Field Crops Research 293 (2023) 108831

Contents lists available at ScienceDirect

Field Crops Research

journal homepage: www.elsevier.com/locate/fcr

Value of heterogeneous material and bulk breeding for inbred crops: A pea
case study
P. Annicchiarico a, *, L. Russi b, M. Romani a, T. Notario a, L. Pecetti a
a
Council for Agricultural Research and Economics (CREA), Research Center for Animal Production and Aquaculture, Viale Piacenza 29, 26900 Lodi, Italy
b
University of Perugia, Department of Agricultural, Food and Environmental Science, Borgo XX Giugno 74, 06121 Perugia, Italy

A R T I C L E I N F O A B S T R A C T

Keywords: Context: Genetically heterogeneous materials of inbred crops, such as evolutionary populations (EPs) and line
Breeding scheme mixtures, could reduce the genetic erosion and increase the yield stability of crops. EPs may also be exploited for
Evolutionary breeding bulk-based selection of inbred lines as an alternative to single-seed descent (SSD)-derived lines.
Intercropping
Objective: This study focused on pea to compare: (a) biparental EPs from three connected crosses and their
Line mixture
Yield stability
combination, and their bulk-derived inbred lines and 6-line and 12-line static mixtures, for grain yield, yield
stability, and reliability (which combines mean yield and yield stability); (b) bulk-based vs. SSD-based selection
of inbred lines.
Methods: Bulk-derived lines were mass-selected after four-year natural selection and underwent further multi-
environment selection of elite lines and components of mixtures. EPs, selected bulk-derived lines, and static
mixtures were evaluated across eight environments of Northern and Central Italy encompassing organic or
conventional management and pure stand or intercropping with cereals. Bulk-based vs. SSD-based selection of
inbred lines were compared across three environments. Farmers’ acceptability and various adaptive traits were
also recorded. We also assessed shifts for adaptive traits between EPs or derived lines for the target region and
material selected for a non-target region featuring severe terminal drought.
Results: In a comparison based on best-performing material, a 6-line mixture exhibited 4–7 % greater yield or
yield reliability, and EP material showed a modest penalty (actually nil in its area of initial development),
compared with bulk-derived lines. Most of this material tended towards greater yield, yield stability and reli­
ability (in a context of large genotype × environment interaction), competitive ability in intercropping and
farmers’ appreciation than the best-performing commercial cultivar. Bulk-derived lines displayed 24 % higher
mean yield, 2.3–2.5 greater yield gain per selection cycle and greater farmers’ acceptability than SSD-derived
lines, and were the only lines that allowed for genetic progress over commercial cultivars. Their advantage
was associated with greater tolerance to low winter temperatures and taller plant stature that favored their
adaptation to intercropping or organic farming. The EPs selected in contrasting environments or their derived
lines displayed rapid shifts in adaptation pattern and associated traits.
Conclusions: Evolutionary breeding issued highly valuable populations, inbred lines and line mixtures. EPs dis­
played quick adaption to specific environments for a low selection cost.
Implications: Our results have implications for small and large breeding programs, to increase the efficiency of
phenotypic selection and favor the optimal integration of genomic selection.

1. Introduction keep pace with the population growth, while concurrently adapting to
climate change. This challenging effort can be jeopardized by genetic
Farming systems require a sustainable increase of crop production to erosion and loss of biodiversity, which already exceeded a safe boundary

Abbreviations: ANOVA, analysis of variance; EP, evolutionary population; GEI, genotype × environment interaction; MS, mixed stand (alias intercropping); PS,
pure stand; RIL, recombinant inbred line; SSD, single-seed descent.
* Corresponding author.
E-mail addresses: [email protected] (P. Annicchiarico), [email protected] (L. Russi), [email protected] (M. Romani), tommaso.notario@crea.
gov.it (T. Notario), [email protected] (L. Pecetti).

https://doi.org/10.1016/j.fcr.2023.108831
Received 5 September 2022; Received in revised form 12 December 2022; Accepted 20 January 2023
Available online 6 February 2023
0378-4290/© 2023 Elsevier B.V. All rights reserved.
P. Annicchiarico et al. Field Crops Research 293 (2023) 108831

for human development (Steffen et al., 2015). Modern agriculture (2016) observed shifts for adaptive traits across organic and conven­
caused a progressive genetic erosion of cultivated material by (a) tional crop management, after just five or six generations in small-grain
restricting the number of crops grown in space and time and (b) cereals.
reducing the allele richness and allele evenness within each crop Only a few studies compared EP vs. line mixture breeding strategies.
through the replacement of landraces of inbred crops with a small Patel et al. (1987) reported higher mean yield of EPs over dynamic
number of genetically homogeneous cultivars grown across vast areas mixtures of barley under natural selection, attributing this result to
(van de Wouw et al., 2009). The latter trend could be counterbalanced greater potential for segregation, recombination and heterosis of EPs. In
by the re-introduction of genetically heterogeneous cultivars of inbred Legzdiņa et al. (2022), EPs of barley tended to show higher mean yield in
crops, if they proved to be at least as valuable as pure line cultivars. organic and stress-prone environments, and greater Type I stability, than
Heterogeneous material may include evolutionary population (EP) and dynamic mixtures. In Allard (1961), EPs of lima bean exhibited greater
cultivar (or breeding line) mixture material (Ceccarelli and Grando, mean yield and yield stability than static mixtures. Döring et al. (2015)
2020; Wuest et al., 2021). EPs originated by pooling equal amounts of reported in wheat a somewhat greater yield advantage over the mean of
progeny seed issued from various crosses and evolved across many contributing parents for static mixtures compared with EPs (3.6 % vs.
generations under natural selection in a target environment were pro­ 2.4 %) along with a trend towards greater Type I yield stability of EPs
posed by Harlan and Martini (1929) and recommended by Suneson over mixtures, which resulted in a similar yield reliability (as expressed
(1956), who named these selection strategies “evolutionary plant by an index combining mean yield and yield stability) of these variety
breeding”. EPs may also be selected by combining mass selection with types.
natural selection (Murphy et al., 2005; Phillips and Wolfe, 2005; Cec­ EPs may also be used as a germplasm source for pure line selection
carelli and Grando, 2022). Cultivar mixtures are mostly static, i.e., (Phillips and Wolfe, 2005; Raggi et al., 2017), through procedures of
assembled by pooling same seed amounts of their component lines ordinary bulk breeding in the case of populations originated by two
before each cropping year, but may also be dynamic, i.e., subjected to parents. Likewise, bulk breeding may originate valuable biparental EPs
evolutionary adaptation through changes of original genotype fre­ (Merrick et al., 2020), although a higher number of parents is usually
quencies due to natural selection under specific cropping conditions adopted (Döring et al., 2011). Some reports (Döring et al., 2015;
(Ceccarelli and Grando, 2020). For long, EU regulations allowed just the Brumlop et al., 2017) suggested that EPs originated from many parents
marketing of static mixtures of registered cultivars. The new EU regu­ may fail to display sufficient yielding ability compared with locally
lation for organic farming allows for marketing all kind of EPs and best-adapted parent pure lines, because of the genetic load caused by
mixtures starting from 2022. suboptimally-adapted parent material. Compared with single-seed
The agronomic value of cultivar mixtures has mostly been studied in descent (SSD), bulk breeding schemes are less expensive, may increase
cereal crops. In a meta-analysis, Reiss and Drinkwater (2018) reported the chance of identifying favorable trait recombinations via a higher
2.2 % greater yielding ability of mixtures over the mean of their number of advanced inbred lines, and may increase the frequency of
component lines, along with a trend towards greater yield stability of genotypes adapted to a target environment due to natural selection; on
mixtures (especially across cropping years) in terms of Environmental the other hand, they delay the production of inbred lines by preventing
variance, i.e., according to the so-called static (or Type I) stability the performance of off-season generations (Simmonds, 1979; Ranalli
concept (Becker and Léon, 1988). An earlier review by Smithson and and Cubero, 1997). Witcombe and Virk (2001) recommended bulk
Lenné (1996) reported (a) a mixture advantage over the mean yield of breeding-based selection within just a few crosses between
component lines in the range 0.7–5.4 % depending on the species, (b) carefully-selected, elite parent genotypes. Another important option for
the rare occurrence of mixture advantage over the best-performing bulk-based breeding is growing the selected inbred lines as intra-specific
component line, and (c) a trend towards greater yield stability of mix­ mixtures instead of pure lines (Murphy et al., 2005; Fig. 1).
tures over pure lines due to lower genotype × environment interaction The ability of bulk breeding to produce top-yielding lines, and that of
(GEI), i.e., according to the dynamic (or Type II) stability concept EPs to achieve high yielding ability, require a positive relationship be­
(Becker and Léon, 1988). The advantage of mixtures over pure lines tween individual plant fitness under intra-specific competition and crop
tended to increase with increasing severity of a major biotic or abiotic yielding ability (Hamblin and Rowell, 1975). A loose or even a negative
stress (Finckh et al., 2000; Reiss and Drinkwater, 2018). The reported relationship may occur in the presence of large within-population
number of components of grown or experimentally-tested mixtures was variation for plant stature, because taller genotypes display greater
mostly in the range 3–6 (Smithson and Lenné, 1996; Finckh et al., 2000), intra-specific competition for light (Goldringer et al., 2001; Knapp et al.,
with a trend towards greater yield progress of mixtures including at least
four components (Reiss and Drinkwater, 2018).
Murphy et al. (2005) and Phillips and Wolfe (2005) considered EPs
of special interest for low-input environments, although their agronomic
value was first shown by Suneson (1956) for barley in
agriculturally-favorable environments. EPs exhibited greater yield and
yield stability than the mean yield of their parent lines in a pioneer study
on lima bean by Allard (1961). Later studies focusing on cereal crops
revealed greater Type I or Type II yield stability along with similar or
somewhat lower mean yield of EPs relative to top-performing parent or
commercial pure lines (Soliman and Allard, 1991; Döring et al., 2015;
Brumlop et al., 2017; Raggi et al., 2017; Bocci et al., 2020; Goldringer
et al., 2020; Merrick et al., 2020; van Frank et al., 2020; Baresel et al.,
2022). This result was confirmed for wheat intercropped with pea
(Timaeus et al., 2022), a scenario that may maximize the yielding
ability, resilience and resource use efficiency of low-input systems
through exploitation of inter-specific diversity (Bedoussac et al., 2015).
Phillips and Wolfe (2005) envisaged up to 15 generations of natural
selection to select well-performing EPs. However, Rasmusson et al. Fig. 1. Schematic representation of pea breeding schemes into focus: single-
(1967), Bocci et al. (2020) and Legzdiņa et al. (2022) reported successful seed descent-based selection of a pure line, and bulk breeding-based selection
adaptation to specific growing environments, and Bertholdsson et al. of a pure line, a mixture of pure lines or an evolutionary population.

2
P. Annicchiarico et al. Field Crops Research 293 (2023) 108831

2020) while possessing lower grain yielding ability due to lower in­ Miladinovič et al., 2011; Meena and Kumar, 2012; Khosla et al., 2019).
vestment in reproductive development (Khalifa and Qualset, 1975). This study focuses on field pea with the main objective of comparing
Such a drawback may be limited by avoiding to cross parent genotypes different breeding strategies summarized in Fig. 1, namely, SSD-based
with contrasting plant stature, such as tall and semi-dwarf types. Natural pure line selection, and bulk breeding-based selection of EPs, pure
selection towards a taller stature within either of these plant types may lines and static line mixtures. The comparison was based on grain yield
actually by positive, especially for organic systems, to increase the crop and yield stability across Italian target environments belonging to
competitive ability against weeds (Döring et al., 2011). Northern or Central Italy, which encompassed organic or conventional
Greater cultivation of high yielding and resilient grain legumes has management and pure stand (PS) or mixed stand (MS), alias intercrop­
strategic importance in Europe and other regions, to increase the sus­ ping, with cereals. Test plant material derived from a common genetic
tainability of agriculture in terms of crop diversity, soil fertility, energy base obtained by crosses among a limited number of elite,
efficiency, greenhouse gas emissions and pollution on the one hand, and geographically-diversified parent cultivars, envisaging EPs and mixtures
to decrease the huge dependency from international markets for high- of different genetic diversity. Additional objectives of this study were to
protein feedstuff on the other (Foyer et al., 2016; Pilorgé and Muel, assess (a) shifts for adaptive traits between the EPs or their derived lines
2016; Watson et al., 2017). It also affects crucially the expansion of for the target region and EPs or derived lines selected from the same
organic farming (Barbieri et al., 2021). Pea (Pisum sativum L.) has special genetic base for a non-target region featuring severe terminal drought,
interest for Southern Europe because of greater production of grain and and (b) the intrinsic advantage of heterogeneity in large mixtures of
feed energy per unit area than other cool-season grain legumes bulk-derived or SSD-derived lines.
(Annicchiarico, 2008). Various studies (Annicchiarico and Iannucci,
2008; Pecetti et al., 2019) support a wide-adaptation strategy for pea 2. Materials and methods
breeding targeting Italy, because GEI for grain yield is more affected by
year-to-year climatic variation than by geographical distance of growing 2.1. Parent cultivars and generation of inbred lines and evolutionary
sites. Information on the value of evolutionary breeding strategies or the populations
comparison of EPs vs. line mixtures is extremely limited for this or other
grain legumes. Comparisons of bulk vs. SSD (or single-pod descent) The genetic base for this study derived from paired crosses between
breeding schemes for ability to produce top-yielding inbred lines are not three semi-dwarf, semi-leafless cultivars of different phenological type,
numerous for these crops, and their results are largely inconsistent namely, Attika (a European cultivar described as a spring-type), Isard (a
(Haddad and Muehlbauer, 1981; Kumar et al., 1992; Gill et al., 1995; French winter-type cultivar), and Kaspa (an Australian cultivar of

Fig. 2. Timing and stages of generation for each of three pea connected crosses of (a) single-seed descent (SSD)-derived inbred lines (in red), (b) evolutionary
populations (EPs), EP-derived inbred lines and mixtures of EP-derived lines for the target region (Northern and Central Italy) (in blue), and (c) EPs and EP-derived
inbred lines for a non-target region (severely drought-prone Mediterranean region) (in green). Two shades of blue in (b) distinguish generation and testing of EP-
derived inbred lines and line mixtures (dark blue) from those of EPs (light blue).

3
P. Annicchiarico et al. Field Crops Research 293 (2023) 108831

Mediterranean type). These cultivars revealed high and stable grain random sample of 396 plants per plot on the ground of plant seed yield
yield and modest difference in crop maturity across environments of previously dried for a few weeks in a non-heated glasshouse, selecting
Northern and Southern Italy in the evaluation of 53 modern cultivars 15 plants per plot.
bred in Europe or Australia (Annicchiarico, 2005; Annicchiarico and Each EP for the non-target region was sown manually around mid-
Iannucci, 2008). Progeny material from crosses of these parent lines is November at 100 seeds/m2 density in 30 neighboring grids, each
indicated hereafter as A×I, K×A and K×I from the initials of the relevant including 10 plants arranged in two rows of five plants at 0.1 m spacing
parents. between and within rows (sowing density of 100 seeds/m2), applying
Our development of inbred lines, line mixtures and EPs is summa­ increasing within-grid selection intensity across generations (30 %, 20 %
rized in Fig. 2. From each cross we generated a recombinant inbred line and 10 % selected fraction in the first, second and third generation,
(RIL) population of 60 F6 lines via SSD (path a in Fig. 2). Concurrently, respectively) based on plant seed yield dried in a non-heated glasshouse
bulked seed provided in equal amount by 60 F2 plants was used for each (path c in Fig. 2). Drought stress increased as well across generations, as
cross to develop: one EP for the target region (Northern and Central a consequence of water amounts in the first, second and third generation
Italy) through four cycles of natural selection under autumn-sown field equal to 250, 200 and 180 mm, respectively, over October-May, and
conditions in Lodi, Northern Italy (45º19′ N, 9º30′ E), and 30 bulk- equal to 20, 15 and 10 mm, respectively, over April-May (the latter
derived F6 inbred lines through a final stage of plant mass selection amounts being drastically lower than those provided by rainfall to the
(path b in Fig. 2 b); one EP for the non-target region represented by dry EPs for the target region: Table 1). We pooled the seed of mass-selected
Mediterranean-climate conditions through three cycles of mass selection plants after each cycle (thereby allowing for population evolution also as
under managed severe drought in an autumn-sown rain-out sheltered a function of the different yielding ability of the drought-tolerant
environment equipped with micro-sprinklers, and 30 bulk-derived F5 selected plants). Other differences between field and managed
inbred lines through a final selection stage (path c in Fig. 2 c). The EPs growing environments were the milder low-temperature stress in winter
for the non-target region, which underwent much greater terminal and the greater high-temperature stress in late spring of the managed
drought stress compared with those for the target region, contributed to environment that arose from its plastic cover. All growing environments
assess the effect of the evolution environment on the adaptive response received pre-sowing mineral fertilization (24 kg/ha N, 72 kg/ha P2O5,
and a few adaptive traits of the resulting material. and 72 kg/ha K2O).
EP material was always grown in two plots per cross. Great care was
taken to favor a balanced intra-specific competition between plants 2.2. Assessment of SSD-derived and bulk-derived inbred lines, global line
across EP generations. We ensured consistent spacing and 3 cm depth of mixtures and evolutionary populations: layout of experiments (Exp. 1, 2
sown seed by using a pneumatic seed drill (Hege 95) under field con­ and 3)
ditions and manual sowing in the managed drought stress environment,
leaving at least one row of non-harvested border plants under any This study took place in the cropping seasons 2013–14 and 2014–15
circumstance. Weeds were controlled chemically by Stomp® 330 E (a.i. after a stage of seed multiplication of inbred lines in 2012–13. It
Pendimethalin at 307 g/L) at 4.5 L/ha and occasionally by additional included, for each of three crosses, 60 individual SSD-derived lines and
weeding. Each EP for the target region was sown at the beginning of their global mixture, 30 individual RILs obtained by mass selection from
November at 90 seeds/m2 density in one plot of 12 rows spaced 0.2 m the EP for the target region and their global mixture, 15 individual lines
and 2.5 m long in the first year, and two plots of 18 rows spaced 0.2 m obtained by mass selection from the EP for the non-target region
and 4 m long in the following three years. Rainfall and temperature (randomly chosen out of the 30 ones available), the EP for the target
information for these cropping years is given in Table 1. Mass selection region, and the EP for the non-target region (Fig. 2). The line mixtures
within each EP for the target region was performed in the last year on a (obtained by bulking equal seed amount from each component line)

Table 1
Climatic variables and mean pea grain yield of test environments used for early cycles of natural selection of evolutionary population (EP) material for Northern and
Central Italy, evaluation experiments for inbred line selection and other material, and experiments aimed to compare inbred lines, cultivars, line mixtures and EPs.
Item Locationa Year Managementb Rainfall Jan.- Rainfall Apr.- Abs. min. daily Mean of max. daily No. of Mean grain
Mar. (mm) May (mm) temp. (◦ C) temp., May (◦ C) frost days yield (t/ha)c

EP early natural
selection cycles
- F2 progeny plants Lodi 2007–08 Conv, PS 131 156 –6.8 22.8 29 –
- F3 progeny plants Lodi 2008–09 Conv, PS 274 181 –13.6 28.1 54 –
- F4 progeny plants Lodi 2009–10 Conv, PS 285 171 –10.5 23.1 56 –
- F5 progeny plants Lodi 2011–12 Conv, PS 15 181 –17.0 24.0 68 –
Line evaluation
experiments
- Experiment 1 Lodi 2013–14 Org, PS 343 122 –5.7 23.2 35 6.31
- Experiment 2 Perugia 2013–14 Org, PS 280 179 –3.6 23.4 9 2.91
- Experiment 3 Lodi 2014–15 Conv, PS 198 147 –11.6 23.9 34 4.64
Comparison of
selections
- Experiment 4 Lodi 2014–15 Org, PS 198 147 –11.6 23.9 34 1.13
- Experiment 5 Perugia 2014–15 Org, PS 223 73 –2.9 26.0 15 2.97
- Experiment 6 Lodi 2015–16 Org, PS 258 147 –12.0 21.8 26 0.94
- Experiment 7 Perugia 2015–16 Org, PS 194 217 –5.4 22.7 13 1.74
- Experiment 8 Lodi 2018–19 Conv, PS 75 233 –12.0 19.3 54 2.87
- Experiment 9 Lodi 2018–19 Conv, MS 75 233 –12.0 19.3 54 1.16
- Experiment 10 Lodi 2019–20 Conv, PS 126 66 –10.9 23.7 34 3.60
- Experiment 11 Lodi 2019–20 Conv, MS 126 66 –10.9 23.7 34 1.12
a
Lodi, Northern Italy; Perugia, Central Italy.
b
Conv = conventional; Org = organic; PS = pure stand; MS = mixed stand with cereals.
c
Averaged across individual test entries (as represented by inbred lines, cultivars, mixtures or evolutionary populations); yield data for MS doubled to be reported to
same unit area as PS.

4
P. Annicchiarico et al. Field Crops Research 293 (2023) 108831

aimed to test the intrinsic effect of genetic heterogeneity. The total 327 method, expressing the variation for the first two components in terms of
entries, and the three parent lines, were evaluated in Lodi (representa­ genetic coefficient of variation (CV) through division of their square root
tive of the subcontinental climate typical of Northern Italy) for two years values by the mean value of the relevant set of lines and reporting CV
and in Perugia (43◦ 06′ N, 12◦ 23′ E; characterized by a cool Mediterra­ values averaged over crosses. The relative efficiency of these line se­
nean climate that is widespread in Central Italy and inland Southern lection strategies was assessed in terms of yield gain of their 10 % top-
Italy) in 2013–14. The evaluation encompassed the two major crop yielding germplasm over the best-performing parent line, i.e., the six
managements for pea in Italy, namely, organic (the two sites in top-yielding SSD-derived lines vs. the three top-yielding bulk-derived
2013–14) and conventional (Lodi in 2014–15). lines, both for single crosses and over crosses.
Each experiment was laid out in a randomized complete block design All statistical analysis in this report were carried out using SAS
with three replications. The six EPs, the six line mixtures and the three (2011) statistical software.
parent lines were replicated thrice within each block. Each plot had
0.96 m2 size and included four rows 1.2 m long, 0.2 m apart. Fifteen 2.4. Assessment of elite inbred lines, line mixtures and evolutionary
seeds per row (overall seed density = 62.5 seeds/m2) were sown at 3 cm populations: layout of experiments (Exp. 4–11)
depth by a mechanical seed drill (Hege 80). Sowing took place in late
October or November following seedbed preparation with ploughing This assessment neglected substantially the SSD-derived lines,
and harrowing. Pre-sowing mineral fertilization (24 kg/ha N, 72 kg/ha because of their definite inferiority compared with bulk-derived lines
P2O5, and 72 kg/ha K2O) and chemical weed control [(Stomp® 330 E (a. that emerged from results of Exp. 1–3. We assessed, for each of the three
i. Pendimethalin at 307 g/L) at 4.5 L/ha] were applied only in the crosses, the following materials: (a) the two top-yielding lines based on
conventionally-managed environment. The crop was always harvested each of two criteria, namely, mean yield across Lodi and Perugia in the
by combine (Wintersteiger Nursery-master Elite plot combine) within season 2013–14 (Exp. 1–2), and mean yield across three test environ­
the first ten days of June, visually-assessing for the global mixtures ments (Exp. 1–3); (b) one 6-line static mixture and one 12-line static
(which included lines of different maturity) the harvest date that could mixture, including the six and 12 top-yielding lines across Exp. 1–2,
maximize the grain yield. The following traits were recorded on a plot respectively, whose maturity date did not differ for more than two days
basis: (a) winter plant survival, based on plant counts at the onset and (a condition satisfied by the vast majority of the lines, owing to the
the end of winter; (b) onset of flowering, as the number of days after levelling of maturity caused by terminal drought stress); (c) the EP for
April 1 when 50 % of plants in the plot had at least one fully open flower; the target region, using EP seed collected over replicates from Exp. 1 for
(c) plant height at onset of flowering; (d) lodging susceptibility, visually Lodi and Exp. 2 for Perugia in the first test year (Exp. 4 and 5), and
assessed at maturity on a 5-level scale ranging from 1 = lodging limited location-specific EP seed collected in the most recent preceding pea pure
to the basal part of the stem to 5 = complete lodging; (e) dry grain yield, stand experiment in the following years (hence, allowing for progressive
after assessing seed moisture on a random sample of 250 seeds oven- location-specific population evolution). Maturity date of line mixture
dried at 90 ◦ C for four days; (f) individual dry seed weight, assessed components was taken into account, to avoid yield losses due to mark­
on the seed sample used for seed moisture determination. Two addi­ edly dishomogeneous maturity in the combine-harvested crop. In
tional traits were observed in two environments, namely, dry aerial addition, we tested mixture and EP material featuring greater genetic
biomass, recorded only in Lodi, and farmers’ acceptability score, diversity because it was obtained by combining germplasm from the
attributed in the season 2013–14 by nine organic farmers from Northern three crosses, namely: (a) one 6-line static mixture and one 12-line static
Italy in Lodi and nine from Central Italy in Perugia a few weeks before mixture including, respectively, the two top-yielding lines per cross and
crop maturity by a scale ranging from 9 = very high to 1 = very low the four top-yielding lines per cross across Exp. 1–2, with a maturity
(allowing for half-unity values). This evaluation score proved highly range not exceeding three days; (b) one EP obtained by pooling same
predictive of the future crop performance in independent years in a seed amount from each EP developed from each cross, starting from EP
previous study (Annicchiarico et al., 2019b). In each site, farmers were seed collected from Exp. 1 for Lodi and Exp. 2 for Perugia in the first year
subdivided into three groups of three farmers each, and each group and using location-specific seed of this all-cross EP in the following
assessed one experiment replication. Diseases or pests never achieved a years. We also assessed the top-yielding RIL across Exp. 1–2 out of all the
meaningfully recordable level in these or following experiments. RILs from the three crosses (which provided a further albeit limited
assessment of the value of SSD-derived lines), the three parent cultivars
2.3. Assessment of SSD-derived and bulk-derived inbred lines, global line of the crosses, and three recent commercial cultivars (Fraser, Pepone,
mixtures and evolutionary populations: data analysis Spacial) that displayed excellent adaptation to organically-managed
Italian environments of Northern, Central and Southern Italy (Pecetti
An analysis of variance (ANOVA) including the fixed factors envi­ et al., 2019).
ronment, cross and plant material and the random factor block within The plant material was evaluated in eight autumn-sown test envi­
environment aimed primarily to compare for trait mean value the ronments corresponding to Exp. 4–11 in Table 1. They included four
following plant material generated for each cross: (a) individual RILs; organically-managed environments resulting from the combination of
(b) global mixture of RILs; (c) individual lines derived from the EP for Lodi or Perugia locations by 2014–15 or 2015–16 cropping years, and
the target region; (d) global mixture of lines derived from the EP for the four conventionally-managed environments in Lodi resulting from the
target region; (e) individual lines derived from the EP for the non-target combination of pure stand or intercropping with cereals by 2018–19 or
region; (f) EP for the target region; (g) EP for the non-target region. Data 2019–20 cropping years. Exp. 4 and 5, which were performed in the
of each plant material was averaged within block before data analysis. same year as the conventionally-managed Exp. 3 that concurred to
Another ANOVA including the factors cross, plant material and block define the two top-yielding lines across three selection environments
was performed to compare plant material within each environment. (Exp. 1–3) for each cross, actually included a higher number of inbred
Parent line data were excluded from these analyses. lines per cross (8− 12), to ensure that they would comprise the finally-
Other statistical analyses focused on the grain yield comparison of resulting top-yielding lines across Exp. 1–3. Exp. 4–7 (organically-
SSD-derived lines vs. lines derived from the EP for the target region managed) were laid out according to a randomized complete block
(representative of bulk breeding). An ANOVA including the fixed factor design with three replications, adopting 6.5 m2 plot size and higher
environment and the random factors genotype (here relative to all lines sowing density than other trials, i.e., 105 seeds/m2, to favor crop
belonging to each of these line groups) and block within environment competition against weeds. These experiments were sown by a me­
was performed for each cross to estimate the variance components chanical plot seeder. We recorded dry grain yield, onset of flowering,
relative to genotype, GEI and pooled experimental error using a REML plant height at onset of flowering, lodging susceptibility and individual

5
P. Annicchiarico et al. Field Crops Research 293 (2023) 108831

dry seed weight as described earlier, and farmers’ acceptability score Another set of ANOVAs was carried out to compare the following
attributed to all plots by six organic farmers from Northern Italy in Lodi types of plant material: EPs, 6-line mixtures and 12-line mixtures
and six from Central Italy in Perugia. selected within or across crosses, top-yielding inbred lines (considering
The four conventionally-managed experiments (Exp. 8–11) were two top-performing lines across Exp. 1–3 for each cross), parent lines,
sown in Lodi in early December. In both cropping years, PS and MS and elite control cultivars. These analyses, which were performed on
environments represented the main plots of a split-plot design with three data of grain yield or other agronomic traits previously averaged across
replications holding pea material on subplots, to reduce the effect on pea the relevant plant material in each block, held material as fixed factor,
plant responses of environmental factors other than the presence or and environment and block within environment as random factors.
absence of inter-specific competition. The cereal tester was represented Multiple mean comparisons for genotypes or plant materials were al­
by the mixture of the semi-dwarf barley cultivar Atlante with the tall ways performed according to Duncan’s test.
early wheat cultivar San Pastore in 2018–19, and the mixture of Atlante A comparison among breeding strategies was based on the genetic
with the semi-dwarf early wheat cultivar Spada in 2019–20. All plots gain over the best-performing control or parent cultivar that was dis­
had 4.5 m2 size and included 6 rows, blending pea and cereal seeds on played by best-performing germplasm of EPs, line mixtures and inbred
each row in MS (as done ordinarily by local farmers for pea-cereal in­ lines according to mean yield and/or yield reliability values.
tercrops). The experiments were sown by a mechanical plot seeder. The Environment-specific ANOVAs further compared EPs vs. bulk-derived
seed rate of pea in MS was half of that adopted in PS (40 vs. 80 seeds/ inbred lines based on mean yields of the EPs and the top-yielding lines
m2). The cereal seed rates in MS were 75 seeds/m2 for barley and 100 across Exp. 1–3 averaged over crosses, to investigate the possible effect
seeds/m2 for wheat, corresponding to 25 % of the ordinary rate in the of temporal changes for EPs that evolved from material harvested in Exp.
region for PS of each species (which implied a halved seed rate for the 1 for Lodi and Exp. 2 for Perugia and then underwent subsequent evo­
whole of the cereal tester in MS relative to the ordinary rate in PS). The lution cycles in following experiments at each site.
pre-sowing fertilization for PS and MS included 50 kg/ha of N, 75 kg/ha
P2O5 and 100 kg/ha K2O. Chemical weed control [(Stomp® 330 E (a.i. 3. Results
Pendimethalin at 307 g/L) at 4.5 L/ha] was applied only to PS, to limit
the relatively large weed growth expected in this condition. The traits 3.1. Assessment of SSD-derived and bulk-derived inbred lines, global line
recorded on PS plots included dry grain yield, onset of flowering, plant mixtures and evolutionary populations
height at onset of flowering, lodging susceptibility, and individual dry
seed weight. Those recorded on MS plots comprised dry grain yield of The three EPs for the target region underwent serious winter low-
pea and the pooled cereal tester, the total (pea + cereal) dry grain yield temperature stress during the second, third and fourth cycles of natu­
of the mixture and the proportion of pea dry grain yield on total grain ral selection in Lodi (Table 1), in which we observed visually sizeable
yield (which was used as an estimate of pea competitive ability), winter plant mortality of EP stands. Greatest stress occurred in the last
computed after harvesting the plot fresh seed and using a seed sample of cycle, which was used for mass selection of bulk-derived lines. Winters
100 g for separation and dry weight assessment of the relative propor­ in these selection cycles were distinctly colder than those displayed by
tion of pea and cereal components. Pea grain yield in MS was doubled Exp. 1 of Lodi and all experiments performed in Perugia (Exp. 2, 5 and 6)
prior to statistical analysis, to express it with respect to same growing (Table 1).
area (assuming halved area for pea in MS relative to pea in PS). The ANOVA for grain yield of plant materials listed in Table 2
indicated significant variation (P < 0.001) for material, crosses, and
2.5. Assessment of elite inbred lines, line mixtures and evolutionary material × environment interaction (Supplementary Table 1 A). On
populations: data analysis average, the EP material for the target region out-performed any other
material in the mild-winter Exp. 1 and 2, while performing comparably
An ANOVA including the fixed factor genotype (here relative to 27 to its bulk-derived lines and their global mixture in the cold-prone Exp. 3
entries encompassing inbred lines, line mixtures, EPs, parent cultivars (Table 2). On average, the bulk-derived lines for the target region out-
and control cultivars) and the random factors environment and block yielded the SSD-derived lines in each of the three test environments
within environment was performed on plot data of pea grain yield in (P < 0.05), showing 24 % higher yield over environments (5.43 vs.
eight environments (Exp. 4–11). Our wide-adaptation breeding prospect 4.36 t/ha; Table 2). The global mixtures of lines exhibited a modest,
supported by earlier studies (Annicchiarico and Iannucci, 2008; Pecetti non-significant yield advantage over the mean value of their component
et al., 2019) justified the investigation of GEI effects in terms of yield lines, which averaged 3 % for bulk-derived lines and 6 % for SSD-
stability over environments according to the static stability concept as derived lines (Table 2). The effect of the contrasting evolution envi­
measured by the Environmental variance and the dynamic stability ronments on the yield response of the EPs was remarkable: on average,
concept as measured by Shukla’s Stability variance (Shukla, 1972a), the EPs for the target region or their derived lines out-yielded the EPs for
rather than investigating genotype adaptive responses to specific envi­ the non-target region or their derived lines in each test environment
ronments by other techniques. Paired tests of the most stable genotype (P < 0.05), showing a yield advantage over environments of 35 % for
vs. each other one were carried out by means of Ekbohm’s (1981) test for the EPs (5.84 vs. 4.31 t/ha) and 43 % for the EP-derived lines (5.57 vs.
the former stability measure, and Shukla’s (1972b) test for the latter 3.89 t/ha; Table 2). The mass-selected lines from the EP for the non-
measure. Mean yield and yield stability according to either criterion target region displayed even greater misadaptation to the target re­
were combined into an index of yield reliability as proposed by Eskridge gion than the EP they derived from (Table 2).
(1990), estimating for each genotype the lowest yield expected in 80 % Variation for plant material was observed for all other traits except
of cases based on its mean yield and the relevant stability measure. tolerance to lodging, while material × environment interaction was
Other ANOVAs including the same factors were carried out to compare found for onset of flowering and seed weight (P < 0.001; Supplementary
genotypes for grain yield or farmers’ acceptability score in four Table 1 A). On average, the bulk-derived lines for the target region,
organically-managed environments (Exp. 4–7), pea yield in compared with the SSD-derived lines, exhibited 28 % higher aerial
conventionally-managed PS environments (Exp. 8 and 10), pea, cereal biomass, nearly 6 cm taller stature, 4.7 days later onset of flowering, 4 %
and total yield and pea proportion on total yield in MS environments smaller seed, and 14 % greater farmers’ acceptability over test envi­
(Exp. 9 and 11), and onset of flowering, plant height at flowering onset, ronments (P < 0.05; Table 2). They also displayed lower proportion of
lodging susceptibility and individual dry seed weight in six PS envi­ winter plant mortality (0.204 vs. 0.288) in Exp. 3, the only environment
ronments (Exp. 4–8 and 10). Trait interrelationships were investigated that featured material variation in cold tolerance (P < 0.05) and sub­
by Pearson’s correlation analysis. stantial winterkilling (Table 2). In contrast, the bulk-derived lines for the

6
P. Annicchiarico et al. Field Crops Research 293 (2023) 108831

Table 2
Grain yield and agronomic traits of one evolutionary population (EP) for Northern and Central Italy (target region, TR) initially evolved in Lodi, one EP mass-selected
under dry Mediterranean-climate conditions (non-target region, NTR), 30 inbred lines derived from the EP for TR and 60 recombinant inbred lines (RILs) grown
individually or in mixture, and 15 inbred lines derived from the EP for NTR, evaluated in up to three test environments. Material generated for each of three connected
crosses, with evaluation results averaged across crosses.
Grain yield (t/ha)a Mean valuea,b

Material Lodi, Perugia, Lodi, Mean Aerial Onset of Plant height at Individual Farmer Proportion of
2013–14 2013–14 2014–15 value biomass flowering (dd onset of seed weight acceptability winter mortality,
(t/ha) from Apr. 1) flowering (cm) (mg) (score 1–9) Lodi 2014–15

EP for TR 7.55 a 3.87 a 6.09 a 5.84 a 12.86 a 14.0 c 62.5 ab 0.189 c 4.82 ab 0.190 c
Mixture of 6.83 bc 3.44 b 6.43 a 5.57 13.16 a 15.9 b 64.5 a 0.192 c 4.87 a 0.208 c
lines from ab
EP for TR
Mean of lines 6.91 b 3.23 bc 6.15 a 5.43 b 12.55 a 16.9 a 64.4 a 0.190 c 4.95 a 0.204 c
from EP for
TR
Mixture of 6.56 bcd 2.99 cd 4.37 b 4.64 c 9.93 b 10.6 e 60.2 bc 0.198 b 4.44 b 0.249 b
RILs
Mean of RILs 6.07 de 2.81 de 4.19 bc 4.36 c 9.78 bc 12.2 d 58.5 c 0.198 b 4.33 b 0.288 b
EP for NTR 6.23 cde 3.13 bc 3.58 bc 4.31 c 9.25 bc 9.9 f 59.1 c 0.200 ab 4.09 b 0.269 b
Mean of lines 5.87 e 2.59 e 3.20 c 3.89 d 8.46 c 10.9 e 58.8 c 0.203 a 4.04 b 0.371 a
from EP for
NTR
LSD 0.60 0.29 0.97 0.38 1.31 0.6 2.4 0.003 0.39 0.032
(P < 0.05)
a
Lodi, Northern Italy; Perugia, Central Italy. Column means followed by different letter differ at P < 0.05 according to Duncan’s test.
b
Across Exp. 1–2 in Table 1 for aerial biomass and farmer acceptability; across Exp. 1–3 in Table 1 for the other traits.

non-target region exhibited 1.3 days earlier flowering, somewhat larger selection cycle of bulk breeding on the basis of gains of genotypes pooled
seed and greater proportion of winterkilled plants (0.371 vs. 0.288) over crosses (37.8 % vs. 16.1 %), and over 2.5-fold greater efficiency of
compared with the SSD-derived lines (P < 0.05; Table 2). The effects of bulk breeding according to gains averaged over the three crosses (28.0
the evolution environment observed for bulk-derived lines were also % vs. 11.0 %; Table 3). SSD-based breeding failed to produce a yield gain
manifest for EP material, with greater aerial biomass, taller stature, later over the best reference cultivar in the cross K×A (Table 3). For both
flowering, smaller seed, greater farmers’ acceptability and lower winter bulk-derived and SSD-derived lines, the three crosses ranked in the order
plant mortality exhibited by the EPs for the target region relative to K×I > A×I > K×A for yielding ability of their top-performing lines
those for the non-target region (P < 0.05; Table 2). (Table 3). Crosses differed for mean value of yield or any other agro­
The bulk-derived lines for the target region, compared with the SSD- nomic trait except lodging susceptibility (Supplementary Table 1).
derived lines, displayed same genetic CV but lower GEI CV for grain
yield across environments (Table 3), indicating the ability of bulk-based 3.2. Assessment of elite inbred lines, line mixtures and evolutionary
breeding to produce lines tending not only towards greater yielding populations
ability but also towards greater dynamic yield stability (Table 3). The
comparison of bulk-derived vs. SSD-derived lines in terms of yield gain The 27 entries that underwent evaluation across Exp. 4–11 exhibited
over the best-performing parent line Isard (based on the 10 % top- highly significant (P < 0.001) ANOVA variation for genotype mean
yielding lines per group) indicated over 2.3-fold greater efficiency per value and GEI of pea yield. Mean yields of these genotypes are reported
in Table 4. The two top-yielding bulk-derived lines selected over three
Table 3 selection environments (Exp. 1–3) coincided with those selected over
Genetic and genotype × environment interaction (GEI) coefficient of variation two environments (Exp. 1–2) for the cross A×I. They partly differed for
for grain yield across three test environments (Exp. 1–3 in Table 1) of inbred the other crosses, but with a sizeable yield advantage over independent
lines derived from evolutionary population (EP) material and recombinant environments of three-environment selection over two-environment
inbred lines (RILs) averaged across three connected crosses, and yield and ge­ selection only for the cross K×A (Table 4). The absolute top-ranking
netic yield gain of 10 % top-performing EP-derived lines and RILs pooled over SSD-derived line over Exp. 1–2 (AI_21) exhibited a slight yield disad­
crosses or separated for each cross. vantage relative to the top-yielding bulk-derived line within its cross
Item Lines from EPa RILs (AI_L22; Table 4). In general, heterogeneous material tended towards
Lines of three crosses (average value) greater yield stability than commercial cultivars (Table 5). The top-
Genetic CV (%) 11.9 11.9 ranking genotype for static yield stability was the mixture featuring
GEI CV (%) 12.9 18.7 the highest genetic diversity, i.e., the 12-line mixture based on lines
Lines pooled over crosses from all crosses, whose value of Environmental variance was signifi­
Top 10% yield (t/ha) 7.20 6.06
Yield gain (%)b 37.8 16.1
cantly lower (P < 0.05) than that of two other mixtures, three inbred
Lines of cross Attika × Isard lines, and four parent or commercial lines (Table 4). The top-ranking
Top 10% yield (t/ha) 6.53 5.99 genotype for dynamic yield stability was the EP issued from the cross
Yield gain (%)b 25.2 14.7 A×I (followed closely by the EP issued from all crosses, two line mix­
Lines of cross Kaspa × Attika
tures and the cultivar Fraser), but its value of Stability variance was
Top 10% yield (t/ha) 6.03 4.95
Yield gain (%)b 15.6 –5.2 significantly lower only with respect to two other entries (Table 4). The
Lines of cross Kaspa × Isard 12-line and 6-line mixtures issued from all crosses, and the inbred lines
Top 10% yield (t/ha) 7.47 6.45 and 6-line mixture issued from the cross K×I, exhibited high yield reli­
Yield gain (%)b 43.1 23.5 ability according to both yield stability concepts (Table 4).
a
EP targeted to Northern and Central Italy. Additional ANOVAs for genotype yield in subsets of environments
b
Relative to the top-yielding parent cultivar (Isard; mean yield = 5.22 t/ha). indicated, besides genotype variation, (a) GEI across the four

7
P. Annicchiarico et al. Field Crops Research 293 (2023) 108831

Table 4
Mean grain yield (GY), yield stability and yield reliability according to Environmental variance (EV) and Shukla’s Stability variance (SSV) across eight environments,
competitive ability as expressed by pea proportion in intercropping with cereals across two environments, and farmer acceptability (FA) across four environments, for
evolutionary population (EP) material selected in Lodi and Perugia, pea inbred lines and line mixtures selected across Lodi and Perugia for each of three connected
crosses and over crosses, parent cultivars, and elite control cultivars.
Yield stabilityb Yield reliabilityc
a,b 2 2
Material Cross GY (t/ha) EV (t/ha) SSV (t/ha) EV-based (t/ha) SSV-based (t/ha) Pea proportion in MS FAa,da,e (score 1–9)

EP A×I 1.67 1.13 a 0.06 a 0.82 0.70 0.164 3.55

EP K×A 2.13 a 1.11 a 0.15 a 1.29 1.13 0.225 a 3.96
EP K×I 2.06 a 1.15 a 0.19 a 1.20 1.04 0.205 4.58
EP All 2.08 a 1.24 a 0.07 a 1.19 1.11 0.213 a 4.23
6-line mixture A×I 1.78 a 1.61 0.07 a 0.77 0.81 0.100 4.11
6-line mixture K×A 1.86 a 1.05 a 0.08 a 1.04 0.88 0.147 4.29
6-line mixture K×I 2.42 a 1.09 a 0.41 1.59 1.34 0.233 a 5.60 a
12-line mixture A×I 1.97 a 1.47 a 0.14 a 1.00 0.97 0.123 4.78
12-line mixture K×A 1.88 a 1.69 0.12 a 0.84 0.89 0.241 a 4.15
12-line mixture K×I 2.04 a 1.15 a 0.07 a 1.19 1.07 0.235 a 5.19 a
6-line mixture All 2.31 a 1.13 a 0.08 a 1.46 1.34 0.114 4.65
12-line mixture All 2.22 a 0.66 a 0.21 a 1.56 1.20 0.273 a 4.51
Line AI_21f A×I 1.84 a 1.64 0.08 a 0.81 0.86 0.144 4.06
Line AI_L22g,h A×I 1.93 a 1.83 0.15 a 0.85 0.93 0.154 4.18
Line AI_L33g,h A×I 1.60 1.99 0.17 a 0.48 0.60 0.055 3.87
Line KA_L16h K×A 1.89 a 1.33 a 0.28 a 0.97 0.85 0.110 4.37
Line KA_L23g K×A 2.18 a 1.00 a 0.34 a 1.38 1.12 0.293 a 4.69
Line KA_L39g,h K×A 1.97 a 0.73 a 0.13 a 1.28 0.97 0.176 4.26
Line KI_L16h K×I 2.31 a 1.06 a 0.20 a 1.48 1.29 0.220 a 5.65 a
Line KI_L22g K×I 2.27 a 0.86 a 0.22 a 1.53 1.25 0.214 a 5.41 a
Line KI_L34g,h K×I 2.30 a 1.00 a 0.19 a 1.50 1.28 0.240 a 5.44 a
Attika (parent) 1.39 0.98 a 0.14 a 0.60 0.39 0.140 2.75
Isard (parent) 1.41 1.55 0.12 a 0.41 0.42 0.041 3.79
Kaspa (parent) 1.76 1.20 a 0.15 a 0.88 0.76 0.192 4.03
Fraser (control) 1.85 a 1.45 a 0.07 a 0.87 0.88 0.175 3.41
Pepone (control) 1.51 1.69 0.53 0.47 0.40 0.133 1.70
Spacial (control) 1.80 a 1.58 0.09 a 0.79 0.82 0.139 4.20
LSD (P < 0.05) 0.40 − − − − 0.070 0.65
a
Column means followed by letter ‘a′ do not differ from the top-ranking mean at P < 0.05 according to Duncan’s test; genotype × environment interaction used as
the error term for GY and FA.
b
Across Exp. 4–11 in Table 1. Stability values followed by letter ‘a′ do not differ from the lowest value (most stable entry) at P < 0.05 based on Ekbohm’s test for EV
and Shukla’s test for SSV.
c
Across Exp. 4–11 in Table 1. As lowest yield expected in 80 % of cases based on mean yield and relevant stability measure.
d
Across Exp. 9 and 11 in Table 1; cereal tester formed by mixing one barley cultivar and one bread wheat cultivar.
e
Across Exp. 4–7 in Table 1.
f
Top-yielding SSD-derived line across two environments in one test year (Exp. 1–2 in Table 1) out of 180 lines from three crosses.
g
Belonging to the top-yielding pair of EP-derived inbred lines for the relevant cross across three environments in two evaluation years (Exp. 1–3 in Table 1).
h
Belonging to the top-yielding pair of EP-derived inbred lines for the relevant cross across two environments in one evaluation year (Exp. 1–2 in Table 1).

organically-managed environments (Exp. 4–7) mainly due to genotype (Supplementary Table 2 B). On average, the bulk-derived lines out-
× year interaction (P < 0.001) with absence of genotype × location performed the commercial and parent cultivars for mean yield
interaction (Supplementary Table 2 C), and (b) GEI across the four (P < 0.10), yield reliability and competitive ability (P < 0.05), while
conventionally-managed environments (Exp. 8–11) due to interaction of performing comparably to EPs, 6-line and 12-line mixtures derived from
genotypes with PS or MS conditions and years (Supplementary Table individual crosses (Table 5). The 12-line mixture and the EP derived
3 A). Pea genotypes were severely out-competed by cereal companions from all crosses out-yielded the sets of commercial and parent cultivars
in MS, averaging a pea proportion of 0.174 on total (pea + cereal) grain (P < 0.05; Table 5), and were top-ranking in this order for yield reli­
yield. Genotype yield over organically-managed environments (which ability according to either stability concept and competitive ability
implied severe competition against weeds) was correlated with geno­ (Table 5). On average, the elite commercial cultivars showed a trend
type yield in intercropping (r = 0.58, P < 0.01). In contrast, no corre­ towards lower yield stability than EPs, mixtures and inbred lines,
lation emerged for genotype yield across PS and MS conditions over thereby increasing their production gap in terms of yield reliability
2018–19 and 2019–20 in Lodi (r = –0.06), or across organically- relative to these materials (Table 5).
managed (Exp. 4 and 6) and conventionally-managed (Exp. 8 and 10) The yield comparison of top-performing bulk-derived lines vs. EPs
PS environments in Lodi (r = 0.07). Genotype yields across these subsets from single crosses in each experiment is summarized in Fig. 3, which
of environments are reported in Supplementary Table 4. Additional also recalls the evolutionary path of EP material starting from Exp. 1 for
ANOVAs for genotype data in MS (Exp. 9 and 11) indicated genotype Lodi and Exp. 2 for Perugia. Its inconsistent results across environments
variation for pea and total yield and pea proportion (P < 0.01) but no did not express a clear temporal trend due to progressively better
variation for yield of the associated cereal (Supplementary Table 3 C). yielding ability and evolutionary adaptation of EP material. However,
Pea yield and pea proportion (i.e., competitive ability) in MS were they may relate to adaptive responses to different levels of winter low
highly correlated (r = 0.98, P < 0.001) and were both positively temperatures. The inbred lines, which derived from EPs evolved in Lodi
correlated with total yield of the genotypes (r ≥ 0.66, P < 0.01). that were eventually selected across Exp. 1–3 (i.e., in two mild-winter
The ANOVA for pea grain yield over eight environments of the nine environments out of three), tended to be consistently higher-yielding
plant materials listed in Table 5 revealed significant (P < 0.01) variation in the mild-winter site of Perugia (significantly at P < 0.05 in two
for material mean value and material × environment interaction years out of three) and out-performed the EPs in Lodi in the mild-winter

8
P. Annicchiarico et al. Field Crops Research 293 (2023) 108831

Table 5
Average values of mean grain yield (GY), yield stability and yield reliability according to Environmental variance (EV) and Shukla’s Stability variance (SSV) across
eight test environments, competitive ability as expressed by pea proportion in intercropping with cereals across two environments, and farmer acceptability (FA) across
four environments, for evolutionary population (EP) material selected in Lodi (Northern Italy) and Perugia (Central Italy) and bulk-derived inbred lines and line
mixtures selected across Lodi and Perugia for each of three connected crosses and over crosses, parent cultivars, and elite control cultivars.
Yield stabilityb Yield reliabilityb
a,b
Material Description of material GY (t/ EV (t/ SSV (t/ EV-based (t/ SSV-based (t/ Pea proportion in FAa,d
ha) ha)2 ha)2 ha) ha) MSa,c (score
1–9)

EP, low diversity 3 EPs, one per cross 1.95 ab 1.13 0.13 1.10 0.96 0.198 bc 4.03 ab
EP, high diversity EP from EPs of 3 crosses 2.08 a 1.24 0.07 1.19 1.11 0.213 b 4.23 ab
Line mixture, low 3 6-line mixtures, one per cross 2.02 ab 1.25 0.19 1.13 1.01 0.160 bcd 4.67 a
diversity
Line mixture, fair 3 12-line mixtures, one per 2.08 a 1.32 0.09 1.16 1.10 0.200 bc 4.71 a
diversity cross
Line mixture, fair 6-line mixture, 2 lines per cross 1.97 ab 1.46 0.14 1.00 0.97 0.114 d 4.65 a
diversity
Line mixture, high 12-line mixture, 4 lines per 2.22 a 0.66 0.21 1.56 1.20 0.273 a 4.51 a
diversity cross
Inbred line 6 lines, 2 lines per cross 2.04 ab 1.23 0.20 1.17 1.03 0.189 b 4.64 a
Parent cultivars of 3 3 parent lines of connected 1.52 c 1.24 0.14 0.63 0.52 0.124 d 3.52 bc
crosses crosses
Elite control cultivars 3 cultivars 1.72 bc 1.59 0.23 0.71 0.70 0.149 cd 3.11 c
LSD (P < 0.05) 0.31 − − − − 0.051 0.70
a
Column means followed by different letter differ at P < 0.05 according to Duncan’s test; material × environment interaction used as the error term for GY and FA.
b
Across Exp. 4–11 in Table 1, averaging the results of relevant entries reported in Table 4. Yield reliability as lowest yield expected in 80 % of cases based on mean
yield and the relevant stability measure.
c
Across Exp. 9 and 11 in Table 1; cereal tester in MS formed by mixing one barley cultivar and one bread wheat cultivar.
d
Across Exp. 4–7 in Table 1.

environments emerged for all other traits (P < 0.01) except lodging
susceptibility, concurrently with GEI (Supplementary Table 2 A and 2
B). The top-performing set of genotypes for farmers’ acceptability
included the 6-line and 12-line mixtures and the three inbred lines is­
sued from the cross K×I (Table 4). On average, the bulk-derived inbred
lines and line mixtures exhibited greater farmers’ acceptability than
commercial and parent lines (P < 0.05), whereas EP material exhibited
intermediate acceptability values relative to these germplasm sets
(Table 5). In general, breeding materials featured later flowering and
taller plant stature than parent or commercial lines, with the greatest
differences shown by bulk-derived lines and line mixtures (Supple­
mentary Table 5). Differences for seed weight were modest between
breeding material and parent lines, all of them featuring smaller seed
than the mean value of the commercial lines (Supplementary Table 5).
Both pea competitive ability and GEI leading to relatively better pea
yield in MS relative to PS (as indicated by the ratio of the pea yield
values in MS and PS given in Supplementary Table 4) were correlated to
taller stature of the genotypes (r ≥ 0.58, P < 0.01), which was closely
associated with later onset of flowering (r ≥ 0.74, P < 0.001).
Genetic gains for mean yield or yield reliability of best-performing
material within each germplasm type are given in Table 6 with
respect to Fraser, which was the top-performing cultivar for yield and
Stability variance-based yield reliability and was nearly top-ranking for
Environmental variance-based yield reliability (Table 4). The assess­
Fig. 3. Mean pea grain yield of three evolutionary populations (EPs; black ment, which was based on data reported in Table 4, included also
symbols) initially evolved in Lodi (Northern Italy) and then evolved separately competitive ability and farmers’ acceptability, owing to the importance
in Lodi and Perugia (Central Italy) and six inbred lines (white symbols) of these traits. The comparison included two bulk-derived lines (both
including the pair of bulk-derived top-yielding lines across three test environ­ from the cross K×I) since no single line was top-ranking for all perfor­
ments (Exp. 1–3 in Table 1) for each EP, evaluated in 10 test environments mance criteria, and the top-yielding SSD-derived line in early testing
(Exp. 1, 2, 4–11 in Table 1). Circle and square symbols are yields in pure stand
(Exp. 1–2) as a further assessment for this germplasm type. The genetic
and in intercropping with cereals, respectively; vertical bars are least significant
progress displayed by most selections was substantial (Table 6). The
difference values (P < 0.05); dotted lines indicate environments providing EP
material to the next cropping year in each location. gains for mean yield and both yield reliability measures were maximized
by the 6-line mixture from the cross K×I, which showed advantages over
the top-performing bulk-derived inbred lines in the range 4–7 % for
Exp. 1 (P < 0.05); whereas the EPs out-yielded (P < 0.05) the inbred
mean yield or yield reliability (Table 4). Compared with this mixture,
lines in Lodi’s cold-winter years of 2014–15 and 2018–19 in PS (with a
the best bulk-derived lines and the 12-line mixture issued from all
trend towards higher yield also in MS in the latter year).
crosses exhibited 19–35 % lower genetic gain for mean yield, but just
Variation among the 27 genotypes or the nine plant materials in PS
11–16 % lower gain averaged across all traits (Table 6). The 6-line

9
P. Annicchiarico et al. Field Crops Research 293 (2023) 108831

Table 6
Genetic gain (%) for grain yield (GY), yield reliability according to Environmental variance (EV) and Shukla’s Stability variance (SSV), competitive ability as expressed
by pea proportion in intercropping with cereals and farmer acceptability (FA), of top-performing evolutionary population (EP) material selected in Lodi (Northern
Italy) and Perugia (Central Italy), bulk-derived inbred line and line mixture germplasm and one single-seed descent (SSD)-derived line selected across Lodi and Perugia.
Yield reliability

Material Description of material GY (t/ha) EV-based (t/ha) SSV-based (t/ha) Pea proportion in MS FA Average gain
(score 1–9)

EP, low diversity EP from cross K×A 15.1 48.3 28.4 28.6 16.1 27.3
EP, high diversity EP from EPs of 3 crosses 12.4 36.8 26.1 21.7 24.0 24.2
Line mixture, low diversity 6-line mixture from cross K×I 30.8 82.8 52.3 33.1 64.2 52.6
Line mixture, fair diversity 12-line mixture from cross K×I 10.3 36.8 21.6 − 34.3 52.2 31.0
Line mixture, fair diversity 6-line mixture, 2 lines per cross 24.9 67.8 52.3 34.9 36.4 29.3
Line mixture, high diversity 12-line mixture, 4 lines per cross 20.0 79.3 36.4 56.0 32.3 44.8
Inbred line Bulk-derived line KI_L16 24.9 70.1 46.6 25.7 65.7 46.6
Inbred line Bulk-derived line KI_L22 22.7 75.9 42.0 22.3 58.7 44.3
Inbred line SSD-derived line AI_21 − 0.5 − 6.9 − 2.3 − 17.7 19.1 − 1.7

The table includes top-performing EP and bulk-derived mixture and inbred line material out of three crosses based on GY or yield reliability measures. Gain relative to
the top-performing cultivar Fraser, based on data reported in Table 4.

mixture from all crosses maximized the Stability variance-based yield The difference between broadly-based and narrowly-based EPs or line
reliability but showed no progress for competitive ability and, along mixtures related, anyway, to material derived from just three or two
with the remaining EP or mixture material, featured much lower genetic parental cultivars. Likewise, Merrick et al. (2020) reported no difference
gain averaged over traits (Table 6). Compared with best bulk-derived in yield stability between biparental EPs and EPs originated from three
lines, the top-performing EP (issued from the cross K×A) exhibited cultivars.
8–9 % lower mean yield and 10–16 % lower yield reliability (Table 4). Several elite inbred lines derived from biparental EPs tended towards
The SSD-derived inbred line was the least valuable selection, as it failed higher yield stability than commercial cultivars, and comparable yield
to achieve any genetic progress for all traits except farmers’ accept­ stability relative to EPs and line mixtures. The same result was reported
ability (Table 6). by Raggi et al. (2017) for EPs and EP-derived line material of barley.
Three reasons may have contributed to the high yield stability of elite
4. Discussion bulk-derived lines in our study: (a) the prior choice of stable-yielding
parent cultivars; (b) the observed trend of bulk-derived lines towards
4.1. Genotype × environment interaction, yield stability and adaptation greater dynamic yield stability than SSD-derived lines based on GEI
patterns results from Exp. 1–3, which could be the consequence of within-EP
natural selection across cropping years featuring wide variation in
Our test environments (Exp. 4–11), which aimed to represent the winter low temperatures; (c) the selection of these lines on a
diversity of cropping conditions featuring pea in Northern and Central multi-environment basis (across Exp. 1–3). Accounting for yield stability
Italy, indicated high GEI for pea grain yield that regional breeding differences through yield reliability measures influenced the germplasm
programs have to cope with. In this region, the fairly modest pea crop­ comparison as in earlier studies (Döring et al., 2015; Bocci et al., 2020),
ping area (about 16,000 ha), and the greater impact on GEI of the year leading to substantially greater agronomic value and genetic gain of elite
factor than the location factor (as represented by the geographically- heterogeneous material or bulk-derived lines over top-performing
distant sites of Lodi and Perugia) confirmed by current ANOVA results commercial germplasm in comparison with results based on mean
in agreement with earlier studies (Annicchiarico and Iannucci, 2008; yield (Table 6).
Pecetti et al., 2019), prompt to select pea lines and static line mixtures Large GEI arose not only from year-to-year climatic variation (whose
for wide geographical adaptation and greater yield stability via extent is bound to increase under climate change) but also from crop
multi-environment selection (as assumed in this study). The selection of management factors, as indicated by the lack of correlation for genotype
EPs for the target region may actually be more environment-specific yield responses across PS and MS conditions or across organically-
than that of inbred lines or static mixtures depending on the available managed and conventionally-managed PS environments of Lodi
seed amount of early-segregating material and considering, anyway, (considering, in the latter case, that different test years for the two
that the GEI pattern may limit the exploitation of site-specific managements may have contributed to the large GEI). Poor correlation
adaptation. across PS and MS conditions emerged in an earlier study based on a
The current detection of yield stability differences was complicated larger sample of independent pea genotypes (Annicchiarico et al., 2021),
by high sampling errors associated with the availability of only eight test while pea genetic variation in competitive ability against cereals
environments, which may be considered as the minimal number for emerged in that study and other reports (e.g., Hauggaard-Nielsen and
yield stability assessments (Piepho, 1998). However, statistically sig­ Jensen, 2001; Baxevanos et al., 2017), in the general context of severe
nificant differences were detected especially with respect to the Envi­ competitive disadvantage of pea in MS with cereals. Current ANOVA
ronmental variance, a measure that is closely related to that for crop and correlation results for pea, cereal and total yield and pea proportion
resilience proposed by Zampieri et al. (2020). The current trend of EPs in MS indicated that greater pea competitive ability led to greater total
and line mixtures towards greater yield stability than most commercial yield of the intercrop due to greater pea yield in the absence of a con­
cultivars according to either or both stability concepts confirmed for a current yield decrease of the associated cereals. This result, which
legume crop the results of several earlier studies on cereals (e.g., Döring reinforced the importance of pea competitive ability, may occur for
et al., 2015; Raggi et al., 2017; Reiss and Drinkwater, 2018; Merrick legume species at severe competitive disadvantage in intercrops, as a
et al., 2020; van Frank et al., 2020; Legzdiņa et al., 2022; Baresel et al., consequence of more nitrogen made available for cereals by a more
2022). No definite relationship emerged between wider genetic di­ competitive legume and other complementarity effects (Annicchiarico
versity within heterogeneous material and greater yield stability, et al., 2019a). The positive relationship of greater competitive ability in
although the line mixture and the EP with greatest genetic diversity MS with taller plant stature emerged consistently in earlier studies on
displayed high static and high dynamic stability, respectively (Table 5). pea (Hauggaard-Nielsen and Jensen, 2001; Annicchiarico et al., 2021)

10
P. Annicchiarico et al. Field Crops Research 293 (2023) 108831

and other legumes (Annicchiarico et al., 2019a), owing to the crucial spp., or low winter temperatures). The current importance of natural
importance of a taller stature in inter-specific competition for light. The selection for cold tolerance was highlighted by the substantial reduction
ability of a taller stature to also enhance the pea competitive ability of winter plant mortality displayed by bulk-derived lines, to which the
against weeds observed by McDonald (2003) would justify the positive concurrent shift of this material towards delayed flowering contributed
correlation for pea genotype yield observed across organically-managed as a mechanism of cold stress escape (Lejeune-Hénaut and Wery, 1994).
PS cropping and conventionally-managed intercropping. Likewise, the shift towards earlier flowering shown by bulk-derived lines
for the non-target region may concur to pea adaptation to severe ter­
4.2. SSD-derived vs. bulk-derived inbred lines minal drought as a major drought stress escape mechanism (Turner
et al., 2001; Annicchiarico et al., 2017).
This study confirmed the high value of EPs as sources of inbred lines The shift towards taller plant stature of bulk-derived lines compared
put forward (inter alia) by Phillips and Wolfe (2005) and Ceccarelli and with SSD-derived lines, which occurred along with a shift towards
Grando (2020), by revealing the definite superiority of bulk-derived greater aerial biomass, was already observed in lentil (Haddad and
lines for the target region over SSD-derived lines. This conclusion Muehlbauer, 1981) and cereal crops (Goldringer et al., 2001; Knapp
derived from: (a) 24 % higher mean yield, 2.3–2.5 greater yield gain per et al., 2020) as a result of intra-specific competition for light. Taller plant
selection cycle, greater farmers’ acceptability, lower GEI indicative of a stature within the semi-dwarf genetic background was not detrimental
trend towards greater dynamic yield stability of bulk-derived lines, and to tolerance to lodging on the ground of the lack of genetic variation for
the inability for SSD-derived lines to realize any yield progress over the lodging susceptibility that was consistently observed in this study, while
best-performing parent cultivar for one cross (K×A), based on data from enhancing the ability to compete with associated cereals or with weeds
Exp. 1–3; and (b) the inability of the top-performing SSD-derived line under organic management. The fairly modest shift towards lighter seed
across Exp. 1 and 2 to realize any genetic progress over the displayed by bulk-derived lines could be associated with greater geno­
best-performing commercial cultivar for yield, yield reliability and type fitness in evolving populations if it also implied a larger number of
competitive ability based on data from Exp. 4–11, in sharp contrast with seeds per genotype (Döring et al., 2011), an hypothesis that could not be
the remarkable genetic gains exhibited by the best-performing bulk-­ verified. The effect of natural selection on seed size is reportedly vari­
derived lines. The latter result, albeit clear-cut, may be viewed as able, and a shift towards larger seed has occasionally been observed
somewhat circumstantial, since it was not based on the comparison of (Horneburg and Becker, 2008). As a matter of fact, both seed size and
same numbers of selected bulk-derived and SSD-derived lines. The very seed number may be reduced by plant evolution under intra-specific
low outcrossing rate reported for pea (Polowic et al., 2002) makes it competition as a consequence of greater resource allocation to vegeta­
unlikely that the yield advantage shown by bulk-derived lines may arise tive development (Goldringer at al., 2001).
from greater residual heterosis due to larger outcrossing in bulks than
SSD (whereas, anyway, also SSD lines may have undergone outcrossing 4.3. Bulk-derived inbred lines vs. line mixtures vs. evolutionary
because they were multiplied without isolation from pollinators). The populations
over two-fold greater efficiency per selection cycle exhibited by
bulk-based breeding supports the superiority of this selection strategy This study revealed the high agronomic value of elite genetically
even when considering yield gains per unit time under the assumption of heterogeneous material of pea. The modest heterogeneity advantage
two generations per year (of which one off-season) for SSD-based indicated by 3–6 % higher yield of two global line mixtures over the
breeding. Based on our findings, SSD-based selection could display mean value of their component lines across Exp. 1–3, which was slightly
greater gains per unit time only when adopting rapid generation tech­ higher than the mean advantage of mixtures in earlier studies summa­
nology alias speed breeding for SSD lines, which may allow for up to five rized by Smithson and Lenné (1996) and Reiss and Drinkwater (2018),
generations per year (Mobini and Warkentin, 2016). In the absence of took place despite the likely occurrence of some yield penalty caused by
such a technology, the adoption of bulk-based breeding is reinforced by the diversity in maturity time of the component lines. This drawback
its much lower cost relative to SSD line development. was excluded for 6-line and 12-line mixtures that were selected and
Earlier comparisons of SSD-based vs. bulk-based breeding for inbred evaluated in Exp. 4–11, by imposing no more than 2–3 days of maturity
grain legume species provided inconsistent results, showing an advan­ difference for line selection. Our development of line mixtures by
tage of SSD in Kumar et al. (1992) for pea, Haddad and Muehlbauer assembling top-yielding lines evaluated separately is frequently envis­
(1981) for lentil and Miladinovič et al. (2011) for soybean, an advantage aged, because of the prohibitive cost implied by the preliminary
of bulk in Khosla et al. (2019) for soybean and Meena and Kumar (2012) assessment of many possible line combinations (Wuest et al., 2021). We
for chickpea, and no difference in Gill et al. (1995) for mungbean. Some did not assess the heterogeneity advantage of 6-line or 12-line mixtures
of these studies were only based on single-environment results. Meena over their component lines because of the high number of involved lines,
and Kumar (2012) contemplated the acquisition of bulk-derived lines but aimed to assess the agronomic value of these variety types relative to
also by means of within-bulk mass selection, showing the advantage of inbred lines and EPs.
visually mass-selected lines relative to unselect material. Our prior A major result from this study was the sharp diversity in adaptation
performance of severe within-bulk yield-based mass selection (about 3.8 pattern and/or key adaptive traits shown by EPs originated from the
% selected fraction) for line acquisition, the presence of a key environ­ same genetic base that evolved or were mass-selected in contrasting
mental stress promoting natural selection for regional adaptation such environments. The 35 % greater yield, 39 % greater aerial biomass, over
as winter low temperatures, and great attention paid to ensuring a 4 days later flowering and 34 % lower winter mortality shown on
balanced intra-specific plant competition and minimizing average by the EPs for the target region relative to those mass-selected in
micro-environmental variation during EP generation, probably were the drought-prone managed environment (Table 2) is remarkable when
important determinants of the current definite advantage of bulk-based considering that these differences emerged after few generations of
over SSD-based breeding. These circumstances, and the careful choice of evolution or selection (four for the EPs for the target region, and three
elite parent lines as advocated by Witcombe and Virk (2001), contrib­ for those for the non-target region) out of the relatively modest genetic
uted to achieve the remarkable breeding progress over elite commercial variation represented by the progenies of a biparental cross. The
germplasm that was displayed by best-performing bulk-derived lines application of natural or mass selection onto early segregating material
(which recently led to the successful variety registration in Italy as (from F2 plants onwards) and the impact of two major abiotic stresses
’Pantera rosa’ of one of these lines). Bulk breeding was successfully used that led to divergent selection for phenology and adaptation pattern
by Singh et al. (1994) in chickpea for low-cost selection under severe (low winter temperatures and terminal drought) probably were impor­
levels of a prevailing biotic or abiotic stress (Ascochyta blight, Fusarium tant determinants of this result. Earlier reports of rapid shifts of EP

11
P. Annicchiarico et al. Field Crops Research 293 (2023) 108831

adaptation were provided by Rasmusson et al. (1967), who observed a geographically-distant locations or different crop managements (e.g.
57 % yield increase after six generations of evolution under severe pure stand or intercropping), possibly by a decentralized
drought for a barley EP that pooled genotypes of a world collection, and farmer-participatory approach that may contemplate additional mass
Bocci et al. (2020), who found successful site-specific adaptation after selection (Murphy et al., 2005; Ceccarelli and Grando, 2020; Goldringer
five generations across geographically contrasting Italian sites for wheat et al., 2020). However, it is important to ensure sufficient starting seed
EPs that pooled F2 to F4 plants from crosses among 256 parents. Even our for each environment, considering that the number of preserved geno­
mass selection of bulk-derived F5 plants in a cropping year characterized types is proportional to the initial number of plants (Kervella and
by particularly low winter temperatures (2011–12: Table 1) had a clear Fouilloux, 1992). Several hundreds if not a few thousand initial plants
bearing on the characteristics of the resulting materials, which showed are recommended to keep sufficient genetic variation within bulks
nearly 3 days later flowering and lower adaptation to mild-winter en­ (Simmonds, 1979), and additional plants would be needed as border
vironments than the EPs they derived from (Table 2; this shift was partly plants to ensure a balanced intra-specific plant competition. Each EP
compensated by the following selection of elite bulk-derived lines across originated from 540 F2 seeds in this study. A solution in the current
Exp. 1–3, which featured two mild-winter environments). On the other context could be the development of specific EPs for different sites or
hand, results in Fig. 3 indicated the poor ability of biparental EPs that cropping conditions starting from F3 seed produced in one environment
included highly inbred component lines (from F7 onwards) to progres­ as representative as possible of a target region.
sively increase their yielding ability relative to elite bulk-derived lines in Our study reported rapid genetic shifts for a few morphophysio­
Lodi or Perugia. This result, to which their relatively narrow genetic logical traits displayed by EPs for the non-target region and their mass-
base may have contributed, implied a consistent yield penalty in Peru­ selected inbred lines, but could not assess the adaptation of these ma­
gia, where the EPs were brought to from Lodi. Anyway, shifts for terials to their drought-prone target environments. On-going research
adaptive traits such as onset of flowering occur more rapidly than those work in Algeria and Morocco will aim to compare EPs vs. bulk-derived
for site-specific crop performance, and improved adaptation by EPs may lines of pea for these environments.
be reflected by greater yield stability rather than greater mean yield, as
shown by Goldringer et al. (2006) and van Frank et al. (2020) for wheat 4.4. Conclusions
EPs in France.
The comparison among EPs, line mixtures and bulk-derived inbred This study confirmed the high value for cultivation of genetically
lines based on data averaged over crosses highlighted general differ­ heterogeneous material also for a major inbred legume such as pea.
ences between these materials, such as the particularly high yielding Compared with bulk-derived lines, static line mixtures exhibited some
ability, yield reliability and competitive ability of the line mixture advantage, and EP material a modest penalty (actually nil in its area of
featuring greatest genetic diversity, and the comparable performance of initial development), on the basis of yield or yield reliability data of best-
most of the remaining materials (Table 5). The comparison based on performing material. Interestingly, this result occurred across environ­
best-performing germplasm within crosses, which may have greater ments that lacked severe biotic stresses, that is, when the important
practical interest for breeding strategies, emphasized the high agro­ potential advantage of functional diversity for disease and pest reduc­
nomic value of specific material within crosses, such as the 6-line tion (Finckh et al., 2000; Phillips and Wolfe, 2005) could not express
mixture and two lines issued from the cross K×I, while confirming the itself. Our study did not explore the effect of quite contrasting levels of
value of the 12-line mixture issued from all crosses. Our comparison of intra-population diversity, but its top-performing heterogeneous mate­
EPs vs. line mixtures differed from those by Patel et al. (1987) and rial was not the one with greatest genetic diversity, indicating that
Döring et al. (2015), because the current lines were extracted from the excellent EPs or line mixtures may be obtained from progenies of a
EPs, and underwent individual multi-environment testing before being biparental cross between carefully chosen parent lines. Biparental EPs
selected from inclusion in the mixtures. Operationally, the selection and exposed to natural or mass selection displayed rapid adaptation to
commercial production of a 6-line static mixture derived from one elite contrasting target environments and rapid shifts of adaptive traits. They
cross is less resource-committing than that of a broadly-based 12-line proved quite valuable also for mass selection of bulk-derived inbred
mixture. The greater production cost of a 6-line mixture relative to a lines and line mixtures. The current inclusion of organic farming or
top-performing bulk-derived inbred line ought to be weighted against its intercropping among the target growing conditions, and our account for
4–7 % estimated advantage for mean yield or yield reliability that yield stability beside mean yield in performance assessments, empha­
emerged here. sized the agronomic value of EPs, inbred lines derived from EPs and
The development of biparental EPs issued from carefully-chosen mixtures of these lines in comparison with elite commercial inbred lines,
parent genotypes generated agronomically valuable germplasm, in indicating that taller plants generated by these breeding strategies have
agreement with findings by Merrick et al. (2020) for wheat. The current special interest in the prospect of more environment-friendly and
yield disadvantage of the best-performing EP relative to the agroecology-based cropping systems.
best-performing bulk-derived lines (8–9 % for mean yield, and 10–16 % On the whole, our results suggested that among the four breeding
for yield reliability) ought to be weighted against the much lower se­ strategies summarized in Fig. 1, the conventional one based on SSD-
lection cost of EP germplasm. Top-performing inbred lines previously derived lines could be the least efficient in terms of genetic yield gains
selected from EPs reportedly performed comparably to EPs in barley generated by top-performing material, except when considering gains
(Raggi et al., 2017). A major drawback of EPs in this study was the lower per unit time under the scenario of three or more SSD generations per
acceptability expressed by organic farmers relative to best-performing year (which has, anyway, much greater costs). The implications of our
inbred lines (Table 4), for reasons not made explicit by the assessment results may differ depending on the size, infrastructure, resource allo­
procedure. Our multi-environment assessment of EPs was influenced by cation and objectives of breeding programs. They encourage small
the consistent yield penalty of this germplasm relative to elite inbred breeding programs to develop EPs from carefully chosen parents in a few
lines in Perugia (Fig. 3). Indeed, the best-performing EP (issued from the environments that contrast for GEI pattern within a target region, with
cross K×A) exhibited nearly the same mean yield (2.02 vs. 2.03 t/ha) as the perspective either to derive mass-selected inbred lines from elite EPs
that of the best-performing inbred lines (KI_L16 and KI_L22), when the for further selection and marketing of individual lines or line mixtures,
assessment was limited to Lodi’s environments (Exp. 4, 6, 8–11). One or to release different EPs (whose registration in the EU is currently
may speculate that the earlier development of the EPs in Perugia may possible as Organic Heterogeneous Material). Developing EPs, either for
have resulted in better local adaptation of this material. More generally, registration or for spread via informal seed systems, may also be
the rapid shifts of adaptation pattern exhibited by early-segregating attractive for public breeding programs aimed to support seed systems
material supports the early development of EPs in that feature modest specific breeding, as it may be the case for organic

12
P. Annicchiarico et al. Field Crops Research 293 (2023) 108831

systems, or for conventional systems in developing countries. Our Ponzini and V. Vizioli (Associazione Italiana per l′ Agricoltura Biologica)
findings can also be inspiring for large breeding programs with good and performed by A. Baroni, I. Begliomini, S. Brambilla, L. Brambilla, F.
infrastructure aimed to implement cost-efficient solutions for inbred line Bossi, A. Briatti, R. Ciechi, T. Comolli, R. De Cielli, G. De Paolis, E. Di
selection, also with respect to incorporation of genomic selection. This Porzio, P. Galuffo, G. Garuffi, R. Lovati, D. Mocchiutti, S. Pettinacci, E.
selection strategy recently showed high potential value for pea yield Pietromarchi, A. Pitton, C. Rocca, M. Rubeca, E. Tavazzani, C. Vailati
improvement both for moderately favorable and drought-prone target and F. Viganò,
regions (Annicchiarico et al., 2019c, 2020), but its cost-efficient
implementation is constrained by the time needed for prior model Appendix A. Supporting information
construction based on phenotyping data of a training germplasm set.
Our results suggest to exploit rapid generation technology (Mobini and Supplementary data associated with this article can be found in the
Warkentin, 2016) to timely produce a training set representative of the online version at doi:10.1016/j.fcr.2023.108831.
target genetic base to be subjected to multi-year phenotyping, while
exploiting the advantages of bulk breeding for the concurrent generation References
of inbred lines onto which the genomic selection model will be applied.
When relying solely on phenotypic selection, breeding programs may Allard, R.W., 1961. Relationship between genetic diversity and consistency of
envisage solutions that combine the advantages of bulk-based selection performance in different environments. Crop Sci. 1, 127–133.
Annicchiarico, P., 2005. Scelta varietale in pisello e favino rispetto all’ambiente e
with the need for early generation of inbred lines in different ways, for all’utilizzo. Inf. Agr. 61 (49), 47–52.
example by applying bulk breeding to just a few early generations of Annicchiarico, P., 2008. Adaptation of cool-season grain legume species across
segregating material followed by speed breeding or off-season genera­ climatically-contrasting environments of Southern Europe. Agron. J. 100,
1647–1654.
tion of its mass-selected lines. In any case, EPs let evolved under specific Annicchiarico, P., Iannucci, A., 2008. Adaptation strategy, germplasm type and adaptive
conditions could represent a precious dynamic management of genetic traits for field pea improvement in Italy based on variety responses across
resources for future selection. Breeding line mixtures is hardly climatically contrasting environments. Field Crops Res. 108, 133–142.
Annicchiarico, P., Nazzicari, N., Pecetti, L., Romani, M., Ferrari, B., Wei, Y., Brummer, E.
contemplated by mainstream breeding programs as yet, but the sizeable
C., 2017. Accuracy of GBS-based genomic selection for pea grain yield under severe
increase of performance displayed by mixtures currently selected by the terminal drought. Plant Genome 10. https://doi.org/10.3835/
simple procedure of assembling top-yielding lines with a compatible plantgenome2016.07.0072.
Annicchiarico, P., Collins, R.P., De Ron, A.M., Firmat, C., Litrico, I., Hauggaard-
phenology shows promise, especially when selecting for high yield
Nielsen, H., 2019a. Do we need specific breeding for legume-based mixtures? Adv.
reliability across target environments that feature large GEI. Agron. 157, 141–215.
Annicchiarico, P., Russi, L., Romani, M., Pecetti, L., Nazzicari, N., 2019b. Farmer-
Funding participatory vs. conventional market-oriented breeding of inbred crops using
phenotypic and genome-enabled approaches: a pea case study. Field Crops Res. 232,
30–39.
The different stages of this research work were funded by the projects Annicchiarico, P., Nazzicari, N., Pecetti, L., Romani, M., Russi, L., 2019c. Pea genomic
Core Organic II ‘Coordinating Organic Plant Breeding Activities for Di­ selection for Italian environments. BMC Genom. 20, 603.
Annicchiarico, P., Nazzicari, N., Laouar, M., Thami-Alami, I., Romani, M., Pecetti, L.,
versity (COBRA)’, ‘Foraggi, mangimi, breeding e biodiversità in sistemi 2020. Development and proof-of-concept application of genome-enabled selection
zootecnici biologici (ZooBio2systems)’ and ‘Esplorazione della bio­ for pea grain yield under severe terminal drought. Int. J. Mol. Sci. 21, 2414.
diversità vegetale alla ricerca di alleli superiori da inserire nei pro­ Annicchiarico, P., Nazzicari, N., Notario, T., Monterrubio Martin, C., Romani, M.,
Ferrari, B., Pecetti, L., 2021. Pea breeding for intercropping with cereals: variation
grammi avanzati di miglioramento genetico a sostegno dell’agricoltura for competitive ability and associated traits, and assessment of phenotypic and
nazionale (ESPLORA)’ granted by the Italian Ministry of Agricultural, genomic selection strategies. Front. Plant Sci. 12, 731949.
Food and Forestry Policies, the project ‘Redesigning European cropping Barbieri, P., S., Pellerin, V., Seufert, V., Smith, L., Ramankutty, N., Nesme, T., 2021.
Global option space for organic agriculture is delimited by nitrogen availability. Nat.
systems based on species mixtures (REMIX)’ funded by the European
Food 2, 363–372.
Union (Grant Agreement no. 727217), and core budget by the University Baresel, J.P., Bülow, L., Finckh, M.R., Frese, L., Knapp, S., Schmidhalter, U., Weedon, O.,
of Perugia. 2022. Performance and evolutionary adaptation of heterogeneous wheat
populations. Euphytica 218, 137.
Baxevanos, D., Tsialtas, I.T., Vlachostergios, D.N, Hadjigeorgiou, I., Dordas, C.,
CRediT authorship contribution statement Lithourgidis, A., 2017. Cultivar competitiveness in pea-oat intercrops under
Mediterranean conditions. Field Crops Res. 214, 94–103.
P. Annicchiarico: Funding acquisition, Conceptualization, Meth­ Becker, H.C., Léon, J., 1988. Stability analysis in plant breeding. Plant Breed. 101, 1–23.
Bedoussac, L., Journet, E.P., Hauggaard-Nielsen, H., Naudin, C., Corre-Hellou, G.,
odology, Germplasm resources, Investigation, Formal analysis, Writing - Jensen, E.S., Prieur, L., Justes, E., 2015. Ecological principles underlying the
original draft. L. Russi: Funding acquisition, Investigation, Writing - increase of productivity achieved by cereal-grain legume intercrops in organic
review & editing. M. Romani: Germplasm resources, Investigation. T. farming. A review. Agron. Sustain. Dev. 35, 911–935.
Bertholdsson, N.O., Weedon, O., Brumlop, S., Finckh, M.R., 2016. Evolutionary changes
Notario: Investigation. L. Pecetti: Investigation, Writing - review & of weed competitive traits in winter wheat composite cross populations in organic
editing. and conventional farming systems. Eur. J. Agron. 79, 23–30.
Bocci, R., Bussi, B., Petitti, M., Franciolini, R., Altavilla, V., Galluzzi, G., Di Luzio, P.,
Migliorini, P., Spagnolo, S., Floriddia, R., Li Rosi, G., Petacciato, M., Battezzato, V.,
Declaration of Competing Interest Albino, A., Faggio, G., Arcostanzo, C., Ceccarelli, S., 2020. Yield, yield stability and
farmers’ preferences of evolutionary populations of bread wheat: a dynamic solution
The authors declare that they have no known competing financial to climate change. Eur. J. Agron. 121, 126156.
Brumlop, S., Pfeiffer, T., Finckh, M., 2017. Evolutionary effects on morphology and
interests or personal relationships that could have appeared to influence agronomic performance of three winter wheat composite cross populations
the work reported in this paper. maintained for six years under organic and conventional conditions. Org. Farming 3,
34–50.
Ceccarelli, S., Grando, S., 2020. Evolutionary plant breeding as a response to the
Data availability
complexity of climate change. iScience 23, 101815.
Ceccarelli, S., Grando, S., 2022. Evolutionary plant breeding with an introduction to
Data will be made available on request. participatory plant breeding. Mim Edizioni Srl, Sesto San Giovanni, Italy.
Döring, T.F., Annicchiarico, P., Clarke, S., Haigh, Z., Jones, H.E., Pearce, H., Snape, J.,
Zhan, J., Wolfe, M.S., 2015. Comparative analysis of performance and stability
Acknowledgements among composite cross populations, variety mixtures and pure lines of winter wheat
in organic and conventional cropping systems. Field Crop. Res. 183, 235–245.
We gratefully acknowledge the technical contribution by S. Proietti, Döring, T.F., Knapp, S., Kovacs, G., Murphy, K., Wolfe, M.S., 2011. Evolutionary plant
breeding in cereals - into a new era. Sustainability 3, 1944–1971.
A. Passerini, C. Monterrubio Martin, F. Vecchietti and S. Vergoni, and Ekbohm, G., 1981. A test for the equality of variances in the paired case with incomplete
the farmer-participatory evaluation coordinated by C. Micheloni, D. data. Biom. J. 23, 261–265.

13
P. Annicchiarico et al. Field Crops Research 293 (2023) 108831

Eskridge, K.M., 1990. Selection of stable cultivars using a safety-first rule. Crop Sci. 30, Murphy, K., Lammer, D., Lyon, S., Carter, B., Jones, S., 2005. Breeding for organic and
369–374. low-input farming systems: an evolutionary–participatory breeding method for
Finckh, M., Gacek, E., Goyeau, H., Lannou, C., Merz, U., Mundt, C., Munk, L., Nadziak, J., inbred cereal grains. Renew. Agric. Food Syst. 20, 48–55.
Newton, A., de Vallavieille-Pope, C., Wolfe, M.S., 2000. Cereal variety and species Patel, J.D., Reinbergs, E., Mather, D.E., Choo, T.M., Sterling, J.D.E., 1987. Natural
mixtures in practice, with emphasis on disease resistance. Agronomie 20, 813–837. selection in a doubled-haploid mixture and a composite cross of barley. Crop Sci. 27,
Foyer, C.H., Lam, H.-M., Nguyen, H.T., Siddique, K.H.M., Varshney, R.K., Colmer, T.D., 474–479.
Cowling, W., Bramley, H., Mori, T.A., Hodgson, J.M., Cooper, J.W., Miller, A.J., Pecetti, L., Marcotrigiano, A.R., Russi, L., Romani, M., Annicchiarico, P., 2019.
Kunert, K., Vorster, J., Cullis, C., Ozga, J.A., Wahlqvist, M.L., Liang, Y., Shou, H., Adaptation of field pea varieties to organic farming across different environments of
Shi, K., Yu, J., Fodor, N., Kaiser, B.N., Wong, F.-L., Valliyodan, B., Considin, M.J., Italy. Crop Pasture Sci. 70, 327–333.
2016. Neglecting legumes has compromised human health and sustainable food Phillips, S.L., Wolfe, M.S., 2005. Evolutionary plant breeding for low input systems.
production. Nat. Plants 2, 16112. J. Agric. Sci. 143, 245–254.
Gill, J.S., Verma, M.M., Gumber, R.K., Brar, J.S., 1995. Comparative efficiency of four Piepho, H.P., 1998. Methods for comparing the yield stability of cropping systems - a
selection methods for deriving high-yielding lines in mungbean [Vigna radiata (L.) review. J. Agron. Crop Sci. 180, 193–213.
Wilczek]. Theor. Appl. Genet. 90, 554–560. Pilorgé, E., Muel, F., 2016. What vegetable oils and proteins for 2030? Would the protein
Goldringer, I., Enjalbert, J., Raquin, A.L., Brabant, P., 2001. Strong selection in wheat fraction be the future of oil and protein crops? OCL 23, D402.
populations during ten generations of dynamic management. Genet. Sel. Evol. 33, Polowic, P.L., Vandenberg, A., Mahon, J.D., 2002. Field assessment of outcrossing from
S441. transgenic pea (Pisum sativum L.) plants. Transgenic Res. 11, 515–519.
Goldringer, I., Prouin, C., Rousset, M., Galic, N., Bonnin, I., 2006. Rapid differentiation of Raggi, L., Ciancaleoni, S., Torricelli, R., Terzi, V., Ceccarelli, S., Negri, V., 2017.
experimental populations of wheat for heading time in response to local climatic Evolutionary breeding for sustainable agriculture: selection and multi-
conditions. Ann. Bot. 98, 805–817. environmental evaluation of barley populations and lines. Field Crop. Res. 204,
Goldringer, I., van Frank, G., Bouvier d′ Yvoire, C., Forst, E., Galic, N., Garnault, M., 76–88.
Locqueville, J., Pin, S., Bailly, J., Baltassat, R., Berthellot, J.-F., Caizergues, F., Ranalli, P., Cubero, J.I., 1997. Bases for genetic improvement of grain legumes. Field
Dalmasso, C., de Kochko, P., Gascuel, J.-S., Hyacinthe, A., Lacanette, J., Mercier, F., Crops Res. 53, 69–82.
Montaz, H., Ronot, B., Rivière, P., 2020. Agronomic evaluation of bread wheat Rasmusson, D.C., Beard, B.H., Johnson, F.K., 1967. Effect of natural selection on
varieties from participatory breeding: a combination of performance and robustness. performance of a barley population. Crop Sci. 7, 543-543.
Sustainability 12, 128. Reiss, E.R., Drinkwater, L.E., 2018. Cultivar mixtures: a meta-analysis of the effect of
Haddad, N.I., Muehlbauer, F.J., 1981. Comparison of random bulk population and single- intraspecific diversity on crop yield. Ecol. Appl. 28, 62–77.
seed-descent methods for lentil breeding. Euphytica 30, 643–651. SAS Institute, 2011. SAS/STAT® 9.3 User’s guide. SAS Institute Inc., Cary, NC.
Hamblin, J., Rowell, J.G., 1975. Breeding implications of the relationship between Shukla, G.K., 1972a. Some statistical aspects of partitioning genotype-environmental
competitive ability and pure culture yield in self-pollinated grain crops. Euphytica components of variability. Heredity 29, 237–245.
24, 221–228. Shukla, G.K., 1972b. An invariant test for the homogeneity of variances in a two-way
Harlan, H.V., Martini, M.L., 1929. A composite hybrid mixture. Agron. J. 21, 487–490. classification. Biometrics 28, 1063–1072.
Hauggaard-Nielsen, H., Jensen, E.S., 2001. Evaluating pea and barley cultivars for Simmonds, N.W., 1979. Principles of Crop Improvement. Longman, London.
complementarity in intercropping at different levels of soil N availability. Field Singh, K.B., Malhotra, R.S., Halila, M.H., Knights, E.J., Verma, M.M., 1994. Current
Crops Res. 72, 185–196. status and future strategy in breeding chickpea for resistance to biotic and abiotic
Horneburg, B., Becker, H.C., 2008. Crop adaptation in on-farm management by natural stresses. Euphytica 73, 137–149.
and conscious selection: A case study with lentil. Crop Sci. 48, 203–212. Smithson, J.B., Lenné, J.M., 1996. Varietal mixtures: a viable strategy for sustainable
Kervella, J., Fouilloux, G.A., 1992. Theoretical study of the bulk breeding method. productivity in subsistence agriculture. Ann. Appl. Biol. 128, 127–158.
Euphytica 60, 185–195. Soliman, K.M., Allard, R.W., 1991. Grain yield of composite cross populations of barley:
Khalifa, M.A., Qualset, C.O., 1975. Intergenotypic competition between tall and dwarf effects of natural selection. Crop Sci. 31, 705–708.
wheats. II. In hybrid bulks. Crop Sci. 15, 640–644. Steffen, W., Richardson, K., Rockström, J., Cornell, S.E., Fetzer, I., Bennett, E.M.,
Khosla, G., Gill, B.S., Sharma, P., 2019. Comparison of different breeding methods for Biggs, R., Stephen, C.R., de Vries, W., de Wit, C.A., Folke, C., Gerten, D., Heinke, J.,
developing superior genotypes in soybean. Agric. Res. J. 56, 628–634. Mace, G.M., Persson, L.M., Ramanathan, V., Reyers, B., Sörlin, S., 2015. Planetary
Knapp, S., Döring, T.F., Jones, H.E., Snape, J., Wingen, L.U., Wolfe, M.S., Leverington- boundaries: guiding human development on a changing planet. Science 347,
Waite, M., Griffiths, S., 2020. Natural selection towards wild-type in composite cross 1259855.
populations of winter wheat. Front. Plant Sci. 10, 1757. Suneson, C.A., 1956. An evolutionary plant breeding method. Agron. J. 48, 188–191.
Kumar, R., Niwas, R., Dahiya, B.S., 1992. Comparison of selection methods in dwarf field Timaeus, J., Weedon, O.D., Finckh, M.R., 2022. Harnessing the potential of wheat-pea
pea (Pisum sativum L.). II. Effectiveness for yield and its components. Int. J. Trop. species mixtures: evaluation of multifunctional performance and wheat diversity.
Agric. 10, 172–179. Front. Plant Sci. 13, 846237.
Legzdiņa, L., Bleidere, M., Piliksere, D., Ločmele, I., 2022. Agronomic performance of Turner, N.C., Wright, G.C., Siddique, K.H.M., 2001. Adaptation of grain legumes (pulses)
heterogeneous spring barley populations compared with mixtures of their parents to water-limited environments. Adv. Agron. 71, 194–233.
and homogeneous varieties. Sustainability 14, 9697. van de Wouw, M., Kik, C., Van Hintum, T., Van Treuren, R., Visser, B., 2009. Genetic
Lejeune-Hénaut, I., Wery, J., 1994. Influence du froid sur la survie des plantes. In: Ney, erosion in crops: concept, research results and challenges. Plant Genet. Resour. 8,
B., Duchêne, E., Carrouée, B., Angevin, F. (Eds.). Agrophysiologie du pois 1–15.
protéagineux. UNIP-ITCF-INRA, Paris, 139–144. van Frank, G., Rivière, P., Pin, S., Baltassat, R., Berthellot, J.-F., Caizergues, F.,
McDonald, G.K., 2003. Competitiveness against grass weeds in field pea genotypes. Weed Dalmasso, C., Gascuel, J.-S., Hyacinthe, A., Mercier, F., Montaz, H., Ronot, B.,
Res. 43, 48–58. Goldringer, I., 2020. Genetic diversity and stability of performance of wheat
Meena, H.P., Kumar, J., 2012. Relative efficiency of different breeding methods for population varieties developed by participatory breeding. Sustainability 12, 384.
improvement of yield and yield components in chickpea (Cicer arietinum L.). J. Food Watson, C.A., Reckling, M., Preissel, S., Bachinger, J., Bergkvist, G., Kuhlman, T.,
Legumes 25, 165–170. Lindström, K., Nemecek, T., Topp, C.F.E., Vanhatalo, A., Zander, P., Murphy-
Merrick, L.F., Lyon, S.R., Balow, K.A., Murphy, K.M., Jones, S.S., Carter, A.H., 2020. Bokern, D., Stoddard, F.L., 2017. Grain legume production and use in european
Utilization of evolutionary plant breeding increases stability and adaptation of agricultural systems. Adv. Agron. 144, 235–303.
winter wheat across diverse precipitation zones. Sustainability 12, 9728. Witcombe, J., Virk, D., 2001. Number of crosses and population size for participatory and
Miladinovič, J., Burton, J.W., Tubič, S.B., Miladinovič, D., Djordjevič, V., Djurič, V., classical plant breeding. Euphytica 122, 451–462.
2011. Soybean breeding: comparison of the efficiency of different selection methods. Wuest, S.E., Peter, R., Niklaus, P.A., 2021. Ecological and evolutionary approaches to
Turk. J. Agric. 35, 469–480. improving crop variety mixtures. Nat. Ecol. Evol. 5, 1068–1077.
Mobini, S.H., Warkentin, T.D., 2016. A simple and efficient method of in vivo rapid Zampieri, M., Weissteiner, C.J., Grizzetti, B., Toreti, A., van den Berg, M., Dentener, F.,
generation technology in pea (Pisum sativum L.). Vitr. Cell. Dev. Biol. Plant 52, 2020. Estimating resilience of crop production systems: from theory to practice. Sci.
530–536. Total Environ. 735, 139378.

14