Advances in plant transgenic technology in the 20th century overcame the major hurdle for transfer of genetic material between species. This not only enabled fundamental insights into plant biology, but also revolutionized commercial agriculture. Adoption of transgenic plants in industrial agriculture has reduced pesticide application, while bringing significant increase in crop yields and farmers' profits. The progress made in transgenic technology over the last three decades paved the way mainly for simple single-gene insect and herbicide tolerance (HT) trait products. Modern agriculture demands stacking and pyramiding of complex traits that provide broad-spectrum insect and HT with other agronomic traits. In addition, more recent developments in genome editing provide unique opportunities to create precise on-demand genome modifications to enhance crop productivity. The major challenge for the plant biotech industry therefore remains to combine multiple forms of traits needed to create commercially viable stacked product. This review provides a historical perspective of conventional breeding stacks, current status of molecular stacks and future developments needed to enable genome-editing technology for trait stacking.
A defining challenge of modern agriculture is to ensure global food security by feeding an increasingly populated, urban and affluent planet without compromising the diverse and crucial ecosystems . Given that world population is projected to increase to 9 billion in next 30 years, with the current agricultural productivity an additional ∼1 billion hectares of cultivated land will be required by 2050 to meet the food demand . A potential solution to address this global agriculture challenge without expanding environmental footprint is to improve productivity by maximizing crop yield to its full biophysical potential. The adoption of advances made in breeding and agricultural practice in the 1960s led to the green revolution, which provided food security to millions of people . Introduction of novel traits by conventional breeding is limited to germplasms that are sexually compatible.
The species barrier of trait transfer was overcome by transgenic trait technology in the mid-1990s, which was a major milestone in commercial agriculture. Transgenic technology made a huge impact in improving agriculture productivity through mainly single-gene herbicide and insect traits in key crops like maize, soybean and cotton. The next-generation trait products entail stacking multiple insect and disease resistance and herbicide tolerance (HT) genes to provide control of a broad range of insect pests and weeds . In addition, pyramiding of 2–3 genes is needed against a single pest to provide multiple modes of action for robust and durable control . Rapid advances made in genomics coupled with high-throughput phenotyping are greatly expanding the potential of identifying networks of native genes and alleles controlling disease resistance and agronomic traits . Manipulation of agronomic traits depends on precise engineering of complex metabolic pathways, which requires concerted expression of multiple genes .
Designed nucleases have become a powerful gene editing tool to create targeted DNA double-strand breaks (DSBs) at specified genomic locations . Currently, there are four different types of designed nucleases being used, namely meganucleases, zinc finger nucleases (ZFNs), transcription activator-like effector nucleases (TALENs) and clustered, regularly interspaced short palindromic repeat (CRISPR)-associated endonucleases. These tools have successfully been used for targeted genome editing of multiple plant species. However, editing of multiple genes or alleles is generally needed to create a commercially viable trait. For example, the disease control conferred by single disease resistance (R) gene is often overcome within a few seasons by pathogen evolution, and multiple R genes are needed to provide durable resistance in plants .
Given that a majority of the stacked biotech products currently in the market were developed by conventional crossing, trait stacking using breeding is clearly a straightforward and viable option. However, the breeding approach is limited to a few multiple independently segregating loci, and the method requires substantial time and resources for sorting and deregulation of multiple unlinked transgenes [4,10]. This review will focus on different trait stacking approaches including conventional plant breeding, transgenic molecular stack and gene editing with a major emphasis on application of designed nucleases as a tool for trait stacking.
Conventional trait stacks
Trait stacks are mainly created using two approaches: molecular stack and breeding stack. Molecular stacks are made via plant transformation, incorporating two (or more) genes from a single transgene construct into the plant genome. Molecular stacks are physically linked and therefore co-segregate, thus making subsequent introgression and conversion fairly simple. The first generation of commercial molecular stacks mainly contained single-gene insect resistance (IR) and HT trait, which include Syngenta's event Bt11 with pat and cry1Ab, TC1507 with pat and cry1F jointly developed by Dow AgroSciences and Pioneer, Monsanto's MON 89034 with cry1A.105 and cry2Ab2 . While molecular stacks with two to three genes are straightforward to make, more genes with multiple modes of action are desired in a commercial product to have a resilient management system for robust tolerance to target pests. With increased number of genes, assembly of the gene cassettes becomes complex and challenging, especially with repetitive use of a limited number of regulatory elements in a stack, which could potentially lead to gene silencing or spontaneous transgene rearrangements . Furthermore, with the addition of each new sequence, the size of the transgene is increased while the number of permutations for vector construction increases exponentially. Plant transformation quickly becomes a limiting factor with the large size of the transgene and the number of constructs [7,11,12] leading to significant costs in transformation and characterization of transgenic events. In addition, when multiple gene candidates are present to control the same pest on one stack, the efficacy trials to parse out the functionality of each gene become very complex.
Breeding stacks are developed using traditional plant breeding methods where each parental line contains one or more previously incorporated transgene(s). These parental lines are crossed with each other to create offspring that contain all the traits. Monsanto in 1997 introduced the first breeding stack, which was produced by crossing two cotton events: Bollgard™ (producing the Bt toxin cry1Ab) and Roundup Ready™ (producing epsps enzyme: 5-enolpyruvyl-shikimate synthase) . Following commercial success of breeding stacks, biotech seed companies invested heavily in developing trait stacks to meet specific market needs. Based on USDA survey data, stacked corn occupied 76% of corn acres and stacked cotton reached 80% of cotton plantings by 2016 (https://www.ers.usda.gov/data-products/adoption-of-genetically-engineered-crops-in-the-us/recent-trends-in-ge-adoption.aspx). Breeding stacks with previously deregulated events are straightforward and flexible when a few trait loci are needed to make a commercial product. In some countries, such as in the U.S.A. and Canada, no additional safety assessment studies are required for registration of new breeding stacks created using previously deregulated commercialized events. Even when a new approval is needed, the process is expedited as the biology of the existing germplasm and transgenic events is well characterized. Those independently registered events are randomly integrated into chromosomes, most likely unlinked, offering the flexibility of ‘fit for purpose’.
Different trait stacks can be developed to meet the market needs with specific pest spectra, such as in SmartStax™ and PowerCore™ technologies jointly developed by Dow AgroSciences and Monsanto. In SmartStax™, four independent transgenic loci (events MON 89034, TC1507, 88017 and DAS 59122) were brought together via breeding in a single product with multiple modes of action covering a broad spectrum of insect pests [14,15]. Meanwhile, PowerCore™, with MON 89034, TC1507 and NK603, is a unique combination of three events, providing wide spectrum protection against above-ground pests for areas with minimal corn rootworm pressure . One distinct advantage of breeding stacks is their flexibility for cross-licensing. This was clearly demonstrated with PowerCore™, the product mentioned above, which has two traits, MON 89034 and NK603 from Monsanto, and one, TC1507 from Dow AgroSciences.
While breeding stacks do provide some advantages, with increased number of events, deployment of stacks for these unlinked genetic loci quickly becomes a huge burden to the breeders. Based on Mendelian segregation, ∼19 000 F2 individuals need to be planted and analyzed to obtain one homozygote from six transgenic events containing unlinked segregating loci . With increased number of unlinked loci, additional time is needed for the conversion of desired traits into elite germplasms. For example, complete conversion might take <4 years for one to three loci, while >5 years are needed for creating a product containing four or more loci. This would lead to an estimated ∼1.2% hidden loss in genetic gain with every year lost for the conversion . Furthermore, to take the advantage of diverse genetic pool of germplasm, transgenic traits are often introduced into thousands of elite lines, which would demand ever-increasing field resources and analytics to track and ensure product stewardship. In addition, with every locus, there is also concern of carrying unwanted linkage drag which could negatively affect overall product performance.
Molecular stacks using gene targeting
Creation of new molecular stacks brings some challenges as conventional transformation methodologies such as Agrobacterium or biolistics do not target transgenes into specific locations in the genome. As these methodologies insert transgenes at essentially random locations, large numbers of transgenic events need to be screened in order to eliminate events with negative characteristics. These can include epigenetic modification of the transgene, thus affecting expression, incomplete or rearranged transgenes at insertion, disruption of native genes, insertion at undesired genomic locations, agronomic effect resulting from transgene insertion and insertion of vector backbone [18,19]. Additionally, many governmental regulations are associated with the negative implications of the random nature of conventional transgene insertions. To overcome challenges associated with random insertions, genome-editing tools could be used for precise insertion of transgenes into predefined desired genomic locations. This process of gene addition could be continued for multiple generations of product development by sequential targeting of trait genes within or in close proximity to an existing transgene (Figure 1).
Different gene targeting approaches for trait stacking in plants.
The advent of multiple tools for the nuclease-mediated creation of DSBs in the genome has enabled site-specific genome modification. The demonstration that homologous recombination could be augmented by two orders of magnitude through the creation of DSB in the genome [20,21] was the initial step in the development of our modern toolbox of genome-editing tools in plants. Genomic DSB can be repaired by non-homologous end-joining (NHEJ) or homology-dependent repair (HDR). The process of HDR is limited to S and G2 phases of the cell cycle, while NHEJ is the primary DNA repair pathway in the somatic cells . Given this, it is understandable that the prevalence of NHEJ targeted mutagenesis or ‘knockout’ using genome-editing tools has become routine, while HDR-mediated targeted integration or ‘knockin’ still remains more challenging. The type of repair utilized by plants (HDR vs. NHEJ) is a function of many factors including the presence of repair template and cell type . At present, there are multiple tools which can foster the creation of DSBs in the genomes of plants (Table 1). The use of these tools has been described in detail previously elsewhere [24–26].
|Technique .||Type of gene stacking .||Crop species .||Reference .|
|Meganuclease||Stacking near transgene||Cotton|||
|Nonfunctional marker gene||Corn|||
|ZFN||Nonfunctional marker gene||Tobacco|||
|Nonfunctional marker gene||Tobacco|||
|Technique .||Type of gene stacking .||Crop species .||Reference .|
|Meganuclease||Stacking near transgene||Cotton|||
|Nonfunctional marker gene||Corn|||
|ZFN||Nonfunctional marker gene||Tobacco|||
|Nonfunctional marker gene||Tobacco|||
Rare cutting enzymes (such as I-SceI) and meganucleases were among the first tools used to demonstrate transgene integration into the genome of plants [20,21,27–29]. With the demonstration of the I-SceI-induced DSB and site-specific integration of a transgene into a pre-engineered site of maize , the initial step of precise trait stacking mediated by DSB was taken. While the basic path for precise gene targeting and stacking was laid out, hurdles for routine deployment still existed. Gene targeting still relied upon the prior integration of a random event, as well as pre-engineering of nuclease recognition sites. These difficulties were addressed through the development of engineered meganucleases that could be designed to recognize novel sequences [24,31]. Meganuclease technology has been used to create transgenic trait stacks in both maize and cotton. Interestingly, demonstration of precise trait stacking in cotton involved single-chain I-CreI–based engineered homing endonuclease for targeted integration of two herbicide tolerance genes proximal to a pre-existing, high-performance transgene locus (Figure 1A) . Based on independently transformed embryogenic callus lines, ∼2% targeting frequency and Mendelian inheritance of targeted traits were reported.
Zinc finger nucleases
ZFNs, consisting of a DNA-binding domain attached to a nuclease domain, are another early tool developed for induction of DSBs in plant genomes. The DNA-binding domain can be designed to recognize almost any DNA sequence and thus enable targeted double-strand break creation . The utility of ZFNs was first shown in the model plant species Arabidopsis via targeted mutagenesis of a pre-engineered sequence . Demonstration of targeted mutagenesis of endogenous loci was later accomplished in Arabidopsis [35,36] and in soybean . Gene targeting via homologous recombination (HR) induced by ZFNs was first accomplished using reconstitution of a nonfunctional selectable marker in a pre-engineered target (Figure 1B) inserted in tobacco [38–40] and Arabidopsis [38,41]. The HR-mediated gene targeting using ZFNs in a crop plant was later demonstrated through the insertion of the pat (phosphinothricin acetyl transferase) gene  into the endogenous IPK1 locus in maize . The plants obtained from this targeted insertion exhibited both herbicide resistance and reduced phytate levels. ZFNs have also been used to induce targeted gene exchange in the tobacco BY2 cell reporter system .
Use of ZFNs for stacking of transgenic traits in maize has been shown through multiple systems. Engineered trait landing pads with pre-engineered ZFN recognition sites were used to demonstrate targeting of a herbicide tolerance gene (aad1) into an existing transgene containing the pat herbicide tolerance gene (Figure 1C). This was accomplished through the co-transformation of the second marker gene, aad1 encoding aryloxyalkanoate dioxygenase , flanked by sequences homologous to the integrated trait landing pad along with a corresponding ZFN. The result was a molecular stack of two herbicide resistance genes with demonstrated activity of both genes . Since every cycle of gene stacking requires a new selectable marker, the limited number of available markers poses a major hurdle for sequential targeting and trait stacking. This limitation was overcome in a recent study where new transgenic stacks were loaded in successive gene targeting cycles through the exchange of selectable marker genes in maize (Figure 1D) . Cassette exchange was accomplished using the homology of an intron downstream from a promoter driving the selectable marker gene cassette and a trait landing pad homology at the 3′-end. Targeting rates of up to 30% of generated events were reported using the pat selectable marker gene. Unlike previous gene targeting methods (described above) that utilize defective or partial genes for selecting targeted events, this system offers distinct advantages as it exchanges a fully functional selectable marker in every iteration, thus allowing for recycling of selectable markers needed to accomplish sequential targeting in multiple generations of trait stacking.
Transcriptional activator-like effector nucleases
TALENs, which are derived from transcriptional activator-like effectors of the plant pathogen Xanthomonas spp., have been shown to be versatile for design with sequence recognition, which can be used to create DSBs at practically any genomic locus [48–50]. Genome editing via TALENS has been demonstrated in multiple plant species including Arabidopsis thaliana protoplast  and transient expression assays in Nicotiana benthamiana leaves . Additionally, site-directed mutagenesis using TALENs has been reported in crop plants such as potato , tomato , soybean , rice [55,56], maize , wheat  and barley . Targeted gene editing using TALENs has been achieved in rice . Precise gene targeting via TALENs has been shown in tobacco protoplast , barley cells , rice  and tomato plants .
Clustered, regularly interspaced short palindromic repeat
The newest and widely popular genome-editing tool is the engineered CRISPR/CRISPR-associated (Cas) systems . Cas9 (CRISPR-associated) endonuclease is guided to a target DNA sequence adjacent to the PAM (protospacer adjacent motif) sequence by the complex of a CRISPR RNA (crRNA) with a trans-activating crRNA (tracrRNA). Later, a fusion of the dual-tracrRNA : crRNA complex was engineered creating a single-guide RNA (sgRNA) to direct Cas9-mediated targeted DNA cleavage . The design process of this system is extremely easy since Cas9 protein does not require any re-engineering while specificity of the nuclease is defined by the sgRNA. CRISPR-Cpf1 (CRISPR from Prevoltella and Francisella1) has recently been reported as a new type of genome-editing tool . Unlike the Cas9 system, Cpf1 is a type V CRISPR effector that recognizes thymidine-rich PAM sequence and induces cohesive DSBs at target site guided by a single crRNA [66–68]. Since discovery in 2012, the CRISPR system technology has expanded quickly with rapid demonstration of genome modification occurring in multiple crops such as rice [64,69,70], maize [57,71], soybean [72,73] and wheat [58,64]. The initial demonstration of targeted gene insertion utilizing the CRISPR system included rice  and Arabidopsis [74,75]. Gene targeting via homologous recombination has been achieved using the CRISPR/Cas technology in maize , soybean  and tomato .
In reality, all the technologies described above perform the same core function, which is to create DSBs in the genome of a plant. Therefore, factors such as ease of design, fidelity and intellectual property (IP) need to be considered when choosing the most appropriate technology for their use. When evaluating the ability to effectively design a nuclease, both CRISPRs and TALENs are generally considered quite favorably in terms of design and ease of use. However, CRISPR has some design limitations [65,66], which are not present with technologies such as ZFNs. In terms of fidelity, ZFNs utilizing high-fidelity FokI nuclease requiring hetrodimerization have been evaluated in clinical trials and deemed to be safe . In addition, the IP rights surrounding CRISPRs are quite complex, one must consider all facets of IP when making a choice of technologies [78,79].
Intragenomic homologous recombination and trait stacking
Although gene editing is a quantum technological leap in the field of plant biotechnology, there are yet bottlenecks to its full deployment in trait stacking for crop improvement. While targeted mutagenesis to create gene edits using designed nucleases is becoming a routine approach, targeted transgene integration remains a huge challenge due to its reliance on genetic transformation of plants and on a DNA repair process (HDR) less prevalent in plants. Significant advances in plant transformation have been made in last ∼20 years; however, the majority of plant species and genotypes are still recalcitrant to transformation and regeneration .
Intragenomic homologous recombination (IGHR) provides an attractive opportunity to address limitations of low transformation efficiency by utilizing the DNA recombination system of a cell to replicate randomly inserted donor transgene for subsequent IGHR-mediated insertion and stacking of the donor into the target site. As described in Figure 2, this approach needs a very small number of primary transgenic plant lines that contain stable integration of donor and the matching designed nuclease transgene. Unlike a few cells that contain donor and designed nuclease transgene in direct transformation-mediated precise gene stacking described above, all the cells in the whole plant contain donor and nuclease leading to intrachromosomal recombination between target and donor throughout the life cycle of plant development. The germinal targeted events containing donor sequence stacked to the target site could be screened in the next generation. A similar approach using a site-specific recombinase (FLP) and a site-specific enodnuclease (I-SceI) was used in Drosophila for modification of the yellow locus [81–83].
Gene stacking in plants using IGHR.
Plant scientists were quick to propose this method for plant gene targeting and stacking ; it was demonstrated in Arabidopsis  using endonuclease (I-SceI) and CRISPR/Cas system . The IGHR-mediated GT in a model system paved the way for application of this method in maize by somatic ectopic recombination and tissue culture selection [86–88]. The recent demonstration of this technique in maize using zinc finger nucleases  provides a robust gene stacking strategy for commercial product development. A flexible and modular gene stacking design described in the report utilized nuclease-mediated cassette exchange (NMCE) between target and donor to activate a promoterless selectable marker in a donor sequence . This approach provides flexibility to strategically place different traits in modular donor or target format, which could be mixed and match meeting the changing market needs
Conclusions and future direction
Designed nuclease-mediated targeted mutagenesis is becoming routine in plants [43,45,56–58,64,71,89–91] to create desired traits by introducing knockout mutations. However, to provide broad spectrum of insect and disease resistance, herbicide tolerance and wide-ranging agronomic enhancements, we envision that future biotech crops will contain trait packages consisting of native, gene-edited and transgenic traits stacked together in one genome. Recent advances in gene targeting via IGHR have generated huge potential for the application of designed nucleases for gene editing and stacking of edited traits with native and transgenic traits. Unlike conventional methods that depend on an efficient transformation system, the IGHR approach utilizes the conventional breeding technique for precise gene stacking, which makes this system very attractive for gene editing and stacking in a broad group of economically important crop plants that remain recalcitrant to genetic transformation. The method could further be improved by using viral replicons to amplify nuclease and repair a template for enhanced gene targeting . In addition, recent development toward targeted gene insertion and replacement using the NHEJ pathway , which avoids homology restrictions, will further enable deployment of genome-editing technology for trait stacking.
Modern agriculture demands stacking and pyramiding of multiple genes to provide diverse agronomic trait package with multiple modes of action for robust and durable control against pest.
A vast majority of the stacked biotech products currently in the market were developed by the conventional breeding approach, which requires substantial time and resources for sorting and deregulation, and the method is only practical for limited number of independently segregating loci.
Designed nucleases have become a powerful genome-editing tool to create targeted DNA double-strand breaks at specified genomic locations.
Recent advances made in gene stacking using genome-editing tools have created a huge opportunity for application of this technology in commercial agriculture.
CRISPR from Prevoltella and Francisella1
clustered, regularly interspaced short palindromic repeat
gene of interest
intragenomic homologous recombination
nuclease-mediated cassette exchange
protospacer adjacent motif
phosphinothricin acetyl transferase
transcription activator-like effector nucleases
S.K., W.C. and S.N. drafted the manuscript. All authors read and approved the final manuscript.
We are grateful to Michelle Smith, Otto Folkerts, Lakshmi Sastry-Dent, Steve Webb and Rodrigo Sarria for reviewing the manuscript and providing valuable suggestions.
The authors in this publication certify that they are employees of Dow AgroSciences LLC, a subsidiary of The Dow Chemical Company. As employees, the authors may have some company stock (The Dow Chemical Company), but no other personal financial interest in the subject matter of the materials discussed in this manuscript.
Bollgard™, Roundup Ready™, SmartStax™ and PowerCore™ are trademarks of Monsanto Technology LLC.