ORIGINAL RESEARCH

published: 18 September 2019

doi: 10.3389/fpls.2019.01111

Proposal of a New Hybrid Breeding

Method Based on Genotyping,
Inter-Pollination, Phenotyping and
Paternity Testing of Selected Elite
F1 Hybrids
Katarina Rudolf-Pilih 1, Marko Petkovšek 2, Jernej Jakše 1, Nataša Štajner 1, Jana Murovec 1
and Borut Bohanec 1*
1Biotechnical Faculty, University of Ljubljana, Ljubljana, Slovenia, 2 Faculty of Mathematics and Physics, University of
Ljubljana, Ljubljana, Slovenia

Testing inbred lines for their combining ability is, due to high numbers of line to line testing needed
Edited by: for determination of hybrid performance, the most limiting factor in the F1 hybrid breeding
Ryo Fujimoto,
Kobe University, Japan
procedure. We propose a novel method of F1 hybrid breeding that enables evaluation of large
Reviewed by:
number of line to line crosses for their hybrid performance. Inbred lines (preferably doubled
Zhanguo Xin, haploid - DH) are produced from heterozygous populations, genotyped and maintained. A
Department of Agriculture,
group of lines is inter-pollinated randomly and their progeny examined. To identify elite F1
United States
Radu E. Sestras, hybrids, these individual plants are selected by their superior phenotypic characteristics.
University of Agricultural Sciences Finally using paternity testing only of selected hybrids, the origin of paternal lines is revealed.
and Veterinary Medicine of
Cluj-Napoca, Romania
To predict the number of F1 offspring needed in relation to the number of inbred lines being
*Correspondence:
inter-pollinated, a mathematical formula was developed. For instance, using this formula for
Borut Bohanec the inter-pollination of 60 distinct lines, the probability of obtaining all descendants of paternal-
[email protected]
parent lines in a maternal-parent row represented at least once is achieved with 420 F1 plants
in a row (p = 0.95). In a practical experiment with white cabbage, DH lines were produced
Specialty section:
This article was submitted to using microspore culture; plants were grown to maturity and genotyped at eight polymorphic
Plant Breeding, SSR loci. Two groups of lines (36 and 33 lines per group) were inter-pollinated by two
a section of the journal
Frontiers in Plant Science methods, either using cage pollination with bumblebees or using open pollination in isolated
Received: 14 June 2019 field. A total of 9,858 F1 plants were planted and based on their phenotypic characteristics
Accepted: 13 August 2019 213 were selected as elite phenotypes. 99 of them were genetically diverse and 5 of them
Published: 18 September 2019
were selected as super elite. Selected plants were analysed by the same SSR markers and
Citation:
the paternal origin of selected F1 plants was determined. Out of 213 selected elite plants 48
Rudolf-Pilih K, Petkovšek M,
Jakše J, Štajner N, Murovec J and were reciprocals thus exhibiting power of selection based on single plant. We demonstrate
Bohanec B (2019) Proposal of a that this new approach to hybrid development is efficient in white cabbage and we propose
New Hybrid Breeding Method Based
on Genotyping, Inter-Pollination, breeders to test it in various vegetable and crop species. Moreover, some other aspects of the
Phenotyping and Paternity Testing of proposed technique need to be tested and verified both for practical and economic criteria.
Selected Elite F1 Hybrids.
Front. Plant Sci. 10:1111. Keywords: F1 hybrid breeding, doubled haploids, testing of combining ability, paternity determination, simple
doi: 10.3389/fpls.2019.01111 sequence repeats markers

Frontiers in Plant Science  |  www.frontiersin.org 1 September 2019 | Volume 10 | Article 1111

Rudolf-Pilih et al. Innovative F1 Hybrid Breeding Method

INTRODUCTION alternative to testing for combining ability termed “reverse

breeding” was proposed by Dirks et al. (2009) and elaborated
Hybrid breeding methods have long been used to efficiently by Wijnker et al. (2012). In their method, superior individual
produce varieties with superior performance and are today heterozygous plants are first identified within a population,
considered the optimal choice due to the expressed heterotic then due to achiasmatic chromosomes, formed by the
effect, uniformity, and fast trait combination and as a form insertion of genes preventing crossing over, gametes with
of intellectual property protection. Although causal factors limited recombination frequency are formed. The following
and genetic mechanisms of heterosis remain not completely steps aim to create an identical F1 hybrid as the original
understood (Berlan, 2018; Miyaji and Fujimoto, 2018; heterozygote required haploid induction and the monitoring
Govindaraju, 2019) hybrid breeding is now applied in majority of individual chromosomes by molecular markers. The
of crops. method has several limitations: it can only be used in species
Several textbooks, such as those by Fehr (1993), Simmonds and with small chromosome numbers and the transgenic status of
Smartt (1999) or Acquaah (2012), and research papers (Nishio and intermediate generation can pose an obstacle.
Kitashiba, 2017; Dimitirijevic and Horn, 2018) describe methods Due to these limitations, improvements to existing protocols
for breeding hybrid cultivars of various species. To produce an seem crucial for further F1 breeding progress. Here, we provide
F1 hybrid variety, several putative parental lines obtained from theoretical and practical examination of the new protocol
heterozygous sources are first made homozygous by several aimed at providing breeders with the ability to perform a higher
generations of inbreeding from a genetically heterogeneous number of line to line testing compared to existing protocols. The
genepool. Recently, haploid induction followed by chromosome plant species chosen for practical evaluation of the protocol was
doubling has frequently replaced selfing and is now recognised as white cabbage (Brassica oleracea L.), which is due to sporophytic
the most convenient method to produce inbred lines (Murovec incompatibility cross-pollinated species expressing inbreeding
and Bohanec, 2012). Alternatively, by accelerated selfing to up depression when made homozygous. However, the described
to six generations per year, inbreeding can be achieved by the protocol can be implemented also in hybrid breeding programs
process termed as “speed breeding” (Watson et  al., 2018). This of self-pollinated species in which several approaches to prevent
method does not lead to complete homozygosity but has so far self-pollination.
been developed for several predominantly cereal and legume The manuscript is composed of three parts; in the first
species. Once inbred lines are created, breeders attempt to part, the new breeding scheme is proposed; in the second
recognise which crossing combination of lines give superior part, theoretical calculations are provided for the number of
offspring, a well-known effect termed heterosis. F1 progenies needed to achieve a significant number of cross
Although genetic mechanisms that cause heterosis are still combinations; and in the third part, the whole procedure was
not completely understood (Chen, 2013), several methods tested in practical experiment with white cabbage using DH
used by breeders are well established and are focused on the lines, their genotyping, maintenance, inter-pollination, field
identification of lines with optimal “combining ability” (Acquaah, selection and paternity determination.
2012). Combining ability is the phenomenon that comes forth
only when some inbred lines are crossed with each other,
complementing each other in desired traits. Plant breeders often MATERIAL AND METHODS
tend to produce a large number of inbred lines in a breeding cycle,
one thousand or more being very common. It is well known that Breeding Scheme
it is impossible to perform all cross combinations, for practical A new breeding method (Figure 1) was developed, supported
reasons. To perform all possible combinations without reciprocals, by mathematical calculations and tested in a practical breeding
n(n-1)/2 crosses are required (n being the number of inbred lines). protocol as described here.
Testing the combining ability of each of a hundred lines would Cabbage cultivars and DH were obtained from our cabbage
require 4,950 non-reciprocal crosses. Therefore, in a standard F1 hybrid breeding program, the list of genotypes included is
breeding scheme, the first step is testing for “general combining given in Supplementary Tables S1, S2, S3 and S4. In 2016,
ability”, meaning testing of a large number of lines with one or doubled haploid lines were induced from highly heterozygous
more “tester lines”. Based on progeny performance, the best are donor plants (Supplementary Tables S1 and S2) by microspore
chosen for “specific combining ability (SCA)” testing, i.e., line- culture, as described earlier (Rudolf-Pilih et al., 2018).
to-line crossing in a single or reciprocal way. Particularly in Genetically diverse DH cabbage seedlings were cultivated on
vegetable breeding, F1 hybrid breeding protocols tend to exclude a sterilised clay substrate (Klasmann-Deilmann) fertilised
testing for general combining ability; therefore, the performance with Osmocote exact (15-9-12+2MgO+TE) in a greenhouse
of a limited number of previously determined superior seed lines and exposed to low temperatures (0–5°C) for three months.
is predominantly tested by various pollen parents. Such search During this period, each plant was genetically characterised
for optimal combining ability in relation to metabolites was by eight SSR markers as described in the next paragraph.
presented in white cabbage by Singh et. (2018). Vernalised plants were induced to flowering in 2017. DH lines
Testing inbred lines for their combining ability is the most were maintained by selfing, which was efficient in only 6 of 69
limiting factor in the F1 hybrid breeding procedure. Therefore, lines due to self-incompatibility. For this reason, the majority
some attempts were made to overcome this bottleneck. An of lines were maintained by micropropagation. Briefly, axillary

Frontiers in Plant Science  |  www.frontiersin.org 2 September 2019 | Volume 10 | Article 1111

Rudolf-Pilih et al. Innovative F1 Hybrid Breeding Method

FIGURE 1 | A method for combining ability testing of F1 hybrids by inter-pollination among genetically characterized DH lines and by revealing paternal origin of
identified elite individuals within intercrossed progeny.

buds were sterilised in 1.6% dichloroisocyanuric acid, washed a cage at the beginning of flowering with bumblebees
in sterile doubled-distilled water and placed on MS medium (Bombus  terrestris) obtained from Koppert B.V. (Berkel
containing 20 g l-1 sucrose, 8 g l-1 agar, 2 mg l-1 indolebutyric en Rodenrijs, The Netherlands) (Figure 2A) until the end
acid and 3 mg-1 benzylaminopurine. Shoots were subcultured of flowering. In the second procedure (open pollination),
on the same medium, while roots were induced on half 33 lines, each represented by a single plant, were placed
strength MS medium lacking growth regulators. on an isolated field and exposed to natural pollinators,
For inter-pollination, two separate groups of DH lines predominantly bees. At maturity, pods were collected, dried
were formed and inter-pollinated randomly by two different and then seeds were scored and stored at 4°C until the next
procedures. In the first procedure (cage pollination), 36 season. The selection of elite F1 hybrid plants was performed
lines, each represented by a single plant, were placed into in 2018. Seeds were sowed at the end of March into a plug tray

Frontiers in Plant Science  |  www.frontiersin.org 3 September 2019 | Volume 10 | Article 1111

Rudolf-Pilih et al. Innovative F1 Hybrid Breeding Method

FIGURE 2 | Illustration of selection process: (A) cage pollination using bumblebees (B) maternal-parent rows at transplanting (C) maternal parent rows at
marketable maturity (D) head- cross section of representative hybrid: HL, head length; Hwi, Head width; ICL, inner core length.

with 160 cells (60 trays in total) and seedlings were planted on polymerase; 0.15 µM of each primer (forward tailed primer
the experimental field at the beginning of May and grown until and reverse primer). Each forward SSR primer has an universal
marketable maturity. Fertilisation, irrigation and pest control M13 18 bp tail sequence added at 5’ end (5’-TGT AAA ACG
were performed according to general agricultural practice. ACG GCC AGT-3’) (Schuelke, 2000). Four different fluorescent
In total, 9,858 seedlings were grown with known maternal dyes (6-FAM, VIC, NED, and PET) were used to label the four
origin. All F1 hybrid descendants of a single maternal-parent universal M13 primers which were included in single locus PCR
plant were planted in a single row (Figure 2B). As previously at a concentration of 0.2 µM.
described (Rudolf-Pilih et al., 2012) at maturity (Figure 2C), The cycling conditions were as follows: 95°C for 5 min; 10
the following criteria were used to select elite F1 individual cycles of 30 s at 95°C, 30 s at 65°C, which was lowered by 1°C in
plants: head weight, head length, head width, length of inner each cycle, and 30 s at 72°C; 25 cycles of 30 s at 95°C, 30 s at 55°C,
core, head firmness and head shape (Figure 2D), while and 30 s at 72°C; and a 5-min extension step at 72°C. Samples
other characteristics like field resistance to Xanthomonas were kept at 4°C until analysis.
campestris, maturity, leaf colour and wax coating were also The PCR products amplified with 4 different dyes were
recorded. A leaf sample from each selected elite F1 plant was mixed together and same amount of formamide was added with
collected for DNA isolation and determination of paternity by GeneScan 600 LIZ size standard, heat denatured, chilled on
SSR markers. DNA was isolated following the CTAB method ice and run on a capillary electrophoresis system ABI 3730XL
(Doyle and Doyle 1987). analyser (Applied Biosystems). Resulting electropherograms
were analysed using GeneMapper 4.0 (Applied Biosystems) or
Target SSRs and Design of Multiplex PeakScanner software (Applied Biosystems), where the length of
Primers for Genotyping alleles was recorded.
Eight microsatellite loci were included in the paternity test
(Table  1), based on the power of distinction evaluated in the Paternity Assignment
genotyping study of 352 cabbage genotypes (data available Using likelihood ratios, determination of paternity by CERVUS
upon request). Amplification of eight microsatellites Kholilatul 3.0.7 software was calculated (Kalinowski et al., 2007). Based
et al. (2014) (Table 1) was performed in a total volume of 15 μl on genotyping data with eight microsatellites, the following
containing 15 ng of DNA template, 1x PCR reaction buffer, parameters of genetic variability and information content were
3.0 mM MgCl2, 0.8 mM of each dNTP, 0.45 unit Taq DNA calculated for all 282 genotypes analysed.

Frontiers in Plant Science  |  www.frontiersin.org 4 September 2019 | Volume 10 | Article 1111

Rudolf-Pilih et al. Innovative F1 Hybrid Breeding Method

TABLE 1 | SSR markers and characteristics of primers used in genetic analysis encountered, carrying a highly desirable set of allelic combinations.
of white cabbage (Kholilatul et al., 2014).
These individual F1 hybrid plants, being heterozygous and unique,
Locus/Marker Sequences 5’-3’ Motif are identified by their phenotypes. Since each inbred line is
genetically characterised both parents are identified and confirmed
1. BoESSR825-for GGACAGCGACACATTGAGTG
by molecular marker analysis.
BoESSR825-rev GGGAAGAGGTTCCCAAACAT (CCG)7
2. BoESSR391-for GCGACCTGTTGAAGAAGGAG The scheme of the method for breeding hybrid plants is given
BoESSR391-rev TTCTCCGCAAGAAATACAAGG (GAT)7 in Figure 1; a more detailed explanation of each step is given
3. BoESSR632-for CCCTGCAATTGAAAACCAGT below:
BoESSR632-rev AAACCGTCCAAGGATCATCA (TGT)7 Step 0: Establishment of heterozygous starting population due
4. BoESSR492-for GCGCAGAATCCAGATCATAG
BoESSR492-rev GGCTGGAGTATGAGCGAGAC (GA)9
to the breeding goals.
5. BoESSR338-for TGTAGCCGAAAGGGAATGAG Step 1: Producing essentially homozygous donor lines from
BoESSR338-rev GTGCTTGCATCCAGAAACCT (AC)10 starting population. This can be optimally achieved by induction
6. BoESSR484-for ACCCATACGTCCACGTCAAT of doubled haploids, but other methods of inbreeding can also
BoESSR484-rev GCAATCGTCTTTCCACCAAT (AGA)7 be considered.
7. BoESSR087-for GTTTCCTCTTCCACCACCAA
BoESSR087-rev AATCTATCAAGAGGGCCAAGG (TCC)7
Step 2: Each DH line is genetically characterised by means
8. BoESSR053-for TTTGCCAAGAAGCCTGAAGT of molecular markers to obtain a unique genetic profile for each
BoESSR053-rev TGTACCAGCTGCAACCTCTG (GAA)7 line. For each group of lines entering step 3, a unique database is
formed.
Step 3a: Each DH line is maintained either by selfing (if
Parentage analysis was performed in two steps, in the simulation
achievable) or clonally propagated by various means including
of parentage analysis and actual parentage analysis. Simulation
micropropagation.
was run to estimate the resolving power of a series of SSR loci;
Step 3b: Inbred lines are induced to flower simultaneously
simulation parameters were as follows: 10,000 progenies, the
and random inter-pollination is stimulated by various methods
number of candidate parents set to 36 for cage pollination and 33
to obtain F1 hybrid progeny.
for open pollination, the proportion of sampled parents assumed
Step 4: F1 hybrid progeny is sown and the maternal origin
to be one, proportion of loci typed 0.998 and proportion of loci
for each seed is recorded (»maternal-parent rows«). Elite F1
mistyped 0.002. Actual parentage analysis was performed to test
hybrid individuals are identified by phenotyping among the F1
candidate parents against offspring and, for each offspring tested,
progeny;
to assign the most-likely candidate parent with a pre-determined
Step 5: Paternity of selected elite individual plants is
level of confidence. Finally, paternity analysis was done as maternal
determined by molecular markers using specific software by
genotypes were available and 100% confirmed with parentage
comparison to the DH line database formed in Step 2.
analysis. The objective was to assign a male parent to each offspring.
The number of lines entering inter-pollination is limited by
The overall likelihood ratios were expressed as logarithm of odd
practical reasons such as pollination characteristics of plants and
(LOD) scores (natural logarithm of the overall likelihood ratio) and
the ability of breeders to maintain lines and perform analysis.
were assigned to each possible parent and parent pair. When the trio
For most cases, we propose that the number of lines within one
LOD score had a probability >95% (strict) or >80% (relaxed) based
group will not exceed 100. In the case that more lines need to be
on confidence derived from prior parentage simulation, and there
tested, a two-step procedure is proposed. In the first step, a group
were no parental marker discrepancies, the candidate parents were
of, say, 100 lines are tested according to the scheme in Figure 1,
assigned to the progeny plant. CERVUS only assigned paternity if
and at the same time additional groups are tested in the same way.
at least eight out of eight loci were scored.
Identification of a particular DH that produced numerous selected
hybrid plants in cycle 1, would be an indication of positive general
Mathematical Calculations combining ability. These good combining DHs can be intercrossed
For calculation of line/offspring probabilities, Wolfram
in a second cycle to identify superior hybrids.
Mathematica v. 11 was used, as discussed in Results point B.
B. Theoretical Calculation of the Size of
RESULTS F1 Families Needed to Obtain a Given
Probability of Cross Combinations
A. Breeding Scheme “Inter-Pollination/ Following Inter-Pollination Among
Progeny Testing” Inbred Lines
A breeding scheme, aimed to increase the efficiency of inbred testing Inter-pollination among selected inbred lines by previously
for hybrid development is proposed. Briefly, from a heterozygous explained methods produces F1 progeny that should in the
starting population, DH lines are produced and genotyped. optimal case represent all possible combinations. The probability
A selected group of diverse lines (determined by molecular of achieving all crossing combinations is related to the number of
markers) is inter-pollinated and their F1 progeny evaluated in lines entering inter-pollination and the number of seeds obtained
field experiments. It is expected that, by chance, in the progeny and sown by each inbred line. We developed a mathematical
of random crosses between inbred lines individual elite plants are model that can be used for the calculation of these probabilities.

Frontiers in Plant Science  |  www.frontiersin.org 5 September 2019 | Volume 10 | Article 1111

Rudolf-Pilih et al. Innovative F1 Hybrid Breeding Method

We also examined consequences of the number of F1 seeds needed TABLE 2 | Probabilities Q(n, k, m) for n = 61 and m = 60, 54, 48.
in case the breeder is satisfied with some missing combinations - k Q(61, k, 60) Q(31, k, 54) Q(31, k, 48)
for instance with 90 or 80% one-way crossing representation per
maternal-parent row. 60 0.000 0.000 0.000
120 0.000 0.216 0.969
Assumptions: A set L = {v1, v2,…, vn} of n distinct inbred lines of
180 0.032 0.979 1.000
a plant species is given. For each i {1,2,…,n}, k plants of line vi are 240 0.304 1.000 1.000
grown and successfully pollinated by a mixture containing pollen 300 0.650 1.000 1.000
from each of the lines v1, v2, ..., vn . It is assumed that the probability 360 0.854 1.000 1.000
of successful pollination of vi by vj is independent of i and j. 420 0.943 1.000 1.000
480 0.979 1.000 1.000
Question: We wondered what the probability is for n lines 540 0.992 1.000 1.000
in the F1 offspring that, in an individual row with the same 600 0.997 1.000 1.000
maternal-parents, at least m of the paternal-parents are present 660 0.999 1.000 1.000
at least once. More precisely, what is the probability Q(n, k, m) 720 1.000 1.000 1.000
that a fixed maternal-parent line vi will be pollinated by at least m
paternal-parent lines different from vi?
To answer this, we model our experiment by the process of
TABLE 3 | Probabilities Q(n, k, m) for n = 31 and m = 30, 27, 24.
selecting, uniformly at random, an element from the set Lk of all
strings of length k over the alphabet L. First, we fix a subset Lj ⊆ L \ k Q(31, k, 30) Q(31, k, 27) Q(31, k, 24)
{vi} of j paternal-parent lines, and enumerate the set Mj of all those
30 0.000 0.000 0.003
strings of length k over Lj  {vi} in which each of the lines from Lj
60 0.004 0.340 0.925
appears at least once. Clearly Mj = M’  M’’ where M’ is the set of 90 0.173 0.947 1.000
those strings from Mj which do not contain vi, and M’’ the set of 120 0.540 0.999 1.000
those strings from Mj which do contain vi. Then, M’ is in a one-to- 150 0.799 1.000 1.000
one correspondence with the set of all surjective maps from the set 180 0.921 1.000 1.000
210 0.970 1.000 1.000
{1, 2,…, k} onto Lj, and M’’ is in a one-to-one correspondence with 240 0.989 1.000 1.000
the set of all surjective maps from {1, 2,…, k} onto Lj  {vi}, so: 270 0.996 1.000 1.000
300 0.998 1.000 1.000
|M j | = |M'| + |M''| = j ! Sk , j + (j + 1)! Sk , j +1 330 0.999 1.000 1.000
360 1.000 1.000 1.000

= j ! (Sk , j + (j + 1)Sk , j +1 ) = j ! Sk +1, j +1 (1)

Tables 2 and 3 exhibit some values (rounded to three decimal
where Sk,j denotes the Stirling number of the second kind, places) obtained from formula (2) for the probability that given a
enumerating the set of all partitions of a k-element set into fixed maternal-parent line, vi, at least m paternal-parent lines will
j non-empty blocks. Since a set Lj ⊆ L \ {vi} can be chosen in appear in the F1 seeds.
 n −1 
  ways, formula (1) implies that the cardinality of the
 j 
C. Elaboration of »Interpollination/
set Mj,k,n of all those strings Lk in which exactly j lines different Paternity Testing« Method in White
from vi appear is: Cabbage
Interpollination Among DH Lines
 n −1 
|M j ,k ,n| =    j!S Interpollination was successful in both pollination methods. It
 j  k +1, j +1 should be noted that since each DH was represented just as a
single plant, suboptimal representation of each genotype can be
Therefore, the number of strings from Lk in which at least m
expected. We found that from 36 and 33 different DHs in cage
lines different from vi appear is equal to:
or open field selections, respectively, 23 and 18 paternal parents
were represented in their elite F1 progeny. It is most likely that
 n −1 
∑ progenies of some paternal DHs were not chosen as elite because
n −1
  j!S
j =m
 j  k +1, j +1 of undesired phenotypic characteristics.

and the probability that at least m paternal-parent lines

distinct from vi will be represented in the offspring is: Selection Procedure
In total, 5,018, and 4,840 F1 hybrid plants resulting from cage
 n −1 
1
∑ pollination or open pollination, respectively, were planted,
n −1
Q(n,k,m) = k   j!S (2)
n j =m
 j  k +1, j +1 of which 126 and 87 were selected based on their phenotypic
characteristics (Table 4 and Table 5). In general, major differences
which is the answer to the question posed. in phenotypic characteristics were found but, as expected for

Frontiers in Plant Science  |  www.frontiersin.org 6 September 2019 | Volume 10 | Article 1111

Rudolf-Pilih et al. Innovative F1 Hybrid Breeding Method

TABLE 4 | List of maternal-parents and the determined paternal-parents within TABLE 4 | Continued
F1 progeny, logarithm of odd (LOD) score and number of selected plants with the
DH plant line line Determined LOD No. of selection
same parent line for cage pollination experiment.
No. (maternal paternal parent with the same
DH plant line line Determined LOD No. of selection parent) parental lines
No. (maternal paternal parent with the same
65 7.62 1
parent)
79 6.27 1
1 11 5.80 1 *rec
249 6.39 1
28/281 5.17 1
11 11 7.58 1 261 7.19 9
79 6.27 2 342 11 5.79 1
192 8.67 1 *rec
*reciprocal.
28 1 7.58 2
59 5.51 1
79 6.27 1 half-sib families, also some similarities among plants within
272 7.71 3
maternal-parent rows (Figure 2).
311 9.46 1 *rec
40 275 9.51 3 In both pollination groups, a relatively large number of
43 11 5.79 1 plants selected in the field according to their phenotypic
341 5.85 6 *rec characteristics had identical parents (Table 4 and Table 5).
48 1 7.57 1 Also, several reciprocal genotypes were found (48 in total),
272 7.71 1
meaning that plants with the same genotype but spread across
311 9.45 1
52 43 6.79 1 the selection field were recognised as elite within rows with
261 7.19 2 different maternal-parents. This finding confirms, that based
342 6.16 1 on phenotypic selection of individual plants at least in white
65 52 6.63 1 cabbage selection is efficient.
121 7.90 4
121 261 7.19 1
Morphological characteristics of six chosen superior hybrids
311 9.45 2 *rec and the standard variety ‘Presnik F1’ are given in Table 6.
341 5.85 1
189 79 6.27 1 Determination of Paternity
275 9.51 1 Eight SSR loci used for genotyping produced distinct patterns
192 11 5.80 3 among 36 and 33 inbred lines of both cage and open pollinated
79 6.27 1 groups, respectively. The only exceptions were two lines (line
311 9.46 2 *rec
28 and line 281) which exhibited the same allelic structure
210 11 5.79 3
311 9.45 1 *rec (Supplementary Tables S5 and S6).
236 261 7.19 1 Variability parameters were calculated for eight loci on the
311 9.45 1 whole set of data; microsatellites were successfully amplified
249 79 6.27 1 in all 282 genotypes, and a total of 29 different alleles were
121 7.90 2
341 5.85 2 *rec
detected. The average number of alleles per locus was 3.625,
261 79 6.27 1 and, the number of amplified alleles at each locus varied from 2
341 5.85 19 *rec (locus 6) to 5 (loci 2, 4, 5). The highest observed heterozygosity
272 1 7.58 1 (0.631) was found at locus 5 and the lowest at locus 6 (0.089)
121 7.90 1
(Supplementary Table S7). In spite of low heterozygosity, locus
236 8.66 1
274 121 7.90 1
6 has additional value in distinguishing all of the parent lines
275 121 7.90 1 in combination with the other seven loci.
276 249 6.39 1 Parentage analysis based on 8 microsatellite markers resulted in
281 1 7.58 1 a panel of parent/offspring trios LOD (Table 4, Table 5). The LOD
11 5.80 5
score value was used for determination of the threshold that validates
79 6.27 1
275 9.51 1 the hypothesis and for discarding trios with high allelic discrepancies.
311 9.46 2 *rec In the case of cage pollination, the strict LOD threshold (95%
311 11 5.80 2 confidence) was at LOD >5.35 and the correct paternity assignment
28/281 5.17 2 was achieved in all of them. In the case of open pollination, the LOD
104 7.89 1
121 7.90 4
threshold was at LOD >5.28 and the correct paternity assignment
192 8.68 1 was also achieved in all of them.
210 5.58 1 Among the 213 elite F1 plants tested, 99 were found to be
249 6.39 2 genetically diverse, while others were identical, including
346 8.50 1 those formed by reverse combination. Out of these, 5 hybrids
341 43 6.79 3
52 6.63 1
were selected as super elite. These super elite F1 hybrids are
now entering multiplication and additional field testing to be
(Continued)
registered as new varieties.

Frontiers in Plant Science  |  www.frontiersin.org 7 September 2019 | Volume 10 | Article 1111

Rudolf-Pilih et al. Innovative F1 Hybrid Breeding Method

TABLE 5 | List of maternal-parents and the determined paternal-parents within Using molecular markers, individual plants with distinguished
F1 progeny, logarithm of odd (LOD) score and number of selected plants with the
genotypes can be clearly identified. In our case, eight SSR markers
same parent line for open pollination experiment.
were sufficient to discriminate between all but one included
DH plant line Determined LOD No. of selection with cabbage inbred lines. Since maternal-parent origin of lines was
line No.(maternal paternal parent the same parental
parent) lines
recorded, only paternal-parent lines of elite hybrids needed to
be identified.
7 164 6.67 1 Paternity analysis is used extensively in molecular
260 6.92 1
evolution, molecular ecology and forensic science. For
349 11.30 1
15 49 7.71 1
this purpose, software applications were developed such as
176 5.88 2 CERVUS. The aim of this test is to identify paternal identity.
247 6.29 1 This approach was used to identify paternal origin of Chinese
75 171 6.82 2 Holstein cattle (Tian et al., 2008). Nine selected polymorphic
306 8.01 1
SSR markers efficiently discriminated the parental origin of
349 11.30 1
85 9 7.95 2 330 genotypes.
49 7.71 5 The idea of using paternity testing in plant breeding
103 176 5.88 1 schemes was first presented (Lambeth et al., 2001) in loblolly
260 6.92 1 pine (Pinus taeda L.). This method was later applied in
123 49 7.71 2
176 5.88 1
several other tree species, for instance in Eucalyptus sp.
273 9.76 2 (Cupertino et al., 2009) and in olives (Baruca-Arbeiter et al.,
349 11.30 4 2014). Paternity testing was also used to identify pollen
164 9 7.95 1 contamination rate, for instance in loblolly pine (Vidal et al.,
168 26 9.47 1 2015). Under the term “breeding without breeding” paternity
247 6.29 1 *recip
292 6.70 1
determination was proposed for forest tree breeding protocols
176 35 7.66 1 (Wang et  al., 2010). It is more difficult to identify paternal
247 6.29 1 *recip origin in tetraploid allogamous plants; nevertheless, such
247 9 7.95 1 an approach was achieved in some forage plant species,
168 8.15 2
particularly with the help of software applications developed
176 5.88 2
253 8.31 1
specifically for these needs. Such applications were developed
273 9.76 1 (Spielmann et al., 2015) and used in autotetraploid species
253 260 6.92 1 such as alfalfa. Paternity testing was also proposed in alfalfa
304 8.91 1 breeding programs (Riday, 2011; Riday et al., 2013). It
306 8.01 3 *recip
should be noted that in alfalfa polycross breeding, producing
344 6.90 1
260 273 9.75 1 *recip synthetic cultivars but no F1 hybrids is practiced. For these
292 6.70 3 *recip reasons, alfalfa and other forage crop breeders do not select
273 253 8.31 3 individual homozygous lines or genotypes in search for F1
260 6.92 1 hybrid performance, as discussed in this manuscript.
292 6.70 1
292 49 7.71 1
Despite the evident success of F1 hybrid varieties in crop
164 6.67 1 plants, vegetables and ornamentals, the major features of
260 6.91 6 breeding methods have not been changed for several decades.
306 8.01 2 So far, alternative attempts like the already described “reverse
349 11.30 1 *recip breeding” (Dirks et al., 2009) have not gain major attention.
304 13 7.14 1
247 6.29 2
In these standard methods, testing for combining ability
264 7.68 3 is often the limiting factor. As stressed by Acquaah (2012),
306 164 6.67 2 even testing for specific combining ability among 50 lines
253 8.31 2 requires 2,450 crosses, which is too laborious and technically
292 6.70 1
demanding to be performed. Experimental data support this
344 306 8.01 3
347 9 7.95 2 assumption. For instance in a large trial with Artemisia annua
49 7.71 1 (Townsend et al., 2013) attempting to perform diallele cross
349 49 7.71 2 among 30 lines, the authors succeeded in obtaining only 366
*reciprocal. that yielded enough seed to be grown up for screening (156
of these were reciprocals) from 870 possible reciprocal cross
combinations. Several other authors report on even much
DISCUSSION lower numbers of SCA testing.
In our “inter-pollination/paternity testing” method, inter-
The main characteristic of the method presented is built pollination among all inbred lines is an option, since the
around paternity determination of the best selected individual origin of offspring is later determined by genetic profiling.
F1 plants originating from the inter-pollination of inbred lines. The method is much less laborious since genetic profiling is

Frontiers in Plant Science  |  www.frontiersin.org 8 September 2019 | Volume 10 | Article 1111

Rudolf-Pilih et al. Innovative F1 Hybrid Breeding Method

TABLE 6 | Selected potential super elite F1 plants based on head morphological characteristics, compared to the released cultivar ‘Presnik F1’.

Mother plant Male parent Average head weight Average length of Average head shape* Average Figure
(kg) inner core (cm) head
firmness**

261 341 3.78 9.3 0.7 5

192 11 3.97 8.0 0.6 5

341 52 3.13 6.5 1.0 5

253 304 3.35 8.0 0.7 5

85 49 3.67 7.5 0.7 5

Hybrid Presnik 2.60 7.5 0.9 5

F1

*quotient between height and width number lower than 1 - flattened, over 1 - round.
**firmness: 1 (low) 5 (highest).

done only on inbred lines entering inter-pollination and only In cabbage, accidental self-pollination is not frequent due to
on previously phenotypically selected elite F1 hybrids. This strong sporophytic self-incompatibility. For instance, in our test,
was also shown in our experiment with cabbage, where 69 we found only three inbreds within 104 detected experimental
parental lines and only 213 F1 hybrids from 9,858 evaluated hybrids (data not shown). It is a key characteristic of inbred lines
needed to be genotyped in total. to be less vigorous than hybrids. For this reason it was also shown

Frontiers in Plant Science  |  www.frontiersin.org 9 September 2019 | Volume 10 | Article 1111

Rudolf-Pilih et al. Innovative F1 Hybrid Breeding Method

in our experiments where no accidentally self-pollinated DH line expected, the probability p = 1.000 is achieved with 240 or 180
was selected among elite F1 hybrids. F1 plants, respectively.
In the case that the proposed breeding strategy will be It should be noted that probabilities are given per maternal-
implemented for self-pollinating species (hybrids already parent line while reciprocal effects are not considered. If we
dominate market in rice and became important in wheat and neglect »maternal effects« caused by differences in cytoplasmic
barley), the optimal pollination method used will differ from inheritance, the probability that the same combination of
species to species. Various solutions can be proposed. To DH lines occurs within other rows because of reciprocity
perform interpollination of self-pollinating species we propose is significant. If this is taken into account for practical
pollen collection from each inbred line followed by and manual considerations, the number of F1 plants examined might be
pollination of emasculinated flowers with pollen mixture. even lower.
It might be discussed whether selection based on individual It is a characteristic of white cabbage that all experimental
F1 plant performance is adequate. Major traits with high conditions needed for evaluation of the presented breeding
heritabilities, such as those used in our trial with cabbage, are method are well established, namely the procedure of DH
optimal for selection. Although the overall performance of induction yields large numbers of haploid embryos that frequently
selected individual F1 hybrid combinations will be adequately spontaneously double (Hong-Hao et al., 2014), DH lines are
tested in the next growing seasons, two indications show routinely selfed or maintained by micropropagation, the SSR
that phenotypic selection among individuals is likely to be marker analysis is well established (Wang et al., 2012) resulting in
efficient. Namely, paternity analysis showed, that in both a high level of polymorphism, and phenotypic selection of most
pollination methods several F1 hybrids with the same parental valuable characteristics can be performed. In our experiments,
inbred lines were selected (Tables 4 and 5). Also, in 48 cases, we found no major obstacle and all lines constituting selected F1
the same F1 combination was selected, but in a reciprocal way, hybrids were maintained.
meaning that the selection of these F1 most likely based on A more complex polycross pollination method, which
phenotype was adequate. These observations would be an usually contains 10 or more repetitions of each genotype,
indication of good general combining ability, which can be would certainly increase the representation of genotypes, but
tested by intercrossing among DHs in subsequent studies. would require an additional year to micropropagate each DH
Advances in phenotypic analysis achieved during recent years line. The benefits of both options need to be further evaluated.
using image analysis methods usually called plant phenomics
(Afonnikov et al., 2016) might even improve the selection process.
Using specific software and computer image analysis, individual CONCLUSION
plants are tested (among others) concerning development, water use,
architecture, shapes and reflectance at a wide range of wavelengths, Here we propose a novel F1 hybrid breeding method which is
from visible light to heat imaging. This testing can be performed in composed of already known elements such as DH extraction,
both controlled and field conditions (White et al., 2012). genotyping, intercrossing, line maintenance and paternity
Using a formula that was developed to calculate how many testing. New is the combination of these elements that provide
F1 offspring have to be screened after such random crosses, it higher efficiency and is less labour intensive.
is now possible to predict the number of F1 offspring needed Selection based on individual plants can result in the
per maternal-parent row. It would be optimal if all of the lines overestimation of some selected F1 plants because of their
were crossed among each other and their progeny evaluated. favourable position. Although the selection of elite plants
Even if we assume that pollination was equally successful and by chance cannot be excluded and could be tested only by
that all plants were inter-pollinated, the probability that each repeated pollination and testing of selected parental lines,
progeny is presented at least once in a maternal-row increases two indications suggest otherwise. The fact that we collected
with the number of lines being inter-pollinated. For instance, several elite hybrid plants with identical parents and the fact
according to Table 2, when 30 DH lines are inter-pollinated, that in addition to this several of them were found across the
the probability p = 0.996 that all possible combinations will selection field as being reciprocals indicate that selection based
occur among F1 progeny is achieved with 270 F1 plants per on individual plants by phenotype is most likely efficient.
maternal-parent row. However, if the breeder accepts that Besides reducing the time and labour needed for crossing,
only about 90 or 80% of possible combinations are expected the proposed breeding scheme favours the selection of parent
in F1 progeny, the probability p = 0.999 or p = 1.000 that lines which flower simultaneously, are compatible and produce
this criterion is satisfied is reached already with 120 or 90 higher amounts of seeds, all of which is very important in
F1 plants, respectively. Similar conclusions can be drawn in breeding programs.
the case when 60 DH lines are intercrossed (Table 3). The We believe that the proposed scheme could be easily adopted
probability p = 0.992 that all possible combinations will occur for other vegetable or crop species. Of course all aspects of the
among F1 progeny is reached with 540 F1 plants per maternal- proposed technique need to be widely tested and verified both
parent row. If only 90% or 80% of possible combinations are for practical and economic criteria.

Frontiers in Plant Science  |  www.frontiersin.org 10 September 2019 | Volume 10 | Article 1111

Rudolf-Pilih et al. Innovative F1 Hybrid Breeding Method

DATA AVAILABILITY and the funding of the project by the Ministry of Agriculture,
Forestry and Food (No. 08-6-72/2017).
All datasets generated for this study are included in the
manuscript/Supplementary Files.
ACKNOWLEDGMENTS
AUTHOR CONTRIBUTIONS The authors thank Adriana Podržaj MSc, Viktorija Dolenc
and Kristina Marton PhD for managing field experiments and
JJ, KR-P and BB designed the research; MP, KR-P, BB, NŠ and data extractions.
JM performed the research and analysed the data; BB wrote the
manuscript with input from MP, KR-P, NŠ and JM.
SUPPLEMENTARY MATERIAL
FUNDING
The Supplementary Material for this article can be found online
The authors acknowledge the financial support from the at: https://www.frontiersin.org/articles/10.3389/fpls.2019.01111/
Slovenian Research Agency (research core funding No. P4-0077) full#supplementary-material

REFERENCES Murovec, J., and Bohanec, B. (2012). “Haploids and Doubled Haploids in
Plant Breeding,” in Plant breeding. Ed. I. Y. Abdurakhmonov (London,
Acquaah, G. (2012). Principles of Plant Genetics and Breeding. Malden: Wiley- UK :INTECH), 87–106. doi: 10.5772/29982
Blackwell. doi: 10.1002/9781118313718 Nishio, T., and Kitashiba, H. (2017). “F-1 hybrid breeding using genome
Afonnikov, D. A., Genaev, M., Doroshkov, A. V., Komyshev, E. G., and information,” in The Radish Genome. Eds. N. Takeshi and K. Hiroyasu. (Cham:
Pshenichnikova, T. A. (2016). Methods of high-throughput plant phenotyping Springer) doi: 10.1007/978-3-319-59253-4
for large-scale breeding and genetic experiments. Russ. J. Genet. 52, 688–701. Riday, H. (2011). Paternity testing: a non-linkage based marker-assisted selection
doi: 10.1134/S1022795416070024 scheme for outbred forage species. Crop Sci. 51, 631–641. doi: 10.2135/
Baruca-Arbeiter, A., Jakše, J., and Bandelj, D. (2014). Paternity analysis of the olive cropsci2010.07.0390
variety “Istrska belica” and identification of pollen donors by microsatellite Riday, H., Johnson, D. W., Heyduk, K., Raasch, J., Darling, M. E., and Sandman,
markers. Sci.World J., 2014, 1–6. doi: 10.1155/2014/208590 J. M. (2013). Paternity testing in an autotetraploid alfalfa breeding polycross.
Berlan, J. P. (2018). Hybrid corn and the unsettled question of heterosis. J.Genet. Euphytica 194, 335–349. doi: 10.1007/s10681-013-0938-7
97, 1075–1082. doi: 10.1007/s12041-018-1037-2 Rudolf-Pilih, K., Bohanec, B., and Šuštar-Vozlič, J. (2012). Cabbage genebank and
Chen, Z. J. (2013). Genomic and epigenetic insights into the molecular bases of breeding of hybrid varieties (Brassica oleracea L. var. capitata L. f. alba). Acta
heterosis. Nat. Rev. Genet. 14, 471–482. doi: 10.1038/nrg3503 Agricult. Slovenica 99, 387–391.
Cupertino, F. B., Leal, J. B., Vidal, P. O., and Gaiotto, F. A. (2009). Parentage testing Rudolf-Pilih, K., Potokar, U. K., and Bohanec, B. (2018). Improvements of doubled
of hybrid full-sib families of Eucalyptus with microsatellites. Scand. J. Forest haploid production protocol for white cabbage (Brassica oleracea var. capitata
Res. 24, 2–7. doi: 10.1080/02827580802613073 L.). Folia Hortic. 30, 57–66. doi: 10.2478/fhort-2018-0006
Dimitirijevic, A., and Horn, R. (2018). Sunflower hybrid breeding: from markers Schuelke, M. (2000). An economic method for the fluorescent labelling of PCR
to genomic selection. Front. Plant Sci. 8, 2238. doi: 10.3389/fpls.2017.02238 fragments. Nat. Biotech. 18, 233–234. doi: 10.1038/72708
Dirks, R., van Dun, K., de Snoo, C.B., van den Berg, M., Lelivelt, C. L. C., Voermans, Simmonds, N. W., and Smartt, J. (1999). Principles of crop improvement. Oxford:
W., et al. (2009). Reverse breeding: a novel breeding approach based on engineered Blackwell Science Ltd.
meiosis. Plant Biotechnol. J. 7, 837–845. doi: 10.1111/j.1467-7652.2009.00450.x Singh, S., Bhatia, R., Kumar, R., Sharma, K., and Dash, S., and Dey, S. S. (2018)
Doyle, J. J., and Doyle, J. L. (1987). A rapid DNA isolation procedure for small Cytoplasmic male sterile and doubled haploid lines with desirable combining
quantities of fresh leaf tissue. Phytochem. Bull. 19, 11–15. ability enhances the concentration of important antioxidant attributes in
Fehr, W. R. (1993). Principles of Cultivar Development Vol. 1. Ames: Macmillan Brassica oleracea. Euphytica 214 (11): 207. doi: 10.1007/s10681-018-2291-3
Publishing Company. Spielmann, A., Harris, S. A., Boshier, D. H., and Vinson, C. C. (2015). ORCHARD:
Govindaraju, D. R. (2019). An elucidation of over a century old enigma in genetics- Paternity program for autotetraploid species. Mol. Ecol. Resour. 15, 915–920.
Heterosis. PLoS Biol. 17 (4), e3000215. doi: 10.1371/journal.pbio.3000215 doi: 10.1111/1755-0998.12370
Hong-Hao, L. V., Qing-Biao, W., Li-Mei, Y., Zhi-Yuan, F., Yu-Mei, L., Mu, Z. et al. Tian, F., Sun, D., and Zhang, D. (2008). Establishment of paternity testing system
(2014). Breeding of cabbage (Brassica oleracea L. var. capitata) with fusarium using microsatellite markers in Chinese Holstein. J. Genet. Genomics 35, 279–
wilt resistance based on microspore culture and marker-assisted selection. 284. doi: 10.1016/S1673-8527(08)60040-5
Euphytica 200, 465–473. doi: 10.1007/s10681-014-1197-y Townsend, T., Segura, V., Chigeza, G., Penfield., T., Rae, A., Harvey, D., et al.
Kalinowski, S. T., Taper, M. L., and Marshall, T. C. (2007). Revising (2013). Use of combining ability analysis to identify elite parents for Artemisia
how the computer program CERVUS accommodates genotyping error annua F1 hybrid production. PLoS One 8 (4), e61989. doi: 10.1371/journal.
increases success in paternity assignment. Mol. Ecol. 16, 1099–1106. doi: pone.0061989
10.1111/j.1365-294X.2007.03089.x Vidal, M., Plomion, C., Harvengt, L., Raffin, A., Boury, C., and Bouffier, L.
Kholilatul, N. I., Reflinur, R., and Yang-T., J. (2014). Development of EST-SSR (2015). Paternity recovery in two maritime pine polycross mating designs
markers to assess genetic diversity of broccoli and its related species. Indones. and consequences for breeding. Tree Genet. Genomes 11, 105. doi: 10.1007/
J.Agric. Sci. 17, 17–26. doi: 10.21082/ijas.v17n1.2016.p17-26 s11295-015-0932-4
Lambeth, C., Lee, B. C., O’Malley, D., and Wheeler, N. (2001). Polymix breeding Wang, W. X., Huang, S. M., Liu, Y. M., Fang, Z. Y., Yang, L. M., Wie, H., et al. (2012).
with parental analysis of progeny: An alternative to full-sib breeding and Construction and analysis of a high-density genetic linkage map in cabbage (Brassica
testing. Theor. Appl. Genet. 103, 930–943. doi: 10.1007/s001220100627 oleracea L. var. capitata). BMC Genomics 13. doi: 10.1186/1471-2164-13-523
Miyaji, N., and Fujimoto, R. (2018). “Hybrid vigor: importance of epigenetic Wang, X. R., Torimaru, T., Lindgren, D., and Fries, A. (2010). Marker-
processes and consequences for breeding,” in Plant epigenetics coming of age based parentage analysis facilitates low input ‘breeding without breeding’
for breeding applications, vol. 88 . Eds. M. Mirouze, E. Bucher, and P. Gallusci strategies for forest trees. Tree Genet. Genomes 6, 227–235. doi: 10.1007/
(London: Academic Press) 247–275. doi: 10.1016/bs.abr.2018.10.001 s11295-009-0243-8

Frontiers in Plant Science  |  www.frontiersin.org 11 September 2019 | Volume 10 | Article 1111

Rudolf-Pilih et al. Innovative F1 Hybrid Breeding Method

Watson, A., Ghosh, S., Williams, M. J., Cuddy, W. S., Simmonds, J., Rey, M. D., Conflict of Interest Statement: The authors declare that the research was
et al. (2018). Speed breeding is a powerful tool to accelerate crop research and conducted in the absence of any commercial or financial relationships that could
breeding. Nat. Plants 4, 23–29. doi: 10.1038/s41477-017-0083-8 be construed as a potential conflict of interest.
White, J. W., Andrade-Sanchez, P., Gore, M. A., Bronson, K. F., Coffelt, T. A.,
Conley, M. M., et al. (2012). Field-based phenomics for plant genetics research. Copyright © 2019 Rudolf-Pilih, Petkovšek, Jakše, Štajner, Murovec and Bohanec. This is
Field Crops Res. 133, 101–111. doi: 10.1016/j.fcr.2012.04.003 an open-access article distributed under the terms of the Creative Commons Attribution
Wijnker, E., van Dun, K., de Snoo, C. B., Lelivelt, C. L. C., Keurentjes, J. J. B., License (CC BY). The use, distribution or reproduction in other forums is permitted,
Naharudin, N. S., et al. (2012). Reverse breeding in Arabidopsis thaliana provided the original author(s) and the copyright owner(s) are credited and that the
generates homozygous parental lines from a heterozygous plant. Nat. Genet. original publication in this journal is cited, in accordance with accepted academic practice.
44, 467–470. doi: 10.1038/ng.2203 No use, distribution or reproduction is permitted which does not comply with these terms.

Frontiers in Plant Science  |  www.frontiersin.org 12 September 2019 | Volume 10 | Article 1111