2.1. Natural auxotrophic markers
Natural auxotrophy is a simple and widespread UTBA technique commonly applied to prevent the releasing and proliferation of GMYs outside the production facility. In general terms, natural auxotrophy can be defined as a nutritional deficiency induced by mutated genes resulting in an GMY strain that depends on the addition of the nutrient on its growth media [
39]. The yeast
S. cerevisiae has been used as a model for auxotrophic markers due to the facility of inducing deletion and/or point mutations in different genes associated with nutrient metabolism [
39]. In this sense, genes linked to the metabolism of amino acids like L-histidine, L-leucine, L-tryptophan, and L-methionine [
39], and nitrogen bases (e.g., adenine and uracil) [
40] have been used as auxotrophic markers for decades.
Considering nitrogen base auxotrophies, the allele
ade2-1, which contains a nonsense mutation (Glu64STOP) [
41], is widely found in different laboratorial
S. cerevisiae strains and confer adenine dependence. A major phenotypic characteristic of
ade2-1 is the development of a red ochre color due to the intracellular accumulation of oxidized adenine-associated metabolic intermediates [
40,
42], a phenotype that can be useful for red-white colony screening [
43], redox biology [
44] and drug discovery [
40].
The depletion of adenine reserves in yeast cells increase trehalose synthesis and lead to cell cycle arrest, recapitulating the protective effects observed for desiccation stress tolerance. As a consequence, the adenine-deficient cells become viable for a longer period of time and evade biocontainment [
42]. Complementally, it was observed that adenine auxotrophy increases mutagenesis rate in yeast cells [
45], a condition that is observed for other auxotrophic markers like
leu2 and/or
lys2 alleles. Yeast cells carrying
leu2 and/or
lys2 alleles, when subjected to L-lysine or L-leucine starvation, display an increase in the number of respiratory deficient cells (
rho- cells) due to the accumulation of mutations in mitochondrial genome, a condition termed “adaptive mutation” [
46]. Additionaly, adaptive mutations are linked to auxotrophy marker reversion to the wild-type state, as observed for
his4 marker in yeast under selective pressure [
47].
In order to avoid the reversion of auxotrophic markers, a series of deletion gene markers have been generated for
S. cerevisiae [
48], as well as for other non-
Saccharomyces species, like
Kluyveromyces lactis [
49]. However, it has been observed in some non-
Saccharomyces species that auxotrophic gene deletion leads to a bradytroph/leak auxotroph phenotype, like those observed for
PHA2 in
Pichia pastoris, which is linked to L-phenylalanine biosynthesis [
50]. In this case, the auxotrophic
phe- cells are able to survive under L-phenylalanine starvation due to alternative L-phenylalanine biosynthesis mechanisms [
50].
Considering the impact of auxotrophy in yeast metabolism, the effect of adaptive mutation in the selection of prototrophic cells, and bradytrophy due to the presence of poorly described biochemical pathways in non-
Saccharomyces species, natural auxotrophy
per se has strong limitations regarding its use as a biocontainment strategy. In addition, the presence of other microorganisms in the environment and/or production facility can provide the auxotrophy-dependent nutrients for GMY mutants, bypassing the auxotrophy requirement [
51].
A solution to overcome the natural auxotrophies limitations is the use of synthetic amino acids and nitrogen bases not available outside the production facility and can not be biologically supplied. This approach, termed “xenobiology” or “synthetic auxotrophies”, make use of top-down synthetic biology techniques to engineering GMYs and create new biocontainment strategies [
35]. Another approach to circumvent the limitations of natural auxotrophies is the use of a so-called “conditional gene essentiality”, which employs promoter engineering to modify genes linked to nutrient metabolism and strictly regulate their expression by using synthetic molecules and/or orthogonal RNA polymerases [
52].
2.2. Xenobiology and synthetic auxotrophies
Xenobiology focuses on the development of synthetic biological devices and systems that utilize non-canonical amino acids (ncAAs), nucleic acids with non-standard sugar backbone (xeno-nucleic acids or XNA) and non-natural nitrogen base pairs for different purposes, including the design of synthetic metabolic processes (neo-metabolism) [
53], and biocontainment [
52]. Many of xenobiology devices are designed by principle using orthogonality where synthetic components (e.g., proteins, RNAs, DNAs, and small molecules) are engineered for a purposed function and will not interfere with the natural biochemistry of a host cell [
53,
54]. The orthogonality also ensures that these components will not be used by natural biological systems, making it useful as a biocontainment strategy [
55,
56].
Biocontainment-based xenobiology is mostly centered on the use of different ncAAs for protein synthesis for both GEM and GMY. As example, the application of genetic code expansion (GCE) or orthogonal translation systems (OTSs) techniques [
57,
58] has been used with success in
Escherichia coli biocontainment [
56].
In bacteria, GCE/OTS relies on a orthogonal pyrrolysyl-tRNA synthetase/tRNA
PylCUA (PylRS/tRNA
PylCUA) pair derived from
Methanosarcina barkeri, M. mazei or
Methanocaldococcus jannaschii to incorporate ncAAs in proteins [
59,
60], in the reassignment of the UAG amber codon, a rare stop codon in both
E. coli and
S. cerevisiae [
61], and the deletion of the release factor 1 (RF1). Considering GCE for
E. coli biocontainment, the genome of this bacterium was refactored by the introduction of a reassigned UAG codon into 22 essential genes together with a
Methanocaldococcus jannaschii PylRS that is able to incorporate L-phenylalanine derivatives into tRNA
PylCUA [
56], generating a synthetic auxotrophy. Data from this work indicated that the synthetic
E. coli auxotroph cells were dependent on the addition of L-phenylalanine derivatives; moreover, these synthetic auxotrophs have undetectable escape frequencies in both solid and liquid culture media [
56]. Additional work related to the creation of
E. coli synthetic auxotroph strains have been made [
38], pointing to the feasibility of this biocontainment technique for bacteria.
The generation of yeast ncAAS-dependent synthetic auxotrophs was achieved with very limited success [
58,
62]. Yeast are naturally able to incorporate ncAAs into proteins [
61,
63] and a GCE/OTS technique for
S. cerevisiae was developed by Chin et al. [
62]. In this work, the authors engineering orthogonal codons, anticodons and tRNA synthetase, including an
E. coli tyrosine-tRNA synthetase (TyrRS) and a amber suppressor tRNA
TyrCUA, to generate a library of TyrRS mutants that pairs only with ncAAs [
62]. Once this TyrRS library was transformed into a
S. cerevisiae strain, five different ncAAs were incorporated into human superoxide dismutase 1 protein (hSOD1) [
62].
Other relevant works related to the development of GCE/OTS for
S. cerevisiae and
Pichia pastoris have been made by using the
E. coli TyrRS/tRNA
TyrCUA or leucyl- (Leu)RS/tRNA
Leu5CUA pairs for ncAA incorporation into proteins [
60,
64,
65,
66,
67,
68,
69,
70,
71,
72]. However, the use of GCE/OTS in yeast have some major challenges, including the low expression of tRNA
PylCUA in yeast due to the absence of intragenic promoter sequences A- and B-boxes [
58,
60]; moreover, the eukaryotic release factor 1 (eRF1), codified in yeast by the
SUP45 essential gene [
73], compose the Sup45p-Sup35p complex that is necessary to end translation by binding into all three stop codons in yeast cells [
74]. Comparatively, the
E. coli RF1, which recognizes the UAG/UAA codons, can be deleted without major physiological impacts into the cell due to the functional superimposition with the release factor 2 (RF2) [
75].
In order to circumvent the limitations inherently associated with the implementation of a GCE/OST-based biocontainment in GMY, different approaches were applied, like (i) prospecting new archeal PylRS with higher ncAA incorporation efficiency in
S. cerevisiae [
76], (ii) selecting yeast strains with specific mutations (e.g.,
yil014c-aΔ and
alo1Δ) for increasing ncAAs incorporation into tRNA
PylCUA [
77], (iii) improving
E. coli TyrRS and LeuRS with enhanced ncAAs polyspecificity and efficiency by using random mutagenesis and directed selection [
78], (iv) the use of an OTS based on recognition of quadruplet codons by an engineered orthogonal ribosome that allows to expand the genetic code to 256 codons; this quadruplet codon-base OTS have been implemented with more or less success in
E. coli [
58,
79,
80] and mammalian cells [
81], and (v), the use of synthetic/unnatural nitrogen base pairs (UBPs) to expand the genetic code and increase the repertoire of ncAAs that can be used for protein synthesis in
E. coli [
82].
There are different approaches when considering the use of UPBs and XNAs for xenobiology and GMY/GEM biocontainment [
83]. In fact, it is expected that XNA technology could be efficient for GMY/GEM biocontainment since the building blocks of XNAs (
e.g., nucleobases, sugar moieties and phosphate-modified groups) can not be find in natural environments and the GMY/GEM cells should be able to incorporate these building blocks into new XNA polymers by the usage of specialized polymerization enzymes and transmembrane proteins able to uptake this precursors [
84,
85]. Thus, UBPs and XNAs can be like a “genetic firewall” [
84], where HGT events could be avoided to the restrained aspects of XNA technology.
The developments on XNA technologies follows two major mainstreams: (i) the use of unnatural nucleobases, sugar moieties and phosphate-modified groups (XNA substrates) to incorporate into XNAs or hybrid XNA/DNA/RNA polymers by canonical DNA and/or RNA polymerases [
55,
86] and (ii) design new DNA and/or RNA enzymes (XNAzymes) able to metabolize XNA polymers and/or substrates with an “alien” chemistry, like threose nucleic acid (TNA), cyclohexenyl nucleic acid (CeNA), arabino nucleic acid (ANA), 2’-fluoro-arabino nucleic acid (FANA), glycol nucleic acid (GNA), and locked nucleic acid (LNA) [
87,
88,
89,
90,
91,
92,
93,
94]. In all cases, the XNAs should display orthogonality
in vivo with little or, preferentially, none interaction with the canonical components of DNA/RNA metabolism [
95]. Unfortunately, many different XNA technologies were not implemented
in vivo, which renders its usage for GMY/GEM biocontainment still far from technical and practical viewpoint [
96]. Finally, both GCE/OTS and XNA technologies were not tested in an open and uncontrolled environment, making its behavior unpredictable for real biocontainment applications [
26].
Another approach related to the induction of synthetic auxotrophies in
E. coli is based on the selection of essential proteins whose structure and activity is dependent on the presence of a small molecule ligands, like the so called “synthetic auxotrophs based on ligand-dependent essential genes” (SLiDE) technique [
97]. In this work, the authors were able to select five essential genes in
E. coli by applying protein engineering and saturation mutagenesis and generate proteins dependent on benzothiazole. The authors related a very low escape frequency (< 3 x 10
-11) under laboratory assays [
97]. Similar works on bacterial synthetic auxotrophs have been made, including phosphite-dependent
Synechococcus elongatus [
98] and
Pseudomonas putida [
99]. In
S. cerevisiae it has been identified a series of mutations in
CDC10 gene, which codify an essential septin protein that can be rescued in the presence of small molecules, like guanidinium ion [
100,
101]. Although the authors of this work have not applied the
CDC10 conditional mutants for biocontainment, the data may indicate new techniques based on chemical rescue for GMY biocontainment.
2.5. Nuclease-based kill switches
Kill switches are defined as gene circuits activated by specific environmental inputs, resulting in the expression of lethal genes that lead to cell death [
113]. Different kill switch designs have been used for GEM biocontainment, including gene circuits that respond to environment changes by activating/deactivating a type II toxin-antitoxin system CcdB/CcdA [
113]. Besides it, specific and unspecific nucleases, auxotrophies, as well as type I toxin-antitoxin pairs have been employed to engineer kill switches circuits for biocontainment purposes, many of them in a MTBA format (
Table 1) [
56,
114,
115,
116,
117,
118] in order to reduce the GEM escape frequency.
Considering nucleases for kill switch design, it has been shown its usefulness for bacteria biocontainment. For example, the expression of
EcoRI endonuclease-
EcoRI methylase pair combined with conditional gene essentiality in
E. coli considerably lowers the escape frequency [
117]. In this sense, combining
EcoRI and mf-Lon protease in a strictly regulated gene circuit allowed the development of the “Deadman” kill switch, which considerably lowered the escape frequency associated with a high genetic stability [
114]. Other non-specific nucleases, like
nucA from
Serratia marcescens [
119] or nuclease A from
Staphylococcus aureus have been employed in different bacterial biocontainment projects [
118].
In
S. cerevisiae, a kill switch based on the conditional expression of
S. marcescens nucA under the control of a glucose-repressed
ADH2/GAPDH hybrid promoter has been proposed [
120]. This work showed that under non repressible conditions, the
nucA was efficiently expressed leading to yeast cell death in both laboratorial and soil microcosm assays [
120]. Interestingly, type II restriction enzymes have been expressed in yeast to study DNA damage and repair mechanisms [
121,
122,
123]. Data from the expression of type II restriction enzymes in yeast cells point to a low survival and DNA damage tolerance, specially when blunt ends are induced by type II restriction enzymes (e.g.,
PvuII) [
123]; however, the use of type II restriction enzymes as part of a kill switch mechanism in yeast cells for biocontainment purposes was not described until now.
In addition to conventional exo- and endonucleases, the CRISPR-associated nucleases (Cas) that have been used in
E. coli to design kill switches. These kill switches gene circuits make use of Cas3 [
124] or Cas9 [
125] and were successfully used to biocontain
E. coli cells, both displaying an escape frequency ≤ 10
-8 cells. A major advantage of the Cas-based kill switches is the use of guide RNAS (gRNAs) to select specific DNA sequence targets or microorganism strains, allowing to selectively eliminate the target strain from microbiome [
125,
126]. Until now, no Cas-based kill switches were reported for GMY.
2.6. Kill switches based on type I and II toxin-antitoxin systems
Another kill switch extensively applied in GEM biocontainment is based on type I and II toxin-antitoxin (TA) systems. These kill switch-based TA systems efficiently lead to controlled cell death outside the production facility and can be easily engineered in GEM/GYM cells. What makes the TA systems so attractive for biocontainment applications is its pleiotropy, which targets different molecular mechanisms, like DNA and mRNA synthesis, cell cycle, nucleotide synthesis, and protein translation [
127,
128].
The prokaryotic TA system is classified into seven types named from I to VII [
130], which act in different cell mechanisms. For example, type I toxin-antitoxin system prevents toxin RNA translation through the binding of sRNA [
129], while type II toxin-antitoxin system is based on the interaction of an endoribonuclease and an inhibitor, forming a stable complex [
131]. By its turn, type III toxin-antitoxin system consists of the inhibition of the toxin protein by an antitoxin RNA [
132]. Type IV toxin-antitoxin system is composed of two proteins that do not form a complex; instead, the antitoxin acts as an antagonist of the toxin at its cellular targets [
133]. The type V toxin-antitoxin system differs from the others by having its antitoxin cleave the toxin mRNA [
134]. Type VI is composed of an antitoxin-based proteolytic adapter that degrades the toxin-based protein [
135], while type VII is composed of an antitoxin that enzymatically neutralizes the toxin by post-translational modification [
130].
Type I and II are historically employed for biocontainment gene circuit design, especially for
E. coli [
136]. For example, the type I
hok-sok pair have been applied for the design of conditional suicide of plasmid-containing
E. coli cells in phosphate- [
137] or tryptophan-limited [
136] environments. Additionally, type I TA systems have been used in synthetic biology projects related to GEM biocontainment [
118]. Similarly, type II TA systems have been applied into kill switch circuits based on the use of
ccdB-ccdA pair regulated by a bi-stable cI/Cro memory switch (‘essentializer’ circuit) and by a cold-inducible promoter (‘cryodeath’ circuit) [
113]. Finally, type II TA systems have been applied for the design of ‘plasmid addiction’ systems to prevent the lost of plasmids under production facility [
138]
In addition to GEM, the use of type I and I TA systems was proven to be effective for GMY biocontainment, like the expression of the
relE-relB system in
S. cerevisiae [
139]. When not being neutralized by the RelB antitoxin, the RelE toxin inhibits protein synthesis by cleaving mRNA that are being translated on the ribosome [
140]. When RelB is neutralizing RelE it displaces the toxin´’s C-terminal and forms the RelBE complex [
141]. To express this TA system in
S. cerevisiae two recombinant plasmids were constructed, one containing the
relE toxin gene under the control of galactose-induced
GAL1 promoter (pKP727) and another containing
relE regulated by
GAL1 promoter and the
relB antitoxin gene under the control of methionine-repressible
MET25 promoter (pKP1006). Yeast cells transformed with the plasmid pKP727 showed visible growth inhibitory effects induced by the
relE toxin; on the other hand, yeast transformed with the plasmid pKP1006 showed a higher growth rate. This data indicates that the toxic effects of
relE can be minimized by
relB antitoxin in yeast [
139]. Similar to the
relE-relB system, the ribonuclease-associated toxin Kid (killing determinant) and its antitoxin Kis (killing suppressor) [
142] have been used in
S. cerevisiae as a potential biocontainment system. For this purpose, an expression system containing the antitoxin Kis controlled by the
MET25 promoter and the Kid toxin controlled by the Cu
2+-induced
CUP1 promoter was cloned into the recombinant integrative plasmid pRS303. In the presence of both methionine and Cu
2+, the expression of Kis antitoxin is inhibited while the expression of Kid toxin is induced, leading to cell death [
143].
Another TA system already tested in yeast is based on the ε-ζ (
epsilon-
zeta) genes from the gram-positive bacteria
Streptococcus pyogenes [
144]. The ε-ζ genes are organized in an operon together with the ω (
omega) gene, which codify a repressor that modulates the transcription of the ζ toxin, ε antitoxin and its own [
144]. This TA system was expressed in a yeast two-hybrid system without including the ω sequence, through which the toxicity of the ζ protein was shown as well as the efficiency of the ε protein in antagonizing its toxin [
145]. Despite the efficiency of ε-ζ pair in to induce yeast cell death, its use as a biocontainment system for GMY was not attempted until now.
Finally, it was observed that the expression of the
chpK-chpI TA pair from
Lepstospira interrogans in
E. coli and
S. cerevisiae modulates their cell growth [
146,
147]. However, the application of
chpK-chpI TA pair as a biocontainment system was not reported until now. In addition to
S. cerevisiae, the expression of endoribonuclease-associated toxin
mazF from the
E. coli mazEF module [
148] in
Pichia pastoris result in cell death and can be useful as a counter-selectable marker for genetic modification [
149].