The development of breeding materials is shown in Figure 1 . Two white curd varieties, namely, Pusa Ashwini (PA) and Pusa Kartiki (PK) (as female parent) of the early maturity group of Indian tropical cauliflower, were crossed with purple cauliflower line KTPCF-1 (PPCF-1) (as male parent) of the snowball group. The snowball group forms curd at 10°C–16°C, while the early group forms at 20°C–27°C. Unlike tropical varieties ‘PA’ and ‘PK’, ‘PPCF-1’ does not flower satisfactorily in plains; hence, F₁s were generated at IARI Regional Station, Katrain, Kullu, HP, in 2019–2020. The synchronisation in flowering between ‘PPCF-1’ and recipient varieties ‘PA’ and ‘PK’ was obtained by adjusting the sowing time.

The F₁ seeds (‘PA’ × ‘PPCF-1’; ‘PK’ × ‘PPCF-1’) were transported to IARI, New Delhi, and sown in the nursery in August. Two F₁ plants from each set were selfed by bud pollination to generate F₂s, and also F₁ was backcrossed (as female parent) with ‘PA’ and ‘PK’ (as male parent) to obtain BC₁F₁. The F₂ and selected purple BC₁F₁ plants were advanced to F_2:3 and BC₁F₂, respectively.

The F₂ and BC₁F₂ plants were first selected for i) purple curd colour, followed by morphological traits, i.e., ii) semi-erect leaf, iii) October–November curd maturity, and iv) tropical flowering habit (flowering in December–January) (Singh and Kalia, 2020). White curding plants were also advanced for evaluation purposes. A total of 40 F₂ plants were selected in both sets (i.e., PA/PPCF-1 and PK/PPCF-1). Similarly, 30 plants in BC₁F₂ were also selected for marker analysis.

The procedure for marker-assisted selection is presented in Figure 2 . The F₂ plants were first observed for purple curd colour, followed by selection on the basis of criteria associated with the plant type of tropical cauliflower, i.e., semi-erect leaf, October–November curd maturity, and tropical flowering habit. A total of 48 F₂ plants were selected, including 40 purple curding and eight white curding plants for MAS. The F_2:3 progenies from these selected homozygous (PrPr) F₂ plants were advanced to F_3:4. Similarly, the 18 morphologically selected BC₁F₂ plants from PA and 12 plants from PK were also taken for foreground selection with Pr gene-linked markers. The identified homozygous (PrPr) BC₁F₂ plants were advanced to BC₁F_2:3. The progenies from these 70 plants were tested in F_2:3 and BC₁F_2:3 generations for seedling apical colour and curd colour to confirm their zygosity level.

The genomic DNA from the leaf samples was isolated using the cetyltrimethylammonium bromide (CTAB) method (Murray and Thompson, 1980). The extracted DNA was treated with RNase as described by Rakshita et al. (2021). The DNA quality and quantity were checked using a 0.8% agarose gel and Nanodrop (Shimadzu, Kyoto, Japan). Three polymerase chain reaction (PCR)-based Pr gene-linked primers, namely, BoMYB2m, BoMYB3m, and BoMYB4m (from Chiu et al., 2010), were synthesised from Integrated DNA Technologies (IDT) through JP Scientific, India ( Supplementary Table S1 ). These primers were first tested in three parental lines and seven random purple curd F₂ plants. Furthermore, the primers were screened using the genomic DNA of all 48 F₂ plants by following the PCR programme as described by Singh et al. (2022). The amplicons were resolved through gel electrophoresis using a 3% agarose gel and visualised in a gel documentation system (Bio-Rad, Hercules, California, USA).

The populations (parental lines, F₁, F₂, and BC₁F₁) were evaluated in the Research Farm, IARI, New Delhi, during 2021–2022. The crop was raised following standard crop-growing practices (Singh and Sharma, 2001). The location of the research farm is 28.63 N, 77.15 E, and is 228.61 m above mean sea level. The monthly mean temperature during the crop period ranged from 12.2°C to 25.0°C. Similarly, all the F_2:3 progenies and BC₁F₂ were raised for evaluation in the subsequent year (i.e., 2022–2023) by following the standard crop-growing practices (Singh and Sharma, 2001).

A total of 13 MAS-derived homozygous F_2:3(purple) progenies, including six from PA × PPCF-1 and seven from PK × PPCF-1, were raised in paired rows (15 plants/row) in a randomised block design with three replications. The observations on seven morphological traits, i.e., leaf length, leaf width, plant height, plant spread, gross plant weight, marketable curd weight, and marketable curd yield, were recorded from 10 random plants. Furthermore, days taken for curd initiation, marketable curd maturity, bolting, flowering, and seed harvesting stages were recorded from the remaining 10 plants to assess their closeness for developmental transitions to the tropical varieties PA and PK. Parental lines (PA, PK, and PPCF-1) and ‘Graffiti’ (a purple cauliflower variety, Syngenta) were used as the references. The mean value was used for comparison.

Curd samples measuring approximately 50 g were harvested per selected plant and homogenised, from which 500 mg per plant was analysed for total anthocyanin content using the differential pH method as described by Scalzo et al. (2008).

The χ² method was used to test the goodness of fit of the observed ratios for seedling apical colour and curd colour to the theoretically expected ratio in the segregating populations (F₂, BC₁, and F_2:3) (Panse and Sukhatme, 1967). The data from morphological and developmental transitions were analysed for the level of significance using the ‘agricolae’ package of RStudio. The normal distribution curve for anthocyanin content in plant materials was analysed using population size (N = 90) and the Microsoft Excel^® software (Singh et al., 2020).

The F₁ plants produced purple colour seedling apical leaves and marketable curd. The intensity of colour was less than that of the parental line ‘PPCF-1’, and it was absent in both ‘PA’ and ‘PK’ ( Table 1 ).

Genetics of purple apical colour: The F₂ segregation for both crosses was not significantly different from the expected 3:1 ratio ( Table 1 ). The F_2:3 families were split into those that continued to segregate, which similarly followed the expected 3:1 ratio and those in which apical leaves were uniformly purple ( Table 1 ). No segregation was observed in the F_2:3(green) progenies. Both backcross-derived crosses followed the expected 1:1 ratio at BC₁F₁ and the 3:1 ratio at BC₁F₂ ( Table 1 ).

Genetics of purple curd colour: Segregation for curd colour followed similar patterns. Both F₂ populations followed the expected 3 (purple):1 (white) segregation ratios ( Table 1 ). A number of F_2:3 families were fixed for either purple or white curd and showed no segregation, but the progenies within segregating families also followed the 3:1 ratio ( Table 1 ). For backcross-derived material, both BC₁F₁ crosses followed the expected 1:1 ratio, while the BC₁F₂ progenies segregated as 3:1. All seedlings with apical colour developed into plants with purple curd, showing that the Pr gene controls purple pigmentation throughout plant development.

Plants with bicolour curd phenotype (having two colours in a single curd) were observed in both sets of F₂ populations ( Supplementary Figure S1 ). PK × PPCF-1 had a greater number of such plants (11) than PA × PPCF-1/PA (9). PK/PPCF-1 had eight plants with purple + white curd and three plants with intense purple + light purple. There were five plants with purple + white and four plants with intense purple + light purple for PA/PPCF-1. The bicolour curd plants also had purple apical colour pigmentation.

The results of PCR amplification of three Pr gene-linked markers (BoMYB2m, BoMYB3m, and BoMYB4m) in three parental lines, two F₁s, and seven random purple curding F₂ plants are presented in Figure 3 . All three markers were polymorphic for the Pr locus. BoMYB2m and BoMYB4m were codominant, while BoMYB3m was dominant ( Figures 4A–C ). The screening of 40 selected early maturing purple curding F₂ plants with BoMYB2m ( Figure 4A ) and BoMYB4m ( Figure 4C ) revealed 19 as common homozygous and 16 as heterozygous for the Pr gene. BoMYB4m revealed 23 F₂ plants as homozygous. Five plants (Nos. 1, 3, 10, 32, and 34) did not amplify a proper amplicon with BoMYB2m. BoMYB2m amplified two amplicons of 850 bp (in purple) and 400 bp (in white), while BoMYB4m produced 450-bp (in purple) and 600-bp (in white) amplicons. BoMYB2m amplified a prominent band of 850 bp in purple homozygous plants. Heterozygous purple plants amplified a prominent band of 400 bp along with a faint band of 850 bp. BoMYB2m amplified an amplicon of 400 bp in eight white plants, and BoMYB4m amplified an amplicon of 600 bp in five white curding plants. BoMYB3m amplified an amplicon of 650 bp only in all the purple plants ( Figure 4B ). Out of 19 MAS-derived homozygous F₂ plants, only 13 plants could produce sufficient seeds (F_2:3). Furthermore, 30 BC₁F₂ plants having purple curds and matched with the recurrent parents (PA and PK) for the key morphological traits (i.e., semi-erect leaf, October–November curd maturity, and tropical flowering habit) were screened using BoMYB4m markers and identified 11 plants from PA/PPCF-1 cross and 10 plants from PK/PPCF-1 as homozygous for the Pr locus ( Supplementary Figure S2 ). Furthermore, the observations from the F_2:3 and BC₁F_2:3 progenies of these 70 plants for seedling apical colour and curd colour confirmed their zygosity level and validated the marker results ( Supplementary Table S2 ).

The 13 MAS-selected homozygous F_2:3 progenies showed a significant variation in morphological and yield traits ( Supplementary Table S2 ). All the plants from these 13 progenies had purple curds, confirming their homozygous state. Leaf length ranged from 36.2 to 48.0 cm and leaf width from 10.1 to 18.3 cm. Plant height (55.7 cm) and plant spread (53.3 cm) were maximum in PC2304-21. This progeny also had maximum gross plant weight (1,574.7 g) and marketable curd weight (817.7 g). The marketable curd yield level ranged from 15.8 to 25.7 t/ha. Two F_2:3 progenies from each PK/PPCF-1 and PA/PPCF-1 had yield levels higher than those from PK and PA, respectively.

The selected F_2:3(purple) progenies from both sets, PA × PPCF-1 and PK × PPCF-1, showed significant earliness than ‘Graffiti’ and ‘PPCF-1’. The progenies were closer to the recipient varieties PA and PK ( Table 1 ). It was similar for the case of both F₁s. ‘Graffiti’ had the maximum days for each of the observed plant stages in the Delhi condition, followed by the immediate donor parent ‘PPCF-1’. The progenies from PA/PPCF-1—PC6704-16 (109.3 ± 2.5 DAS), PC2304-66 (113.3 ± 2.5 DAS), and PC2303-21 (117.0 ± 2.0 DAS)—were the earliest for curd maturity.

Anthocyanin in F₂ population: In F₂ populations, anthocyanin analysis in 90 F₂ plants (only with purple curd) from PA × PPCF-1 resulted in a range of 6.5 to 94.6 mg/100 g fw, while it varied from 2.34 to 97.5 mg/100 g fw in F₂ plants from PK × PPCF-1 ( Figures 5A, B ). A total of 22 and 33 F₂ plants from both sets had anthocyanin content higher than 43.0 mg/100 g fw. In white varieties, it was negligible, i.e., from 0.2 to 0.83 mg/100 g fw, and in both parental lines PA (0.3 mg/100 g fw) and PK (0.2 mg/100 g fw).

Anthocyanin in F_2:3 progenies: Anthocyanin content in 13 selected F_2:3 progenies is shown in Figure 6 . The most promising progenies for anthocyanin in the curd portion were PC2304-93 (93.78 mg/100 g fw) and PC2304-65 (92.34 mg/100 g fw) from PK/PPCF-1, and PC6704-36 (87.37 mg/100 g fw) and PC6704-16 (70.8 mg/100 g fw) from PA/PPCF-1. The selected progenies had 57.5 to 115.8 times higher anthocyanins in the PK/PPCF-1 set and 37.9 to 72.2 times higher anthocyanins in the PA/PPCF-1 set than white curding recipient varieties PA and PK, respectively. Furthermore, 11 progenies had higher anthocyanin than the donor parent PPCF-1 (53.65 mg/100 g fw).

Apical colour pigmentation in F₁ was light purple, and in the F₂, F_2:3, and BC₁F₁ progenies, it ranged from light purple to dark purple, suggesting the role of allelic nature of the Pr locus (Yuan et al., 2009; Chiu et al., 2010) and/or environmental factors (Volden et al., 2009; Chen et al., 2022). In the present study, an attempt was made to segregate the individuals of segregating populations into three classes—purple, purple-green, and green apical—as suggested for the semi-dominant Pr gene (Chiu et al., 2010); however, the visual observations were inadequate to distinguish the plants. The observed ratio of plants did not match the ratio of 1:2:1. The single factor for purple apical and purple curd colour in Indian cauliflower agreed with the earlier report of Chiu et al. (2010). Notably, all the plants having purple apical also produced purple curd, suggesting that the same Pr gene regulated the purple pigmentation of both plant parts. Furthermore, the negligible difference in populations from both sets suggests the absence of the background influence of the Pr gene. This suggests that the Pr gene-regulated seedling purple apical colour can be used as a morphological marker between purple and white types as a quick progeny selection test for curd colour in cauliflower. However, the intensity of apical colour was inadequate to distinguish between homo- and heterozygous purple plants.

The purple pigmentation of apical leaves was gradually replaced by green colour, suggesting the significance of chlorophyll and anthocyanin in photoprotective and photosynthetic processes (Steyn et al., 2002). This developmental colour change pattern is contrary to the leaf pigmentation in red cabbage, wherein the leaves remain purple till the end (Yuan et al., 2009). However, the curd portion was devoid of chlorophyll; hence, it remained purple until the flowering stage, when anthocyanin degraded naturally. Thus, the Pr gene exhibits tissue- and stage-specific patterns of anthocyanin accumulation vis-à-vis degradation. The observations in cauliflower were contrary to those in tomato (Butelli et al., 2008). They reported the green-to-purple conversion of tomato fruits during the mature green stage.

The study observed a significant difference in morphological and yield traits in 13 homozygous MAS-derived F_2:3 progenies. The genetic recombination followed by manual selection of promising F₂ plants performed prior to MAS could be the reason for attaining desirable yield levels in the progenies, probably due to the additive and dominant gene actions for plant weight and curd weight (Singh et al., 2020). The study highlights the significance of morphological trait-based selection for desirable-type plants in segregating populations in the rapid recovery of promising progenies in gene introgression. The promising F_2:3 progenies for curd yield in both sets highlight their breeding use.

The intermediate position of F₁ between Indian and snowball types indicates the quantitative nature of developmental transitions in cauliflower. The F₁s were closer to the mid-late group of cauliflower, supporting the earlier hypothesis that the mid-late group evolved from intercross progenies between the early group of Indian cauliflower and the snowball type (Rakshita et al., 2021; Singh et al., 2020). Dey et al. (2018) also highlighted frequent introgression of genomic regions to evolve mid groups of Indian types. The observed earliness in curding and flowering in F₁ and F₂ than ‘PPCF-1’ but delayed than that in PA and PK suggests the quantitative nature of earliness with a dominant effect. Hulbert and Orton (1984) reported a similar trend in early flowering broccoli genotypes and Singh et al. (2020) in cauliflower/Sicilian Purple combination. The high yield of selected plants suggests that the criteria used for morphological selection were extremely effective in identifying lines with high yield potential. However, it would be interesting to see how this corresponds to retrospective genome-wide genotyping.

Marker-assisted selection for a semi-dominant gene requires codominant marker(s) to distinguish visibly similar-appearing cauliflower plants. Both homozygous and heterozygous Pr gene-carrying plants have purple colour of seedling apical and marketable curds. The purple colour is governed by the Pr gene (Chiu et al., 2010), but the intensity was affected by temperature and moisture stresses (Piccaglia et al., 2002; Volden et al., 2009; Chen et al., 2022), suggesting the significance of molecular markers in the selection of homozygous plants. The present study confirmed the value of foreground selection because i) the donor parent is an agronomically superior and commercial variety in the snowball group, ii) there is no report showing linkage drag during inter-group gene transfer in cauliflower (Singh et al., 2020; Saha et al., 2021), and iii) it helped in attaining the cost-effective precision in breeding programme.

Three PCR-based molecular markers, namely, BoMYB2m, BoMYB3m, and BoMYB4m, linked to the Pr gene were available for breeding use (Chiu et al., 2010). The present study used these three markers, including BoMYB2m and BoMYB4m as codominant markers and one dominant marker, BoMYB3m. Both BoMYB2m and BoMYB4m could effectively distinguish 19 homozygous plants and 16 heterozygous plants. Five plants did not amplify in the first attempt by BoMYB2m; however, their repeat process for PCR analysis resulted in four as homozygous and one as a heterozygous amplicon, matching the amplification pattern of BoMYB4m. Before performing PCR analysis, the selection of plants on the basis of morphological criteria was found to be useful in selecting 40 plants in F₂ and 30 plants in BC₁F₂. Furthermore, the markers were equally effective in the selection of 13 homozygous PrPr plants in large F₂ populations. Singh et al. (2020) also used BoMYB markers while introgressing the Pr gene in cauliflower from Sicilian Purple. Furthermore, the MAS-selected F₂ plants were used as recipient parent for the rapid development of cytoplasmic male sterile lines (CMS) through marker-assisted backcrossing (MABC) in the PK (of DC-23000) background (data not presented).

Anthocyanin accumulation was negligible in white curd cauliflower varieties ‘PA’ and ‘PK’, suggesting a biologically dormant state of the anthocyanin biosynthesis pathway. We analysed total anthocyanin content using a widely followed pH differential method, which revealed a wide variation in F₂ populations ranging from 6.5 to 96.4 mg/100 g fw in PK × PPCF-1 and 2.34 to 97.5 mg/100 g fw in PA × PPCF-1. The observed variation in F₂ agreed with the earlier findings of Singh et al. (2020) in the F₂ population (0.051 to 48.42 mg/100 g fw) from white cauliflower ‘DC 466’/Sicilian Purple ‘PC-1’. Almost a similar range in purple plants from both populations could be due to the relative closeness of the recipient background since both are from the early maturity group and genetically related in the evolutionary process (Rakshita et al., 2021).

The curds of F₁ plants showed intermediate levels of anthocyanin content, which also agreed with the earlier reported values (38.12 mg/100 g fw) (IARI, 2018; Singh et al., 2020). The state of homo- or heterozygosity of the Pr gene and modifiers in the recipient backgrounds may have affected the level of anthocyanin. Scalzo et al. (2008) reported a variation in anthocyanin content from 1.81 to 7.72 mg/100 g fw violet cauliflower landraces from Italy. Negligible amount of anthocyanin (<1.0 mg/100 g fw) in white curding plants agreed with the earlier findings on Pusa Snowball Kt-25 (0.19 mg/100 g fw) (IARI 2018). In white varieties, the traces of anthocyanin could be due to purple pigmentation on the curd surface due to temperature factors and anthocyanin metabolism-related structural genes (Chen et al., 2022). Higher anthocyanin content in 11 progenies could be attributed to the confirmed homozygous state of the Pr locus. Environmental factors also vary for the two sets because the timing of curd maturity is different in the newly derived progenies (October–November) and KTPCF-1 (end of January to February) (Luhana et al., 2023). Volden et al. (2009) and Chen et al. (2022) also reported the influence of environmental factors on anthocyanin content.

Furthermore, none of the plants in F₂ populations and F_2:3 progenies had anthocyanin levels higher than the earlier report value of 375.0 mg/100 g fw in ‘Graffiti’ (Chiu et al., 2010) and 182 mg/100 g fw in red cabbage (Yuan et al., 2009). This may be due to differences in genetic background and edaphic and climatic factors. However, Volden et al. (2009) reported an anthocyanin content of 73.9 ± 2.8/100 g fw in ‘Graffiti’ from Norway. The differences could be due to differences in growing conditions and tissues taken for analysis (Scalzo et al., 2008). The present study found two progenies, namely, PC2304–93 and PC2304-65, from PK × PPCF-1 and PC6704–36 and DC6704–16 from PA × PPCF-1, which are superior to the varieties reported by Volden et al. (2009). The value of anthocyanin content was higher than that of PPCF-1, which could be due to the presence of heterozygous plants (Kumar, pers. comm.). A variation in anthocyanin content in F₂ plants was attributed to the homozygous and heterozygous states of the Pr locus (Butelli et al., 2008; Singh et al., 2020). Differences in the genetic makeup of the recombinants for endogenous promoters (Grotewold et al., 2000; Gonzalez et al., 2008; Chiu et al., 2010) may have affected the Pr allele. The occurrence of bicolour curds in segregating populations was manifested by reaction–diffusion (Turing, 1952) or the role of R2R3MYB activator gene (Ding et al., 2020; Zheng et al., 2021). This also indicates possibilities of gene conversion or epigenetic silencing in somatic cells during the mitotic division of meristems early in curd conversion in bicolour curd phenotypes. Thus, a study is underway for further investigation since the bicolour curd phenotype is a non-conventional trait in cauliflower, which may appeal to consumers.