Introduction

Garden pansies (Viola × wittrockiana Gams.), sometimes referred to as Wittrock’s violets, are popular early spring or fall bedding plants. Their cold tolerance and flowering period early in the season made them a key floricultural crop widely used in both private gardens and public green space (Chandler and Tanaka 2018). Viola × wittrockiana is listed in the Melanium section (Violaceae sect. Melanium Ging.) as a hybrid taxon with compound origin involving multiple species of the Viola genus (Yockteng et al. 2003; Vukics et al. 2008). Due to the hybrid origin of the species, a vast array of flower shapes, colors and sizes are available in the contemporary assortment. The main garden pansy types include rayless violas, with single color flowers devoid of any dark colored streaks, and fancy pansies with characteristic large blotches on the three lower petals (Wittrock 1896). Species in the Melanium section are known to have large variation in base chromosome number (SŁomka et al. 2014), ranging from x = 11 to as low as x = 2 (Erben 1996). Viola × wittrockiana, of which the exact origin is unknown, largely consist of tetraploid varieties with reported chromosome numbers of 48 and 52. High ploidy levels however encumber the breeding for uniform varieties, making it a lengthy and expensive process (Sattler et al. 2016). The seed propagated nature of garden pansies facilitated the introduction of hybrid breeding for this crop (Horn 2002). Many commercial cultivars today are F1 hybrids, which provide good uniformity in growth and flowering. In the past, hybrid breeding goals were mainly focused on plant vigour, early season establishment and abundant continuous flowering (Yoshioka et al. 2006). Today, other morphological characteristics are also becoming more important in pansy breeding: compactness, large circular flowers and straight petals on short peduncles.

Controlling unwanted stretching of seedlings growing in plugs is a primary concern for pansy growers. After flower initiation, pansies react very strongly to temperature which can quickly result in frail and stretched plants with smaller flowers, especially in autumn greenhouse production (Pearson et al. 1995). Several growth regulation techniques have been applied successfully for Viola spp. High DIF (difference of day and night temperatures) is positively correlated with elongation of internodes, peduncles and petioles of Viola × wittrockiana (Niu et al. 2000). Bailey (1998) therefore recommends a constant zero or negative DIF to keep pansies compact. Application of mechanical stress in the form of brushing, is also very effective for Viola tricolor seedlings (Börnke and Rocksch 2018). Although these techniques contribute to environmental friendly and sustainable plant production, chemical growth regulation is, because of its high efficiency, still more widely applied for height control of ornamentals (Lewis et al. 2004; Bergstrand 2017).

The creation of transgenic lines obtained via genetic modification can be used to develop and improve many ornamental plants and could thus substantially reduce the need for chemical growth regulation (Chandler and Sanchez 2012). Genetically dwarfed plants with improved compact growth are often obtained via modification of the endogenous gibberellin metabolism in plants (Bhattacharya et al. 2010; Shao et al. 2020). However, the number of transgenic ornamentals brought into commerce to date are actually far and few between mainly due to strict legislation (Milošević et al. 2015; Chandler and Tanaka 2018). A promising alternative that can lead to sustainable height control of plants is the transformation with wild type rhizogenic agrobacteria. Here it is essential that wild type strains are used, making the resulting plants exempt from the strict GMO (genetically modified organism) legislation (Desmet et al. 2020a). This method makes use of the innate transformation ability of wild type Rhizobium rhizogenes. These bacteria carry a unique type of virulence plasmid, the so-called Ri (root inducing) plasmid, which enables the bacteria to infect a plethora of host plants (Chandra 2012). During this process, the transfer DNA (T-DNA) of the Ri plasmid (pRi) can be integrated into the host plant genome, resulting in the proliferation of characteristic adventitious roots known as hairy roots (HR). The inserted native pRi T-DNA genes have prominent effects on the plant hormone metabolism (Van Huylenbroeck et al. 2019). HR have been an important aspect in many biotechnological applications and scientific research (Ono and Tian 2011). One such application is the use of HR in plant molecular breeding. Plants regenerated from HR are often characterized by a peculiar phenotype, widely known as the Ri phenotype, which is characterized by short internodes, increased branching, dwarfed growth, wrinkled leaves and excessive root formation (Tepfer 1984). Many of these traits are of particular interest where compact growth is highly desired. Plants with the Ri phenotype have already been applied successfully in several ornamental plants such as Kalanchoe blossfeldiana (Lütken et al. 2012), Gentiana spp. (Mishiba et al. 2006) and Digitalis purpurea (Koga et al. 2000). Moreover, these Ri lines are directly integrable within existing breeding programs, facilitating the creation of compact growing varieties that are less dependent on chemical growth regulation.

In this study, Ri lines of multiple Viola × wittrockiana F1 hybrids are created by co-cultivation with wild type R. rhizogenes. A high efficiency co-cultivation is developed and shoot regeneration of hairy roots was successfully obtained. The morphology of 12 unique Ri lines is compared with the original F1 hybrid lines, showing substantial improvement to multiple horticultural traits.

Materials and methods

Tissue culture of Viola × wittrockiana

In vitro shoot tip cultures were established from seeds of 6 Viola × wittrockiana F1 hybrids (denoted v1, v2, v3, v4, v5 and v6) that were kindly provided by Rudy Raes Bloemzaden NV (Destelbergen, Belgium). Seeds wrapped in Miracloth (pore size 22–25 µm, Merck) were surface sterilized as follows: tap water (1 min), sterile water (1 min), ethanol (70%, 1 min), sterile water (1 min), sodium hypochlorite (1.2%, 10 min; technical sodium hypochlorite solution, Carl Roth), 3 × sterile water (3 min each). Next, seeds were dried in air flow for 5 min and transferred to Petri dishes (Ø 90 mm) containing 20 mL of basic medium (BM: 2.2 g L−1 Murashige and Skoog (MS) macro and micro salts, 10 g L−1 sucrose, solidified with 7.3 g L−1 micro-agar, pH 5.7). Seeds were incubated in the dark for 10 days (21 ± 1 °C). A total of 100–200 germinated seedlings were selected at random and transferred to glass jars (Ø 90 mm, 9 cm height) containing 75 mL multiplication medium (MM: 5.5 g L−1 MS macro and micro salts, 20 g L−1 sucrose, 6.5 g L−1 micro-agar, 0.8 g L−1 gelrite, pH 5.7). Growth conditions for shoot tip cultures were 16 h/8 h light/dark photoperiod (Philips TL-D Super 80, 36W/840, PAR: 25.9 ± 9.2 µmol m−2 s−1) at 21 ± 1 °C with a 4 week subculture period, alternating between MM and MM + 1 mg L−1 2-isopentenyladenine (2iP).

Preparation of bacterial inoculum

Wild type rhizogenic strain of the 4 opine types were used for co-cultivation: agropine (Arqua1, ATCC15834, LMG152), cucumopine (NCPPB2659), mannopine (LMG150) and mikimopine (MAFF210266). Bacterial suspensions were prepared according to Desmet et al. (2019). In short, bacteria were cultured in yeast extract glucose agar (YEG) consisting of 10 g L−1 glucose, 10 g L−1 yeast extract, 1 g L−1 (NH4)2SO4 and 0.25 g L−1 KH2PO4. For solid YEG, 15 g L−1 bacto-agar was added. A pre-culture on solid YEG was grown until individual colonies were visible (dark, 28 °C). Single colonies were transferred to 100 mL YEG and incubated (28 °C, 175 rpm) until exponential growth was reached. Adequate growth of each liquid culture was evaluated by measurement of optical density (OD) at 600 nm. Based on the growth curve regression equation per strain, an estimated cell density (colony forming units (CFU) per mL) was calculated. Cultures with an OD correlating with a cell density of at least 1 × 108 CFU mL−1 were used for co-cultivation experiments. For NCPPB2659 and MAFF210266, suspensions with OD600 between 0.3 and 0.4 were used.

Co-cultivation experiments

Seedling hypocotyl was chosen as the explant type. Germinated seeds were transferred to 16 h/8 h light/dark photoperiod, 96 h prior to co-cultivation. The radicle was removed but cotelydons were left intact (Fig. 1a). Explants were incubated in bacterial suspension (25 mL suspension per 50 explants, Ø 90 mm Petri dishes) on an orbital shaker (30 min, 125 rpm). For the control treatment YEG without bacteria was used. Afterwards, explants were blotted on sterile filter paper, dried for 1 min and transferred to co-cultivation medium (CCM: BM with sucrose substituted by 10 g L−1 glucose and supplemented with 20 mg L−1 acetosyringone) and incubated in the dark for 48 h (21 ± 1 °C). Next, the explants were incubated in liquid BM with 500 mg L−1 cefotaxime on an orbital shaker (30 min, 125 rpm). After blotting and drying for 1 min, explants were transferred per 10 to Petri dishes (Ø 90 mm) with each 25 mL subculture medium (SCM: BM with 500 mg L−1 cefotaxime and 100 mg L−1 ticarcillin). Explants were subcultured at 21 ± 1 °C in 16 h /8 h light/dark photoperiod in dimmed light conditions (PAR: 10.0 ± 3.0 µmol m−2 s−1) and the SCM was renewed every 2 weeks. After 2 subculture intervals, root formation was evaluated. The number of explants per treatment showing root formation, tissue necrosis, bacterial regrowth and vital non-reactive explants were recorded. Root formation efficiency (RFE) was calculated as the ratio of the root forming explants divided by the sum of the root forming explants and the vital non-reactive explants.

Fig. 1
figure 1

Co-cultivation and regeneration of shoots from Viola × wittrockiana hairy root tissue (scale bar in all panels = 2 cm). a In vitro germinated seedlings of v5, 10 days after sowing. b Hypocotyl explants of v5 co-cultivated with ATCC15834, 4 weeks post inoculation. c Root line obtained from v2 after ATCC15834 co-cultivation proliferating in BM, 8 weeks post inoculation. d Non-proliferating v3—ATCC15834 root line. e Callus induction in a v1—MAFF210266 root line. f High frequency shoot induction on a v1—MAFF210266 root line

Full size image

In a first experiment, hypocotyl explants of all 6 lines were co-cultivated using ATCC15834 and control treatment. Each treatment-line combination consisted of 10 explants and was repeated 10 times. The line with the highest RFE was selected for use in the second experiment in which all 6 strains and a control treatment were co-cultivated with hypocotyl explants of v1. Each treatment consisted of 10 explants and was repeated 10 times.

Hairy root regeneration

Culture conditions during HR regeneration were 16 h/8 h light/dark photoperiod (PAR: 10.0 ± 3.0 µmol m−2 s−1) and 21 ± 1 °C. Four weeks after inoculation, roots present on hypocotyl explants were excised as individual root lines (RL). The regeneration phase consisted of 4 stages, each with a 4 week duration: (1) proliferation on SCM, (2) proliferation on BM, (3) callus induction and (4) shoot induction. RL without active proliferation and with tissue necrosis were discarded. Callus-induction medium (CIM) is BM supplemented with 3 mg L−1 2iP and 0.3 mg L−1 2,4-dichlorophenoxyacetic acid (2,4-D). Shoot induction medium (SIM) is the BM amended with 3 mg L−1 2iP. Afterwards, roots and calli were maintained on SIM for a total of 4 months, with monthly subculture. Evaluation of regeneration frequency was conducted on a monthly basis per step of the regeneration phase: proliferation (number of RL showing active proliferation during subculture), callus induction (number of RL with visible callus formation) and shoot regeneration (number of RL in which shoots were formed).

Molecular characterization of regenerated shoots

Shoots derived from roots or calli were excised and transferred to MM. After 4 weeks, each shoot was sampled twice: (1) mixed tissue sample (1 cm stem + 1 cm2 leaf tissue) for testing the presence of residual bacteria, (2) and a 100 mg leaf sample for DNA extraction. Sample processing and molecular analysis was performed as described in Desmet et al. (2021). The absence of residual bacteria was determined using virD2 primers of Haas et al. (1995) and DNA quality was confirmed using the its u3 and u4 primers as described by Cheng et al. (2016). Evaluation of the presence of pRi T-DNA genes (rolA, rolB, rolC, rolD, aux1, aux2, rolBTR) was done using different primers pairs (Table 1). Primers for pRiA4 (rolA, rolB, rolD, aux1, aux2 and rolBTR) and pRi1724 (rolA, rolB, rolD) are according to Desmet et al. (2021). For rolC of both pRiA4 and pRi1724, the rolC primers described in Desmet et al. (2020a) and Desmet et al. (2021) respectively, were used. Both quantification cycle (Cq) values and melting curves (melting temperature and profile of the peak) were considered in the analysis.

Table 1 Primer pairs used for amplification of pRi T-DNA genes
Full size table

Morphological characterization of Ri lines

Micro-cuttings (apical cuttings with 2 nodes) of garden pansy Ri and control hybrid lines were placed on hormone free MM and incubated in the dark at 6 °C for 24 h. Next, the cuttings were placed individually in paperplugs (Ø 35 mm) in 84 insert trays in plastic tunnel constructions (95–100% relative humidity) for 10 days, after which the plastic covering was gradually removed over the course of 4 days. Plantlets were grown in a greenhouse for another 2 weeks in paperplug before being transplanted to individual plant plots and spaced to a plant density of 80 plants per m2 (Ø 10.5 cm, peat based substrate with 1.0 kg m−3 fertilizer 14N:16P:18 K + trace elements, Agaris, Belgium). In the greenhouse (latitude 51.0565° N, longitude 3.7990° E), plants were watered twice per week. Fertilizer was applied at watering (EC 120–400 µS cm−1, 3 L m−2 consisting of 50/50 ratio of 6N:18P:36 K and 20N:10P:20 K). Plants were grown under natural day length and temperature until flowering. No plant growth retardants were applied unless specified. The selection of Ri lines for morphological measurements was based on the availability of sufficient tissue culture material of each Ri line.

Experiment 1: Morphology of Ri lines

Cuttings of ATCC15834 derived Ri lines Reg1, 2, 3, 4, 5, 7, 8, 12, 15, 16, 17, 19 and control hybrid lines v1, v2 and v3 were acclimatized as described above from week 4, 2020 until week 10, 2020. Each treatment consisted of 15 plants. Measurements conducted at anthesis included: flowering onset, peduncle length of the first and second flower, diameter of the first and second flower, petiole length of 2 fully developed leaves on the main branch, length of the main stem, number of internodes of the main stem, presence or absence of leaf and flower wrinkling.

Experiment 2: Chemical growth regulation trial

Cuttings of Reg15 and its respective original hybrid line v3 were acclimatized as described above from week 35, 2019 until week 39, 2019. For this experiment, 3 treatment groups were established once the plants entered the greenhouse: (1) v3 grown without chemical growth regulation (v3 − CGR), (2) v3 chemically growth regulated (v3 + CGR) and (3) Reg15 (Ri plant obtained from v3). Each group consisted of 10 plants. V3 + CGR plants received a combination of daminozide (Alar®85, Bayer, Belgium) and chlormequat (750.0 g L−1 chlormequat, Cycocel®75, BASF, Belgium). In total, 7 foliar sprays of 3 g L−1 daminozide and 0.75 g L−1 chlormequat were applied at a volume of 50 mL m−2 during a 7 week period (one treatment per week, starting from week 42 of 2019 until week 49 of 2020). Chlorophyll content (chlorophyll a + b) of each plant was determined by means of spectrophotometry at two time points: 8 and 27 weeks after transplant (coinciding with flowering onset of v3 and Reg15, respectively). From each plant, the youngest most fully expanded leaf was harvested and approx. 50 mg was sampled. Samples were transferred to separate tubes containing 2.0 mL of 96% ethanol and incubated for 24 h in the dark (26 °C, 200 rpm). Next, optical density of each sample was determined at 649 nm and 665 nm. Each sample was measured twice in a measurement volume of 200 µL. Chlorophyll content was calculated according to the formula of Wintermans and De Mots (1965). Morphological measurements were conducted for each plant at anthesis. The following parameters were recorded: flowering onset (number of weeks after transplant to final pot), peduncle length of the first and second flower, diameter of the first and second flower (measured horizontally), petiole length of 2 fully developed leaves on the main branch, length of the main stem, number of internodes of the main stem, presence or absence of leaf and flower wrinkling. Due to the extreme dwarfed growth of Reg15, morphology of v3 was recorded at 2 time points: (1) a first round at anthesis of each individual v3 plant (Reg15 was still in a vegetative stage), and (2) a second observation round when Reg15 also started flowering.

Statistical analysis and software

Statistical test validity based on assumptions of normality and homogeneity of variances was evaluated by means of the Shapiro and Levene’s test respectively. If the assumptions were met, a parametric test (one way or two way ANOVA) was conducted. Otherwise Aligned Rank Transform (ART) was used as non-parametric alternative. Analysis of significant main effects was performed by multiple comparisons using the Tukey’s HSD correction. All plots and statistical analyses were prepared and conducted in R (version 3.6.0) using RStudio (version 1.1.463). Figure collages were assembled using Inkscape version 0.92.

Results

Co-cultivation experiments

Hypocotyl explants started to form roots from 2 weeks after the start of the co-cultivation. Root formation was observed for both ATCC15834 co-cultivated (HR + wild type roots) and control treatments (wild type roots) (Table 2). In both treatments, roots formed directly on the hypocotyl without any visible callus. No distinct morphological differences of the roots in both treatments were obvious: all roots were white, proliferate in all directions and had a ‘hairy’ appearance due to visible root hairs (Fig. 1b). RFE is higher in the co-cultivated treatment for all the garden pansy hybrid lines, but varies between the different lines: the highest RFE for the ATCC15834 treatment was observed for hybrid line v1 (74.9%) which produced on average 3.5 roots per hypocotyl explant whereas the lowest RFE was observed with hybrid line v4. Line v6 produced the highest average number of roots per explant (Table 2). The maximum number of individual roots recorded for a single explant (v6 + ATCC15834) was 17. No visible bacterial regrowth was observed during explant subculture (Supplementary Information S1).

Table 2 Root formation efficiency (RFE, in %) and average number of roots per root forming explant (x̅) of hypocotyl explants of different Viola × wittrockiana hybrid lines co-cultivated with ATCC15834 (values represent mean ± standard deviation, n = 10, values sharing the same letter per column are not statistically different, α = 0.05, RFE: two-way Anova, x̅: ART ANOVA)
Full size table

In the second co-cultivation experiment, line v1 was co-cultivated with different R. rhizogenes strains. Spontaneous rooting in the control treatment was also observed. ATCC15834 performed best, with very similar RFE compared to v1 in the first experiment (Table 3). The mannopine strain LMG150 performs slightly better than Arqua1. The other strains are considerably less efficient in terms of RFE (Table 3). For the hypocotyl explants of v1 co-cultivated with NCPPB2659 the highest explant mortality was recorded, with this strain also being the only one where bacterial regrowth was observed after co-cultivation (Supplementary Information S2).

Table 3 Root formation efficiency (RFE, in %) and average number of hairy roots per root forming explant (x̅) of hypocotyl explants of Viola × wittrockiana hybrid line v1 co-cultivated with different rhizogenic strains (values represent mean ± standard deviation, n = 10, values sharing the same letter per column are not statistically different, α = 0.05, ART ANOVA)
Full size table

Hairy root regeneration

A total of 1255 individual RL were harvested from the co-cultivation experiments, excluding roots from control treatments (Table 4). The majority of these roots originated from ATCC15834 co-cultivation. During subculture, many RL started to proliferate and branched rapidly upon transfer from SCM to BM (Fig. 1c). RL lacking vigorous growth changed to a yellow–brown color (Fig. 1d) followed by tissue necrosis. The proportion of proliferating RL is the highest for v6—ATCC15834 (87%), followed by v1—MAFF210266 and v3—ATCC15834 (74%). Only 15% of v4 RL showed active proliferation. Transferring RL to CIM with 2iP and 2,4-D stimulated the production of numerous vigorously growing and branching roots with strong radial expansion (Fig. 1e). Radial expansion and nodular callus formation occurred in up to 85% (v5—ATCC15834) of the RL submitted to callus induction (Table 4). Shoot formation was observed at different time periods. Shoot primordia were visible as soon as the first month of SIM subculture. Extended subculture of 2, 3 and 4 months was necessary for other root derived calli to initiate shoot formation (Table 4, Fig. 1f). The highest regeneration frequency of 27% was obtained in v3—ATCC15834 calli. Line-strain combinations for which no shoots were obtained are: v4—ATCC15834, v1—Arqua1, v1—LMG152 and v1—NCPPB2659 (Table 4). From both experiments, 54 unique shoots were obtained.

Table 4 Regeneration from Viola × wittrockiana root lines obtained from co-cultivation experiments using different hybrid lines and rhizogenic strains (No. of proliferating roots = root lines with active proliferation and no tissue necrosis after subculture on subculture medium and basic medium, No. of calli = root lines with visible callus formation, No. of unique shoots = callus forming root lines in which shoots are formed after 1, 2, 3 or 4 months of subculture on shoot induction medium, cumulative)
Full size table

Molecular characterization of Ri lines

Shoots regenerated from root lines were individually cultured and analyzed for the presence of residual agrobacteria and pRi T-DNA genes. Of the 54 shoots, 3 had poor growth and had died during culture. In total, 51 regenerated shoots from root lines obtained with the strains ATCC15834 (40/51), LMG150 (7/51) and MAFF210266 (4/51) were maintained in vitro. No amplification of the virD2 gene was observed, indicating that all shoots were devoid of residual bacteria. Independent integration of pRiA4 left T-DNA (TL-DNA: rolA, rolB, rolC and rolD) and right T-DNA (TR-DNA: aux1, aux2 and rolBTR) was observed in ATCC15834 derived shoots: 31 shoots harbor the 4 rol genes and the 3 TR-DNA genes, whereas in the remaining 9 shoots only the 4 rol genes are present (Supplementary Information S3). No shoots were obtained with genes of the TR-DNA only. None of the shoots derived from LMG150 root lines harbored the pRi genes. However, MAFF210266 derived shoots all carry the 4 pRi1724 type rol genes (Supplementary Information S3).

Morphological evaluation of Ri lines

Morphology of Ri lines

In the vegetative stage, all Ri lines displayed leaf wrinkling with very pronounced leaf veins and dark green color. Petiole length was significantly decreased in 11 Ri lines, except for Reg7 (Table 5). Main shoot length was significantly altered in all Ri lines with high inter Ri line variation (Fig. 2). The v3 derived Ri lines had on average the largest reduction ranging from 36 (Reg12) to 49% (Reg17). Shoot length was the least reduced in v1 Ri lines, ranging from 9 to 34% (Table 5). The shoots of v1 and v3 Ri lines counted 1 to 4 internodes less on average. The number of internodes per shoot was the least altered for v2 derived Ri lines. Due to these alterations, Ri lines had a more bushy and compact appearance (Fig. 2). The v2 group started flowering earlier than the v1 and v3 groups (Table 5). In 7 of the 12 Ri lines delayed flowering up to a maximum of 2 weeks on average was observed. All Ri lines carried flowers on significantly shorter peduncles (Table 5). The reduction of peduncle length relative to the control was the most pronounced for v2 and v3 Ri lines, ranging from 36 to 56% reduction. Peduncle length of v1 Ri lines was reduced by 13 to 29% depending on the Ri line (Fig. 2). Flower diameter was significantly decreased in all Ri lines (Fig. 3). The decrease ranged from 24 to 48% and varied between the groups: v3 Ri lines had the highest relative decreased flower size (Table 5). Severe flower petal wrinkling was observed in the v1 Ri lines (Reg1, Reg2, Reg3 and Reg4), and in 3 of the 5 v3 Ri lines (Reg12, Reg15 and Reg16) but not for v2 Ri lines (Fig. 3).

Table 5 Morphology of control hybrid lines v1, v2 and v3 and derived Ri lines (all obtained with ATCC15834) measured at anthesis (values represent mean ± standard deviation, n = 15, values sharing the same letter per parameter per hybrid line (v1, Reg1, Reg2, Reg3 and Reg4; v2, Reg5, Reg7 and Reg19; v3, Reg8, Reg12, Reg15, Reg16 and Reg17) are not statistically different, α = 0.05, ART ANOVA)
Full size table
Fig. 2
figure 2

Morphology of Viola × wittrockiana Ri lines and control hybrid lines. a, b Side and top view of hybrid line v1 (left) and Reg3 (right). c, d Side and top view of hybrid line v2 and Reg5. e, f Side and top view of hybrid line v3 and Reg15. Scale bar in all panels = 10 cm

Full size image
Fig. 3
figure 3

Flower petal morphology of Viola × wittrockiana hybrid lines a v1, b v2 and c v3 and Ri lines d Reg 3, e Reg5 and f Reg15. Scale bar = 5 cm

Full size image

Chemical growth regulation experiment

Reg15 displayed severe dwarfed growth from very early on during plant development, with yellowish pale and wrinkled leaves (Fig. 4a, b). Controls started flowering from 7 to 8 weeks after transplant, at which point a reduction of 44% in chlorophyll a and b in Reg15 was recorded (Table 6). Chlorophyll content was the highest in v3 + CGR. Shoots of the control were twice the length of Reg15 shoots, and had more than double the number of internodes. On average, petiole length was reduced by 1 cm in Reg15 compared to v3 (Table 6, Fig. 4a, c). Reg15 started flowering from 26 weeks after transplant onwards, at which point the controls had reached a mature plant habit (Fig. 4b, d). At anthesis of Reg15, chlorophyll content of Reg15 was still significantly (28%) lower than the controls (Table 7). Leaves of Reg15 were wrinkled and had very pronounced leaf veins. Shoot length was significantly decreased in v3 + CGR and Reg15. The number of internodes was not altered by CGR treatment, but shoots of Reg15 had approx. 50% less internodes (Table 7). Petiole length was significantly decreased in the v3 + CGR and Reg15 groups. The effect on peduncle length was more pronounced; peduncles in both controls were twice as long when compared to Reg15 (Table 7, Fig. 4b, d). The flowers of Reg15 were on average 2 cm smaller, with a more pale-red color but no petal wrinkling. The overall plant habit of Reg15 was ball-shaped, compact with upright flowers.

Fig. 4
figure 4

Phenotype comparison of Viola × wittrockiana Reg15 and v3. Side and top view 8 (a, c) and 27 (b, d) weeks after transplant (from left to right: v3, v3 + CGR and Reg15). Scale bar in all panels = 10 cm

Full size image
Table 6 Morphology of Ri line Reg15 and control hybrid line v3 (with and without chemical growth regulation (CGR)) measured at anthesis of v3 (values represent mean ± standard deviation, n = 10 unless explicitly stated otherwise, values sharing the same letter per column are not statistically different, α = 0.05, ART ANOVA)
Full size table
Table 7 Morphology of Reg15 and control hybrid line v3 with and without chemical growth regulation (WT + CGR and WT respectively) measured at anthesis of Reg15 (27 weeks after transplant, values represent mean ± standard deviation, n = 10 unless explicitly stated otherwise, values sharing the same letter per row are not statistically different, α = 0.05, ART ANOVA)
Full size table

Discussion

Until now, few studies have reported on tissue culture of Violaceae species (Slazak et al. 2015). As pansies are commercially propagated by seed (Horn 2002), seedling price is much lower than the production price per tissue culture plantlet, which eliminates a major factor to the general need for tissue culture protocols of Viola spp. This certainly has an impact on the application of many biotechnological techniques that rely intensively on tissue culture based techniques, such as Agrobacterium mediated transformation. In the present study, we report on co-cultivation with R. rhizogenes and regeneration of hairy roots for large flower garden pansies (Viola × wittrockiana).

Preliminary co-cultivation experiments (data not shown) indicated that conventional explants such as leaf disc and nodal segments were not feasible because of a hypersensitivity reaction upon co-cultivation with rhizogenic agrobacteria. However, by use of hypocotyl explants, abundant root formation was obtained after co-cultivation. Hypocotyls have a favorable morphogenetic profile due to the high juvenile nature of the tissue, making it the explant of choice for genetic transformation in many crops (Piqueras et al. 2010). Root formation on garden pansy hypocotyl was observed in both co-cultivated and control treatments. Additionally, whereas in other plant species hairy roots show a typical morphology (Mugnier 1988), no apparent morphological differences of garden pansy roots could be observed between the treatments. Roots formed on co-cultivated explants thus probably consist of a mixture of HR and wild type roots, although substantially more root formation was noticeable in co-cultivated treatments. Root formation efficiency varied depending on the opine type of the strain used. ATCC15834 was the most virulent, which has been observed as well in other plant species such as Hypericum perforatum (Bivadi et al. 2014) and Dracocephalum kotschyi (Sharafi et al. 2014). The other agropine type strains Arqua1 and LMG152 were significantly less efficient in inducing root formation, which is consistent with the findings recorded for Sinningia speciosa (Desmet et al. 2020b) and Osteospermum fruticosum (Desmet et al. 2021). The mannopine strain LMG150 resulted in the second most efficient root formation, which contrasts sharply with results from the R. rhizogenes host range studies of De Cleene and De Ley (1981) and Porter and Flores (1991). In these studies mannopine type strains LMG63 (equivalent strains ICPB TR7 = NCPPB2626 = ATCC25818) and LMG150 (equivalent strains NCPPB2991 = ATCC11325), were used to test the pathogenicity in Viola tricolor and Viola septentrionalis respectively. They concluded that Viola spp. are not susceptible for R. rhizogenes. In contrast, the current results clearly indicate that some species of the Viola genus are indeed susceptible and are thus encompassed within the R. rhizogenes host range. Moreover, hairy root formation was not limited to mannopine strains, but was also achieved with agropine and mikimopine strains.

Root lines of Viola × wittrockiana showed a varying degree of proliferation during extended subculture on hormone free medium. The proliferation of root lines was dependent on the genetic background of the roots (i.e. hybrid line). Extended subculture and proliferation based selection of individual root lines is likely to create a positive selection for true HR by eliminating wild type roots that lack the autonomous proliferation ability. This selective pressure was also observed in HR cultures of Gentiana scabra (Suginuma and Akihama 1995) and Rehmannia glutinosa (Zhou et al. 2009). Additionally, the inclusion of a subculture period on hormone free medium has proven to be critical to improve the regeneration capacity of HR of Rosa hybrida (van der Salm et al. 1996) and Digitalis purpurea (Koga et al. 2000). Shoot regeneration was readily obtained in root lines of different garden pansy hybrid lines and rhizogenic strains (ATCC15834, LMG150 and MAFF210266). An important prerequisite for shoot regeneration to occur, was the combination of 2iP and 2,4-D in a separate callus-induction phase to confer regeneration potential to the root lines. Similarly, use of 2iP but in conjunction with NAA was also implemented in the regeneration phase of Datura sanguinea and Datura arborea hairy roots (Giovannini et al. 1997).

Mannopine strains and mikimopine type pRi have a single T-DNA fragment that carries the rol genes as the primary oncogenes. Agropine strains on the other hand, next to the rol genes located on the TL-DNA, have additional oncogenes located on the TR-DNA due to their split T-DNA structure (Veena and Taylor 2007). The transfer of the TR-DNA is species dependent and has previously been linked to the regeneration capacity of hairy roots (White et al. 1985; Camilleri and Jouanin 1991). In garden pansies, co-transfer of both TL- and TR-DNA or sole transfer of the TL-DNA region of the pRiA4 plasmid occurs. Co-transfer of both T-DNA fragments occurred more frequently. However, no shoots containing only genes of the TR-DNA were obtained. These results align with the hypothesis of Taylor et al. (1985) who found a selective pressure against auxin gene expression during regeneration of Nicotiana spp. HR. A similar mechanism was proposed by Hegelund et al. (2018) in oilseed rape. This selection effect can also explain as to why there have been no reports of TR-DNA only Ri lines obtained via wild type strains. Shoots obtained via MAFF210266 all harbor the pRi1724 type rol genes. Interestingly, none of the shoots regenerated from root lines obtained with LMG150 contain mannopine (pRi8196) type rol genes. These shoots could thus have originated as shoot escapes, or the root lines obtained from LMG150 co-cultivation were in fact wild type roots.

Plants of all evaluated Ri lines exhibited altered plant morphology compared to the original hybrid lines. Garden pansy Ri lines were more compact with a dwarfed, bushy appearance and characterized by smaller flowers, reduced shoot length, shortened peduncles, and leaf and flower wrinkling. Similar dwarfed Ri lines have been reported in Ipomoea trichocarpa (Otani et al. 1996), Actinidia deliciosa (Yamakawa and Chen 1996) and Hypericum perforatum (Bertoli et al. 2008). In garden pansy, the severity of morphological changes varied significantly between Ri lines and was dependent on the original hybrid line as some changes, such as flower wrinkling, were not observed in v2 derived Ri lines. Relative to the control, v2 and v3 derived Ri lines were the most compact. Genotype-mediated differences in the Ri phenotype have previously also been reported by Mishiba et al. (2006) for Gentiana triflora x Gentiana scabra and G. scabra Ri lines. In addition, inter-Ri line phenotype variation has been described for Ri lines of Nicotiana tabacum (Tepfer 1984), Calibrachoa excellens (Gennarelli et al. 2009) and Prunus avium × Prunus pseudocerasus (Rugini et al. 2015). The plant habit of Ri lines such as Reg3, Reg5 and Reg 15 is promising for breeding compact garden pansy varieties. Moreover, Reg15 displayed a more compact stature than the control hybrid line after chemical growth regulation. Traits that heavily affect the overall ornamental value such as wrinkling of the leaves and smaller flowers are however unwanted. Thus, the garden pansy Ri lines could be implemented as pre-breeding material in introgression breeding or for the creation of homozygous parental lines. Unwanted traits such as leaf wrinkling can be substantially reduced in as little as 2 selection cycles (Lütken et al. 2012; Desmet et al. 2020b). Delayed flowering time is also undesired as this increases production time and cost (Christensen et al. 2008). Interestingly, the time to anthesis is less reduced if the Ri lines were grown in early spring, rather than in late fall. This can be explained by the fact that due to the growth retardation, control v3 and Reg15 start the winter period in different developmental stages. This further reinforces the implementation of Ri lines as pre-breeding material in garden pansy breeding.

In conclusion, in this study a high number of unique garden pansy Ri lines were developed. These Ri lines were characterized by substantial variation both in terms of molecular (inserted pRi T-DNA genes) and phenotypic constitution. This finding underlines the importance of an efficient regeneration protocol, since this facilitates the creation of a large and diverse selection of Ri lines. Furthermore, the large degree of variability between these lines is regarded as beneficial for breeding and selection purposes. These Ri lines are exempt from GMO legislation and can thus be readily implemented into existing breeding programs. Several compact Ri pansy lines have been identified and will be used to develop cultivars that do not rely on chemical growth regulation.