The plant material used in our study were shoots of the cv Berkeley, cultivated on proliferation medium (PM) consisting of WPM basal medium including vitamins (Lloyd and McCown, 1980), 0.2 mgL^-1 of L-glutamine, 3% (w/v) sucrose 3%, and 2.28 µM of zeatin trans-isomers (Duchefa, NL) (pH 5.6, gelled with 0.7% w/v agar - B&V, Italy) and grown under a 16 h photoperiod, provided by Fluora L58W/77 lamps (Osram, Germany), at photosynthetic photon lux (PPF) of 40 μmol m⁻² s⁻¹ and at 24 ± 1°C.

Vaccinium myrtillus pds (Vmpds) coding sequence was detected from the Genome Database for Vaccinium (GDV; https://www.vaccinium.org/ accessed on 1^st december 2021). The PROSITE database (https://prosite.expasy.org/ accessed on 20^th january 2022) was then used to analyze and annotate Vmpds functional domains. Subsequently, using NGPhilogeny online tool (https://ngphylogeny.fr, accessed on 31^st july 2022), the Vmpds coding sequence was aligned using MAFFT Alignment with 44 other plant pds coding sequence ( Supplementary File S1 ), available on NCBI database, to produce the phylogenetic tree. The statistical significance of each branch was achieved by bootstrap analysis with 1,000 iterations. The phylogenetic tree was then visualized using PRESTO (Phylogenetic tReE viSualisaTiOn https://ngphylogeny.fr/data/displaytree, accessed on 31^st January 2022) tool.

To knock-out the Vcpds gene sequence, two specific guide RNA (gRNAs) ( Table 1 ) were designed on the CDS of Vaccinium myrtillus pds, available on the Genome Database for Vaccinium, using CRISPR-P 2.0 online tool (http://crispr.hzau.edu.cn/CRISPR2, accessed on 31^st january 2022).

Two pairs of primers ( Table 2 ) designed on V. myrtillus pds gene were used to amplify V. corymbosum pds cv Berkeley to confirm the gRNAs sequences conservation among species.

The gRNAs were then assembled into a CRISPR/Cas9 vector using the GoldenBraid 4.0 assembly system (Vázquez-Vilar et al., 2021). The Cas9 coding sequence was placed under the control of 35S promoter, the nptII gene for kanamycin resistance was under pNos promoter, while gRNAs were under pU6-26 promoter.

The final vector, named alpha_nptII_Cas9_gRNA1_gRNA2 ( Supplementary File S2 ), was transferred into the Agrobacterium tumefaciens EHA105 strain (Hood et al., 1993) by using the freeze/thaw method (Xu and Li, 2008).

Approximately 1144 leaf explants were excised from 3-week-old in vitro cultured shoots of cv. Berkeley and placed on 44 Petri dishes (26 leaf each one), following the protocol described by Marangelli et al. (2022). Briefly, the leaf explants were placed abaxial side up on the adventitious shoots organogenesis medium (SOM), consisting of WPM including vitamins, sucrose 3%, and 11.4 µM of zeatin riboside (ZR), (pH 5.6, gelled with 0.7% w/v agar - B&V, Italy), supplemented with 100 µM Acetosyringone (AS).

For genetic transformation experiments, A. tumefaciens EHA105-alpha1_NptII_Cas9_gRNA1_gRNA2 was grown overnight in Luria-Bertani (LB) selection medium supplemented with 50 mg L^-1 kanamycin and 50 mg L^-1 rifampicin at 28°C with shaking (100 rpm). Then, an aliquot (0.2-0.5 ml) of actively growing bacterial suspension was resuspended into 10 ml of LB (with kanamycin and rifampicin), under the same conditions, until an OD₆₀₀ = 0.6 was reached. In a next step, centrifugation was used to recover the bacteria (4,000 rpm for 15 min) and the pellet was then resuspended in 10 ml of hormone-free SOM supplemented with 100 µM AS.

After 24 h in dark conditions at 28°C, a few drops of the Agrobacterium suspension were poured over the leaves and, after 10-15 minutes, the excess fluid was removed by absorption with sterile paper discs. The explants were then co-cultured with the engineered Agrobacterium EHA105 on the same medium (SOM + 100 µM AS) for 48, 72 and 96 hours, in dark conditions at 24°C.

After co-cultivation, half of the explants were placed in the same medium above described supplemented with 150 mg L^-1 Cefotaxime (Cfx) alone, or in combination with 50 mg L^-1 kanamycin (Kan). The same media have been used to induce adventitious shoot organogenesis on untreated leaf explants (as control). Since for the explants cultured on the medium enriched with kanamycin, no organogenesis occurs, all the explants have been transferred to the SOM medium containing 200 mg L^-1 cefotaxime alone. The plates with the organogenetic tissues were checked daily for possible contaminations due to Agrobacterium strain used or other contaminants, and to check for the appearance of any mutant phenotypes. Throughout the experiment, regenerated shoots showing mutant phenotypes were isolated and after six weeks from the start of the experiments, all the regenerated shoots have been transferred to a PM supplemented with 200 mg L^-1 Cfx and 50 mg L^-1 Kan. Some of the mutants have been selected for molecular analysis.

To better understand when the shoot organogenesis occurs, early observations have been done by SEM analyses. Samples were fixed overnight at 4°C with 2.5% (v/v) glutaraldehyde + 2% (v/v) paraformaldehyde in 0.1 M cacodylate buffer, pH 7.2. After 3x20 min washings at 4°C in the same buffer, samples were post-fixed with 2% (v/v) osmium tetroxide in 0.1M cacodylate buffer, pH 7.2 for 2 h at 4°C. Specimens were washed in the same buffer (3 changes for 15 min each at 4°C), and then dehydrated in a graded ethanol series. Samples were dried by the critical point method using CO₂ in a Balzers Union CPD 020. Then samples were attached to aluminum stubs using carbon tape and sputter-coated with gold in a Balzers MED 010 unit. The observations were made by a JEOL JSM 6010LA electron microscope.

Genomic DNAs were extracted from 0.1 g of plant tissue obtained through transformation with nptII_Cas9_pds_gRNA1_gRNA2 vector using EZNA^® Plant DNA kit (Omega Bio-tek, Norcross, GA, USA). The kanamycin-resistant shoots include not only the albino phenotype but, in general, all the shoots showing an atypical phenotype and/or resistance to the presence of kanamycin in the growing medium.

Mutation frequencies at the target sites were evaluated through PCR amplification using primers designed on gRNAs flanking regions ( Table 2 ). DNA was amplified using KAPA HIFI Taq (KapaBiosystems, Roche, Basel, Switzerland) and the following PCR program: 95°C/3 min, followed by 30 cycles of 98°C/20 sec, 60°C/20 sec, 72°C/45 sec and 72°C/3 min. The PCR products were purified using DNA/RNA Clean Up E.Z.N.A.^® kit (Omega Bio-tek, Norcross, GA, USA). Samples were sequenced using the Sanger method and the chromatograms obtained were analyzed using the TIDE online software (https://tide.deskgen.com, accessed on 31^st July 2022).

The data was subjected to analysis of variance (ANOVA) performed using the software SPSS version 20.0. Percentage data were transformed as the arcsin of the square root of p/100 before analysis of variance. Data were presented as means of three replications at least. Significance was determined using Duncan’s test (p < 0.05).

The Vmpds coding sequence is 1,737 bp long, and the protein size is 578 amino acids. Within the protein sequence a single amino_oxidase domain is present ( Figure 1A ). The amino_oxidase domain starts at the amino acid position 49 and ends at position 274 and has the molecular function of catalyzing oxidation-reduction (redox) reactions. Available full-length NCBI pds coding sequence orthologues ( Supplementary File 1 ) were used for phylogenetic tree construction, and the resulting unrooted maximum-likelihood tree shown in Figure 1B suggests the phylogenetic relationship of gene sequence between V. myrtillus L. and Rhododendron kiusianum L. (94% bootstrap value). Both are species belonging to the Ericacee family, and they represent species-rich groups within the Ericacee (Rose et al., 2018). Affinity also emerged with Camellia sinensis L. and Diospyros kaki L. which belong to different families but from Ericales order.

The presence of de novo shoot organogenesis was simultaneously observed just 5–6 days after the begin of the experiments on the control, in which single ( Figure 2A ) or grouped bud primordia ( Figure 2B ) were observed, confirming that the induction and differentiation phases in this species take place early as recently highlighted (Marangelli et al., 2022).

Different from other species, shoot organogenesis did not occur frequently from the wounds but on the whole leaf surface, as also reported in Marangelli et al. (2022). These results were also observed by Liu et al. (2010), who described two different patterns during the de novo organogenesis processes of various cultivars of southern highbush blueberry: pattern I, where regenerants emerged mainly from the portion closer to the cutting area, and pattern II, where regeneration occurred directly from the entire leaf blade.

As shown in Table 3 , a total of 936 explants has been used for genetic transformation, and 208 explants used as control (no co-culture with A. tumefaciens strain). Furthermore, in a preliminary experiment we also tested the effect of cefotaxime on the regeneration potentiality of highbush blueberry leaves cv. Berkeley (data not shown). The results of the experiment display that the presence of Cfx at 150 mg L^-1 does not affect the regeneration process. As previously observed some antibiotics can affect morphogenesis, while kanamycin exerts a negative effect in some species, including in highbush blueberry, a positive effect has been highlighted when β-lactams such as cefotaxime are added to the regeneration media (Bosela, 2009; Silvestri et al., 2016).

After about ten days, when the first regenerations were observed in explants cultured on media enriched with only Cfx, no regeneration events were detected on the explants cultured on media also enriched with kanamycin. In general, a significant difference was observed in the regeneration ability of the explants derived from proliferation in glass jars or in PVC-jars, both in the controls and in the Agrobacterium-inoculated leaves. In fact, the average number of leaf-forming shoots was higher for the leaves harvested in PVC-Jars than the ones derived from glass jars (47.1 ± 17.3 vs 35.8 ± 15.5 respectively) ( Table 3 ).

As confirmed by statistical analysis, co-culture with the bacteria resulted in a significant reduction in the regeneration efficiency of leaf explants, irrespective of both the origin (glass- or PVC jars) and the duration of co-culture (48, 72 and 96 hours). On the regeneration media containing only Cfx, the regenerating tissues produce numerous adventitious buds just after a couple of weeks ( Figure 3A ) which develop into tiny shoots (up to 30 shoots per explant) ( Figure 3B ). When they were transferred to the PM supplemented with 200 mg L^-1 Cfx and 50 mg L^-1 Kan, most of the explants continue to grow but, after two weeks, only the explants showing altered phenotypes were able to grow in kanamycin-enriched medium, while most of the shoots stop the growth and, in some cases, they rapidly turn brown. A broad phenotypic variation was observed among the regenerated shoots. In few cases, regenerated shoots showed the expected albino phenotype ( Figures 3C, D ), some shoots were characterized by a uniform red pigmentation ( Figure 3E ), and in many cases, instead, shoots with chimeric phenotype were observed, presenting green, white, and red colorations simultaneously ( Figures 3F-I ), emphasizing the difficulty to obtain fully edited shoots. It appears that explants exhibiting an albino phenotype showed stunted growth also if cultivated in in vitro conditions on an artificial medium. This characteristic could be ascribed to the fact that PDS represents not only an essential plant carotenoid biosynthetic enzyme and a prominent target of certain inhibitors, such as norflurazon, acting as bleaching herbicides (Koschmieder et al., 2017), but also an endogenous plant growth regulator, due to its role in gibberellin biosynthesis (Qin et al., 2007). Indeed, it has been demonstrated how PDS disruption could lead to a GA deficiency. Since the processes of adventitious organogenesis are particularly sensitive to hormonal balances, this aspect should be further investigated.

To confirm the editing event, kanamycin resistant selected samples were sequenced through the Sanger sequencing and the resulting chromatograms were analyzed using the TIDE online software. The editing efficiencies were different among the two gRNAs ( Table 4 ) ranging between 2.1 and 9.6% for gRNA1 and 3.0 and 23.8 for gRNA2. It reinforces the necessity to use different guides simultaneously to ensure an efficient gene knockout (Pavese et al., 2021).

As reported in Table 4 , of the 24 lines analyzed, 16 lines were edited for gRNA1 (66.6%) while 19 for gRNA2 (79%). While for gRNA1 only heterozygous mutants were observed, for gRNA2 6 out of 19 edited lines (31%) resulted chimeric. Our results are in agreement with the results reported by Pavese et al. (2021) for European chestnut, where they found almost all heterozygous lines, contrarily to other woody crops, in which most of the mutants were homologous and biallelic (Nishitani et al., 2016; Charrier et al., 2019; Dai et al., 2021). Chimeric genotypes represent one of the major concerns in genome editing, especially for heterozygous species such as perennial fruit crops characterized by high degree of heterozygosity, the length of their juvenile phase, the long generation times and auto-incompatibility systems (Ding et al., 2020). Anyway, this problem could be minimized through the second round of adventitious shoot organogenesis, as recently reported for highbush blueberry by Han et al. (2022).

The most common mutations observed in the edited lines were single insertions or deletions (+/-1 bp) and, only in one case (line 5) a -8 bp deletion was observed. Previous observations (Bortesi et al., 2016) showed that, in most cases, one bp insertions and deletions are the predominant ones, especially interesting T and A (Bortesi et al., 2016). Anyway, the type and size of mutations could vary considerably, probably due to different factors such as target sequences, vectors, and plant species (Jacobs et al., 2015; Xu et al., 2015; Pavese et al., 2021).

The editing efficiency we reported (up to 23.8%) is in line with the only two manuscripts so far published by other authors on highbush blueberry. The first work described the knockout of the Centroradialis (Cen) gene, selected because it is present in a single-copy gene in the polyploid genome of the highbush blueberry plant. The authors demonstrated for the first time the applicability of this technique to highbush blueberry, obtaining different edited lines with an average mutated allele ratio of 19-22%, on the basis of the selected gRNA. Nevertheless, the editing efficiency of fruit crops is very variable: up to 90% in apple (Osakabe et al., 2018; Charrier et al., 2019), less than 10% in callus of kiwifruit (Wang et al., 2018), 51% in Populus spp. (Fan et al., 2017), 31% in grape (Wang et al., 2018), 35% in orange (Peng et al., 2017) and higher than 90% in Castanea sativa Mill (Pavese et al., 2021).

In our experiments, a complete albino phenotype was observed even with low editing efficiency (23.8%) ( Figure 4A ), as well chimeric mutants showing green islands ( Figure 4B ) or anthocyanins ( Figure 4C ), that often appears with the albinism, as also observed in other species (Henderson et al., 2020). Anyway, the CRISPR/Cas9 in polyploids is more challenging, due to the presence of multiple homoeologous copies (or alleles) of each gene (Schaart et al., 2021), and low editing efficiency could be ascribed to different reasons. In fact, Wilson et al. (2019) demonstrated that by targeting pds gene in octoploid strawberry, some albino shoots were obtained also when wild-type pds alleles were still present and, in the same way, also they observed pale-green shoots containing the 11% of the wild-type alleles, assuming that probably just one allele is responsible for enzyme production and the others are non-functional alleles. Martin-Pizarro et al. (2019), by targeting FaTM6 gene to study the function of the MADS box transcription factor during flower development of Fragaria x ananassa obtained the expected phenotype with modified anthers even though none of the edited lines showed mutations in all present alleles. Similarly, Gao et al. (2020) targeted the Reduced Anthocyanins in Petioles (RAP) gene in octoploid strawberry, obtaining changes in fruit colour without the occurrence of the editing in all the alleles, confirming that the loss of functional alleles is a common phenomenon in allopolyploids (Wendel et al., 2018).

Here we found a very effective adventitious shoot regeneration protocol for highbush blueberry, leading to the regeneration of up to 30 adventitious shoots from each leaf tissue. These results confirmed that zeatin and zeatin riboside are very useful to obtain high-efficient protocols for adventitious shoot organogenesis in some blueberry varieties (Rowland and Ogden, 1992; Cao and Hammerschlag, 2000; Marangelli et al., 2022).

Regarding the experiments of genetic transformation and genome editing, our results indicate that the knockout of pds gene by using CRISPR/Cas9 technology is efficient in highbush blueberry, opening the opportunity to the functional knockout of interesting candidate genes.