin–alfalfa (Medicago sativa) mixtures and under different harvest frequencies

Abstract: Condensed tannins in sainfoin (Onobrychis viciifolia) improve forage digestion and reduce the risk of ruminant bloat caused by grazing alfalfa (Medicago sativa). The objectives of this study were to evaluate the forage dry matter yield (DMY) and persistence of newer sainfoin cultivars in mixtures with alfalfa, and to determine the impact of harvest frequency on the persistence of sainfoin. Sainfoin cultivars ‘AAC Mountainview’, ‘AAC Glenview’, ‘Delaney’, ‘Shoshone’, and six experimental populations were compared with ‘Nova’ and ‘Melrose’ at Lanigan, SK, from 2016 to 2018. Field plots were seeded in either monocultures of sainfoin at 33 kg·ha−1 or sainfoin–‘AC Grazeland’ alfalfa mixtures at 16:9 kg·ha−1 in alternate rows. Forage DMY was greater (P = 0.001) in mixtures than in sainfoin monocultures in all harvests. The proportion of sainfoin in mixtures at Cut 1 declined from 4.1% to 1.3% of total DMY from 2016 to 2018, and 19.0% to 4.8% in Cut 2, which was less than the recommended level to eliminate ruminant bloat risk. A second field trial was established in 2017 to compare responses of ‘AAC Mountainview’, ‘Nova’, and ‘Shoshone’ sainfoin under one-, two- or three-harvest frequencies in 2018 and 2019. The increase of harvest frequency did not reduce sainfoin stand (%). Stand percentage of ‘AAC Mountainview’ (91%) was greater (P = 0.01) than ‘Nova’ sainfoin (62%). Further agronomic studies focusing on weed control in sainfoin stands and the optimum seeding ratios of sainfoin–alfalfa within the Parkland region of Saskatchewan are needed.


Introduction
Sainfoin (Onobrychis viciifolia Scop. ssp. viciifolia) is a non-bloat causing perennial forage legume that is native to Europe and Central Asia. It is cultivated for hay in Europe, from Turkey in the south to Sweden in the north (Bhattarai et al. 2016). The non-bloating characteristic of this forage legume is due to the condensed tannins (CTs), a phenolic compound, which binds with soluble plant protein in the rumen (Gea et al. 2011;Scharenberg et al. 2007). When fed in combination with bloat-prone forages such as alfalfa (Medicago sativa L.), sainfoin reduces or prevents frothy bloat of ruminant (McMahon et al. 1999;Acharya et al. 2013).). Li et al. (1996) reported that about 1.0 mg CT·g −1 dry matter (DM) was needed to prevent the frothy bloat associated with alfalfa. Other benefits of sainfoin for grazing animals is a reduction in parasite load in the digestive tract (Hoste et al. 2012).
Adoption of sainfoin by producers has been limited since its introduction in the 1960s in Canada because it yielded only 85% of the forage production of alfalfa (Goplen et al. 1991). In recent years, there has been a renewed interest in growing sainfoin in mixtures with alfalfa to provide high quality legume mixtures for greater weight gain in cattle without the risk of ruminant bloat caused by alfalfa (Acharya et al. 2013;Bhattarai et al. 2016;Iwaasa et al. 2018). In a recent grazing study, Sottie et al. (2014) found that if a mixture of sainfoin and alfalfa contained at least 30% sainfoin by dry weight, then bloat incidences in grazing cattle were reduced by 98%. Sainfoin growth rate in spring is similar to alfalfa, but regrowth after grazing is slower than alfalfa (Hanna et al. 1972), increasing the bloat risk during the regrowth period (Acharya et al. 2013). New sainfoin cultivars with improved regrowth were developed under alfalfa competition in Canada (Acharya et al. 2013. Though sainfoin cultivars released in the 1980s were more winter hardy than the cultivars from the 1960s, they still exhibited winter kill in Canada (Goplen et al. 1991). Sainfoin may require an uninterrupted growth period prior to the first killing fall frost to prevent winter kill (Saskatchewan Forage Council 2007), and harvest frequency may negatively affect its winter survival. In addition, the demand for sainfoin combined with a lack of commercial seed production for the current cultivars resulted in importation of sainfoin cultivars from western USA to Canada (Gray et al. 2006). However, previous studies indicate that sainfoin cultivars imported from other countries and regions may lack winter hardiness (Hanna et al. 1972;Goplen et al. 1991).
The objectives of this research were to (1) evaluate the forage dry matter yield (DMY) and persistence of new sainfoin cultivars and breeding populations from Canadian and USA breeding programs in monocultures and mixtures with alfalfa, and (2) evaluate the impact of harvest frequency on forage DMY and persistence of sainfoin cultivars.

Materials and Methods
Experiment 1: Forage dry matter yield and botanical composition of mixtures Sainfoin seeds of 'AAC Mountainview', 'AAC Glenview', 'Nova', 'Melrose', and experimental breeding populations LRC05 3900, LRC053901, LRC 3519, LRC3432, LRC 4498, and LRC 4500 were obtained from Agriculture and Agri-Food Canada, Lethbridge Research and Development Centre, Lethbridge, AB, and seeds of 'Delaney' and 'Shoshone' were obtained from the Wyoming Seed Certification Service at Powell, WY. For sainfoin-alfalfa mixtures, alfalfa cultivar 'AC Grazeland' was used. Germination data were used to calculate seeding rates of sainfoin and alfalfa to the recommended rate for monoculture sainfoin at 33 kg·ha −1 and alternate row, sainfoin-alfalfa mixtures at a ratio of 16:9 kg·ha −1 .
The soil at the Lanigan, SK (51°51′N, 105°02′W) site is classified as an Orthic Black Chernozem, Meota-Hamlin association of loamy sand to very fine sandy loam texture (Saskatchewan Soil Survey, 1992) within the Thin Black Chernozem soil zone (Padbury et al. 2002). The experimental design was a randomized complete block design with 24 treatments (12 sainfoin monocultures, 12 sainfoin-alfalfa binary mixtures) in four replications. 'AC Grazeland' monoculture was also seeded as a comparison. The 12 sainfoin entries (populations) were seeded as either monocultures or in alternate rows with 'AC Grazeland' alfalfa. Individual plots were 1.8 m × 6.0 m in size. This location was used for cereal crop production for two years followed by one year of fallow prior to this trial. Soil contained adequate soil nutrients (P, K, and S) for legume production based on a soil test. Prior to seeding, 0.4 L·ha −1 glyphosate was applied to control weeds. Seeding was done on 4 June 2015 using a six-row plot drill with 20 cm row spacing. Weeds such as foxtail barley (Hordeum jubatum L.), narrow leaf hawksbeard (Crepis tectorum L.) and sow thistles (Sonchus spp.) were observed throughout the plots after seeding. To control the weeds, Odyssey (Imazamox, imazethapyr) herbicide was applied at recommended rate of 42 g·ha −1 on 26 June 2015. Additional weed invasion occurred after the initial weed control, which was removed by mowing the whole plot areas on 25 Sept. 2015 before they produced seed. Emergence was slowed by dry weather in June 2015, but stand establishment was visually determined to be adequate for future data collection in Sept. 2015. No fertilizer nor irrigation was applied at the Lanigan site.
On 8 July 2016, 6 July 2017, and 5 July 2018 all plots were harvested with a Wintersteiger forage plot harvester CiBus F (WinterSteiger, Salt Lake, UT) for forage DMY. A 500 g subsample of forage was collected, weighed, dried for 72 h at 60°C, and re-weighed to determine DM content. Then, the DMY was calculated using the DM content. An additional 1 kg subsample was collected from the harvested biomass, separated into sainfoin, alfalfa and weeds by hand, then each component was dried for 72 h at 60°C and weighed to determine botanical composition of the alfalfa and sainfoin by dry weight. On 23 Aug. 2016, 29 Aug. 2017, and 27 Aug. 2018 one random 0.5 m 2 quadrat within each plot was hand clipped and separated into sainfoin, alfalfa, and weed subsamples. The regrowth was harvested at the full bloom stage of the alfalfa. The subsamples were dried for 72 h at 60°C and weighed to determine regrowth DMY and botanical composition. Relative DMY of each plot was calculated from the DMY of the plot and 'AC Grazeland' alfalfa DMY in each replication.
Experiment 2: Effect of harvest frequency on sainfoin stand persistence This experiment was conducted at the University of Saskatchewan, Saskatoon (52°04′N, 108°08′W), SK. The soil at the site is classified as an Orthic Dark Brown Chernozem (Saskatchewan Soil Survey, 1999). No fertilizer was applied as soil analyses on 29 Apr. 2017 indicated that total available N concentration was 115.9 kg·ha −1 , available P was 104.4 kg·ha −1 , K was > 1000 kg·ha −1 , and S was 38.2 kg·ha −1 , which were adequate for forage legume establishment. The plots were seeded on 16 May 2017 at a seeding rate of 17 kg·ha −1 of sainfoin seed. The individual plots were 1.2 m × 6 m in size and consisted of four rows, spaced 30 cm apart. The experiment was a split-plot design with four replications. The main plot factor was harvest frequency (one-, two-, and three-harvest per season) and the subplot factor was sainfoin cultivar ('AAC Mountainview', 'Nova', and 'Shoshone'). Weed control was done by a preseeding application of glyphosate at 1.2 L·ha −1 on 11 May 2017. The plots were mowed on Aug. 2017 to remove weeds. No weed control was carried out in 2018 or 2019. On 10 Oct. 2018, 11:52:0 fertilizer was applied to all plots at 108 kg·ha −1 .
The plots were harvested on 13 June,17 July, and 14 Sept. in 2018 and 24 June, 29 July, and 16 Sept. in 2019 for the three-harvest treatment; 25 June and 15 Aug. in 2018, and 2 July and 20 Aug. in 2019 for the two-harvest treatment; and 29 June 2018 and 29 June 2019 for the single-harvest treatment. The plots were harvested with a Wintersteiger forage plot harvester CiBus F (WinterSteiger, Salt Lake, UT). A 500 g subsample was collected from the harvested forage biomass, weighed, dried at 60°C for 48 h, and re-weighed to determine dry matter (DM) content. The DM content, plot area, and forage mass recorded were used to calculate forage DMY.
Sainfoin samples from Cut 1 of the all treatments were analyzed for crude protein (CP), acid detergent fiber (ADF), and neutral detergent fiber (NDF). The dried plant samples were ground in a Wiley mill (Thomas-Wiley, Philadelphia, PA) and then passed through a 1-mm mesh screen (Cyclone Mill, UDY Mfg., Fort Collins, CO). For CP determination, nitrogen content in the ground samples was analyzed using a Leco 628 Element Analyzer (Leco Corporation, St. Joseph, MI). Neutral detergent fiber and ADF concentrations were analyzed using Ankom 2000 automated fiber analyzer (ANKOM Technology Corporation, NY) according to the manufacturer's instruction. In 2019, visual rating of the stand density (0 to 100% scale) was done by two independent and experienced observers. The average visual stand rating was calculated and used for statistical analysis.

Statistical analysis
For Experiment 1 at the Lanigan site, analysis of variance (ANOVA) was conducted for Cut 1, Cut 2, and total yield parameters using year as a repeated measurement over time. The UNIVARIATE procedure was used to test data for normality and homogeneity of variance.
'AC Grazeland' alfalfa monoculture data were excluded from this analysis. Thus, the model reflected 12 sainfoin entries (cultivars and experimental populations) grown in two mixtures (alternate sainfoin-alfalfa vs. sainfoin monoculture). The final statistical model was: where the observation is sum of the mean (μ), the i replication (β), the j mixture (α), the k entry (τ), the jk mixture × entry interaction (ατ), the ijk main plot error (η′) the l year (ζ), the kl entry × year interaction (αζ), jl mixture × year interaction (τζ), jkl entry × mixture × year interaction (ατζ), and the ijkl error (ε) terms. When the F test was significant (P ≤ 0.05) for a source, means were compared with Tukey's HSD test. The ANOVA was determined using JMP software version 13 (SAS Inc., Cary, NC).
In experiment 2 at the Saskatoon site, year was considered as a repeated measure in time in the combined ANOVA analysis with harvest frequency and sainfoin cultivar as fixed effects. The mixed model for the split plot design was: where the observation is the sum of the mean (μ), the ith replication (β), the jth harvest frequency (α), ijth main plot error (η′), thekth cultivar (τ), the jkth harvest frequency × cultivar interaction (ατ), the lth year (ζ), the klth harvest frequency × year (ζα), the klth cultivar × year (ζτ), the ijkth harvest frequency × cultivar × year (ζατ), and the ijklth error (ε) terms. For Cut 2 and Cut 3 DMY, the number of harvest frequency treatments declined, so these data were analyzed separately.

Experiment 1 Forage dry matter yield
The sainfoin-alfalfa mixtures produced significantly (P < 0.01) greater total DMY (including weeds), legume DMY, and contained less weeds at both Cut 1 and Cut 2 than the sainfoin monocultures (Tables 1 and 2). A greater DMY was expected for mixtures containing alfalfa than for sainfoin monocultures (Goplen et al. 1991), but the difference in this study was more than 50%. This large yield difference may be in part due to less vigorous seedlings of sainfoin resulting in higher weed density than the alfalfa stand. However, weed control by the application of Odyssey and mowing improved sainfoin seedling growth, and stand establishment was visually determined to be adequate in Sept. 2015. The droughts in 2015 and 2017 reduced the DMY of both sainfoin and alfalfa, but the impact was higher for sainfoin. This was reflected by a significant year effect (P < 0.0145) for all variables except total annual relative yield (as % of 'AC Grazeland' alfalfa) and alfalfa proportion at Cut 1 (Table 1). Sainfoin seems less drought tolerant than alfalfa, resulting in a lower yield. In a six-year study, Biligetu et al. (2014) found that sainfoin regrowth was slow after a drought stress, while alfalfa recovered to about 40% of pre-drought yield. Total DMY, legume DMY, and sainfoin DMY declined from 2016 to 2018 for both Cut 1, Cut 2, and annual forage DMY except for Cut 1 DMY of sainfoin from 2017 to 2018 (Tables 3  and 4).
Sainfoin cultivars differed for yield parameters ( Table 1). The differences at Cut 1 were small and Tukey's test was unable to indicate a significant mean separation (Table 5). At Cut 2, DMY of 'AAC Mountainview' was greater (P = 0.0171) than that of 'Delaney' and 'LRC 3432', while the other entries did not differ (Table 5). The total annual sainfoin DMY of 'AAC Mountainview' and 'Nova' were greater (P = 0.0117)   Note: Means within a column and cut followed by different letters differ (P < 0.05) by Tukey's HSD; total DMY: alfalfa, sainfoin, and weed biomass; legume DMY: alfalfa and sainfoin biomass; DMY % of alfalfa, DMY as a percentage of 'AC Grazeland' alfalfa; SE, standard error. than that of 'Delaney', but it was similar to the other 10 entries (Table 5). There was a significant (P = 0.035) cultivar × year interaction for Cut 2 sainfoin DMY (Table 1). 'AAC Mountainview' yielded significantly more forage DMY than 'Melrose', 'Delaney', and three experimental populations at Cut 2 in 2016, but there was no difference among them in 2017 and 2018 (Table 6). 'AAC Mountainview' was selected for improved regrowth (Acharya 2015), but its regrowth may have limited in 2017 due to a drier than normal year. The growing season of 2017 received 70 mm of precipitation from May to Aug. (26% of the long-term average) (Fig. 1). Total annual sainfoin DMY in monocultures declined by 72% from 2016 to 2017 (2609 vs. 723 kg ha −1 ) and by 66% from 2017 to 2018 (723 vs. 244 kg ha −1 ). In addition, low winter temperature in Feb. (-17.7°C vs. -12°C long-term average) and March 2018 (-10.3°C vs. -5.3°C long-term average) may also have had a negative impact on growth in 2018, though we did not record winter survival (Fig. 1). Ditterline and Cooper (1975) recommended that sainfoin be seeded in areas with rainfall amounts of 330 mm.

Proportion of sainfoin in mixtures with alfalfa
In the present study, the proportion of sainfoin in sainfoin-alfalfa mixtures was 2.6% (of DM) at Cut 1 and 10.5% at Cut 2 ( Table 2). The proportion of sainfoin at Cut 1 declined annually from 2016 to 2018, while the weed proportion increased each year (Table 3). At Cut 2, the proportion of sainfoin declined from 2016 to 2018, while the proportion of alfalfa increased. The percentage of sainfoin in this study was lower than the 20% to 30% (of DM) needed to avoid ruminant bloat of grazing cattle as reported by Iwaasa et al. (2018). Alfalfa dominated (P = 0.001) the botanical composition of the sainfoinalfalfa mixtures (Table 2), especially during the regrowth period (P = 0.001), increasing from 81% in 2016 to 95% in 2018 (Table 4). This suggests that the impact of alfalfa competition on sainfoin growth increased over time. Alternate row seeding did not show significant increase in sainfoin proportion under drought in this study, due to high weed pressure. Though it was not tested in this study, the same row seeding of sainfoin and alfalfa, a common practice, may further reduce the sainfoin proportion compared with an alternative row seeding Note: Means within a column and cut followed by different letters differ (P < 0.05) by Tukey's HSD; total DMY: alfalfa, sainfoin, and weed biomass; legume DMY: alfalfa and sainfoin biomass; DMY% of alfalfa, DMY as a percentage of 'AC Grazeland' alfalfa; SE, standard error.
because it would increase plant competition (Hanna et al. 1977;Jefferson et al. 1994).
proportions, if not improved it. 'AAC Mountainview' sainfoin had been selected for persistence in competition with alfalfa for grazing (Acharya 2015). However, the improved performance of 'AAC Mountainview' sainfoin compared with 'Nova' for grazing in mixtures with alfalfa as demonstrated by Sottie et al. (2014) in southern Alberta was less evident in the Parkland region of Saskatchewan in this study due to differences in climatic and soil conditions. In this study, the proportion of sainfoin was undoubtedly reduced by weed competition. Weed control, therefore, is an important aspect for sainfoin establishment and stand performance over time. While we did not test the effect of seeding rate, an alfalfa seeding rate of 9 kg·ha −1 , which is close to the full recommended seeding rate in the region, might be too high. The reduction of alfalfa seeding rate may be an important factor for improving sainfoin establishment and performance.

Weed density
The proportion of weeds was greater (P = 0.0001) in the sainfoin monocultures at Cut 1 than in sainfoinalfalfa mixtures in all three years, increasing from 55% in 2016 to 88% in 2018 (Table (4). In the first year of stand establishment, Moyer (1985) found weeds made up more than 90% of DMY in sainfoin fields without any weed control measures. This suggests that effective weed control that promote sainfoin seedling establishment in the presence of alfalfa competition is essential to establish satisfactory stands of sainfoinalfalfa mixtures. Bhattarai et al. (2016) described more than 10 herbicides registered for use in sainfoin stand. The application of glyphosate has a 10-fold greater negative impact on alfalfa DM yield than it does on sainfoin (Peel et al. 2013), which may be an option for weed control in the mixture.

Experiment 2
Forage dry matter yield 'AAC Mountainview' produced more Cut 1 DMY (P = 0.0003) and total DMY (P = 0.0187) than 'Nova' (Table 8). It also exhibited lower NDF and ADF concentrations (P = 0.004 and 0.0006, respectively) than 'Nova' sainfoin and similar CP concentration. 'AAC Mountainview' sainfoin also had higher visual stand rating in 2019 than 'Nova' sainfoin (Table 8). 'Shoshone' sainfoin was intermediate in stand rating between 'AAC Mountainview' and 'Nova'. The decline in 'Nova' sainfoin stand rating in 2019 may be attributable to its low adaptation to multiple-cut management coupled with low Feb. 2019 temperature (−24.1°C compared with long-term average temperature of −11.4°C) with below average snow cover for insulation (Fig. 2). Jefferson et al. (1994) reported that low winter temperature stress can result in reduced plant survival and production in the following growing season and that 'Nova' sainfoin was susceptible to low winter temperature stress at Swift Current, SK.
'AAC Mountainview' sainfoin was selected for persistence in mixtures with alfalfa and improved regrowth yield (Acharya 2015). We observed greater Cut 1 and total annual forage DMY and persistence of 'AAC Mountainview' compared with 'Nova'. Bhattarai et al. (2018) reported that 'Nova' sainfoin was similar in winter survival to 'Shoshone' (88% vs 80% survival) and superior to many other germplasm. They harvested forage once per year followed by seed production from the regrowth, which was a more lenient harvest regime than was used in the current study. 'Nova' sainfoin had been selected for improved winter survival compared with other cultivars available in the 1980s (Hanna 1980). 'Nova' was also reported to exhibit lower regrowth yield compared with multi-cut cultivars from other countries (Hanna 1980). 'Shoshone' sainfoin was selected in Wyoming for improved persistence through increased resistance to root-knot nematode (Gray et al. 2006). It appears to have some potential adaptation in regions where nematodes have not been reported to reduce the persistence of sainfoin.

Harvest frequency and year
The year effect was significant (P < 0.05) for all variables ( Table 7). The harvest frequency significantly (P = 0.0024) affected total forage DMY of sainfoin, where two and three harvests per year yielded more total forage DMY than one harvest (Table 8). Two harvests per year yielded more Cut 2 DMY than three harvests per year. There was no difference in stand rating due to harvest frequency (P = 0.42, Table 7) despite the low temperature observed in Feb. 2018 (Fig.2). In fact, total forage DMY was increased in 2019, mainly because of an increase in Cut 2 and Cut 3 DMY compared with 2018 (Table 8). The above average precipitation in June and July 2019 (Fig. 2) contributed to increased regrowth and forage DMY. Similarly, CP concentration was greater (P = 0.0001) and ADF and NDF concentrations were lower (P = 0.001 for both) in 2019 compared with 2018. There was no significant (P = 0.8844) cultivar × harvest frequency interaction for stand rating in 2019. In our study, the three harvest per year treatment was harvested in Sept. 2018 and there was a 'killing frost' on 29 Sept. (−4.9°C), but there was no evident reduction in stand rating in 2019 due to harvest frequency. As has been reported for alfalfa, the importance of critical fall harvest period may depend on the winter hardiness of the cultivar for the region (Jefferson and Gossen 1992).
However, our study was conducted at one location over three years. Therefore, a multi-location study over a longer period of time will be necessary to further verify fall harvest management of sainfoin stands for persistence and winter survival.

Conclusions
Sainfoin productivity and persistence in mixtures with alfalfa are required traits to provide successful bloat-safe grazing of sainfoin-alfalfa pastures by cattle. A sufficient proportion of sainfoin biomass must be present in the mixture to ensure sufficient intake of  condensed tannins to bind the soluble protein from alfalfa in the anaerobic digestion process of the rumen and reduce the production of stable foam that traps methane gas and causes bloat. The minimum proportion of sainfoin in sainfoin-alfalfa mixtures has been reported to be 20% to 30% of DM, yet none of the tested sainfoin-alfalfa mixtures in the study met this requirement at Lanigan, SK, from 2016 to 2018. Weed control and low alfalfa seeding rate that promote sainfoin seedling establishment in the presence of alfalfa competition is essential to establish satisfactory stands of sainfoinalfalfa mixtures. The selection for persistence of sainfoin in competition with alfalfa has improved in 'AAC Mountainview', but further genetic improvement for the Parkland region of Saskatchewan is needed.
Sainfoin stand declined both at Lanigan and Saskatoon under different growth environments can be attributed to weed and alfalfa competitions, water stress, low winter temperature stress, or their combination. 'AAC Mountainview' sainfoin was more persistent than the current standard, 'Nova' sainfoin, but this was not related to harvest frequency at the Saskatoon site. We conclude that more testing of 'AAC Mountainview' and 'AAC Glenview' sainfoin is needed in the Parkland region of Saskatchewan to confirm their adaptation and use for bloat-safe alfalfa grazing.