Horizontal or lateral gene transfer (HGT) is the stable and heritable acquisition by an organism, of genetic material that did not originate from a parental donor (Keese, 2008). Any DNA sequence, including endogenous sequences or foreign DNA introduced into a genetically modified (GM) organism, has the potential to undergo HGT. This potential is only fulfilled when the genetic material stably integrates into the genome of the recipient and is then transmitted to its offspring (Hülter and Wackernagel, 2008; Brigulla and Wackernagel, 2010; Huang, 2013). HGT can benefit the recipient by enabling the acquisition of a beneficial pre-existing trait from another organism, regardless of phylogenetic distance. It thereby, like vertical gene transfer, accelerates evolution (Fournier et al., 2015).

In Australia, the Gene Technology Regulator (the Regulator) receives applications for the intentional environmental release of GM plants and, as part of the assessment process of these applications, must consider the risks to human and animal health and to the environment from gene technology posed by the proposed activities. GM plants may have genetic elements sourced from other organisms imparting desired traits, e.g., increased nutritional value; drought, pest and disease resistance; or increased productivity. While gene transfer is most likely to occur to sexually compatible relatives through vertical gene transfer, the likelihood of gene transfer to non-sexually compatible organisms via HGT also needs to be considered as part of the risk assessment.

Similarly, in Europe, the Commission Regulation (EU) 503/2013 of 3 April 2013 (on applications for authorisation of genetically modified food and feed) states that “The applicant shall assess the probability of horizontal gene transfer from the product to humans, animals and microorganisms and any potential associated risk when intact and functional nucleic acid(s) remains in the genetically modified food and feed.” In the United Kingdom, the independent Advisory Committee of Releases into the Environment (ACRE) also considers HGT in their assessment for application for the release of genetically modified organisms (GMOs) (ACRE, 2013). Other regulatory authorities may also need to consider HGT before issuing an authorisation or licence. In this review, we discuss the recent advances in detecting HGT events and present updated evidence of the likelihood, factors, barriers and pathways for HGT to take place from GM plants to a variety of other organisms.

In Australia, Regulations 9A and 10 of the Gene Technology Regulations 2001 (OGTR, 2020) specify the risks and matters that must be considered in the risk assessment for an environmental release of GMOs (Figure 1). Considerations relating to gene flow are 1) the potential for spread and persistence of a GMO’s genetic material in the environment and 2) the potential of the GMO to transfer genetic material to another organism. The risk assessment seeks to evaluate the level of risk from the activities with a GMO if HGT from the GMO into other organisms was successful, compared to the status quo.

In Australia, the Risk Analysis Framework (RAF) (OGTR, 2013), in accordance with Australia’s Gene Technology Act 2000 and Regulations, outlines the approach that the Regulator takes to conduct the risk assessment of proposed activities with a GMO. Such activities include the proposed environmental release of the GMO. The RAF describes the three essential components needed for a scenario (set of circumstances) that might give rise to harm as a result of activities conducted with the GMO. The three components are: 1) a source of potential harm, which may be a new or altered property/trait of the GMO; 2) a potential harm to people or the environment; and 3) a plausible causal linkage between components 1) and 2). If a plausible causal linkage or potential harm cannot be described, then the source of potential harm poses no risk (OGTR, 2013).

While HGT per se is not considered a risk, it fits into the pathway component of risk as it can link the introduced or modified DNA to a potential harm to people or the environment. Therefore, the likelihood of HGT occurring determines the potential for any harm. As the number of steps in a pathway leading to harm increases, the likelihood of harm occurring decreases. The Regulator considers the likelihood of occurrence of HGT and the severity of adverse outcomes if HGT was successful. If the level of risk is increased compared to the status quo, then the Regulator may include specific risk management measures to interrupt steps in a pathway and reduce the likelihood of harm occurring or refuse to approve the proposed intentional environmental release. For example, as a measure to limit the likelihood of gene flow by vertical gene transfer from GM plants, an exclusion zone can be imposed where sexually compatible plants are not permitted to be grown.

It should also be noted that risk analysis on a proposal for the release of GM plants occurs in the context of the receiving environment. For example, if a GM (transgenic) DNA sequence was sourced from a ubiquitous bacteria or fungi, then this forms part of the context as the DNA sequence is already in the environment. If HGT from the bacteria or fungi is more likely than HGT from the proposed activities with the GM plant, then the potential of HGT resulting in harm from the release of the GM plant can be no greater than the risk from the parent organism. Similarly, the likelihood of harm occurring as a result of the GM plant release is compared to that of harm occurring in the absence of the genetic modification, i.e., the non-GM plant.

Until recently, comparative analysis for detection of HGT events relied on very limited databases of manually annotated genes (Dupont and Cox, 2017). However, the expansion of next generation sequencing has allowed an ever-increasing number of whole genomes from a vast range of species to be readily available for multiple genome comparisons. As a result, detection of HGT events for all annotated genes can now be performed bioinformatically. By using this approach, HGT can be inferred by either the parametric or phylogenetic methods (Ravenhall et al., 2015). The parametric methods look for sections that significantly differ from the average composition; these include GC content and codon usage (Ravenhall et al., 2015). The phylogenetic methods compare the evolutionary histories of a gene of interest and its homologues across multiple species, to identify conflicting phylogenies (Ravenhall et al., 2015; Soucy et al., 2015; Wybouw et al., 2018).

Advances in computational algorithms have also helped in the identification of additional HGT events from genomes that were previously analysed. For example, Wybouw et al. (2018) using refined bioinformatics parameters, identified 25 additional horizontally transferred genes in the spider mite (Tetranychus urticae) genome, seven years after HGT was first analysed in this species (Grbic et al., 2011). In the grass species, Alloteropsis semialata, initially two genes were identified to be horizontally transferred. Using next generation genome resequencing and strict phylogenetic comparisons amongst 146 other grass species, Dunning et al. (2019) were able to detect 57 additional horizontally transferred genes, some of which are associated with disease resistance and abiotic stress response loci (Dunning et al., 2019). In another example, ToxA, a fungal virulence protein, which is associated with diseases in wheat and barley was shown to reside on a horizontally transferred genomic cluster. The presence of this virulence protein provides a selective advantage to the fungus. Using long-read sequencing technologies, with the available genomes of fungal pathogens in the Pleosporales order, McDonald et al. (2019) were able to confirm the HGT origin of ToxA and define the boundaries of the transferred genomic cluster.

Other examples of HGT events that have been identified between species, using both bioinformatics and experimental methods, include: HGT of genes originating from bacteria, fungi, and plants to bdelloid rotifers (Gladyshev et al., 2008); HGT of carotenoid genes from fungi to pea aphids, causing a red colour polymorphism that provides selective advantage to avoid parasitism compared to green aphids (Losey et al., 1997; Moran and Jarvik, 2010); HGT of a photoreceptor gene from hornworts to ferns, allowing ferns to thrive under low-light conditions (Li et al., 2014); HGT from bacteria and fungi to silkworms of genes thought to confer disease resistance, nutrient and energy metabolism and toxin degradation (Zhu et al., 2011); HGT of an antifreeze protein gene between fish living in icy seawater (Graham et al., 2008), which the authors propose to naturally occur by sperm-mediated HGT during external fertilisation, where the sperm “absorbs” “naked” DNA from the environment; HGT of mitochondrial DNA from the parasite Trypanosoma cruzi to humans (Hecht et al., 2010); and HGT from humans to the strictly human pathogen Neisseria gonorrhoeae (Anderson and Seifert, 2011). Although far less frequent than HGT between bacteria, HGT from bacteria to eukaryotes has also been described, including the transfer of genes from the bacterium Wolbachia to insects and nematodes (Dunning Hotopp et al., 2007; Nikoh et al., 2008), from bacteria and fungi to plant parasitic nematodes (Noon and Baum, 2016) and from Agrobacterium to plants (Matveeva and Otten, 2019). It is to note that the aforementioned HGT examples have non-neutral or advantageous impacts, see section 5 below. A horizontally transferred gene is unlikely to be maintained in a population if it has a negative impact in the recipient.

Given the automated nature of genome data collection and gene prediction annotation in the contemporary setting, it is impractical to manually validate all genes within a genome. Therefore, concerns relating to whether the inferred HGT events in eukaryotes are statistically supported have been raised (Dupont and Cox, 2017). In addition, the short reads produced by many modern sequencing platforms raises concerns about microorganism contamination, especially involving the putative HGT between these microorganisms (Boothby and Goldstein, 2016; Wickell and Li, 2019). Even though new techniques allow HGT in eukaryotic genomes to be detected with greater frequency than a few years ago, HGT in complex eukaryotes is relatively rare when compared with the observed rates in simpler organisms such as viruses or prokaryotes (Andersson, 2005; Keese, 2008; Vogan and Higgs, 2011; Crisp et al., 2015; Qiu et al., 2016; Sieber et al., 2017).

Direct or vector-mediated pathways can facilitate HGT. In direct pathways, the recipient organism either “takes up” DNA from another living cell or uptakes “free” or “naked” DNA present in the environment. Vector-mediated pathways are those where DNA is first taken up from the donor by an intermediate recipient that acts as a vector, such as a virus or prokaryote, and then passed on to a different recipient.

There are a number of factors that affect the likelihood of the introduced or modified DNA sequences in GM plants being successfully horizontally transferred and then retained in the final recipient. These include: the proportion of introduced DNA in the GM plant as a source for HGT; the availability and integrity of the introduced DNA sequence; the physical proximity of introduced DNA and a potential recipient organism; whether the recipient organism has a dedicated mechanism for uptake of DNA; whether homologous DNA sequences are present in the recipient organism; whether the donor and recipient are genetically compatible; and whether the horizontally transferred GM DNA sequence gives an advantage to the recipient organism. These factors will be discussed in the following sections.

As previously discussed, while the acquisition of a new gene by HGT is not considered harm per se, it has the potential to lead to genetic variation within a population, and thus, its impact on driving the evolutionary function of organisms has been considered (Keese, 2008; Boto, 2010). With respect to GM plants, for HGT-induced harm, the acquisition of the genetic material must result in a non-neutral change for the recipient, be maintained in the population and result in an adverse outcome to humans, animals and/or the environment (Keese, 2008).

Occasionally, the function of the transferred genes could strongly affect the severity of the adverse consequences or the likelihood of a HGT event (Keese, 2008). When a population is under strong selective pressures or environmental stresses, HGT can be stimulated. In the recipient, the transferred gene can either confer a detrimental, neutral, or advantageous trait. If this novel trait is advantageous, the recipient can overcome its pressures and stresses, outcompete its neighbours or adapt to a new ecological niche (van Elsas et al., 2003; Keeling, 2009; Raz and Tannenbaum, 2010; Vogan and Higgs, 2011; de la Casa-Esperón, 2012). For example, under intensive agricultural production, the coffee berry borer beetle, Hypothenemus hampei, horizontally acquired a mannanase gene from bacteria that helped it adapt to enzymatically digest the polysaccharides of coffee beans, converting it into an invasive pest (Acuña et al., 2012). Similarly, the whitefly, Bemisia tabaci, horizontally acquired a phenolic glucoside malonyltransferase gene from plants allowing it to neutralise the plant-produced phenolic glycosides that would otherwise kill the whitefly after herbivory. This to our knowledge is the first known example of a HGT event between a plant and an animal and is thought to have occurred ∼86 million years ago (Xia et al., 2021). There are also instances that illustrate how novel genes acquired by HGT contributed to parasite adaptation to a new host in different organisms, for example: HGT of cellulase genes, allowing cellulase activity, from several microbial donors to nematodes, which enhances their parasitism and pathogenicity of plants (Danchin et al., 2010); HGT of genes associated with disease resistance and abiotic stress response in spider mites (Grbic et al., 2011; Wybouw et al., 2018); and HGT of fungal genes involved in the metabolism of plant sugars and genes involved in the breaking down of cell walls to increase the parasitism of oomycetes (Richards et al., 2011).

Even when an organism horizontally acquires a gene that leads to a selective advantage, the advantage might not have a short-term impact on the recipient’s ecology, and changes might only be significant when considered in an evolutionary timescale (Hiltunen et al., 2017). Similarly, the ecological benefits of an adaptation acquired by a sporadic HGT event could dissipate over time (Hiltunen et al., 2017). In these cases, the impacts to the ecosystem of a casual HGT event are difficult to assess, and risk assessments cannot rely on considering just HGT frequency, as this is not a good prognostic tool for long term effects of HGT (Pettersen et al., 2005).

Many adverse effects owing to the potential of HGT of the transgene from GM plants to other organisms, including humans, are gene dependent. These effects include their potential role in allergenicity, pathogenicity, virulence, toxicity and other environmental effects. Therefore, these potential effects should be evaluated on a case-by-case basis in the context of the proposed activities in the risk assessment of each GM plant if occurrence of HGT is considered more likely than when dealing with the non-GM plant.

For example, one of the most common concerns regarding GM plant safety is the potential environmental and health consequences of HGT with regards to prokaryotic-derived antibiotic resistance genes, which are predominantly used as markers to select for the transformation event. Should HGT and subsequent integration occur to the gut microflora of consumers, including humans, the concern relates to the proliferation of antibiotic resistant strains of harmful bacteria which would be harder to control (Rizzi et al., 2012). These concerns have been previously assessed (see EFSA (2017) and references within Woegerbauer et al. (2015)) and there is no account of such an event occurring from a GM plant source under field conditions (Demanèche et al., 2008; Pilate et al., 2016; EFSA et al., 2017; Tsatsakis et al., 2017; Hily et al., 2018). In addition, the presence of these resistance genes has not significantly increased antibiotic resistance in the clinical setting (Breyer et al., 2014). Furthermore, natural antibiotic and herbicide resistance genes are found in widely dispersed soil microorganisms, often on mobile genetic elements (Pontiroli et al., 2007) and the HGT of these naturally occurring genes has previously been described by Domingues et al. (2012). Therefore, these naturally occurring microorganisms, as well as the vast pool of available antibiotic resistance genes already naturally present in the intestinal microflora in the human GIT, are far more likely to be the HGT source for these resistance genes than GM plants (Pontiroli et al., 2007; Tothova et al., 2010; Huddleston, 2014).

That stated, the evaluation of antibiotic resistance genes in GM plants needs to be considered on a case-by-case basis and regulators at the European Food Safety Authority (EFSA) have adopted the classification of these genes into three risk groups. These groups class antibiotic resistance genes based on their abundance in the environment and their significance to human and veterinary medicine, reviewed in De Santis et al. (2018). Regulators in Europe have also issued a directive to phase out antibiotic resistance genes used in GM plants that have adverse effects on human health and the environment [Directive 2001/18/EC of the European Parliament and of the Council (on the deliberate release into the environment of genetically modified organisms)] (Garcia, 2006). As such, biotechnologists are encouraged to develop GM plants without the use of antibiotic resistance gene markers (Breyer et al., 2014). In Australia, the risk assessment process considers the background presence of the gene(s) used in the genetic modification “Antibiotic resistance marker genes commonly used in the selection process for generating GM plants are derived from soil bacteria abundant in the environment. Therefore, exposure to an antibiotic resistance gene, or to the protein encoded by such a gene, derived from a GMO, may or may not be significant against the naturally occurring background” (OGTR, 2013).

HGT is most prominent in prokaryotes, as many lack sexual recombination and thus employ HGT as a mechanism for adaptation to the environment. For example, the beneficial acquisition of antibiotic resistance through HGT in a clinical setting, and several other cases of HGT have been researched and documented in this context. Despite the significant role that HGT has played in the evolution of eukaryotic genomes, nuclear HGT events between multicellular eukaryotes are considered scarce, when compared to those between prokaryotes and in either direction between prokaryotes and eukaryotes (Richardson and Palmer, 2007; Keeling and Palmer, 2008; Keeling, 2009; Bock, 2010; Huang, 2013; Gao et al., 2014; Schönknecht et al., 2014). Specifically, scarcity of HGT between plants can be attributed to the need of a vector to facilitate the transfer (Mahelka et al., 2017).

Overall, the frequency of HGT for all organisms, including viruses and bacteria, is orders of magnitude lower than gene transfer by sexual or asexual reproduction. This is due to HGT needing to overcome numerous barriers, such as those related to the transfer, incorporation, and transmission of the DNA between organisms. In eukaryotes, additional barriers are needed to be overcome, where the DNA may first need to be transferred from the somatic to germ cell line and then be transferred to the recipient’s offspring.

The advances in whole genome sequencing and comparative genomics demonstrates that, although historical, HGT events in eukaryotic organisms may have previously been underestimated. (Bock, 2010; Crisp et al., 2015; Drezen et al., 2017; Quispe-Huamanquispe et al., 2017; Sieber et al., 2017; Matveeva and Otten, 2019). However, by the publication date of this manuscript, there have been no reports of adverse impacts on human health or environmental safety as a direct or indirect result of HGT from GM plants. Moreover, in the Australian context, GM plants approved for environmental release often contain genes and regulatory sequences that originate from naturally occurring organisms that are already present in the environment. Therefore, the potential for unintended adverse effects of HGT of the inserted genetic material is unlikely to be greater than those by its naturally occurring genetic counterparts. As such, the potential for adverse effects to human and animal health and to the environment as a result of HGT from GM plants already authorised for environmental release in Australia are highly unlikely. However, it is worth considering that, by exchanging the host of the introduced genetic material, a closer physical association to a potential recipient might be enabled, potentially increasing the likelihood for HGT. Furthermore, with recent advances in genome editing, non-food plants, such as N. benthamiana, are likely to be genetically modified with DNA sequences encoding components for production of pharmaceuticals and vaccines (Bally et al., 2018). The DNA sequences in such GM plants may be novel or synthetic, and therefore are unlikely to already be present in the environment. This would pose new challenges to gene technology regulators in conducting their risk analysis, as a direct comparator is not immediately apparent. The Australian approach provided by the Regulator’s Risk Analysis Framework (OGTR, 2013) would still allow the risk assessment of any novel or synthetic DNA sequence in the GM plant to be conducted. Similarly, the risks associated from gene-edited plants would also follow the above processes, noting that in Australia, organisms modified via Site Directed Nucleases (SDN) without guide RNAs (SDN-1) are organisms that are not genetically modified organisms under Schedule 1 of the Gene Technology Regulations 2001 for regulatory purposes following the changes made to the Australian Regulations in 2019 (OGTR, 2020; O’Sullivan et al., 2022).