Targeting a conserved functional motif in the PDS gene enables efficient CRISPR/Cas9 editing in banana

Abstract

Incomplete editing and chimeric phenotypes are major challenges in CRISPR/Cas9-mediated genome editing of polyploid crops. In this study, a single guide RNA (gRNA) was designed to target a conserved dinucleotide-binding motif within exon 3 of the phytoene desaturase (PDS) gene in ‘Grand Naine’ banana. The gRNA was carefully selected for GC content, guanine residues near the PAM, and predicted secondary structure to enhance Cas9 cleavage efficiency. Agrobacterium-mediated transformation of embryonic cell suspensions produced 102 putative transgenic plants, all exhibiting altered phenotypes, with 91% displaying albino and 9% pale green coloration, indicating efficient PDS gene knockout and absence of chimerism. Sequencing confirmed tri-allelic editing, with all edited plants consistently showing two identical and one distinct mutation. Notably, small in-frame deletions of two to six amino acids within the conserved motif were sufficient to abolish PDS function, confirming its critical role in carotenoid biosynthesis. This strategy is adaptable to clonally propagated polyploid crops, providing a practical framework for achieving high-efficiency, uniform genome edits and supporting the development of precise, non-chimeric CRISPR/Cas9 editing approaches.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-51803-5.

Introduction

Banana (Musa spp.) is a globally important fruit crop, serving as a staple food for over 400 million people and providing significant income in tropical and subtropical regions. It ranks among the top five fruit crops worldwide in terms of production and trade. Most edible bananas are seedless, triploid hybrids (AAA, AAB, or ABB), which are sterile and propagated vegetatively¹. Among commercial cultivars, ‘Grand Naine’ (AAA genome group, Cavendish subgroup) is particularly significant, occupying more than 60% of the area under banana cultivation. This cultivar dominates global trade due to its high yield, uniformity, and strong consumer preference. However, its narrow genetic base and susceptibility to major diseases, such as Fusarium wilt Tropical Race 4 (TR4) and Black Sigatoka, pose serious threats to sustainable production².

The complete sterility of ‘Grand Naine’ limits the applicability of conventional breeding strategies, such as backcrossing, for its genetic improvement. This reproductive constraint, combined with its polyploid genome, renders conventional breeding approaches time-consuming, technically challenging, and often ineffective in introducing traits such as disease resistance or improved fruit quality^3,4. In this context, genome editing technologies, particularly CRISPR/Cas9, offer promising alternatives by enabling precise, site-specific genetic modifications without the integration of foreign DNA.

CRISPR/Cas9-mediated genome editing has been successfully employed in several clonally propagated and polyploid crops, including grape (Vitis vinifera)⁵, strawberry (Fragaria × ananassa)⁶ and blueberry (Vaccinium corymbosum)⁷. However, success rate of editing in polyploid species remains a major challenge due to the presence of multiple homologous / homeologous gene copies, which complicates the complete knockout of target gene/s⁸. Several studies have reported low knockout efficiency in polyploid crops, with edited plants often displaying wild-type or chimeric phenotypes^9,10,7. Ntui et al.¹¹ also observed chimerism in regenerated banana plants, where CRISPR/Cas9-induced mutations occurred at different developmental stages, resulting in a mixture of edited and unedited cells. This is likely due to the sustained activity of the genome editing machinery post-transformation, which continues to introduce mutations until the target sites are mutated¹².

Phytoene desaturase (PDS) is a key enzyme in the carotenoid biosynthesis pathway, and its disruption of this gene impairs chlorophyll synthesis, leading to an albino phenotype. This albino phenotype serves as a convenient visual marker for successful gene editing¹³. PDS gene has been targeted to confirm the editing efficiency of CRISPR/Cas9 in banana. However, less than 55% of the edited plants in these studies exhibited the expected albino phenotype¹⁴, which has been attributed to mono/diallelic editing¹⁴ and/or mis-folding of sgRNAs (gRNAs) reducing their binding affinity to Cas9 and consequently lowers cleavage efficiency¹⁵. Although two guide RNAs were used to target the PDS gene to improve editing efficiency in banana, only one gRNA proved effective, generating nearly 65% albino plants^16,11. Similar observations have been reported in other crops, like highbush blueberry⁷ and chilli pepper¹⁷, where one sgRNA outperformed the other, reinforcing the importance of optimizing sgRNA for efficient editing. This suggests that sgRNA design still requires further optimization.

Jung et al.¹⁸ evidenced that cleavage ability of the Cas9-sgRNA complex is one of the key factors influencing CRISPR/Cas9 editing efficiency. The secondary structure of single-guide RNAs (sgRNAs) plays a crucial role in facilitating effective interaction with the Cas9 protein, thereby enhancing cleavage efficiency. For example, Dang et al.¹⁹ showed that modifications to the sgRNA scaffold such as duplex extension and targeted mutations can significantly increase knockout efficiency. Disruption of the repeat: anti-repeat pairing within sgRNAs has been found to abolish Cas9 activity entirely^20,21. Moreover, even minor alterations in the sgRNA scaffold structure can impact editing efficiency by up to tenfold, depending on the sequence context and target site²². These findings underscore the precise sgRNA design, particularly in optimizing secondary structure, to achieve efficient and consistent genome editing results.

Additionally, gene knockout experiments have proved that sgRNAs targeting the DNA sequences encoding essential protein domains often produce stronger dropout phenotypes than those targeting non-critical regions of the same gene²³. This is likely because CRISPR-Cas9 predominantly introduces small insertions or deletions (indels), which can lead to either frameshift or in-frame mutations. Frameshift mutations typically disrupt protein function entirely, while in-frame indels resulting in the addition or deletion of amino acids may or may not impair protein activity, depending on their location. In-frame insertions or deletions within non-essential domains of a protein may not disrupt its overall activity. However, when such mutation occurs within functionally essential domains, they are likely to disrupt structural motifs or active sites, leading to reduced or abolished protein function²⁴. Therefore, selecting conserved functionally vital target regions is essential for achieving consistent and effective knockout phenotypes.

For perennial and clonally propagated polyploid crops, achieving homozygous knockout mutations is crucial for both functional studies and crop improvement¹⁰. While the albino phenotype serves as a visual marker for PDS editing, many genes of agronomic traits do not produce visible phenotype when edited, this presents a significant challenge in identifying fully edited mutants. Therefore, improving genome editing efficiency particularly for generating complete knockouts in all alleles of polyploid genomes such as banana is critical. This highlights the need for optimized genome editing protocols that maximize editing efficiency to improve success rates in polyploid crops.

In the present study, we aimed to improve CRISPR/Cas9 editing outcomes in the triploid banana cultivar ‘Grand Naine’ by targeting a conserved and functionally critical region of the PDS gene. Special emphasis was given on the gRNA design to maintain the optimal secondary structure of sgRNA and selection of the CRISPR/Cas9 vector with a promoter optimized for high Cas9 expression in banana. This strategic approach resulted in a high proportion of albino phenotypes, indicating effective tri-allelic editing in this triploid cultivar.

Results

Target selection and guide RNA design

PDS (Phytoene desaturase) is a single-copy gene, reported in Cavendish banana¹⁶ and other plants such as cassava²⁵, melon²⁶, and onion²⁷. To achieve efficient gene knockout in banana PDS gene in cv. Grand Naine, belongs to Cavendish group of AAA genome, we prioritized both target sequence selection and optimized guide RNA design. For target sequence selection, conserved motifs were identified through multiple sequence alignment of the Cavendish (Williams) PDS protein sequence with PDS sequences from the Rasthali AAB genome (MF574096.1 and MF574097.1), along with sequences from five monocot species (rice, wheat, maize, ginger, and onion) and six dicot species (tomato, cucumber, melon, papaya, Arabidopsis, and Nicotiana benthamiana). Three key conserved motifs were identified: the dinucleotide-binding motif, the putative carrier motif, and the carotenoid-binding motif (Fig. 1).

Fig. 1.

Multiple sequence alignment of PDS proteins from various plant species, including Musa PDS. Colored boxes highlight conserved functional motifs and domains across the aligned sequences. The red box indicates the dinucleotide-binding (DB) motif (GAGLAGLSTAKYLADAGHKPIVLEARDVLGG), the yellow box denotes the putative substrate/carrier (PSC) motif, and the blue box represents the carotenoid-binding (CB) motif, which are essential for the catalytic activity of the PDS protein.

An optimal guide RNA (gRNA) was designed within the conserved region of the dinucleotide-binding motif using the CRISPR-P 2.0 web-based tool²⁸. The selected gRNA had a GC content of 60%, guanine residues proximal to the PAM sequence, and a gRNA sequence that preserved the sgRNA’s essential stem-loop structure with minimum off-target activity (Fig. 2). The chosen guide RNA targeted the dinucleotide-binding motif located within the third exon of the PDS gene (positions 418–437 in the coding region) (Fig. 3).

Fig. 2.

Secondary structure of PDS sgRNA.

Fig. 3.

Schematic representation of the PDS gene and CRISPR/Cas9 construct used for editing banana. (a) Genomic organization of the banana PDS gene. Green boxes represent exons (numbered) and introns are indicated by line. (b) Target site within the PDS gene. Guide RNA (gRNA) is indicated by red arrow. Target sequences are shown in blue, with PAM sequences highlighted in red. The black arrow indicates the region flanking the target sequences to detect mutations. (c) Schematic diagram of the CRISPR/Cas9 binary vector pRGEB32-MaPDSgRNA used for banana transformation.

The complete PDS gene in Musa acuminata (gene ID: Ma08_g16510), comprising 14 exons and spanning 28,118 bp, was retrieved from the Banana Genome Hub (Fig. 3). To confirm the absence of allelic variation within the target region, an 869 bp fragment of the PDS gene from cv. Grand Naine was amplified, spanning from the start codon through the third exon. This amplified fragment included 226 bp of exon 1, 84 bp of intron 1, 130 bp of exon 2, 210 bp of intron 2, 89 bp of exon 3, and 125 bp of partial intron 3.

The amplified product was cloned into the pJET1.2/blunt vector, and 12 clones were randomly selected for sequencing. The sequencing results showed that all clones were nearly identical to the reference PDS gene sequence of Musa acuminata, with three nucleotide polymorphisms detected: an A/T variation at position + 347 and T/C variations at positions + 385 and + 389. At position + 347, 66% of the clones contained an ‘A’ and 34% a ‘T’; a similar frequency distribution was observed at position + 385 and + 389 for the ‘T’/‘C’ variation (Fig. S2). These results confirm that the PDS gene of cv. Grand Naine possesses two different alleles. These allelic variations are consistent with findings in another Cavendish group cultivar ‘Williams,’ as reported by Naim et al.¹⁶

Notably, all clones of the PDS fragment confirmed the absence of allelic variation at the target site (positions 418–437). The selected 20-bp gRNA (GAGGCTAGAGATGTCCTGGG) was cloned into the pRGEB32 vector digested with BsaI. The clone was confirmed by sequencing and designated as pRGEB32-MaPDSgRNA.

Transformation and regeneration of banana plants

The Cas9 with sgRNA expression cassette (pRGEB32-MaPDSgRNA) was introduced into embryogenic cell suspensions of banana cv. Grand Naine via Agrobacterium-mediated transformation, following the protocol of Khanna et al.²⁹. Transformed cells were selected on hygromycin-containing medium (12.5 mg/L) and subsequently regenerated into plantlets, resulting in 105 putative transgenic lines. These lines exhibited distinct phenotypes: 93 albino, 9 pale green and 3 with a normal green phenotype. Plantlets were sub-cultured at every three weeks. Albino and pale green plants exhibited slow and stunted growth, compared to wildtype plants. Subsequently, all plants were transferred to rooting medium, where root development was observed to be similar to wild-type plants (Fig. 4).

Fig. 4.

Development of PDS edited plants in cv. Grand Naine using the CRISPR/Cas9 (pRGEB32-MaPDSgRNA) construct. (a) Transformed embryogenic cell suspension (ECS) on hygromycin selection medium, (b) embryo formation from transformed ECS, (c) embryos germinating on selective regeneration medium, (d) regenerated plantlets exhibiting an albino phenotype, (e and f) Albino plantlets on shoot and root induction medium, (g) representative albino, (h) pale green and (i) non transformed wild-type plantlets.

To confirm the presence of the transgene, regenerated plants (albino, pale green, and green) were validated by PCR analysis using Cas9-specific primers. As expected, all transgenic events exhibiting albino and pale green phenotypes showed amplification of a 486 bp fragment, confirming the presence of the Cas9 gene. While all three green plants failed to show amplification, indicating the absence of the transgene and suggesting that these plants were likely escapes. After successful root establishment, the plantlets were transferred to potting mixture and maintained in a growth chamber. All pale green plants survived, whereas all albino plants eventually died.

Genotype of PDS edited plants

To evaluate the types of mutations generated in transgenic plants, PCR was performed on 9 representative transgenic plants (7 albino and 2 pale green) using primers flanking the gRNA target site. The resulting 455 bp PCR products were cloned into the pJET 1.2 vector for sequencing. PCR was performed to identify indels (insertions/deletions). Based on band shifts several colonies from each plant were sequenced to confirm the presence and characterize the nature of the mutations. Among the 9 lines, 3 showed a reduced amplicon size compared to the wild type, suggesting larger deletions. The other 6 lines showed amplicon sizes consistent with the wild type, indicating the presence of small indels (insertions or deletions).

Sequencing results confirmed mutations in all the albino (ALB) and pale green (PG) plants. In ALB1, two different mutations were observed, a deletion of 51 nucleotides in the target region, which spans both coding and intron regions, and an insertion of 1 nucleotide in the target region 4 bp upstream of the PAM sequence. Likewise, ALB4 also showed two different mutations, a 51-nucleotide deletion and a 1-nucleotide insertion. Whereas in ALB6, the mutations observed were 51 nucleotide deletion, 3 nucleotide deletion and single nucleotide substitution 3 bp upstream of the PAM sequence (Fig. 5).

Fig. 5.

Genotyping of cv. Grand Naine PDS edited plants. Sequence alignment of seven representative albino (ALB) and two pale green (PG) edited plants and wild-type (WT). For each PDS-edited (albino and pale green) plant, only two types of mutations were identified through colony sequencing and are shown. The number of colonies corresponding to each variant is indicated on the right side of the sequences. The CRISPR/Cas9 target site is shown in blue letters, with the protospacer adjacent motif (PAM) indicated by bold underlined letters. DNA deletions are marked with red dashes, and deletion sizes (nucleotides) are indicated on the right side of each sequence. DNA insertions and substitutions are marked in red and green letters, respectively.

In ALB2 and ALB3, the same mutation pattern was observed, a 6 nucleotide deletion and a 1 nucleotide insertion at 3 bp upstream of the PAM sequence. In ALB7, a 2 nucleotide deletion at 3 bp upstream of the PAM sequence and 1nucleotide deletion at 4 bp upstream of the PAM sequence (Fig. 5).

In case of PG1, a 3 nucleotide deletion and a 2 nucleotide deletion at 3 bp upstream of the PAM sequence was observed. In PG2, a 3 nucleotide deletion and a 1 nucleotide insertion, both occurring 3 bp upstream of the PAM sequence respectively were observed (Fig. 5). These results demonstrate that only two distinct mutation types were consistently observed in each albino and pale green plant.

Naim et al.¹⁶ reported allelic variation in the PDS gene of triploid Cavendish banana (cv. ‘Williams’), with two identical alleles and one distinct. In the present study, we observed a similar allelic arrangement in cv. ‘Grand Naine’. In all PDS-edited (albino and pale green) plants only two types of mutations were observed.

The correlation between nucleotide mutations and the resulting changes at the protein level was examined in all edited lines. ALB1, ALB4, and ALB6 contained an in-frame 51 bp deletion (comprising 18 bp within the coding region and 33 bp in the intron) in two alleles, maintaining the reading frame and allowing for protein formation, and a 1 bp insertion in the third allele, which led to premature termination of the gene.

In ALB2 and ALB3, an in-frame 6 bp deletion in the coding region was present in two alleles, maintaining the reading frame. while a 1 bp insertion in the coding region of the third allele caused a frameshift and early stop codon. In ALB5, a 5 bp deletion in two alleles and a 1 bp insertion in the third allele resulted in frameshifts and premature stop codons in all alleles. Similarly, ALB7 showed 2 bp and 1 bp deletions across all three alleles, leading to truncated protein.

Protein alignment of mutant PDS sequences with the wild-type PDS revealed that an in-frame deletion of six amino acids (VLGGKV) in ALB1, ALB4, and ALB6 is associated with the albino phenotype. Similarly in ALB 2 and ALB 3, the deletion of two amino acids (VL) also resulted in the albino phenotype. These amino acids are located at the end of the dinucleotide-binding motif (GAGLAGLSTAKYLADAGHKPIVLEARDVLGGKV). In contrast, in the pale green lines (PG1 and PG2), deletion of only a single leucine (L) correlated with partial loss of gene function (Fig. 6).

Fig. 6.

Alignment of PDS Protein Sequences from Edited and Wild-Type cv. Grand Naine Plants. Protein sequence alignment of amino acid sequences translated from the edited MaPDS genes (albino and pale green phenotypes) compared to the wild-type PDS protein. Underlined regions in the wild-type (WT) sequence indicate the amino acid regions corresponding to the gRNA target site and the dinucleotide motif is highlighted in yellow colour. Amino acid deletions are indicated by red dashes, with deletion sizes (in amino acids) shown on the right side of the sequence. Amino acid substitutions are marked in green letters, and asterisks (*) represent stop codons. (Two identical mutations and a distinct mutation in the third allele were observed in each edited line).

Structural analysis of PDS and mutant PDS proteins

Structural analysis of wild-type and mutant PDS proteins was performed using the Robetta structure prediction tool (https://robetta.bakerslab.org/). This analysis focused on structural differences in PDS mutant proteins. The predicted structures included the wild type and the following mutant variants: ALB1 (a 6-amino-acid deletion, lacking residues 144 to 149: VLGGKV), ALB2 & ALB3 (a 2-amino-acid deletion at positions 144 and 145: Valine and Leucine), ALB6 (a 1-amino-acid deletion at position 144: Valine, and a substitution of Leucine 145 with Glycine), and PG (a 1-amino-acid deletion at position 145: Leucine). These structures were generated using Robetta (Fig. 7a).

Fig. 7.

Structural Modeling and Comparison of Mutant PDS Proteins with wild-type cv. Grand Naine. (a) Predicted 3D structures of the wild-type PDS protein and representative mutant variants: (1) Wild type, (2) ALB1_PDS, (3) ALB6_PDS, (4) PG_PDS, (b) Superimposition of mutant PDS protein structures (ALB1_PDS, ALB6_PDS, and PG_PDS) onto the wild-type PDS structure to visualize structural differences resulting from gene editing, (c) Structural representation of the dinucleotide-binding motif (amino acids 117–147) in the wild-type and each mutant (ALB1_PDS, ALB6_PDS, PG_PDS) protein, highlighting potential alterations in this conserved region.

The ERRAT scores for all predicted models were above 94%, indicating high-quality structures with minimal disallowed regions. Superimposition of wild-type and mutant models revealed high structural similarity, with root mean square deviation (RMSD) values indicating only minor global conformational changes (Fig. 7b).

The dinucleotide-binding motif (residues 117 to 147) of wild-type PDS and the mutant variants ALB1, ALB6, and PG was visualized. In ALB1, a small additional β-sheet was observed at the N-terminal flank of the dinucleotide-binding motif, whereas no structural changes were evident in ALB6 or PG (Fig. 7c). Although the deletion occurs within a conserved functional motif, the predicted mutant models retained a broadly similar global fold to the wild-type protein. This observation reflects preservation of overall structural topology in the in-silico models rather than direct evidence of structural robustness. The deletion may induce localized structural rearrangements that are not readily apparent from static structural predictions but could nonetheless influence protein function.

The phenotypic changes observed in the genome-edited plants suggest that the deletions may affect protein function likely through altered binding affinity or catalytic efficiency rather than through major structural instability. While in silico modelling provides valuable structural insights, it may not fully capture dynamic conformational shifts or subtle energetic perturbations critical for ligand binding and enzymatic activity.

Furthermore, in its native configuration, PDS functions as a membrane-bound tetramer. Structural alterations in the mutant variants, although minor at the monomeric level, may affect tetramerization or membrane binding. These potential disruptions in quaternary structure could influence subsequent enzymatic reactions by altering the spatial arrangement of active sites or membrane interactions. Thus, while the overall fold appears preserved, the mutations may impact functional assembly and efficiency in vivo.

Quantification of chlorophyll content

Total chlorophyll contents were quantified in the leaves of albino, pale green, and non-transformed control (wild type) lines. The levels of chlorophyll varied among these phenotypes. In albino plants, chlorophyll a ranged from 0.32 to 0.39 µg/ml, and chlorophyll b from 0.49 to 0.69 µg/ml, representing approximately 92% and 81% reductions, respectively, compared to wild-type plants, which contained 4.46 µg/ml chlorophyll a and 3.10 µg/ml chlorophyll b. Pale green plants showed chlorophyll a levels ranging from 0.68 to 0.78 µg/ml and chlorophyll b from 1.02 to 1.28 µg/ml, corresponding to reductions of about 84% and 63%, respectively. Both albino and pale green plants exhibited a significant decrease in chlorophyll and total chlorophyll content compared to the wild type (Fig. 8). Generally, chlorophyll a content is higher than chlorophyll b in wild-type plants, whereas in mutant plants (both albino and pale green), chlorophyll b levels exceed those of chlorophyll a.

Fig. 8.

Pigment quantification in edited PDS and Wild-Type plants in cv. Grand Naine. Chlorophyll a and chlorophyll b levels were quantified in three phenotypically distinct plant types: Albino (ALB1–7), pale green (PG1–2) and wild-type. Bars represent the mean ± standard deviation (SD) of three independent biological replicates (n = 3).

Discussion

CRISPR/Cas9 has emerged as a highly efficient and versatile genome editing platform, facilitating both functional genomics research and crop enhancement^30,31. In polyploid and clonally propagated crops, CRISPR-based editing offers a transformative approach for genetic improvement. Studies in various polyploid and clonally propagated crops, including banana^14,16, pear^32,33, and strawberry⁶ have frequently reported a high percentage of variegated (chimeric) plants alongside partially or fully albino phenotypes. This indicates, a major challenge remains in developing efficient editing protocols that can simultaneously disrupt all allelic copies of a target gene to achieve a complete knockout.

Phytoene desaturase (PDS) serves as an excellent visible marker gene and its disruption results in a distinct albino phenotype. PDS knockout has been successfully demonstrated in various polyploid crops, including potato³⁴ and wheat³⁵, and many fruit crops such as apple¹⁰, grape⁵, strawberry⁶, and banana¹⁴. However, many of these studies achieved only moderate albino frequencies. Furthermore, variegated (chimeric) plants were obtained even when two guide RNAs were targeting the PDS gene in many polyploidy crops such as banana^16,11, highbush blueberry⁷. This strongly indicates that polyploidy poses a significant challenge to genome editing, as partial edits frequently fail to generate distinct phenotypic outcomes^36,8,37.

This challenge is particularly significant when targeting genes that do not produce visible phenotypes upon disruption. In such cases, distinguishing fully edited plants from partially or non-edited ones becomes difficult, hampering downstream applications and functional analyses. This highlights the urgent need for strategies that not only improve editing efficiency but also increase the probability of achieving complete gene knockouts across all alleles. In the present study, we addressed these limitations by designing guide RNAs (gRNAs) to target a conserved functional motif of the PDS gene, aiming to induce complete phenotypic alterations. Additionally, guide sequences were selected based on optimized sgRNA secondary structure to enhance Cas9 cleavage activity, thereby ensuring high-efficiency genome editing in a polyploid background.

The PDS gene encodes a key enzyme in the carotenoid biosynthesis pathway, characterized by conserved domains such as a dinucleotide-binding motif, a carrier motif, and carotenoid-binding regions essential for its activity. Notably, the dinucleotide-binding motif at the N-terminus of the PDS sequence is a signature characteristic of all known carotenoid desaturases^38,39. Hence, in this study, the gRNA was designed to target this critical dinucleotide-binding motif of the PDS gene in banana cv. ‘Grand Naine’.

As Grand Naine belongs to Cavendish group (AAA) genome, confirmed allelic variations of the wild-type PDS gene through sequencing prior to gRNA design. Sequencing a short segment of the PDS gene in ‘Grand Naine’ (wild type) revealed naturally occurring polymorphisms, specifically at position + 347 (A/T variation) and at positions + 385 and + 389 (T/C variations). These allelic variations are consistent in another Cavendish group cultivar, ‘William’, as reported by Naim et al.¹⁶. The presence of these two distinct patterns of allelic variation within the PDS gene confirms that, although a single-copy gene, it exists as three alleles in the triploid Cavendish genome, where two alleles are identical, and one is distinct.

Agrobacterium-mediated transformation of ‘Grand Naine’ embryogenic cell suspensions was performed using the pRGEB32-MaPDSgRNA construct to precisely target the dinucleotide-binding motif within exon 3 of the PDS gene. This resulted in the regeneration of 102 putative transgenic lines, with 91% exhibiting a complete albino phenotype and 9% being pale green. Notably, no variegated (chimeric) phenotypes were observed, an outcome that contrasts with previous reports in banana^14,16,11 and other horticultural crops such as onion and blueberry^27,7, where variegated plants were commonly observed along with albino individuals. Variegation is typically attributed to incomplete editing, or limitations in the functional efficiency of the Cas9/sgRNA complex within initial transformed cells¹⁴, or due to chimeric mutations occurring at different stages of plant regeneration¹¹. Moreover, the high frequency of albino plants in our study was comparatively higher than previously reported. This high frequency of fully albino plants in ‘Grand Naine’ edited lines strongly suggests an exceptionally efficient and uniform functional activity of the Cas9/sgRNA complex during the embryogenic cell stage, leading to comprehensive editing before plant regeneration.

To further validate the functional knockout of the PDS gene, sequencing was performed in nine representative lines (seven albino, two pale green). Sequencing results revealed a diverse array of insertions and deletions (indels) near the protospacer adjacent motif (PAM) site, predominantly involving small indels or in-frame deletions (Fig. 5). Among these, six lines (ALB 1–5 and PG2) exhibited both nucleotide deletions and insertions in either of the alleles, whereas ALB 7 and PG1 showed only deletions in all the alleles. Interestingly, ALB 6 displayed both a deletion and a substitution within the same allele, indicating the occurrence of multiple mutation types during the editing process. This is consistent with the error-prone non-homologous end joining (NHEJ) repair pathway, which follows CRISPR/Cas9-induced double-stranded DNA breaks and often leads to characteristic mutations such as insertions or deletions, as observed in numerous CRISPR-edited crops like rice⁴⁰, maize⁴¹, wheat⁴², tomato⁴³, soybean⁴⁴, and cucumber⁴⁵.

Among the three colonies sequenced for each putative transgenic plant, two colonies exhibited the same mutation, while the third colony displayed a different mutation which suggests that two types of mutations occurred in each albino and pale green plants (Fig. 5). As, two type of allelic pattern was observed in the PDS gene of wild type Cavendish bananas (two identical, one distinct) in the present study (cv. Grand Naine) and an earlier report in cv. Willliams¹⁶, it is suggested that the identical alleles acquired the same type of mutation, while the distinct allele acquired a different mutation in all the edited plants. Although the underlying reason for this differential mutation pattern is yet to be determined, this observation unequivocally confirms that the CRISPR/Cas9 system effectively targeted all three alleles of the PDS gene in the triploid Grand Naine (AAA), indicating efficient tri-allelic editing.

To elucidate the functional consequences of targeted mutations in the PDS gene, the protein sequence of the PDS edited plants were analysed. This revealed that two edited lines had frameshift mutation in all alleles while the remaining seven lines had both in-frame deletions and frame shift mutation in either of the alleles in the conserved dinucleotide-binding motif of PDS gene. For example, ALB line 7 (− 2 bp in two alleles and − 1 bp in one allele) and ALB line 5 (− 5 bp in two alleles, + 1 bp in the third) exhibited indels that disrupted the reading frame of the PDS gene, resulting in frameshift mutations. These mutations subsequently introduced premature stop codons, leading to the early termination and non-functionality of the PDS enzyme. The remaining albino (ALB1, 2, 3, 4, 6) and pale green (PG 1 and 2) lines also carried both in-frame nucleotide deletions in one or two allele/s, while the other alleles of these lines had frameshifts that introduced stop codons and resulted in premature protein termination.

Significantly, the in-frame deletion of 51 bp (18 bp in exon and 33 bp in intron) in ALB 1, 4, and 6, and a 6 bp deletion in ALB 2 and 3, maintained the reading frame but removing six amino acids (VLGGKV) and two amino acids (VL) respectively. Despite being in-frame, these deletions critically disrupted the conserved dinucleotide-binding (DB) motif, thereby abolishing PDS enzymatic function and leading to albinism (Fig. 6). Thus, it is speculated that mutations occurring within the DB motif directly disrupt its interaction with the flavin adenine dinucleotide (FAD) cofactor, which is essential for PDS catalytic activity⁴⁶. This disruption effectively breaks the downstream pathway of carotenoid biosynthesis that converts phytoene to other carotenoids, resulting in a complete pigment loss and proving the key role of this motif in the PDS gene’s involvement in the carotenoid biosynthesis pathway.

Interestingly, PG1 and PG2 edited lines, which had a 3 bp deletion (absence of Leucine alone), resulted in pale green phenotypes, while ALB 2 and 3, with a 6 bp deletion (absence of a Valine-Leucine dipeptide), resulted in albino phenotypes. This revealed that the severity of the observed phenotype in our edited lines correlated precisely with the nature of the in-frame deletions within the PDS gene (Fig. 6). Our findings strongly proved that the Valine-Leucine (VL) dipeptide of the dinucleotide-binding (DB) motif plays a critical role in the PDS enzyme’s interaction with its essential FAD cofactor. It is speculated that specific disruption in the VL dipeptide of this motif might directly impair the enzyme’s catalytic ability to desaturate phytoene in the carotenoid biosynthesis pathway. In contrast, Naim et al.¹⁶, who observed that 21% of edited Cavendish plants showed no phenotypic alteration despite having in-frame deletion in at least one of the three PDS alleles. This revealed that targeting the functional domains is crucial than those targeting non-essential regions of the same gene to induce more pronounced loss-of-function phenotypes.

In addition, in silico structural predictions were carried out for the wild-type PDS protein and mutant variants (ALB1, ALB2&3, ALB6, and PG) containing in-frame deletions. The observed phenotypic changes in the genome-edited plants thus imply significant functional consequences of even small deletions, possibly due to altered binding affinity or catalytic efficiency, rather than gross structural instability. Koschmieder et al.⁴⁷ and Suarez et al.⁴⁸ also provided evidenced that single amino acid changes or small alterations within the FAD-binding domain or active site have been shown to abolish or significantly impair enzymatic activity, even when the overall protein fold is maintained.

Our study achieved a remarkably high albino frequency of 91%, corresponding to confirmed tri-allelic knockout of the PDS gene in ‘Grand Naine’ banana. This represents a substantial improvement over previous CRISPR/Cas9-based PDS editing studies in banana, which reported maximum albino frequencies of 63%¹⁶ and 67.5%¹¹. In earlier studies, the occurrence of wild-type or variegated phenotypes has frequently been linked to in-frame deletions¹⁶ or mono/bi-allelic editing in polyploid genomes¹⁴. To understand the factors contributing to the elevated frequency of albino phenotypes observed in our study, we conducted a comparative analysis focusing on the target site within the PDS gene, the guide RNA sequence, the predicted sgRNA secondary structure (using CRISPR-P 2.0), and the promoter used to drive Cas9 expression, relative to those employed in previous banana PDS editing studies (Table S1).

Regarding the gRNA sequence, previous studies used gRNAs with GC content of 40%¹⁴ and 45%¹⁶, while Ntui et al.. (2020) used two gRNAs with 40% and 50% GC content. Liu et al.. ⁴⁹ reported that in animal and mammalian systems, GC content significantly influences editing efficiency. In plants, although a broad range (30–80%) of GC content is effective, reduced editing efficiency has been associated with very low or high GC content⁵⁰, while high editing efficiency was observed with GC content exceeding 50%⁵¹ .

Liang et al.⁵⁰ reported that guanine is preferred at positions 1 and 2 near the PAM, and cytosine at position 3, for efficient editing in mammalian systems. Similarly, Chandrasekaran et al.⁴⁵ observed higher editing efficiency in cucumber with gRNAs containing guanine at positions 1 and 2 near the PAM, compared to those with thymine or cytosine at these positions. In the present study, the PDS gRNA contained guanine at both the first and second positions adjacent to the PAM site. In contrast, prior studies in banana employed gRNAs containing guanine at either position 1 or 2 near the PAM site, but not simultaneously at both positions.

The nucleotide composition of the guide RNA is a crucial factor for internal base-pairing between the guide RNA and the scaffold, which significantly alters the secondary structure of the sgRNA and influences cleavage activity^15,52. Based on the crystal structure of Cas9 in complex with guide RNA, Jinek et al..⁵³ and Nishimasu et al..²¹ reported that the secondary structure of stem-loop 1 facilitates the conformational changes in Cas9 which is required for target DNA binding and subsequent cleavage. Stem-loop 1 forms four small loops (or bulges) in its natural configuration, when transcribed and folded correctly. The presence or absence of bulges in Stem-Loop 1 of the single-guide RNA (sgRNA) significantly affects Cas9 recognition and binding⁵⁴. A comparison of the secondary structure of sgRNAs used in earlier banana studies revealed that sgRNAs with four bulges in stem-loop 1 had higher editing efficiency than those with three bulges¹⁶ or two bulges¹⁴. The guide RNA used in this study fulfils the required parameters with respect to nucleotide composition, such as 60% GC content and guanine at positions 1 and 2 near the PAM and also maintains the secondary structure of sgRNA with four bulges in stem-loop 1, which may ensure optimal Cas9 binding and cleavage activity.

Regarding target site selection, earlier studies on banana PDS, such as those by Naim et al. (2018) in cv. ‘Williams’ (targeting exon 1), Kaur et al. (2017) in cv. ‘Rasthali’ (targeting exon 5), and Ntui et al. (2020) in cv. ‘Sukali Ndiizi’ (targeting exons 5 and 6), without prioritizing conserved functional domains. In contrast, our study deliberately focused on a conserved, functionally important region of the PDS gene, thereby likely increasing the probability of a complete functional knockout. This strategic choice is a key reason for the high frequency of albino phenotypes observed, as in-frame mutations within essential protein domains are far more likely to disrupt function than those in non-essential regions^24,55. Our findings underscore the critical role of rational target site selection for maximizing knockout efficiency in polyploid crops like banana.

With respect to the promoter, Kaur et al.. and Ntui et al..¹¹ used the dual CaMV 35 S promoter for Cas9 expression, while Naim et al..¹⁶ tested both dual CaMV 35 S and ubiquitin promoters independently and reported that the ubiquitin promoter had higher editing efficiency than CaMV 35 S. The ubiquitin promoter is widely recognized as a strong and constitutive promoter in banana^56–58. This is in concordance with our present study, suggests that Cas9 abundance might be due Cas9 driven by ubiquitin promoter and enhancing the editing efficiency.

Our study demonstrates the highly efficient application of CRISPR/Cas9 genome editing in banana, achieving 100% editing in all regenerated plants and an exceptional 91% complete albino phenotype. This remarkable success is directly attributed to a synergistic combination of two key factors: precise targeting within the PDS gene’s conserved dinucleotide-binding motif, the use of an optimally structured guide RNA. This study emphasis that care should be taken for the aforesaid key factors while implementing the gene editing techniques to improve agronomically important traits, such as yield, enhanced nutritional content, increased resilience to biotic and abiotic stresses.

Materials and methods

Development of Embryogenic cell suspension culture

Male flower buds were collected from healthy banana plants of the cv. Grand Naine (AAA) grown in the Theni region of Tamil Nadu and used as explants for developing embryogenic cell suspension (ECS). Grand Naine ECS was established as per the protocol described by Kumaravel et al.⁵⁹. This ECS culture was subsequently used for genetic transformation experiments.

Guide RNA designing and cloning in CRISPR/Cas9 binary vector

The CRISPR/Cas9 binary vector pRGEB32⁶⁰ was used in the present study. It features a rice snoRNA U3 promoter for guide RNA expression, a rice ubiquitin promoter for Cas9 expression, and a Hygromycin phosphotransferase (HygR) gene under the CaMV 35 S promoter for plant selection.

The Phytoene desaturase (PDS) gene of banana cv. Grand Naine was selected as the target gene. The PDS gene sequence was retrieved from the Banana Genome Hub⁶¹. Furthermore, PDS gene sequences of Cavendish ‘Williams’ (AAA) (GenBank accession no. JQ762260.1) and M. acuminata ‘DH-Pahang’ (AA genome, gene ID: Ma08_g16510) were retrieved and aligned to identify conserved regions.

A 20-nucleotide target sequence was designed within the conserved region of exon 3, corresponding to the functional motif of the Grand Naine PDS gene, using the CRISPR-P 2.0 program. Guide RNA was chosen with minimal off targets and formation of proper secondary structure, as assessed by CRISPR-P 2.0²⁸. To clone the guide RNA into the CRISPR vector pRGEB32, complementary DNA oligonucleotides were synthesized with appropriate adaptor sequences that generate 4-bp overhangs. The forward primer contained a 5’-GGCA-3’ adaptor, while the reverse primer included a 5’-AAAC-3’ adaptor (Table 1). To prepare the oligonucleotide duplex, equimolar amounts of the complementary forward and reverse primers were mixed in a PCR tube, heated to 95 °C for 10 min in a thermal block, and then gradually cooled to 25 °C at a rate of 0.1 °C per second, and the duplex was confirmed through gel electrophoresis (Fig. S1).

Table 1.

PDS gRNA sequence and guide RNA oligos for cloning into the pRGEB32 vector.

Guide RNA for PDS
Target region of PDS gRNA	GAGGCTAGAGATGTCCTGGGTGG
The protospacer adjacent motif (PAM) is marked in bold underlined letters
Guide RNA oligos for cloning into pRGEB32 vector
Banana PDS gRNA1	For – 5’ GGCA GAGGCTAGAGATGTCCTGGG 3’ Rev – 3’ AAAC CCCAGGACATCTCTAGCCTC 5’
The bold colour nucleotides indicate the adapter sequence for cloning into pRGEB32 vector

The oligo-duplex guide RNA was ligated into the BsaI digested pRGEB32- Cas9 vector according to the protocol described by Xie et al.⁶² (Fig. S1). The clones were confirmed by PCR using specific primers (M13 rev and PDS guide RNA rev) and sequencing. The resulting construct was named pRGEB32-MaPDSgRNA.

Mobilization of CRISPR/Cas9 constructs into Agrobacterium tumefaciens

The CRISPR/Cas9 construct, pRGEB32-MaPDSgRNA was mobilized to Agrobacterium tumefaciens AGL1 strain by freeze thaw method⁶³. The transformed colonies were selected on LB medium containing kanamycin (50 mg/l) and rifampicin (25 mg/l). The transconjugants were confirmed by PCR with M13 reverse and PDS guide RNA reverse primers.

Agrobacterium mediated transformation of Grand Naine ECS with the CRISPR/Cas9 constructs

Embryogenic cell suspensions of Grand Naine were co-cultivated with Agrobacterium tumefaciens strain AGL1 harboring the pRGEB32-MaPDSgRNA plasmid construct and incubated for three days in darkness at 24 °C, following the protocol described by Khanna et al. (2004). After co-cultivation, the cell suspensions were washed with M2 medium (liquid cell suspension medium) containing 500 mg/L cefotaxime to remove excess Agrobacterium. They were subsequently transferred to M2 medium supplemented with 12.5 mg/L hygromycin to select transformed cells and 500 mg/L cefotaxime for bacterial suppression. After two weeks, the cells were transferred to regeneration medium (M3) containing 500 mg/L cefotaxime and 25 mg/L hygromycin. The cells were sub cultured every three weeks onto fresh regeneration medium with the same antibiotic concentrations. The embryos proliferated from the regeneration medium were separated and transferred to germination medium (M4 medium). The germinated plantlets were then moved to shoot and root development medium (M5) to develop into plantlets.

Molecular analysis of Genome edited plants

PCR analysis to confirm transgene integration

Genomic DNA was isolated from 100 mg of fresh leaf tissue from regenerated plants (albino and pale green) and wild type (control) using cetyltrimethylammonium bromide (CTAB) method⁶⁴. The quality and the concentration of DNA was measured in a NanoDrop spectrophotometer. 100 ng of genomic DNA was taken for the PCR analysis, the presence of Cas9 transgene was confirmed by PCR using specific primers (Cas9 For: CGGTCACGTATTTCACTTTG and Cas9 Rev: CGCCGGCTACATTGACGGC) with the following PCR cycle conditions: an initial denaturation at 95 °C for 10 min, followed by 35 cycles of 95 °C for 1 min, 58 °C for 30 s, and 72 °C for 30 s, with a final extension at 72 °C for 10 min.

Sequence analysis of gene edited plants

PCR confirmed transgenic plants were genotyped to detect indel polymorphisms using primers flanking gRNA target site within the PDS gene (Fig. 3). PCR was performed using 100 ng of genomic DNA from each sample and the primers flanking the PDS target region, PDS gFor (CTCTGAAGGTTGTCATAGCC) and PDS gRev (CAACCACACATGAGAGTCTG). The resulting PCR amplicons were separated on a 1% agarose gel, and the specific bands were excised, purified, and cloned into the pJET1.2/blunt vector (Thermo Fisher Scientific, Waltham, MA, USA). For each transgenic event, ten colonies were randomly selected and screened by PCR using the same PDS primer pair. Plasmid DNA was extracted from randomly selected PCR confirmed colonies for each transgenic event using GeneJET Plasmid Miniprep Kit (Thermo Scientific) and sequenced. To identify and characterize indel mutations, the obtained sequences were aligned with the wild-type PDS sequence using ClustalW in BioEdit software (Tom Hall, Ibis Biosciences, Carlsbad, CA, USA).

Chlorophyl estimation

Fresh leaf samples from albino, pale green and wild-type plants were collected for the estimation of chlorophyll. A 100 mg of fresh leaf tissue was homogenized in 2 mL of 80% acetone. The homogenate was centrifuged at 10,000 rpm for 15 min at 4 °C, and the resulting supernatant was collected. For analysis, 0.1 mL of the supernatant was diluted with 0.9 mL of 80% acetone⁶⁵. The diluted extract was then analysed using a spectrophotometer (Parkin model) by measuring absorbance at 645 nm and 663 nm. The quantification of chlorophyll-a and chlorophyll-b were calculated using Arnon’s Eq.⁶⁶.

Chlorophyll a = 12.25 (A663.2) – 279 A (646.8).

Chlorophyll b = 21.5 (A646.8) – 5.1 (A663.2).

Chlorophyll total = (Chlorophyll a + Chlorophyll b).

Protein structure prediction in mutant PDS

The amino acid sequence of Musa acuminata phytoene desaturase (PDS) (GenBank: AFK76452.1) was retrieved, and its three-dimensional structure was predicted using the Robetta server (https://robetta.bakerslab.org/), which employs advanced sampling techniques for accurate ab initio structure prediction⁶⁷. Structural models were generated for both the wild-type PDS and its mutant variants identified in albino and pale green plants. The albino mutants included an in-frame deletion of six amino acids (VLGGKV), as well as deletions of two amino acids (VL) within the dinucleotide-binding motif. The pale green mutant included a deletion of a single amino acid (L). These mutant sequences were edited accordingly, and their structures were predicted following the same workflow.

The quality of the predicted protein models was evaluated using the SAVES v6.0 server (https://saves.mbi.ucla.edu/), which integrates various structure validation tools. Among these, ERRAT⁶⁸ was used to assess the statistics of non-bonded atom–atom interactions. Structural visualization was performed using PyMOL. Superimposition of the wild-type PDS structure with the edited lines of two albino and one pale green (designated ALB1, ALB6, and PG) was conducted, and root-mean-square deviation (RMSD) values were calculated to quantify structural differences.

Comparative analysis of guide RNAs

To evaluate the effectiveness of the designed gRNA, we compared its editing efficiency and phenotypic outcome (albino plantlets) with previously reported gRNAs used for PDS gene editing in various banana cultivars using CRISPR P 2.0 web tool. The comparison included Kaur et al.¹⁴ in cv. Rasthali, Naim et al.¹⁶ in cv. Cavendish ‘Williams’, Ntui et al.¹¹ in cv. Gonja Manjia and cv. Sukali Ndiizi. This comparative analysis considered factors such as gRNA sequence, target exon, GC content, vector system, promoter used for gRNA and Cas9, editing efficiency, and observed albino phenotypes (Table S1).

Electronic Supplementary Material

Below is the link to the electronic supplementary material.

Author contributions

J.C. and B.S. conceptualized and designed the study and wrote the manuscript. J.C., E.P.S., and S.S. conducted the experimental work. S.A.C. performed the structure prediction experiments. J.C., B.S., and S.R. contributed to the interpretation of the results and critical revisions of the manuscript.

Funding

Jeyabharathy Chandrasekaran received funding from Department of Science and Technology, Ministry of Science and Technology, India; Grant ID DST/WOS-A/LS-202/2021(G).

Data availability

All relevant data has been comprehensively presented in the main text and supplementary material. No new sequences were generated in this study. Raw data is available upon request.

Declarations

Competing interests

The authors declare no competing interests.