Introgression of the Aﬁla Gene into Climbing Garden Pea ( Pisum sativum L.)

: The pea ( Pisum sativum L.) is one of the most important crops in temperate agriculture around the world. In the tropics, highland production is also common with multiple harvests of nearly mature seeds from climbing plant types on trellises. While the leaﬂess variant caused by the aﬁla gene is widely used in developing row-cropped ﬁeld peas in Europe, its use for trellised garden peas has not been reported. In this study we describe a pea breeding program for a high-elevation tropical environment in the Department of Nariño in Colombia, where over 16,000 hectares of the crop are produced. The most widespread climbing varieties in the region are ‘Andina’ and ‘Sindamanoy’, both of which have high-biomass architecture with abundant foliage. They are prone to many diseases, but preferred by farmers given their long production season. This plant type is expensive to trellis, with wooden posts and plastic strings used for vine staking constituting 52% of production costs. The aﬁla trait could reduce these costs by creating interlocking plants as they do in ﬁeld peas. Therefore, our goal for this research was to develop a rapid breeding method to introduce the recessive aﬁla gene, which replaces leaves with tendrils, into the two commercial varieties used as recurrent parents (RPs) with three donor parents (DPs)—‘Dove’, ‘ILS3575 (cid:48) and ‘ILS3568 (cid:48) —and to measure the e ﬀ ect on plant height (PH) and yield potential. Our hypothesis was that the aﬁla gene would not cause linkage drag while obtaining a leaﬂess climbing pea variety. Backcrossing was conducted without selﬁng for two generations and plants were selected to recover recurrent parent characteristics. Chi-square tests showed a ratio of 15 normal leaved to one aﬁla leaved in the BC 2 F 2 plants, and 31:1 in the BC 3 F 2 generation. Selecting in the last of these generations permitted a discovery of tall climbing plants that were similar to those preferred commercially, but with the stable leaﬂess aﬁla . The method saved two seasons compared to the traditional method of progeny testing before each backcross cycle; the peas reached the BC 2 F 2 generation in ﬁve seasons and the BC 3 F 2 in seven seasons. This is advantageous with trellised peas that normally require half a year to reach maturity. Leaﬂess garden peas containing the aﬁla gene were of the same height as recurrent parents and, by the third backcross, were equally productive, without the high biomass found in the traditional donor varieties. The value of the aﬁla gene and the direct backcrossing scheme is discussed in terms of garden pea improvement and crop breeding.


Introduction
Pisum sativum L. is a multi-functional species used for human consumption as a pulse for its dried seed (field peas) or as a fresh vegetable grown for immature grain (garden/English peas) or pods (snow/snap peas) [1]. Dry field peas are grown during the summer in Australia, Canada, China, India, to have low lodging, since this is based on tendril characteristics and overall plant height (PH), with the goal of plants that can grow tall to put on biomass but will not fall over. Leaf shapes in peas vary from those with tendrils and leaflets, through intermediate types (semi-afila), to those with only tendrils (afila) instead of leaflets. In general, the more tendrils a plant has the lower the disease pressure or lodging of the plant [11].
To date there have been no afila pea varieties developed in the Andes region of South America. However, their use would be promising for reducing the high biomasses of commercial genotypes grown in an intensive trellising system. While the leafless variant caused by the afila gene is widely used in developing row-cropped field peas in Europe, its use for trellised garden peas has not been reported. In this study we describe a pea breeding program for a high-elevation tropical environment using the afila gene. We also indicate the afila trait's advantages in a trellis system used for the support of peas, as is common in Colombia, Ecuador and Peru. As the climbing plant type is expensive to trellis with wooden posts and plastic strings used for vine staking, the afila trait could reduce these costs substantially by creating interlocking plants as they do in field peas, therefore requiring less string, posts and wire.
The specific objectives of this project were: (1) to introduce the afila gene into two climbing-type garden peas for use in southern Colombia; (2) to use backcrossing and intensive hybridization without selfing for three backcross generations (to avoid linkage drag), but with a minimum of time invested per backcross by avoiding progeny testing; and (3) to evaluate the yield effect of this introgression on the climbing garden peas typical of highland South America. The abbreviated backcross method recovered the commercial traits in lines containing the afila characteristic. Backcrossing was conducted with the trellised commercial varieties 'Andina' and 'Sindamanoy' with climbing pea phenotype as recurrent parents (RPs) and three bush-type pea breeding lines having the afila trait as donor parents (DPs).

Parents and Hybridizations
Crosses were made using recurrent parents Andina and Sindamanoy both with full leaflets and long vines of the climbing type. These were hybridized with three donor parents: the variety 'Dove', and the breeding lines 'ILS3575 and 'ILS3568 , both from the Andean Crops Research project of the University of Nariño. Among the donor parents for the afila gene, Dove is a winter pea with determinate architecture that is tolerant to cold temperatures and produces on average 2600 kg/ha of green pods, which is less than the production of indeterminate vine-climbing types. The breeding lines have not been yield tested, but were afila genotypes derived from crosses made at the University of Nariño. Among the recurrent parents, the variety Andina (1.7 to 1.8 m tall) was released by CORPOICA/University of Nariño in 2002, and has an indeterminate growth habit with white flowers. Its grain is round, smooth and green with a white-colored hilum. This variety flowers at 65 days after planting (DAP), can be harvested green at 28 days after flowering (DAF) and is ready for dry grain harvest at 155 DAP. Its green pod yield has averaged 6607 kg/ha while its dry grain yield was 1848 kg/ha in the same study [12]. The variety Sindamanoy is also indeterminate, but even taller (1.9 to 2.3 m) than Andina, requiring a strong trellis. As a climbing pea it has two or three stems per plant and yields an average of 36 pods per plant (PPP), with each pod being 6-8 cm in length and having six to nine seeds. This variety also has white flowers, a round grain and seed that is green and smooth with a white hilum. The seed weight is 33 g per 100 seeds and yields have averaged 6400 kg/ha of green pods and 1600 kg/ha of dry grain in a previous study [12]. Its adaptation is from 2300 to 2900 masl (meters above sea level) in the region known as the Nariño altiplano.

Planting Site
Pea crosses and plantings were made in the field at the Lopez Farm (latitude 1 • 12 N and longitude 77 • 15 W) owned by the Servicio Nacional de Aprendizaje (SENA) in Pasto, Colombia, which is at 2633 masl and has an average temperature of 13 • C and annual total precipitation of 700 mm. The rainfall was bimodal and temperature did not vary during the year due to the equatorial location. Seed of the three crosses between recurrent parents (Andina and Sindamanoy) and the donor parents (Dove, ILS3568 and ILS3575) at the final BC 2 F 2 stage were planted in 20-m long plots of five rows, with 1.2 m between rows and 0.1 m between plants. For the following backcross generation, the BC 3 F 2 , was planted in the same way but with four rows each. The numbers of plants showing the two phenotypic states were then counted per row. Plants were categorized as either with the afila characteristic of tendrils replacing leaflets (representing a double recessive allele) or as plants with normal leaflets. A representative set of three segregant plants from each generation of backcrossing derived from the Sindamanoy × ILS3568 cross were grown to harvest to evaluate productivity in terms of number of pods per plant (PPP), pod length (PL), number of seeds per pod (SPP) and yield per plant (YLD) in grams (g/plant), as well as plant height (PH) in meters (m).

Statistical Analyses
Chi-square statistics [13] were used to estimate the goodness of fit to the expected backcross ratios for a single recessive gene occurring in homozygosity without selection in previous generations. The two ratios tested were 15:1 in the BC 2 F 2 generation and 31:1 in the BC 3 F 2 generation, on a per row basis. Morphological data were averaged and tested for correlations with the number of backcross generations. An analysis of variance for these averages (three generations, 2 df ) provided a test of significance with thresholds of p < 0.5 (*) and p < 0.01 (**). Mean separation was performed with a Tukey's test for least significant differences (α = 0.05).

Results
Our hypotheses of goodness of fit were met for the BC 2 F 2 and BC 3 F 2 generations using chi-square tests, with the basis for expected and observed segregations, as shown in Table 1. In each case the chi-square values for average segregation ratio for the BC 2 F 2 were lower than the tabular chi-square values at p ≤ 0.05 (Table 2). In the Andina (recurrent parent) × Dove (donor parent) population averages of 174 plants were normal-leaved and 14 plants had the afila trait for the five rows; the observed chi-square value of 0.46 was lower than the expected value of 3.84. Meanwhile, in the Sindamanoy (RP) × ILS3575 (DP), similar results were observed, with 120 normal-leaved plants and six plants had the afila trait, for an observed chi-square value of 0.47. In the third cross of Sindamanoy (RP) × ILS3568 (DP), 109 plants had normal leaves and five showed the afila trait with an observed chi-square of 0.67. In all three cases, the null hypothesis of 15:1 segregation was accepted for the observed values compared to the expected values for the BC 2 F 2 generation.
For the BC 3 F 2 generation, the expected ratios were also confirmed. For example, a segregation of 158 normal plants and four plants with afila per row was observed in the population from the Andina (RP) × Dove (DP) cross, resulting in a chi-square value of 0.23 ( Table 3). The cross of Sindamanoy (RP) × ILS3575 (DP) had a proportion of 174 normal plants and two with the afila trait and a significant chi-square value of 2.3. Finally, the Sindamanoy (RP) × ILS3568 (DP) cross showed the same tendency, with averages of 180 normal plants and three with the afila trait per row and a significant chi-square value of 1.32. All the observed chi-square values were lower than the expected value of 3.84 for a 31:1 ratio of non-afila to afila plants for the BC 3 F 2 generation.   1  90  3  2  115  5  3  102  5  4  119  6  5  120  6  Average  109  5  15 1 Chi-square observed (X 2 c) = 0.67 Chi-square test for p ≥ 0.05 (X 2 t) = 3.84 Table 3. Observed and expected segregation ratios and chi-square tests for numbers of plants with and without the expression of the homozygous recessive afila gene in the BC 3 F 2 populations from the backcrosses made with the abbreviated selection method for three climbing × bush garden pea crosses.

Observed Ratios Expected Ratios
Andina All the backcross populations described above were created with an abbreviated method for backcrossing a recessive gene into a new commercial background, with no plant selfing performed on any of the backcross F 1 hybrid plants until the last cycle of backcrossing. Therefore, an F 2 was only generated for the last backcross hybrid. This procedure included BC 2 F 1 to obtain the BC 2 F 2 or/and BC 3 F 1 to obtain the BC 3 F 2 . The second generations were used to select the homozygous plants with the tendrilled phenotype and consequently the recessive homozygous afila afila gene fixed into a more commercial background, which would allow the development of a new improved variety containing the afila gene with a tendril leaflet characteristic.
The chi-square tests showed the segregation of the afila gene as expected in the BC 2 F 2 and BC 3 F 2 generations, allowing us to discover a proportion of afila afila plants in rows of about 200 plants ( Table 1). The method of direct backcrossing onto each hybrid generation without waiting for a self-pollination generation allowed the fixation of the afila gene in five seasons for the BC 2 F 2 -derived lines and seven for the BC 3 F 2 -derived lines. If a conventional scheme of selfing had been used to identify a few tendrilled, homozygous recessive afila afila BC 1 or BC 2 plants to cross with the number of seasons required would have been increased by at least two seasons, to seven and nine seasons, respectively.
Phenotypically, the backcross plants appeared more and more like the recurrent parent ( Figure 1). With each generation of backcrossing, and as the genome contribution from the donor parent was lessened and the genome contribution from the recurrent parent was increased, we observed taller plants with the afila gene. An interesting aspect of this approach for converting the leafed non-afila peas to the leafless afila gene-carrying peas was that there was no linkage drag affecting climbing ability in the cross analyzed. Indeed, after three backcross generations, representative segregants returned to the average height of the recurrent parent Sindamanoy of near 1.7 m, while in second and first generations of backcrossing they were significantly shorter as was the donor parent of afila (ILS3568) at only 0.7 m ( Table 4). The F1 hybrid was 1.4 m, showing co-dominance with slight skewing towards the taller parent and indicating some role of dominant genes. By the BC 3 generation, climbing ability had returned to the level of the recurrent parent, being similar to the robust climber Sindamanoy.   In terms of the measures of yield potential also shown in the table of means, representative plants of each backcross generation had subsequently higher pods per plant (PPP) and yield (YLD), with an intermediate status of the F 1 and BC 1 generations. Although PPP returned to recurrent parent levels slowly, YLD in terms of fresh weight seed production returned quickly to seed levels in the recurrent parent. The backcross plants were significantly closer to the recurrent parent by the BC 2 generation based on Tukey's mean comparisons. Return to recurrent parent phenotype was almost complete by the BC 3 generation, with similar phenotypes and additive responses to the recurrent parent genome contribution. A low percentage of donor parent genome contribution in the derived lined could explain the transgressive segregants observed.
Analyses of variance (ANOVAs) for the morphological traits in the Sindamanoy × ILS3568 backcross were significantly (p < 0.05) or highly significantly (p < 0.01) different among generations (Table S1) for all traits except pod length, for which they were not significant (p > 0.05). Pearson's correlation between the number of generations of backcrossing and the morphological traits varied from 0.96 for PH, to 0.84 for YLD, 0.72 for SPP, 0.64 for PPP and 0.53 for pod length (PL), with all being highly significant (p < 0.01). Because PL was not significantly different between generations, the averages are not shown in Table 4. However, PL was still useful for correlation tests. The number of seeds per pod (SPP) was an interesting trait because the backcross generations BC 1 to 3 were lower in value than the donor parent, while the recurrent parent and hybrid were intermediate, suggesting that combining the climbing ability with the afila trait may have some effect on the number of seeds that can be created per pod. Fortunately, this did not affect yield potential seriously, as it increased in each generation-from 56 g/plant in the F 1 to nearly 72 g/plant on average in the BC 3 -surpassing even the averages of the recurrent parent (71.5 g), as well as the BC 1 (61.8 g) and BC 2 (70 g) generations. This might be explained by the number of pods per plant increasing rapidly from 24 on average in the F 1 to 29, 31 and 40 by the BC 1 , BC 2 and BC 3 generations, respectively. Genetic correlations between traits showed yield per plant and number of pods per plant (PPP) to be positively correlated with plant height (PH) having r-values of 0.999 and 0.935, respectively. Negative correlations were found between PH and PPP (r = −0.833), as well as between PPP and seeds per pod (r = −0.508), with pod length not significant.

Success of Abbreviated Backcrossing the Afila Gene
In this study, we were successful at backcrossing the afila gene for tendrilled leaves into two background cultivars, creating new populations segregating the trait using either two or three backcrosses. We monitored segregation ratios and found them to be as expected in each generation. Given this, no selfing was required to select for afila genotypes in the background of the Colombian varieties. Return to the recurrent parent phenotype was important in the backcrossing technique because of the specialized nature and trellised use of the tall garden pea landraces grown on expense-incurring posts and wires in Colombia. Backcrossing is the most frequently used method for the incorporation of a single gene into an adapted cultivar [14]. As such, the method is most popular for monogenically controlled traits, but can also be useful for oligogenic traits. Backcrossing is the recognized method for introducing a single gene, especially a recessive one, into an adapted cultivar where improvement of a Mendelian trait is required.
The backcrossing technique has been widely used in crop plants to bring in important architectural and other traits from foreign germplasm, not to mention many disease resistances. For example, the bush characteristic of Cucurbita pepo was introgressed into the 'Menina Brasilera' cultivar of the viny Cucurbita moschata L. pumpkin, to make a shrubby cultivar called 'Piramoita' [15]. Heat tolerance was improved in Malaysian chilli Kulai (Capsicum annuum L.) by backcrossing with AVPP0702 (C. annuum L.), a heat resistance cultivar [16]. Tetraploid oilseed Brassica napus was used to introgress male sterility to diploid Brassica rapa through backcrossing [17]. Restorer-of-fertility (Rf ) allele in sweet pepper (Capsicum annuum L.) was transferred from hot pepper line AVPP9905 to several sweet pepper genotypes for cost-effective production of sweet pepper hybrid seeds [18]. Seed germination and emergence rate, fruit development, marketable yield, and postharvest fruit storage longevity was improved in cucumber (Cucumis sativus L.) through backcrossing with chilling-tolerant US breeding lines [19]. These are just a few of the many examples existing in crop breeding programs where backcrossing has been used successfully for the transfer or monogenic or oligogenic traits.
In the case of garden peas, backcrossing has been used for architectural traits as was done here and also for various disease resistance genes. For example, Provvidenti and Niblett [20] used backcrossing of a single recessive gene to integrate the passionfruit woodiness virus into peas in North America. Backcrosssing into the pea cultivar MK10 was used to introgress powdery mildew resistance controlled by the recessive er1 locus [21]. A quality gene, sin-2, that controls fiber content and the stringless trait in pea pods has also been transferred through backcrossing [22]. Fw1 and Fw2 genes were integrated into garden peas through backcrossing for resistance to races 1 and 2 of wilt fungus Fusarium oxysporum f.sp. pisi, as well as En for resistance to pea enation mosaic virus and recessive er1 for powdery mildew resistance [23]. In most of these cases, conventional backcrossing was necessary where self-pollination occurred after the BC 1 F 1 generation to determine if the gene was present and could be fixed into one of the BC x F 2 generations, with the fixed plant then used for further backcrossing. This ensured that the diminishing number of segregants with the desirable gene were actually present in BC 3 F 2 or the above generations [24]. In the traditional backcross method, plants with the recessive gene in homozygosity are crossed with the recurrent parent, resulting in heterozygous BC x F 1 and a 3:1 ratio of segregation by the BC x F 2 .
As seen from discussion above, conventional backcrossing relies on having a quarter of the plants express the desirable allele in a homozygous state by the F 2 generation, and for half to be in a homozygous state by the BC x F 2 generations, with a proportion of the recurrent parent genome dependent on the number of backcrosses. The main advantage of conventional backcrossing is that a plant breeder can generate a valuable BC segregant from a set of RP (RP × DP) crosses without generating large populations; however, waiting one extra season per cycle is necessary. This makes conventional backcrossing convenient for greenhouse work over multiple seasons in plants that are small in stature and do not take up too much space or require significant time to grow. On the other hand, the disadvantage for the conventional technique is found when the crop species or cultivar type is long-season or large in plant size, and therefore the extra time to reach an F 2 generation is a burden on the plant breeding program given that two cycles are needed-one for the BC x F 1 grow out and one for the BC x F 2 grow out. For this reason, we found a modified backcrossing approach to be useful for tall, late-maturing climbing peas.
The modification used was to abbreviate backcrossing by applying hybridization with the BC 1 F 1 generation when the ratio was 1:1 for the desirable allele, and then create multiple BC 2 F 2 and BC 3 F 2 populations in the following backcross generations. This method is most useful when a plant breeder has access to a large field and the chance to create the backcrosses in the field. For peas, the method was used before on a large backcross population generated for long-season, daylight-sensitive peas when selecting single-gene weevil resistance [25]. This technique was considered to save time and was an inspiration for the abbreviated backcross method carried out in our study.

Abbreviated Method for Backcrossing of the Recessive Afila Gene
As mentioned above, when a trait is of simple inheritance and controlled by a single locus, backcrossing takes different forms when the overall field conditions vary. In addition, the conventional method is usually easier when the desired allele is dominant compared to the desired allele being the recessive one. Introgression of a dominant gene can often be done in the first filial generation of each backcross, while introgression of a recessive gene requires either selfing to a second filial generation or beyond, which is time consuming and constitutes the principal problem with traditional backcrossing. For example, in annual crops, the first backcross F 1 must be selfed to create an F 2 where the plants containing the recessive allele in a homozygous form are backcrossed to the commercial parent.
Conventional backcrossing adds to the number of seasons needed to create a new variety, especially as backcross cycles are added to the procedure. This makes backcrossing of recessive traits more expensive and time consuming than transfer or dominant genes, where backcross F 1 s can be crossed without selfing. Therefore, an abbreviated method is desirable for backcrossing, with alternatives mentioned above but especially in the context of introgression of a recessive trait, as was done in this study. In this case, the alternative to conventional methods is abbreviate backcrossing with the screening of consecutively larger segregating populations to detect the relatively lower number of recessive genotypes found in each successive backcross generation when no prior phenotypic or genotypic selection is practiced. This process can be referred to as the abbreviated backcrossing method and is the one we followed.
The abbreviated backcrossing method shown here is comparable to the transfer of a dominant gene in terms of the first generation of selection. Table 1 shows the inheritance of the transferred gene from the donor parents and the accumulation of background genes from the recurrent parents, as well as the segregation ratios for the dominant and recessive afila gene in the second and third backcrosses, namely, 15:1 for the BC 2 F 2 generation and 31:1 for the BC 3 F 2 generation. The abbreviated backcrossing procedure was successful for all three backcrosses made: Andina (RP) × Dove (DP); Sindamanoy (RP) × ILS3575 (DP); Sindamanoy (RP) × ILS3568 (DP). As stated above, all crosses were made in the F 1 generation without any advance to the F 2 generation to make backcrosses, and multiple plants were used for crosses in the field rather than in the greenhouse.
The abbreviated method for backcrossing represents a significant time savings consisting of one season saved per cycle of backcrossing compared to the conventional method for backcrossing. For example, if four backcrosses are used to introduce a recessive trait into a commercial variety, the expected average recovery of the recurrent parent genome will be 98.875% and the savings in time to get to the BC 4 F 2 will be four seasons compared to the traditional method. The abbreviated method does require a greater number of crosses (at least 25 to 50) to be made, and therefore is more applicable to large-flowered species that are easy to cross, such as garden peas and common beans. Legumes are cleistogamous and this ensures selfing and avoids unwanted outcrossing. In the case of peas, about 40 crosses can be made per day, allowing crossing onto a population with lower allele frequencies.

Utility of the Afila Gene for Introgression in Climbing Peas and Effect on Yield Ability
Among the most important results of our work was showing expanded geographic and agronomic use of the afila gene with no determent to its use in climbing peas. From its origins in temperate countries of continental Europe and widespread use for bush-type field peas in the United Kingdom and North America, we show the promise of afilous peas in tropical South America that can grow at high elevations as long-season, trellised plants. As a well-characterized gene for increased tendril production, afila has been recognized in peas for the past 60 years, but has mainly been used in Europe [26,27]. The afila gene transforms lateral leaflets on a composite pea leaf into tendrils instead of leaflets and is known to be of recessive inheritance [28,29].
The afila gene has also been evaluated physiologically in Australia, USA and India to date [30,31], and is known to interact with other genes for leaf morphology, including unifoliata and cochleata, showing even more extreme multi-tendrilled phenotypes [27,32,33]. One major element of the discovery and use of the afila gene was combining it with shorter statures using the Mendelian le gene for a more efficient architecture overall, much of which occurred behind the iron curtain starting in the 1970s in Eastern Bloc Europe [34]. Other pea breeding with the afila gene was done almost simultaneously in the United Kingdom, and resulted in releases such as the cultivars 'Sum' and 'Filby'. Continental European pea breeding with afila continued in France, Holland and Spain through the early 1980s with the release of commercial high-tendril varieties such as 'Ascona', 'Ballet', 'Cea' and 'Elsa', all with yields similar to varieties with conventional foliage, many leaflets and a lower amount of tendrils [35]. Other varieties with the afila gene released by the private sector have included 'Solara' in 1986, 'Alex' and 'Choque' in 1989, 'Amadeus' in 1992 and 'Astuce' and 'Charleston' in 1993 [35], followed by 'Rampart' and 'Stampede' in 1990 and 'Golijat' in 2001 [7]. The Czech Republic also has had success in breeding afilous peas [36]. Expansion of the leafless pea trait to North America occurred when breeders used afila and short statures from European germplasm to create field peas adapted to the Pacific Northwest. The first recombinants with the I/I genes for garden sweet peas were obtained at Cornell University by a researcher from Poland who had immigrated to the USA. In South America, a high-tendril variety with the afila gene was released in Chile as 'INIA Brisca' and was developed from the cross of an individual selection named PS210791 × F2 (Porta × Neuga). No other South American afila peas have been released to our knowledge.
Modern field pea cultivars are very dependent on the afila gene, as they tend to be bush-type tendrilled architecture with low-leaf biomasses, but that are at the same time self-supporting. This both reduces foliar pathogens and plant lodging compared to older non-afila field pea varieties, as well as increases the air circulation in the canopy, reducing disease pressure on the foliage [35,37]. In addition, afila varieties generally have good overall productivity [36,38] and high water-use efficiency (WUE) compared to non-afila varieties [39]. A higher WUE and consequently better drought tolerance of afila peas results from the lower surface area of these varieties due to leaflets being replaced by tendrils, allowing for lower transpiration water loss than non-afila varieties [40]. While the benefits of drought tolerance and disease escape of afilous peas were not part of our study, we did see that afilous backcross lines can surpass the yield of non-afilous peas, and they do so perhaps via larger seeds, despite having fewer seeds per pod. In our research, afilous and non-afilous breeding lines of climbing peas had similar number of pods per plant and plant heights, so yield differences were probably due to some seed physiology mechanism. Further studies can look at seed filling in the lines fixed for the afila gene. With climbing peas, this can be done at different heights within the canopy to determine the tendrils that photosynthetically contribute to pod and seed growth and development during the peas' life cycles. A further advantage of the lower leaf surface area in climbing peas that we worked with was that the planting density can be increased for afila varieties [41].
The physiological advantages and historical breeding achievements of afila peas were the inspiration for the work we carried out in this study. Our results on plant morphology and yield component trait analyses for characteristics such as pods per plant and harvest per plant suggest the polygenic control of both climbing ability and yield in relationship to height and growing period in the trellised garden pea varieties preferred in Andean production like Sindamanoy and Andina and the advanced lines generated by our breeding program. This contrasts with the monogenic inheritance of the afila trait even in the background of climbing-type peas. An interaction with seeds per pod seems likely independent of pod length. Furthermore, the multiple genes for climbing ability and yield found in peas [42] appear to be independent and unlinked with the afila gene for leaf loss and tendril gain. If genes for climbing ability and yield had been linked to the afila locus, it would have been unlikely that the backcross progeny could recover the full climbing ability and pod yield of the recurrent parent varieties, which are vigorous climbers. Linkage drag was not observed with the afila gene in our analyzed cross, but has been seen in lodging studies with afila by Smitchger and Weeden (2019) [43]. Climbing ability and its effects on disease and lodging tolerance would likely be an additive trait with complex inheritance and quantitative trait loci (QTL). The combination of the afila gene in multiple backgrounds of garden peas seems very feasible, as it was for our two climbing garden pea parents. Furthermore, success can be achieved in three backcrosses.
In summary, we made crosses between afila bush-type peas as donor parents and tall (1.5 to 2.5 m), long-season (five to six month) climbing varieties as recurrent parents, and advanced the introgression of the afila gene into the trellised pea background. We were successful at selecting afila-type segregants after two to three backcrosses without using progeny testing, and thus we were able to accelerate the breeding process. Mid-parent heterosis of about 15% was observed for plant height but not for pods per plant or yield. Heterosis is usually not important in a self-pollinating crop like peas, but the additivity of yield component traits in the crop was promising for fixing background adaptation into advanced lines that are tendrilled rather than having leaflets. The taller the plant, the more yielding capacity it had, with taller plants having increased number of pods in each successive backcross. The easy introgression of afila was very important to our breeding program as well as to because of the high investment needed to train and grow viny pea plants on trellises in research and production systems of the Andes. Farmers will certainly appreciate the new genotype once they realize that no yield sacrifice is found while realizing the potential savings to trellising efforts in terms of heavy extra work, wires and twining. On a further practical level, we obtained breeding lines that were equivalent to the recurrent parent in numbers of pods and seeds, but with the leafless trait and capability of climbing to 2 m or above. Since production potential, number of fresh pod harvests and total biomass of garden peas grown in the highlands is related to climbing ability, this is a very important discovery for breeding and agronomy of peas grown in the Andean region or similar parts of the world. The yield potential of the trellised peas with the afila gene will be studied in further studies now that the trait has been successfully introgressed into commercial-type breeding lines adapted to the highland environment, which was the major success of this research.
Supplementary Materials: The following are available online at http://www.mdpi.com/2073-4395/10/10/1537/s1. Table S1: Mean square values for five traits measured on two parents, F1 generation and three Backcrosses generations during the introgression of the afila gene into climbing garden peas (Pisum sativum).