Several protein fusions or hybrids have been devised to control herbivorous insects, including Bt toxins or plant defense proteins as primary fusion partners [29,30,31,37,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87] (Table 1). For instance, a fusion protein integrating the Bt toxins Cry1B and Cry1Ab was engineered to broaden the insecticidal spectrum of Bt toxin-expressing lines derived from tropical maize varieties [30]. Likewise, toxic peptides and enzymes, such as the spider venom neurotoxin HWTX-I or the Beauvaria bassiana protease CDEP2, were fused to Cry1Ac to enhance its pesticidal effects against lepidopteran insects [72,73]. A number of authors have also devised Cry toxin hybrids by protein domain swapping to integrate structural elements of different native toxins within a single chimeric protein sequence. SN19, a fusion protein consisting of specific domains of the Cry1Ba and Cry1Ia Bt toxins, was designed based on this approach [66] and used to produce transgenic potato lines resistant to both coleopteran and lepidopteran pests [31]. Another Cry toxin hybrid active against insects of different orders was developed recently based on Cry1Ab modified with part of the Cry3A variable region [64].

Pesticidal effects for fusion proteins employing plant proteins or protein domains have also been reported. For instance, Mehlo et al. [71] fused a galactose-binding domain from ricin B to Cry1Ac to improve its binding properties in the insect midgut epithelium and to broaden its pesticidal effects against different lepidopteran insects. Translational fusions with different plant lectins and/or protease inhibitors have been devised as a way to improve the biological effects of the individual components [70,84,86,87] or to stabilize protein partners prone to degradation along the plant–insect continuum [84,85,88]. Plant lectins have also been used as fusion partners to deliver toxic polypeptides, such as spider or scorpion venom neurotoxins, into the haemolymph of various insects [79,80,81,82,83]. These promising developments, along with rapid advances in protein structural biology and molecular modeling allowing for the design of potentially effective fusions in silico [89], confirm the potential of fusion protein design as a realistic avenue for the production of recombinant proteins integrating multiple pesticidal functions.

On the other hand, protein domain grafting may alter the structure and function of some proteins, with possible unexpected effects on the resulting protein fusions. For instance, the Allium sativum lectin “ASAL” showed an altered glycosylation pattern, decreased thermal stability and no bacteriocidal activity when fused to the small ubiquitin related modifier (SUMO) peptide [90]. Similarly, the inhibitory activity of rice cystatin I against papain and herbivorous pest Cys proteases was substantially lowered when the 39-amino acids inhibitor of carboxypeptidase A, potato carboxypeptidase inhibitor, was grafted at the C-terminus [87]. A strategy to avoid such effects consists of adding a linker peptide to seperate the protein partners in such a way as to avoid detrimental non-specific physicochemical interactions between the non-cognate polypeptides [91,92]. Computational approaches have been developed to devise linker peptides, such as the program LINKER, which generates synthetic linker sequences from known protein structures based on user-defined requirements [93,94]. Another approach takes into account the important issue of proteolytic lability by a systematic screening of candidate linker sequences against the substrate specificities of proteases listed in the MEROPS peptidase database [95,96].

Empirical assessments with different linker peptides will possibly remain the most robust approach to identify linkers adapted to specific macromolecular contexts, but current knowledge on the properties of natural and synthetic linkers already provides helpful hints for successful outputs [92,97]. Gly-rich peptide motifs, such as the well-characterized pentapeptide motif Gly–Gly–Gly–Gly–Ser, are known to ensure a certain flexibility of protein fusions and are useful in preventing functional disturbance and destabilization of the domain partners [98]. By comparison, Pro-rich motifs confer rigidity to the proteins and are useful in maintaining a minimal distance and preventing potential non-specific interactions between the fusion partners [91,92]. Data are still scarce about the performance of linker peptides in plant biotechnology, but a number of studies suggest the potential of Gly-containing linkers to produce protease-resistant fusions with dual pesticidal activities against insects or pathogens [84,85,86,99,100]. A rigid linker of fungal origin has also been used to produce a bifunctional protein, resulting in a dual inhibitor of herbivorous pest proteases that was partially resistant to proteolytic processing [101].

Despite promising results, fusion proteins integrating linker peptides might not always be the optimal alternative to linker-free, tail-to-tail translational fusions. The choice and number of amino acids used to devise an effective linker are critical because this structural element may itself impact the overall performance of the fusion partners [102,103], fusion protein yield [104], tendency to aggregate [103], or robustness of the fusion product in terms of thermal stability [95,104] and resistance to proteolysis [95,103,105]. Biological context-dependent effects, including the host organism used for expression, further add to the complexity of linker sequence design and stress the need for more empirical data on natural and synthetic linkers. For example, a helical linker that allows for more activity of fusion partners than a flexible linker was also more sensitive to proteolysis in yeast [103] and prone to autocatalytic cleavage under certain pH conditions [105]. A trade-off might be unavoidable, in some cases, between the possible gains obtained with linker peptides in terms of yield and activity of the fusion partners, and different constraints impacting their structural stability in the host environment selected for expression.

An exogenous protease with high specificity towards linker cleavage sites may be supplied as a self-processing component of the same polyprotein as the proteins of interest for cleavage in cis, or from a co-expressed DNA sequence as a viral accessory endoprotease for cleavage in trans (Figure 1a). Examples of viral proteases used in plant biotechnology are the well-characterized tobacco etch virus nuclear inclusion NIa protease [109] and the cowpea mosaic virus 24 K proteinase [110]. These proteases excise themselves from viral polyproteins during natural infections, where they then go on to process the remaining polyprotein fragments to release the individual proteins [111,112]. For instance, the NIa protease specifically recognizes and cleaves the heptapeptidic sequence “Glu–X–X–Tyr–X–Gln–↓[Gly or Ser]”, with the scissile bond located between the Gln–Gly (or Gln-Ser) dipeptide motif [113,114]. In practice, the NIa protease may be co-expressed in planta along with an engineered polyprotein bearing the corresponding protease-cleavable motifs between the C- and N-termini of the proteins to be produced. NIa protease/polyprotein precursors have been expressed in several plant systems, notably allowing for the ectopic implementation of short metabolic pathways in transgenic plant cells [115], the coordinate expression of interacting transcription factors to activate endogenous metabolic pathways [116], and the differential subcellular translocation of recombinant proteins harboring distinct targeting signals after polyprotein processing [117]. NIa/polyproteins have also been used to co-express defense proteins in plants, such as the coat proteins of different plant viruses [118,119] and a number of antimicrobial peptides and proteins [120,121].

An alternative to supplying exogenous protease functions consists of integrating a cleavable, plant protease-sensitive linker between the proteins to express and targeting the resulting polyprotein precursor to the endoplasmic reticulum for processing by endogenous proteases along the cell secretory pathway [122] (Figure 1b). The antidigestive protease inhibitors cowpea trypsin inhibitor and rice cystatin I have been produced in Arabidopsis using this approach by expressing a single translational fusion product for the two proteins separated by a cleavable peptide from the plant metallothionein-like protein, PsMTa [101]. In a similar way, a polyprotein precursor including the antimicrobial defensins DmAMP1 and RsAFP2 linked by a seed protein cleavable sequence from Impatiens balsamina was shown to drive the accumulation of free and active forms of the two defensins in the extracellular milieu of transgenic Arabidopsis leaves [123,124,125]. These findings, along with a recently described approach involving ubiquitin-derived linker sequences for processing by endogenous deubiquitinating proteases [126], underline the potential of cleavable polyproteins sensitive to host endogenous proteases for the coordinate expression and release of multiple recombinant proteins, including pesticidal proteins, in plants.

IRES motifs direct ribosomes to initiate translation in a cap-independent manner, at internal positions within polycistronic transcripts [129] (Figure 2a). A number of IRES motifs from plant [130,131] and animal [132] viruses have successfully been used to direct the expression of multiple recombinant proteins in plants and plant cells. Bicistronic constructs have been used, for instance, to engineer nematode pest resistance in tobacco by the co-expression of cowpea trypsin inhibitor and rice cystatin I [133], or to facilitate the implementation of abiotic stress tolerance in cell cultures and transgenic plants by linking the detection of a marker gene with the expression of a stress-related protein [134,135]. However, IRES-based approaches present practical limitations associated with the highly variable efficacy of IRES motifs in planta, their taxonomic origin and the plant tissue selected for protein expression [131,133,134,136,137]. Perhaps the most prominent limitation preventing the widespread use of IRES motifs in plant biotechnology is their relative inefficiency. Internally initiated translation is generally low compared to cap-dependent translation [133] and is, therefore, hardly compatible with plant protection strategies requiring sufficient and comparable amounts of different pesticidal proteins.

A straightforward way to ensure a balanced translation of polycistron-encoded proteins may be to use the 20-amino acids 2A peptide of foot-and-mouth disease virus [51,138]. This peptide mediates a “ribosomal skip” during viral transcript translation that results in a co-translational, non-proteolytic dissociation of the encoded proteins [139]. This unique biochemical process prevents a peptide bond to form between the last two amino acids of the viral peptide and causes the nascent polypeptide to dissociate from the ribosomal translational complex while allowing mRNA translation to continue [139,140]. Polycistronic protein constructs involving the 2A peptide may be engineered in vitro by insertion of this peptide between the recombinant proteins to be expressed (Figure 2b). The correct processing of 2A peptides in plants was initially reported for transgenic tobacco lines expressing combinations of the model reporter proteins β-glucuronidase, chloramphenicol acetyltransferase and green fluorescent protein [141,142]. Several studies then reported the successful co-expression of useful proteins by this approach, for applications as diverse as the heterologous expression of vaccines and antibodies [143,144,145], the engineering of carotenoid biosynthetic pathways [137,146], storage protein enrichment to increase the nutritional value of plant foods [147], and the expression of stress or pesticidal proteins conferring abiotic stress tolerance [148] or pest resistance [125,149]. The presence of several non-cognate amino acids at the C-terminus of the upstream protein(s) following mRNA translation still represents a practical limitation to the 2A peptide technology [128], but strategies have been devised in recent years to remove this extension, relying on the addition of cleavage sites between the C-terminus and the viral peptide [150,151]. A number of factors should make the 2A peptide expression strategy particularly useful for the implementation of multiple protein resistance traits in plants [138,152]. The most obvious are the possibility of integrating several recombinant proteins and 2A peptide connections within the same construct [153,154], and the relatively efficient self-processing activity of 2A peptide-containing polyproteins in plant systems [142]. Other factors are the greater efficiency of this expression system compared to the IRES motif strategy [137], and the possible targeting of recombinant proteins to different subcellular locations after their co-translational release from the ribosomal complex [153,154,155].