Genetics of Quality Attributes and Powdery Mildew Severity in Garden Pea (Pisum sativum Var. Hortense L.) under Sub Temperate Conditions of North-Western Himalayas

Garden pea (Pisum sativum var. hortense L., 2n=2x=14), a member of Fabaceae family, is one of the principal legume vegetable crops grown throughout the world. It is native of Mediterranean region with Near East and Ethiopia as secondary centres. It is a rich source of protein ranging from 23-33% (Sharma et al., 2019), slowly digestive starch, sugars and amino acids. Besides, it supplies an extraordinarily diverse health building nutrients such as vitamins, minerals and also lysine, a limiting essential amino acid in cereals (Sharma et al., 2019). It is eaten as fresh vegetable, pulse and processed as pickle, canned, frozen or dehydrated to consume during lean period (Sharma et al., 2019). It helps to reduce the cost of production by fixing atmospheric nitrogen (Anjum et al., 2015) and provide the advantage of low input and sustainable farming.
       
Garden pea production has suffered because of attack of a number of fungal and bacterial diseases but powdery mildew disease caused by Erysiphie pisi DC ex. Saint Amans is one of the most important and widely prevalent disease which affect fresh pea production all over the world. The disease also reduces the quality of marketable produce significantly. Under Indian conditions, sweet, long and dark green pods are preferred by the consumers. Therefore, it is pertinent to develop suitable variety(ies) possessing these quality traits coupled with high yield and resistance to important diseases particularly powdery mildew. Despite continuous breeding efforts, its average yield is low due to farmer’s preference for specific cultivars, genetic drift in the popular recommended cultivars, biotic and abiotic stresses and development of the new pathogen races (Sharma et al., 2013). Development of cultivars with inherent resistance to powdery mildew is one of the most effective and economical means of controlling the powdery mildew disease. Sources of resistance to powdery mildew in pea have been identified in germplasm all over the world and have shown monogenic recessive inheritance. The conventional breeding approaches of hybridization followed by selection involving commercial susceptible variety and resistant donor parent has resulted in the development of powdery mildew resistant varieties with light, yellowish green and medium sized pods e.g. Palam Priya. Hence, these varieties are not preferred and could not replace the existing susceptible variety like ‘Azad P-1’. Since, yield is a complex trait resulting from interaction among a number of inherent characters and the environment, it can be improved through indirect selection on the basis of yield components. Increase in one component might have positive or negative effect on other components and this occurrence is direct consequence of their interdependence during ontogenetic development of plants. Favourable combinations of yield contributing characters only can improve yielding capacity.
       
The choice of appropriate breeding method for improvement of traits depends largely on gene action. Hence, an understanding of the inheritance of quantitative traits is essential to develop an efficient breeding strategy. The effect of individual gene cannot be measured and must be considered along with suitable statistical procedure to obtain genetic information. Generation mean analysis which was proposed by Mather and Jinks (1971) belongs to the quantitative biometric methods based on measurement of phenotypic performances of certain quantitative traits in basic experimental breeding generations (parental, filial, backcross and segregating generations). It is a useful technique for the estimation of main gene effects (additive and dominance) and their digenic [additive × additive (i), additive × dominance (j), dominance × dominance (l)], trigenic [additive × additive × additive (w), additive × additive × dominance (x), additive × dominance × dominance (y) and dominance × dominance × dominance (z)] and other higher order interactions responsible for inheritance of quantitative and qualitative traits. Its popularity in plant breeding and genetics continues unabated (Piepho and Mohring, 2010) and helps to understand the performance of the parents used in crosses and potential of crosses to be used either for heterosis exploitation or pedigree selection (Dvojkovic et al., 2010). Although some studies have been conducted in garden pea in the past by using generation means to estimate the variance components, but these were based on only five or six generations and limited to the perfect fit model (Sharma and Sain, 2004). Cavalli (1952) reported that accuracy of gene effects increase with increasing number of segregating generations and number of plants on which observations are to be taken.

Experimental site
 
The Experimental Farm of CSKHPKV, Palampur, Himachal Pradesh, India is situated at an elevation of 1290.8 meters above mean sea level with 32°6¢ N latitude and 76°3¢ E longitude. The area is characterized by humid and temperate climate (Zone-I), having severe winters and mild summers with high annual rainfall of 2500 mm of which 80 per cent is received during June-September. The soil is classified as Alfisols typic-Hapludalf clay and is acidic in reaction (pH 5-5.6).
 
Experimental materials and breeding activities
 
To ascertain the genetics of various horticultural traits, twelve generations viz., P₁, P₂, F₁, F₂, B₁, B₂, B_1S, B_2S, B₁₁, B₁₂, B₂₁ and B₂₂ of three intervarietal crosses were developed by utilizing the four diverse parents namely, Palam Sumool, Punjab-89, Azad P-1 and Palam Priya. The F₁’s and first backcross generations (B₁ and B₂) had already been developed in winter 2011-12 and 2012-13, respectively and were raised during summer 2013 at Highland Agricultural Research and Extension Centre, Kukumseri to develop second backcross generations (B₁₁, B₁₂, B₂₁ and B₂₂) and their selfed progenies (B_1S and B_2S) under open field conditions. The seeds of these generations were multiplied by raising the respective populations at Experimental Farm of CSKHPKV, Palampur during winter 2013-14 under polyhouse conditions. Simultaneously, F₁’s were backcrossed with their respective parents to increase the quantity of seeds of B₁ and B₂ generations. Quantity of seeds of second backcross generations was also multiplied in each cross combination.
 
Experimental layout
 
During rabi, 2014-15, the experimental material comprises of twelve generations viz., P₁, P₂, F₁, F₂, B₁, B₂, B_1S, B_2S, B₁₁, B₁₂, B₂₁ and B₂₂ was evaluated in Randomized Complete Block Design in three replications at the Experimental Farm, Department of Vegetable Science and Floriculture, CSKHPKV, Palampur. The sowing was undertaken by assigning single row to parents and F₁’s, four rows to each backcross generations and six rows to F₂’s and second cycle of backcross generations. The seeds were sown keeping inter and intra-row spacing of 45 cm and 10 cm, respectively in a row length of 2.5 m. All the intercultural operations were carried out in accordance with the recommended schedule (Anonymous, 2009).
 
Data collection and analysis
 
The non-segregating generations consisted of homologous population while segregating comprises of heterogeneous population. Accordingly, the data were recorded on 10 randomly selected competitive plants of each parents and F₁’s, 20 plants in each backcross generations (B₁ and B₂) and second cycle of backcross generations (B₁₁, B₁₂, B₂₁ and B₂₂) and 30 plants in each F₂’s, B_1S and B_2S. In the process of random selection, the border plants were avoided. The quality parameters recorded were total soluble solids (Brix), ascorbic acid content (mg/100g), protein content (%), total sugars (%) and powdery mildew disease severity (%). Standard statistical procedures were used to obtain mean and variance for each generation separately. While calculating variances, the replicate effect was eliminated from total variances to obtain within replication variance. These variances were used to compute the standard error for each generation mean. The simple scaling tests (A, B, C and D) given by Mather (1949) and Hayman and Mather (1955) were followed for the detection of digenic interactions. The A, B, C and D values were calculated by the following formulae:
 
 
 
The expectations of above scaling tests, when equal to zero indicate the absence of non-allelic interactions. The significant deviation of A and B tests from zero indicate the presence of all three types of epistatic interactions viz., additive × additive (i), additive × dominance (j) and dominance × dominance (l) whereas, C scaling test reveals the presence of dominance × dominance (l) type of interaction and D scaling test indicates the significance of additive × additive (i) type of gene interaction. The significant deviation of any of the scaling tests A, B, C and D from zero, indicates the presence of digenic interactions, otherwise adequacy of additive-dominance model was assumed. 
       
Scaling tests for detecting of trigenic and higher order interactions were carried out as per Vander Veen (1959), by using formulae:            



The significant deviation of any of the scaling tests X and Y from zero, revealed the presence of trigenic or higher order interactions. Estimation of various genic effects and test of fitness of appropriate genetic model was done according to Joint Scaling Test of Cavalli (1952), as described in detail by Mather and Jinks (1982). Joint scaling test in general consists of estimating various genetic parameters from means of available type of generations followed by the comparison of observed generation means with expected values, derived from the estimates of genetic parameters (genic effects) using weighted least square technique, taking weights as the reciprocals of squared standard errors of each mean. The tests of goodness of fit for a particular model were carried out by using weighted chi-square analysis. The estimation of genic effects and chi-square test of goodness of fit were carried out, using 3-, 6- and 10-parameter model. In three-parameter model (additive-dominance model or non-epistatic model), the following genic effects were estimated:
m = Inbred population mean.
(d) = Additive.
(h) = Dominance.
       
In six-parameter model (digenic interaction model), the following genic effects in addition to m, (d) and (h) were estimated:
(i) = Additive × Additive
(j) = Additive × Dominance
(l) = Dominance × Dominance
In ten-parameter model (trigenic interaction model), besides the above mentioned effects, the following genic effects were estimated:
(w) = Additive × Additive × Additive
(x) = Additive × Additive × Dominance
(y) = Additive × Dominance × Dominance
(z) = Dominance × Dominance × Dominance
       
First, simple additive-dominance model consisting of (m), (d) and (h) gene effects was tried and the adequacy of this model was tested by the chi-square test. When this model failed to explain variation among generation means, successively non-allelic digenic interaction parameters i.e. (i), (j) and (l) were included in this model. Inadequacy of digenic interaction model led to the successive use of trigenic interaction model consisting of parameters namely, (w), (x), (y) and (z). Thus, all possible models with different combinations of epistatic parameters were tried to identify the best fit model with minimum or non-significant value of chi-square with maximum number of significant parameters as suggested by Mather and Jinks (1982).

The analysis for variance of the 12 generations with three inter-varietal crosses namely, ‘Palam Sumool × Punjab-89’ (C₁), ‘Palam Sumool × Azad P-1’ (C₂) and ‘Palam Sumool × Palam Priya’ (C₃) revealed significant mean squares due to genotypes for ascorbic acid content, protein content, total sugars and powdery mildew disease severity. Thus, it highlighted the creation of sufficient genetic variability in the existing genetic material involving twelve generations of three different intervarietal crosses (Table 1). The F₁ hybrids showed better performance than their respective parents in desirable direction in C₁ cross for total soluble solids and ascorbic acid content and C₁ and C₃ for protein content and total sugars, respectively. The performance of F₁ hybrid for powdery mildew disease severity found to be better than parent 2 in cross C₁ (Table 2).  
 

 

       
In regard to total soluble solids, simple scaling tests revealed the presence of non-allelic interactions in two cross namely, ‘Palam Sumool × Punjab-89’ and ‘Palam Sumool × Palam Priya’ as revealed from the significance of ‘A’, ‘B’ and ‘D’ scales in the former while, ‘B’ and ‘D’ in latter cross (Table 3). In contrary, non-significance of these parameters along with chi-square in ‘Palam Sumool × Azad P-1’ indicated the adequacy of simple additive-dominance model. Trigenic scaling tests (X and Y) found to be non-significant for all the three crosses suggested the absence of higher order interactions. Further, the non-significance of chi-square values revealed the adequacy of digenic interaction model in cross ‘Palam Sumool × Punjab-89’ and ‘Palam Sumool × Palam Priya’. The presence of non-allelic interactions for total soluble solids has also been reported earlier by Sharma et al., (2012). The significance of additive (d) component with positive sign and dominance (h) with negative sign in ‘Palam Sumool × Azad P-1’ suggesting to follow selection in the early generation to isolate segregants with desirable  total soluble solids. The estimates of genic effects revealed the presence of duplicate type of epistasis in ‘Palam Sumool × Punjab-89’. Sharma et al., (2012) also observed duplicate epistasis for total soluble solids. 
 

       
With respect to ascorbic acid content, the estimates of scaling tests for digenic interaction (Table 3) revealed the significance of ‘C’ and ‘D’ scales indicating ‘l’ and ‘i’ type of non-allelic interactions in cross ‘Palam Sumool × Punjab-89’ while, presence of all types of gene interactions (i, j and l) in the other two crosses as revealed from the significance of ‘A’ scale in ‘Palam Sumool × Azad P-1’ and all four scales (A, B, C and D) in ‘Palam Sumool × Palam Priya’. The significance of ‘Y’ scale revealed the presence of trigenic or higher order interactions in ‘Palam Sumool × Palam Priya’ whereas, absence of trigenic and higher order interactions were noticed in ‘Palam Sumool × Punjab-89’ and ‘Palam Sumool × Azad P-1’ as depicted from the non-significant values of ‘X’ and ‘Y’ scales. The genic effects revealed the presence of duplicate epistasis in ‘Palam Sumool × Punjab-89’ and ‘Palam Sumool × Palam Priya’ as revealed from the opposite signs of ‘h’ and ‘l’ type of gene effects. Further, it was observed in cross ‘Palam Sumool × Palam Priya’ that the positive sign of ‘l’ changed to negative value of ‘z’ implying thereby to defer the selection in the later generations to select superior progeny with desirable ascorbic acid content. On the other hand, significant and positive additive (d) genic effect suggested to go for the selection in the early generations in cross ‘Palam Sumool × Azad P-1’ but negative and significant value of additive × additive (i) type indicated the presence of decreaser alleles. This showed that one has to be very careful while adopting breeding strategy to improve this trait. The significance of chi-square values in all the three crosses indicated the inadequacy of digenic model in ‘Palam Sumool × Punjab-89’ and ‘Palam Sumool × Azad P-1’ and trigenic model in ‘Palam Sumool × Palam Priya’. The inadequacy of digenic interaction model suggested the probable presence of digenic linked interactions while presence of linkage among interacting genes or interaction at four or more loci may be the probable reason of inadequacy of trigenic model in cross ‘Palam Sumool × Palam Priya’.  
       
Significance in either one of the scales or in combinations representing the existence of epistatic interactions between the genes involved which means the inadequacy of additive-dominance model for protein content and total sugars (Table 3). Also, the significance of ‘X’ or ‘Y’ or both in all these crosses for both the traits revealed the inadequacy of digenic interaction model and presence of trigenic or higher order interactions. The genic effects revealed the presence of duplicate type of epistasis evident from the opposite sign of ‘h’ and ‘l’ in all the three crosses for protein content and two crosses namely, ‘Palam Sumool × Punjab-89’ and ‘Palam Sumool × Azad P-1’ for total sugars. Duplicate type of epistasis was also reported by Singh and Singh (1996), Narayan et al., (2001), Singh et al., (2006) and Sharma et al., (2012) for  protein content and Narayan et al., (2001) and Bhardwaj and Vikram (2004) for sugar content by using different cross combination involving different sets of parents using six parameter model. The selection intensity in such crosses should be mild in the earlier generations and intense in later as it marks the progress through selection (Sharma and Sain, 2002). Further, it was noticed that the negative sign of ‘l’ was changed to positive for ‘z’ type interaction in ‘Palam Sumool × Punjab-89’ and ‘Palam Sumool × Azad P-1’ for both protein content and total sugars. However, the positive ‘l’ type interaction was changed to negative ‘z’ type interaction in ‘Palam Sumool × Palam Priya’ for protein content. These types of gene interactions suggested to delay the selections to obtain superior progenies having desirable protein and sugar contents. The significance of chi-square in ‘Palam Sumool × Azad P-1’ and ‘Palam Sumool × Palam Priya’ for protein content and all three crosses for total sugars indicated the non-fitness of trigenic interaction model revealing the likelihood of either linkage or differential variability or higher order interactions. None of the digenic and trigenic effects were significant in cross ‘Palam Sumool × Palam Priya’ for total sugars which confirmed that a more complex genetic system of non-allelic interactions involved in the inheritance of this trait in the specific cross (Sharma and Sain, 2004).
       
The yield and quality traits of pea are greatly affected by the incidence of powdery mildew disease and therefore, there is necessity to develop resistant lines with high yield. Based on three parameter model, epistatic interactions were noticed in all the three crosses indicating the inadequacy of this model which was observed from the significance of ‘A’, ‘B’, ‘C’ or ‘D’ scales. Sharma et al., (2012) and Sanwal et al., (2013)   also reported the presence of non-allelic interactions in the inheritance of powdery mildew disease resistance in pea. Further, trigenic or higher order interactions were present which was revealed from the significance of both ‘X’ and ‘Y’ parameters in all the three cross combinations (Table 3). The presence of duplicate type of non-allelic interactions was observed from the opposite signs of ‘h’ and ‘l’ in ‘Palam Sumool × Punjab-89’ and ‘Palam Sumool × Palam Priya’ but the positive and significant ‘l’ type interaction changed to negative and significant ‘z’ type non-allelic interaction which suggested the dominance effects at heterozygous loci in each parent for powdery mildew resistance. The paucity of more gene effects suggested that breeder would make limited and slow progress in selecting for powdery mildew resistant combinations in these crosses (Smith et al., 2009). Duplicate type of epistasis was also noticed by Sanwal et al., (2013) and Sharma et al., (2013) by following six parameter generation mean analysis. However, the non-fitness of trigenic interaction model in these respective crosses as evident from the significant chi-square values. If trigenic model was taken as the best model to explain the variation among the generation means, it was unable to account for all the variations among the generation means for powdery mildew disease severity. Hence, failure of trigenic model showed the complexity of the resistance and suggested the role of either minor or modifier genes (Joshi and Ugale, 2002). However, for cross ‘Palam Sumool × Azad P-1’, the weighted square analysis was not done due to the zero values of standard error of means for various generations, as 100 per cent disease severity was observed in parent ‘Azad P-1’ over replications as a result genic effects and chi-square could not be estimated. 
       
Results of absolute totals of fixable [(d), (i) and (w)] and non-fixable [(h), (j), (l), (x), (y) and (z)], gene effects revealed that non-fixable gene effects were many times higher than the fixable gene effects in all the three crosses (C₁, C₂ and C₃) confirming that non-additive gene effects had a very important role in the inheritance of characters namely, total soluble solids, ascorbic acid content, protein content and total sugars (Table 4). Powdery mildew disease severity had also showed high value for non-fixable gene effect in crosses C₁ and C₃.
 

 
Selection of progenies for desirable horticultural and biochemical traits and powdery mildew disease resistance
 
To meet the objective of isolation of desirable trangressive segregants with resistance to powdery mildew disease, single plant selections along with bulking of seeds were followed in each of the generations among all the three crosses namely, Palam Sumool × Punjab-89’, ‘Palam Sumool × Azad P-1’ and ‘Palam Sumool × Palam Priya’. Keeping in view the desirable pod characteristics, more numbers of pods per plants and resistance to powdery mildew disease, a total of ‘203’ individual plants over the generations (F₂, B₁, B₂, B_1S, B_2S, B11, B₁₂, B₂₁ and B₂₂) in all the three crosses were selected (Table 5). The selected plant progenies were harvested separately to advance their generations. Simultaneously, bulk seeds following single seed descent and bulk methods were also harvested which may be quite effective in isolating desirable segregants based on the type of gene effects and non-allelic interactions observed in the existing material.
 

The presence of non-allelic interactions for majority of quality traits in all the three cross combinations signify the adoption of generation mean analysis. The type of gene actions observed in the present study suggested to adopt a breeding methodology which can capitalize fixable and non-fixable genic effects. Moreover, presence of duplicate type of epistasis in one or the other cross combinations for different biochemical traits directed to have mild selection intensity in the early generations followed by intense in later. The failure of trigenic model revealed the complexity of the character in question and indicated the presence of linkage among interacting genes or minor/ modifier genes or prevalence of still higher order non-allelic interactions at several loci for the character(s) in question Therefore, under such complex type of gene effects and non-allelic interactions for majority of the traits in all the three crosses, population improvement methods may be useful for breaking undesirable linkages through recombination. The other alternative can be to defer selection in later generations by advancing segregating material through bulk pedigree or single seed descent methods with one or two inter-mating like recurrent selection. Keeping in view the desirable horticultural and biochemical traits and resistance to powdery mildew disease, ‘203’ individual plants over the generations in all the three crosses were isolated. Besides, bulk seeds following single seed descent and bulk methods were harvested to isolate the desirable transgressive segregants.