Diversity for morphological traits, flowering time and leaf isoflavone content among ecotypes of Trifolium subterraneum L. subsp. yanninicum and their relationships with site of origin

Keywords: eco-geographic variables, genetic resources, genotypic diversity, leaf marks, morphological traits, oestrogenic compounds, pasture legume, subterranean clover.

Subterranean clover (Trifolium subterraneum L.), native to the Mediterranean Basin and surrounding regions, is the most widespread pasture legume in southern Australia, where it has been sown over 29 Mha (Nichols et al. 2013; Ghamkhar et al. 2015). It is also increasingly sown in other parts of the world with Mediterranean-type climates (Porqueddu et al. 2016). A range of cultivars with diverse maturities enables subterranean clover to be grown successfully in different rainfall environments and soil types (Nichols et al. 1996; Nichols 2013). Also key to its extensive adoption are attributes that enable it to be simply managed by farmers for high production and persistence. These include: tolerance of heavy grazing due to a prostrate growth habit and ability to bury seed-bearing burrs; ability to self-regenerate after long, dry summers; tolerance to many herbicides; and farmer experience from historical usage (Nichols et al. 2013).

Subterranean clover comprises three subspecies, which Katznelson and Morley (1965) classify as: subsp. subterraneum, subsp. brachycalycinum, and subsp. yanninicum. Of these, subsp. yanninicum, comprising 10 of the 53 subterranean clover cultivars, is broadly used in medium and high annual rainfall (450–1200 mm) areas (Nichols et al. 2013; PBR 2021), typically in paddocks with poor drainage. Several agronomic and ecological studies show greater persistence of subsp. yanninicum than of the other two subspecies on soils prone to waterlogging (Reed et al. 1985; Craig 1992; Dear et al. 2003). Glasshouse studies confirm its greater waterlogging tolerance (Marshall and Millington 1967; Francis and Devitt 1969; Devitt and Francis 1972; Francis et al. 1974; Gibberd and Cocks 1997; Enkhbat et al. 2021). Current cultivars of subsp. yanninicum comprise a narrow genetic base (Nichols et al. 2013), and the diversity within the subspecies has been little studied (Ghamkhar et al. 2015). Hence, an understanding of the diversity within subsp. yanninicum could lead to a broadening of the genetic base for plant breeding and thereby enable desirable traits from other ecotypes to be utilised.

Diversity for several subterranean clover growth traits have been examined (Piano 1984; Piano et al. 1993, 1996; Pecetti and Piano 1997, 2002; Nichols et al. 2009; Abdi et al. 2020). Flowering time is highly responsive to environment and is considered the most important trait for adaptation of annual legumes (Nichols et al. 2007, 2009, 2012; Abdi et al. 2020). Petiole length is important for light interception, and large leaves are crucial for biomass accumulation because they increase light capture (Davidson and Donald 1958; Black 1960; Gladstones 1967; Nichols et al. 2009, 2013). Long internodes allow growth farther away from the mother plant crown and increased competition and are important particularly when the growing season is longer (Cocks 1992; Nichols et al. 2009).

Genetic variation also exists for the isoflavone compounds formononetin, genistein and biochanin A in fresh leaves of subterranean clover (Francis and Millington 1965; Morley and Francis 1968; Nichols et al. 2013; Abdi et al. 2020). Of these, formononetin is converted to highly oestrogenic compounds in the rumen, causing infertility in ewes (Bennetts et al. 1946; Millington et al. 1964; Hungerford 1975; Brightling 2006). Collins and Cox (1984) and Davies (1986) suggest that formononetin levels <0.3% on a leaf dry weight basis are ‘safe’ for sheep reproduction. Breeding programs have aimed to select cultivars with ≤0.20% of leaf dry matter (Nichols et al. 2013). Highly heritable morphological traits such as leaf marks are also useful characters for distinguishing subterranean clover genotypes for seed certification (Gladstones 1967; Tan and Collins 1987; Nichols et al. 1996; Abdi et al. 2020).

Germplasm collected from different sites within the native habitat of subterranean clover provides a source of genetic diversity (Ghamkhar et al. 2007, 2015) that can be evaluated for morphological characteristics (Tucak et al. 2009). Phenotypic traits can be used to investigate adaptation characteristics by relating them to climatic/eco-geographic variables (precipitation, temperature, latitude, etc.) from their sites of origin (Erskine et al. 1989; Piano et al. 1996; Mousavi-Derazmahalleh et al. 2018; Abdi et al. 2020). Understanding phenological adaptations corresponding to eco-geographic variables of collection sites is vital for: (i) identifying the distribution and adaptation of genotypes; (ii) guiding future germplasm collections and identifying diversity hotspots; and (iii) enhancing knowledge on how existing variation is linked with eco-geographic site-of-origin variables (Erskine et al. 1989; Hill 1996; Piano et al. 1996; Pecetti and Piano 2002; Rosso and Pagano 2005; Ghamkhar et al. 2007; Mousavi-Derazmahalleh et al. 2018; Abdi et al. 2020).

This study investigated diversity of morphological traits, flowering time and leaf isoflavone content corresponding to eco-geographic variables, in order to improve knowledge of diversity among subsp. yanninicum. The study tested two hypotheses: (i) variation for morphological traits, flowering time and leaf isoflavone content exists within subsp. yanninicum; and (ii) these variations are related to eco-geographic variables at the site of origin. These results, and those from other agronomic and physiological studies, will be used to identify adapted parents with new traits for genetic improvement of subsp. yanninicum to meet the future needs of farmers.

Germplasm (seeds) of 154 subterranean clover genotypes from the Australian genetic resource collection of purported subsp. yanninicum was obtained from the Australian Pastures Genebank (APG), operated by the South Australian Research and Development Institute. These genotypes consisted of material deemed by the APG to be of sufficient quantity and viability to be made available for research; a further 28 genotypes were unavailable. Genotypes obtained comprised 144 ecotypes, selected as individual genotypes on the basis of morphological uniformity; and 10 commercial cultivars: Yarloop, Larisa, Meteora, Trikkala, Gosse, Riverina, Napier, Monti, Rouse and Yanco. Among the ecotypes, 142 were collected from their native habitats in the Mediterranean Basin, and two (Neuchatel and Yabba North) are naturalised strains from Australia. Of the cultivars, Yarloop is a naturalised strain from the Yarloop district of Western Australia (WA), first commercialised in 1947 (Oram 1989; Nichols et al. 1996, 2013), and Meteora (CPI 03927YA) and Larisa (CPI 039313Y) are ecotypes collected in Greece. The other cultivars are all derived from cross-breeding and have Larisa and/or Meteora in their parentage (Nichols et al. 2013; PBR 2021). Of the two naturalised strains, Neuchatel is an early flowering mutant of cv. Yarloop from the Waroona district of WA (Gladstones and Collins 1984) and is also an ancestor of all the crossbred cultivars (Nichols et al. 2013; PBR 2021), whereas Yabba North was collected from Victoria, Australia (Aitken and Drake 1941).

The subspecies identity of the 144 purported subsp. yanninicum ecotypes was determined during the growing season, on the basis of distinguishing morphological features of the three subterranean clover subspecies described by Katznelson and Morley (1965) and Nichols et al. (2013). Genotypes were classified as subsp. yanninicum if they had all of the following features: corolla length similar to calyx tubes; strong positive geotropism with active burr burial; burrs with few whorls of sterile calyces that covered all or most of the pods; leathery and transversely wrinkled pods; and round to slightly ovoid seeds. All genotypes classified as subsp. yanninicum also had glabrous or weakly pubescent stems, petioles and leaf upper surfaces, and the majority had cream–amber-coloured seeds. On the basis of this assessment, it was determined that 36 purported subsp. yanninicum ecotypes from the APG had been misclassified. Hence, in total, 118 genotypes of subsp. yanninicum (108 ecotypes and 10 cultivars) were examined in this study. Among these, four ecotypes had black-coloured seeds; all other genotypes had the usual amber- or cream-coloured seeds. These genotypes are listed in Supplementary data Table S1.

Geographic coordinates were used to determine accurate locations of collection sites. Associated passport data (latitude, longitude and altitude) recorded by the plant collectors were obtained from the former Australian Trifolium Genetic Resource Centre and the Germplasm Resources Information Network (GRIN) global website (https://apg.pir.sa.gov.au/gringlobal), using numbers allocated by the APG for genotype identities (Supplementary data Table S1). Passport data were available for 90 genotypes of subsp. yanninicum and the geographic distribution of their collection sites is shown in Fig. S1.

Standardised temperature and precipitation (BIOCLIM) variables were downloaded for collection sites of known latitude and longitude from the WorldClim (ver. 2) website (http://www.worldclim.org) at 2.5 arc-minutes spatial resolution (Hijmans et al. 2005; Ghamkhar et al. 2015). These were used for mapping and spatial modelling in the R software package (http://www.worldclim.org/formats1). Definition and codes of the 19 standard BIOCLIM variables are provided in Table S2 (Hijmans et al. 2005). Temperature data were divided by 10 to convert to degrees Celsius (http://www.worldclim.org). Table S3 presents a summary of the BIOCLIM variables and associated passport data, and Table S4 provides imported climatic variables for each ecotype.

Prior to sowing, seeds of each genotype were scarified as described in Nichols et al. (2009). Seeds were sown on 21 May 2019 by germinating in Petri dishes containing a single Whatman No. 1 filter paper moistened by 3.5 mL tap water. These were placed in a germination cabinet at 15°C. At 2–3 days after sowing (DAS), two germinated seedlings of each genotype were transplanted into hydrated Jiffy-9 peat pots (Jiffy Products, Stange, Norway) in a glasshouse. Rhizobium leguminosarum bv. trifolii strain WSM1325 (Nodulaid; BASF Australia, Melbourne, Vic.) was applied by watering can (40 mL inoculum to 9 L water).

Peat pots were watered daily to prevent moisture stress, and soluble fertiliser (Thrive; Yates Australia, Sydney, NSW) was applied weekly at a rate of 0.5 g/L. Seedlings were thinned to a single healthy seedling per Jiffy pot at 15 DAS and transferred to outside benches for acclimatisation at 21 DAS. Four healthy seedlings of each genotype (except genotype CPI 039315YB, which had only two replicates) were transplanted into the field at 31 DAS.

The experiment was conducted at The University of Western Australia Field Station at Shenton Park (31°57′S, 115°5′E) under netting to prevent bird damage. The soil at the experimental site is a deep sand with pH(H₂O) 6.5 in the top 20 cm. The field was rotary-hoed and cultivated in March. For weed control, glyphosate (540 g a.i./L) at 2 L/ha was applied in early April, and the field was sprayed with Basta (BASF Australia; a.i. glufosinate-ammonium, 200 g/L) at a rate of 1.5 L/ha 7 days before transplanting the seedlings. Single superphosphate with potash (6.8% P, 12.4% K and 8.3% S) was applied at 300 kg/ha in early May. The area was smudged to level the soil before laying six parallel 1-m-wide, 100 m long polyethylene plastic strips (to restrict weed growth), each separated by 1.5 m of bare soil. Seedlings were transplanted into holes cut into the plastic strips.

The climate at Shenton Park is Mediterranean. The mean maximum and minimum temperatures were 21.9°C and 12.3°C, respectively, with 544.2 mm rainfall during the experimental period (May–November 2019) (www.bom.gov.au). Mean monthly maximum and minimum temperatures and rainfall during the experimental period in the study area are shown in Fig. S2 (www.bom.gov.au). Plants were checked several times each week and supplemental overhead irrigation was supplied as required from 1 October to ensure that plants did not suffer moisture stress, in accordance with Nichols et al. (2009) and Abdi et al. (2020). Irrigation ceased on 20 November to allow plants to senesce.

A randomised complete block (row–column) design with four replicates was used. Replicates were arranged perpendicular to (i.e. across) the six plastic strips, which acted as columns. Plants were spaced 0.85 m apart along the strips to allow growth without competition for light and moisture and to prevent edge effects. The plastic strips were removed from the field, to allow burr burial, on 25 August, when the earliest flowering genotypes had just commenced flowering.

In addition to the 154 purported subsp. yanninicum genotypes planted in the study, 14 genotypes comprising subspp. subterraneum and brachycalycinum were included. Thus, the experimental layout consisted of 672 (i.e. 168 × 4) spaced plants and these were arranged in 112 rows across the six plastic strips, with each replicate made up of 28 rows. Only data for the 118 confirmed subsp. yanninicum genotypes (108 ecotypes and 10 cultivars) are reported here.

Measurements were taken on each plant as described in Nichols et al. (2009) and Abdi et al. (2020). At 62–63 DAS (midwinter), the length of the longest petiole of each plant was measured and the size of the leaf (three leaflets combined) on the longest petiole was measured by using a portable leaf area meter (LI 3000; LI-COR Biosciences, Lincoln, NE, USA).

At 87–88 DAS (end of winter), the length of the longest petiole and leaf size on the longest petiole were measured via the same methods as used for the midwinter measurements. This time was chosen so that all genotypes could be compared simultaneously as late as possible while still in their vegetative phase, before flowering of the earliest genotypes.

The number of days from sowing to the appearance of the first open flower (days to first flowering) was measured, with plants checked every 3 days. The stem (runner) with the first flower was marked with a crossing tag to identify the first flowering node for measurement of morphological traits at first flowering. Leaf size and petiole length on the first flowering node were measured 14–30 days after first flowering. In this case, leaf size was measured by visually comparing leaves with the photographic leaf size plates of Williams et al.(1964) and using their conversion formulae as per Nichols et al. (2009) and Abdi et al. (2020). Internode length and stem diameter were measured between the first and second flowering nodes on the stem with the first flower. Stem diameter was measured with a micrometer (Mitutoyo, Tokyo) with 0.01-mm intervals. These measurements were conducted to compare genotypes at the same physiological stage during the reproductive phase.

Isoflavone contents (formononetin, genistein and biochanin A) were measured by using the semi-quantitative technique of Francis and Millington (1965), as used by Ghamkhar et al. (2012), Nichols et al. (2009) and Abdi et al. (2020). Samples of six leaf discs (6 mm diameter) were collected from healthy, newly fully opened leaves at 71–72 DAS (first and second replicates) and 80–81 DAS (third and fourth replicates) for analysis. Isoflavones were immediately extracted in alcohol and subjected to thin-layer chromatography. Duplicate samples taken for dry weight determination were oven-dried at 60°C for 48 h. Isoflavone contents were calculated as a percentage of dry matter (DM).

Leaf marks were rated at 67 DAS, when they were most prominent owing to cold winter temperatures. Standard semi-quantitative descriptors, based on Nichols et al. (1996) and used by Abdi et al. (2020), were utilised (Table S5). Briefly, leaf marks are described in terms of ‘crescents’, ‘arms’ and ‘bands’, and variation can take the form of absence or presence, size, shape and position of each (Tan and Collins 1987; Nichols et al. 1996). Leaf mark patterns are classified by: (i) widths of crescents, which are central triangular markings (usually pale green) that extend from the leaflet centre towards the margins and range in size from a central dot (rating 1) to all the way from margin to margin (rating 4); (ii) breadth of arms, which are stripes (usually white or pale green) that extend from the leaflet margins towards the centre and range from narrow (rating 1) to very broad (rating 4); and (iii) breadth of bands, which are an alternative marking to crescents and arms, consisting of a pale green stripe extending from margin to margin and varying in breadth from narrow (rating 1) to broad (rating 3). Leaf marks can consist of crescents alone, arms alone, crescents with arms, bands alone or no markings at all. Ratings were taken on plants in only two replicates, owing to the high heritability and consistency of markings among plants of the same genotype (Tan and Collins 1987; Nichols et al. 1996).

Data were analysed by using one-way ANOVA to compare genotypic variation among means of plant traits, using R software (ver. 3.6.3; The R Foundation, Vienna). This was conducted: (i) among all 118 subsp. yanninicum genotypes; and (ii) within and between ecotype and cultivar groups of subsp. yanninicum. Least significant differences (l.s.d.), using Fisher’s protected tests, were used to compare means for each measured trait between the ecotype and cultivar groups of subsp. yanninicum, and to compare means for each trait among all genotypes.

Pearson correlation coefficients and their levels of significance were calculated between traits for the 118 genotypes of subsp. yanninicum. They were also calculated between plant traits and eco-geographic (passport and BIOCLIM) variables of collection sites for all 90 subsp. yanninicum genotypes (88 ecotypes and two cultivars) with known collection site information. Broad-sense heritability (H²) was estimated for each trait across all genotypes of subsp. yanninicum, according to Falconer and Mackay (1996).

Broad variation (Table 1, Table S6) with highly significant genotype differences (Table S7) was observed for all morphological traits, flowering time and leaf isoflavone contents among the 118 genotypes. Ecotypes had greater variation than cultivars for all measured traits.

Table 1.  Summary of mean, minimum (Min.) and maximum (Max.) values of morphological traits, flowering time and leaf isoflavone contents for ecotypes and cultivars of Trifolium subterraneum subsp. yanninicum.

Flowering time was highly heritable (92%) and ranged from 94 to 149 days (Table 1, Table S6) with highly significant (P < 0.001) variation among genotypes (Table S7). Cv. Meteora (149 days) and ecotypes EP059WHITE-F (149 days), CPI 083967 L (148 days), CPI 083948B (147 days) and EP059WHITE-D (147 days) were the last to flower. By contrast, ecotypes Neuchatel (94 days), N4848 (98 days), and DGG041-A, DGG046-B and EP082WHITE (all 100 days) were the earliest flowering. However, mean flowering time of the cultivars was significantly earlier than the ecotypes (Table 1). A frequency distribution of mean flowering time is shown in Fig. 1.

Fig. 1.  Frequency distribution of mean flowering time (number of days from sowing on 21 May to first open flower) for 118 Trifolium subterraneum subsp. yanninicum genotypes, consisting of 108 ecotypes and 10 cultivars, grown at Shenton Park, Western Australia.

Leaf isoflavone content varied substantially among genotypes (all P ≤ 0.001), with ranges 0.05–1.38% (of DM) for formononetin, 0.73–2.33% for genistein, 0.15–2.10% for biochanin A, and 1.11–4.46% for total leaf isoflavone content (Table 1, Tables S6 and S7). Heritability was high for both formononetin (60%) and biochanin A (52%) contents, and lower for genistein (23%) and total isoflavone (15%) contents (Table 1). Mean values for formononetin and genistein differed between ecotype and cultivar groups (both P ≤ 0.001), but not for biochanin A or total isoflavones (Table 1). Highest formononetin contents were found in N4848 (1.38%), cv. Yarloop (1.36%) and EP082WHITE (1.33%). By contrast, CPI 103933, EP059WHITE-F and CPI 039314YB had formononetin contents ≤0.06%. Formononetin content was significantly higher than 0.20% in 65.7% of ecotypes.

For leaf marks, broad variation was found among genotypes (Table S6). The most common leaf mark was a crescent with arms (98 genotypes), followed by a crescent only (11 genotypes) and arms only (eight genotypes); one genotype (CPI 083957N) had no leaf mark. For leaf marks with a crescent, 64% of genotypes had a 2–3 crescent width rating. For leaf marks with arms, 57% of genotypes had a 2–3 arm breadth rating and 31% had a 1–2 arm breadth rating; ecotype N4848 scored 4 because of its very broad arms.

Pearson correlation coefficients between mean values of all variables are shown in Table 2. Leaf size and petiole length were strongly correlated at 63 DAS (r = 0.51; P ≤ 0.001), 88 DAS (r = 0.66; P ≤ 0.001) and first flowering (r = 0.51; P ≤ 0.001). Flowering time was negatively associated with both leaf size (r = –0.46; P ≤ 0.001) and petiole length (r = –0.31; P ≤ 0.001) at 88 DAS. However, at flowering time, later flowering genotypes had longer petioles (r = 0.29; P ≤ 0.01), larger leaves (r = 0.34; P ≤ 0.001), longer internodes (r = 0.27; P ≤ 0.01) and thicker stems (r = 0.20; P ≤ 0.05). Notably, internode length was strongly associated with petiole length at first flowering (r = 0.63; P ≤ 0.001) and at both 63 DAS (r = 0.45; P ≤ 0.001) and 88 DAS (r = 0.23; P ≤ 0.05).

Table 2.  Pearson correlation coefficients and significance between mean values of 10 morphological traits, flowering time and leaf isoflavone contents for 118 Trifolium subterraneum subsp. yanninicum genotypes (108 ecotypes and 10 cultivars).

Biochanin A and genistein contents were unrelated, but both, particularly genistein (r = –0.42; P ≤ 0.001), had a negative association with formononetin content (Table 2). Some associations between leaf isoflavone contents and morphological traits and flowering time were found. Flowering time was negatively associated with formononetin content (r = –0.26; P ≤ 0.01) but positively associated with genistein (r = 0.18; P ≤ 0.05) and biochanin A (r = 0.25; P ≤ 0.01) contents. However, there was no relationship between total leaf isoflavone content and flowering time.

For leaf marks, there were several associations with morphological traits, flowering time and leaf isoflavone contents (Table 2). Crescent size increased with flowering time (r = 0.31; P ≤ 0.001), whereas arm breadth decreased (r = –0.25; P ≤ 0.01). Broad arms were also associated with short internode length (r = –0.33; P ≤ 0.001) at flowering and with short petiole length (r = 0.21; P ≤ 0.01) at 63 DAS. Crescent size had weak associations with low formononetin (r = 0.18; P ≤ 0.05) and high biochanin A (r = 0.22; P ≤ 0.05) contents. There was, however, no association between crescent size and arm breadth.

Pearson correlation coefficients for 10 morphological traits, flowering time and leaf isoflavone contents with their site of origin variables were calculated for the 90 genotypes with passport data (cvv. Larisa and Meteora and 88 other ecotypes). Correlation coefficients are shown in Table 3 for passport data and eight standard BIOCLIM precipitation variables (BIO12–BIO19), and in Table 4 for 11 standard BIOCLIM temperature variables (BIO1–BIO11).

Table 3.  Pearson correlation coefficients (r) and significance for mean values of 10 morphological traits, flowering time and leaf isoflavone contents with 11 eco-geographic variables (passport data and BIOCLIM precipitation variables BIO12–BIO19) for 90 Trifolium subterraneum subsp. yanninicum genotypes with known passport data.

Table 4.  Pearson correlation coefficients and significance for mean values of 10 morphological traits, flowering time and leaf isoflavone contents with 11 BIOCLIM temperature variables (BIO1–BIO11) for 90 Trifolium subterraneum subsp. yanninicum genotypes with known passport data.

Leaf size and petiole length in midwinter (63 DAS) and late winter (88 DAS) were both correlated positively with longitude, indicating that genotypes from the eastern part of the subspecies range had larger plants in winter than those from the western part. Large leaves at 88 DAS also tended to be associated with low altitudes (r = –0.33; P ≤ 0.001). Leaf size and petiole length during winter were positively correlated with the majority of precipitation variables, particularly annual precipitation (BIO12), and with variables related to winter precipitation (BIO13, BIO16 and BIO19). Petiole length at 63 DAS was also positively correlated with summer precipitation variables (BIO14, BIO17 and BIO18).

High altitude and high summer precipitation (BIO14, BIO17 and BIO18) favoured later flowering genotypes with large leaves, long petioles and long internodes at first flowering (Table 3). Early flowering was associated with high winter precipitation (BIO13, BIO16 and BIO19). Notably, flowering time and both petiole and internode lengths at first flowering were negatively correlated with annual mean temperature (BIO1) and with both winter (BIO6, BIO9 and BIO11) and summer (BIO8 and BIO10) temperatures (Table 4). Stem diameter was not correlated with any eco-geographic variables.

The inverse associations between formononetin content and both genistein and biochanin A contents was reflected in their associations with several eco-geographic variables. High formononetin content was associated with easterly longitudes, whereas high genistein and biochanin A contents were both associated with more westerly longitudes. Both total isoflavone and biochanin A contents had positive relationships with altitude, whereas genistein content was associated with latitude. Annual precipitation (BIO12) and variables related to winter precipitation (BIO13, BIO16 and BIO19) were positively correlated with formononetin content but negatively correlated with genistein and biochanin A contents.

Leaf mark crescent size was not correlated with any eco-geographic variable. However, the breadth of leaf mark arms was associated with several variables. Broad arms were associated with both low altitudes and more westerly longitudes (Table 3) as well as drier summers (BIO14, BIO17 and BIO18) (Table 3) and warmer mean annual temperatures (BIO1) and warmer winter (BIO6, BIO9 and BIO11) and summer (BIO10) temperatures (Table 4). By contrast, narrow arms were associated with more easterly longitudes with increasing precipitation.

The wide diversity for all traits measured among subsp. yanninicum genotypes supports the first hypothesis of this study. This shows that there is potential to broaden the gene pool for subsp. yanninicum in southern Australian pastures by introducing new traits. Larger leaves, longer petioles and longer internodes could provide increased competitive ability (Black 1960; Gladstones 1967; Nichols et al. 2009, 2013) beyond that of the current cultivars. Likewise, both the latest and earliest flowering genotypes were found among the ecotypes. A wide range of flowering times enables successful growth of subsp. yanninicum in locations with different growing season lengths (Nichols et al. 2009). Later flowering genotypes, with their longer vegetative period, have the opportunity to accumulate more biomass with larger leaves and longer petioles, which is reflected by higher spring DM yield, making them well suited to areas with long growing season (Rossiter 1959; Chapman et al. 2012; Pecetti and Piano 1993, 1997; Nichols et al. 2009; Abdi et al. 2020; Pecetti et al. 2020). Earlier flowering is crucial for persistence in shorter season environments, because it enables adequate seed production prior to onset of terminal drought (Gladstones 1967; Piano 1984; Nichols et al. 1996; Piano et al. 1996; Piano et al. 1993; Pecetti and Piano 2002). Formononetin content was beyond the level considered safe (i.e. 0.2% of DM) in the majority of ecotypes. Cultivars released since 1975 in Australia have undergone deliberate selection for formononetin levels below this safe level, (Nichols et al. 2013), so it was not surprising that the cultivars had a significantly lower level than the ecotypes. Nonetheless, this experiment revealed other ecotypes with low formononetin levels that could be used as alternative parents for developing non-oestrogenic cultivars. Furthermore, the use of high formononetin ecotypes as parents is not excluded, because the high heritability of formononetin content enables ready selection for low formononetin progeny when crossed with a low formononetin parent (Nichols et al. 2013).

This study supports previous findings of Gladstones (1967), Nichols et al. (2009) and Pecetti and Piano (1997) that earlier flowering genotypes tend to have larger leaves and longer petioles during winter growth, whereas later flowering genotypes have larger leaves and longer petioles at flowering. This confirms that leaf growth and petiole elongation of subsp. yanninicum increases progressively with subsequent nodes until the first flowering node, similar to observations with subsp. subterraneum (Gladstones 1967; Pecetti and Piano 1997).

The observation that earlier flowering genotypes tended to have higher formononetin content, whereas later flowering genotypes tended to have higher biochanin A and genistein contents, agrees with the results of Cocks (1992), Morley and Francis (1968) and Nichols et al. (2009) for subsp. subterraneum. In the present study, there was no relationship between genistein and biochanin A, whereas studies by Morley and Francis (1968), Abdi et al. (2020) and Nichols et al. (2013) across all subspecies found a negative relationship. This suggests different biochemical regulatory pathways may be involved for these compounds in subsp. yanninicum.

The second hypothesis, that variation in plant traits is related to the eco-geography of their collection sites, was also supported. Longitude played an important role to discriminate the collection sites and it was well correlated with annual precipitation of the site of origin, supporting Bennett (2000), Bennett and Bullitta (2003) and Ghamkhar et al. (2007). The association of large leaf size at the end of winter with low altitude differs from the report of Abdi et al. (2020), who found no such association, but is consistent with results from studies of Sardinian (Pecetti and Piano 1997) and Sicilian (Pecetti and Piano 1993) subterranean clover ecotypes.

The differences between the results of this study, which focused on subsp. yanninicum, and other studies that focused on the other two subspecies, is most likely due to differences in subspecies distributions, with subsp. yanninicum being much more restricted geographically and ecologically than the other two subspecies (Katznelson and Morley 1965; Francis 1976; Ghamkhar et al. 2015). The natural habitat of subsp. yanninicum is reported to be low-altitude coastal areas (often flat meadows) with high mean annual rainfall (>500 mm) and mild winter temperatures (Katznelson 1966; Francis 1976; Ghamkhar et al. 2015). This contrasts with the natural habitats of the other two subspecies, which occur at altitudes up to 2190 m a.m.s.l. (for subsp. brachycalycinum) and 2940 m (for subsp. subterraneum) and can be found in areas with as little as <300 mm mean annual rainfall (Ghamkhar et al. 2015).

This study supports the assertion of Abdi et al. (2020) and Pecetti and Piano (1993, 1997) that plants from higher elevations tend to have smaller leaves as an adaptation to cold stress. Earlier flowering genotypes of subsp. yanninicum generally had larger leaves in winter, consistent with Pecetti and Piano (1997) and Abdi et al. (2020). Notably, ecotypes with longer petioles tended to originate from higher rainfall sites. This is associated with their later flowering time and is possibly an adaptation to submergence.

This study confirms previous studies of annual legumes by Piano (1984), Piano et al. (1993, 1996), Pecetti and Piano (2002), Cocks (1992), Nichols et al. (2010), Erskine et al. (1989) and Ehrman and Cocks 1996 that natural selection for flowering time is strongly influenced by eco-geographic variables (latitude, altitude, precipitation and temperature). Flowering time of subsp. yanninicum was correlated with latitude in this study, supporting the results of Ghamkhar et al. (2015) and Abdi et al. (2020). Flowering time of subsp. yanninicum also increased with increasing altitude and rainfall of the collection site, confirming the results of Piano (1984) for Sardinian subterranean clover ecotypes. This is in spite of the restricted elevation of subsp. yanninicum compared with the other two subspecies (Ghamkhar et al. 2015) and its common position lower in the landscape (Katznelson and Morley 1965; Francis 1976; Piano 1984), which tends to collect more water. The correlations of flowering time with winter and annual rainfall in subsp. yanninicum are in agreement with the studies of subterranean clover world (Ghamkhar et al. 2015) and core (Abdi et al. 2020) collections, Sicilian (Piano et al. 1993) and Sardinian (Piano 1984; Piano et al. 1996) ecotypes, and a bulk-hybrid subsp. subterraneum population (Nichols et al. 2009). This can be explained as a plant response to adjust its maturity to soil moisture availability in order to facilitate seedset and burr burial prior to onset of the dry summers experienced in the Mediterranean region (Piano 1984; Piano et al. 1993; Pecetti and Piano 2002). Furthermore, later flowering genotypes can remain vegetative and accumulate biomass with larger leaves and longer petioles for a longer period in favourable rainfall environments (Nichols et al. 2009). The results of this study also agree with the assertion of Abdi et al. (2020) and Pecetti and Piano (1997) that high winter growth is associated with warmer climates.

The observation that sites with high winter rainfall tended to produce subsp. yanninicum genotypes with increased formononetin contents and decreased genistein and biochanin A contents contrasts with the results of Abdi et al. (2020) and Ghamkhar et al. (2015), who found no relationships between rainfall variables and the contents of individual isoflavones or total isoflavone. This difference may reflect the focus of their studies on the other two subspecies of subterranean clover. It also suggests that isoflavone production in subsp. yanninicum may respond to soil moisture conditions differently from that in the other two subspecies, and may be related to its superior waterlogging tolerance (Marshall and Millington 1967; Francis and Devitt 1969; Devitt and Francis 1972; Enkhbat et al. 2021).

The finding in this study that leaf mark arm breadth for subsp. yanninicum was correlated with several eco-geographic parameters at the collection sites was unexpected. Among the subterranean clover core collection, Abdi et al. (2020) found an association of arm breadth with soil pH at the collection site, but not with other eco-geographic parameters. However, neither study found any association of crescent size with eco-geographic parameters. Whereas plant growth traits might be expected to have some adaptive relationships with their sites of origin, the significance of the associations between morphological markers and environmental factors is unclear. Abdi et al. (2020) suggest the genes controlling them may be linked to other genes controlling adaptive traits that are responsive to different environments. The accumulation of particular regional characteristics such as the predominance of broad leaf mark arms in the western part of the subsp. yanninicum range may also be due to genetic drift, as a result of its scant distribution and the difficulty of genetic exchange between widely spaced populations.

This study demonstrated that a range of ecotypes exists that could be used as parents in plant breeding programs to expand the genetic base of subsp. yanninicum, which has been heavily reliant on derivatives of the Greek ecotypes CPI 039313Y (cv. Larisa) and CPI 039327YB (cv. Meteora) and the WA naturalised strain, Neuchatel (Nichols et al. 2013; PBR 2021). Ecotypes, rather than cultivars, had the maximum values for leaf size and petiole length in both midwinter and late winter and at flowering, indicating the potential to increase the leaf size and petiole length of current cultivars.

Flowering time of the ecotypes spanned that of the cultivars, indicating little scope to breed earlier or later flowering subsp. yanninicum cultivars. Indeed, there appears to be no need for earlier or later flowering cultivars in southern Australia. Short-season environments tend not to have prolonged waterlogging and are better suited to subsp. subterraneum cultivars, which flower earlier than the subsp. yanninicum cultivars (Nichols et al. 2013). The marketplace for very late flowering subsp. yanninicum cultivars appears to be small, because it tends to overlap with areas suited to perennial legumes such as white clover (T. repens L.), which can provide year-round green forage. The drying climate with shorter growing seasons in southern Australia also reduces the need for long-season annuals (Revell et al. 2012).

This study focused on diversity for plant growth traits, flowering time and leaf isoflavone content. Other studies are being conducted with this material to identify subsp. yanninicum ecotypes with important agronomic traits for adaptation to farming conditions. Such traits include high biomass and seed production potential, hardseededness, resistance to important pests and diseases, and tolerance to waterlogging and other environmental stresses. This will enable selection of parents with desirable traits to be included in crossing programs to improve productivity and persistence of current cultivars.

A major limitation to the continued improvement of subsp. yanninicum is the size of its genetic resource, being confined to 55 collection sites and comprising only 2.2% of the world collection of 8812 subterranean clover genotypes from 2993 sites (Nichols et al. 2013; Ghamkhar et al. 2015). The value of future selection and breeding activities would be enriched by targeted collecting, with detailed passport data, to broaden the genetic diversity of subsp. yanninicum. In particular, its genetic resource can be diversified by collecting in regions where it is naturally distributed but which are not yet represented in annual legume genebanks, such as Croatia, Slovenia, Albania and Serbia (Zohary and Heller 1984; Ghamkhar et al. 2015).

This study was conducted under highly controlled experimental conditions on undefoliated, weed-free, spaced plants on well-drained and well-fertilised soil. Some aspects of plant performance are likely to differ in swards, particularly under grazing. Further research is also needed to investigate plant performance of different subsp. yanninicum ecotypes under waterlogged conditions, particularly because the subspecies is widely reported as having greater tolerance than subspp. subterraneum and brachycalycinum (Marshall and Millington 1967; Francis and Devitt 1969; Devitt and Francis 1972; Francis et al. 1974; Gibberd and Cocks 1997). Other environmental factors such as low temperatures and nutrient deficiencies are also likely to influence plant growth. The common garden conditions of this study did enable estimation of the genetic basis of the measured traits. Time to first flowering, isoflavone contents and leaf markings are highly heritable and values are likely to be transferable between seasons and locations under similar conditions (Nichols et al. 2013). However, field trials under typical farming conditions at different sites with diverse climates and soils are crucial for understanding the adaptation and agronomic merits of subsp. yanninicum genotypes.

This study demonstrated that although the available genetic resource of subsp. yanninicum is relatively small, there is sufficient diversity to broaden the genetic base of the subspecies and develop new cultivars to increase the productivity of pastures in southern Australia and other regions with Mediterranean-type climates. We also conclude that further genetic improvements may result from targeted collecting, particularly in areas not represented in annual legume genebanks.

Supplementary material is available online.

The data that support this study are available in the article and accompanying online supplementary material.

The authors declare no conflict of interest. Megan Ryan is an Associate Editor of Crop and Pasture Science, but was blinded from peer-review process for this paper.

This research was conducted as part of the Annual Legume Breeding Australia (ALBA) Joint Venture between PGG Wrightson Seeds and UWA. The project is under the International Collaborative Program in Agricultural Sciences between Nagoya University and UWA. G.E. is supported by an Australian Government Research Training Program (RTP) scholarship and a Science Industry PhD Fellowship from the Government of Western Australia. Kevin Foster was funded by the UWA-MLA DC project P.PSH.1171.