The genetic alphabet encodes all biological information, but it is limited to four letters that form two base pairs. To expand the alphabet, we developed synthetic nucleotides that pair to form an unnatural base pair (UBP), and used it as the basis of a semisynthetic organism (SSO) that stores increased information. However, the SSO grew poorly and lost the UBP under a variety of standard growth conditions. Here, using chemical and genetic approaches, we report the optimization of the SSO so that it is healthy, more autonomous, and able to store the increased information indefinitely. This SSO constitutes a stable form of semisynthetic life and lays the foundation for efforts to impart life with new forms and functions.

All natural organisms store genetic information in a four-letter, two-base-pair genetic alphabet. The expansion of the genetic alphabet with two synthetic unnatural nucleotides that selectively pair to form an unnatural base pair (UBP) would increase the information storage potential of DNA, and semisynthetic organisms (SSOs) that stably harbor this expanded alphabet would thereby have the potential to store and retrieve increased information. Toward this goal, we previously reported that Escherichia coli grown in the presence of the unnatural nucleoside triphosphates dNaMTP and d5SICSTP, and provided with the means to import them via expression of a plasmid-borne nucleoside triphosphate transporter, replicates DNA containing a single dNaM-d5SICS UBP. Although this represented an important proof-of-concept, the nascent SSO grew poorly and, more problematically, required growth under controlled conditions and even then was unable to indefinitely store the unnatural information, which is clearly a prerequisite for true semisynthetic life. Here, to fortify and vivify the nascent SSO, we engineered the transporter, used a more chemically optimized UBP, and harnessed the power of the bacterial immune response by using Cas9 to eliminate DNA that had lost the UBP. The optimized SSO grows robustly, constitutively imports the unnatural triphosphates, and is able to indefinitely retain multiple UBPs in virtually any sequence context. This SSO is thus a form of life that can stably store genetic information using a six-letter, three-base-pair alphabet.

The natural genetic alphabet is composed of four letters whose selective pairing to form two base pairs underlies the storage and retrieval of virtually all biological information. This alphabet is essentially conserved throughout nature, and has been since the last common ancestor of all life on Earth. Significant effort has been directed toward the development of an unnatural base pair (UBP), formed between two synthetic nucleotides, that functions alongside its natural counterparts (1⇓–3), which would represent a remarkable integration of a man-made, synthetic component into one of life’s most central processes. Moreover, semisynthetic organisms (SSOs) that stably harbor such a UBP in their DNA could store and potentially retrieve the increased information, and thereby lay the foundation for achieving the central goal of synthetic biology: the creation of new life forms and functions (4).

For over 15 years, we have sought to develop such a UBP (1), and these efforts eventually yielded a family of predominantly hydrophobic UBPs, with that formed between dNaM and d5SICS (dNaM-d5SICS; Fig. 1A) being a particularly promising example (5⇓–7). Despite lacking complementary hydrogen bonding, we demonstrated that the dNaM-d5SICS UBP is well replicated by a variety of DNA polymerases in vitro (7⇓⇓–10), and that this efficient replication is mediated by a unique mechanism that draws upon interbase hydrophobic and packing interactions (11, 12). These efforts then culminated in the first progress toward the creation of an SSO in 2014, when we reported that Escherichia coli grown in the presence of the corresponding unnatural nucleoside triphosphates (dNaMTP and d5SICSTP), and provided with a plasmid-encoded nucleoside triphosphate transporter (NTT2) from Phaeodactylum tricornutum (which we denote as PtNTT2) (13), is able to import the unnatural triphosphates and replicate a single dNaM-d5SICS UBP on a second plasmid (14).

UBPs and transporter optimization. (A) Chemical structure of the d NaM -d 5SICS and d NaM -d TPT3 UBPs with a natural dC–dG base pair included for comparison. (B) Comparison of fitness and [α- 32 P]-dATP uptake in DM1 and the various constructed strains: pCDF and inducible PtNTT2(1–575) (gray); pSC and constitutive PtNTT2(66–575) (blue); and chromosomally integrated and constitutive PtNTT2(66–575) (green). The promoters from which PtNTT2 is expressed are indicated by the labels next to their corresponding markers. Open triangles denote corresponding control strains without PtNTT2. The pCDF plasmids are in E. coli C41(DE3); pSC plasmids and integrants are in E. coli BL21(DE3). All PtNTT2 strains are non-codon-optimized for plasmid-based expression and codon-optimized for chromosomal expression unless otherwise indicated; r.d.u. is relative decay units, which corresponds to the total number of radioactive counts per minute normalized to the average OD 600 across a 1-h window of uptake, with the uptake in DM1 induced with 1,000 µM IPTG set to 1 ( Materials and Methods for additional details). Error bars represent SD of the mean, n = 3 cultures grown and assayed in parallel; the error bars on some data points are smaller than their marker.

Although this first SSO represented an important proof-of-concept, the generality of the expanded genetic alphabet remained unclear, as retention of the UBP was explored at only a single locus and in only a single sequence context. True expansion of the genetic alphabet requires the unrestricted retention of multiple UBPs at any loci and in any sequence context. Moreover, several limitations were already apparent with the nascent SSO (14). First, although expression of the nucleoside triphosphate transporter enabled E. coli to import dNaMTP and d5SICSTP, its expression caused the SSO to grow poorly, with doubling times twice that of the parental strain. Second, the UBP was not well retained during high-density liquid growth or during growth on solid media, presumably due to the secretion of phosphatases that degrade the unnatural triphosphates. Finally, even under optimal conditions, the nascent SSO was unable to retain the UBP with extended growth. Clearly, the ability to robustly grow under the standard repertoire of culture conditions and indefinitely retain the UBP is a prerequisite for true semisynthetic life. Here, we used genetic and chemical approaches to optimize different components of the SSO, ultimately resulting in a simplified and optimized SSO that grows robustly and is capable of the virtually unrestricted storage of increased information.

Results and Discussion

In our previously reported SSO (hereafter referred to as DM1), the transporter was expressed from a T7 promoter on a multicopy plasmid (pCDF-1b) in E. coli C41(DE3), and its toxicity mandated carefully controlled induction (14). In its native algal cell, PtNTT2’s N-terminal signal sequences direct its subcellular localization and are ultimately removed by proteolysis. However, in E. coli, they are likely retained, and could potentially contribute to the observed toxicity. Using the cellular uptake of [α-32P]-dATP as a measure of functional transporter expression and as a proxy for the uptake of the unnatural triphosphates, we found that removal of amino acids 1 to 65 and expression of the resulting N-terminally truncated variant PtNTT2(66-575) in E. coli C41(DE3) resulted in significantly lower toxicity, but also significantly reduced uptake (Fig. S1 A and B), possibly due to reduced expression (15). Expression of PtNTT2(66-575) in E. coli BL21(DE3) resulted in significant levels of [α-32P]-dATP uptake with little increase in toxicity relative to an empty vector control, but the higher level of T7 RNAP in this strain (16) was itself toxic (Fig. S1 A and C).

Fig. S1. The dATP uptake and growth of cells expressing PtNTT2 as a function of inducer (IPTG) concentration or promoter strength, strain background, and presence of N-terminal signal sequences. (A) Uptake of [α-32P]-dATP in strains with inducible PtNTT2. Error bars represent SD of the mean, and n = 3 cultures; r.d.u. is relative decay units, which corresponds to the total number of radioactive counts per minute normalized to the average OD 600 across a 1-h window of uptake, with the uptake of C41(DE3) pCDF-1b PtNTT2(1-575) (i.e., DM1) induced with 1,000 µM IPTG set to 1. Deletion of the N-terminal signal sequences drastically reduces uptake activity in C41(DE3), but activity can be restored with higher levels of expression in BL21(DE3). (B) Growth curves of C41(DE3) strains. Induction of PtNTT2(1-575) is toxic. (C) Growth curves of BL21(DE3) strains. Induction of T7 RNAP in BL21(DE3) is toxic (see empty vector traces), which masks the effect of deleting the N-terminal signal sequences of PtNTT2 on cell growth. (D) Uptake of [α-32P]-dATP in strains that constitutively express PtNTT2(66-575) from the indicated promoters. (E) Growth curves of plasmid-based and chromosomally integrated transporter strains. All PtNTT2 strains are non-codon-optimized for plasmid-based expression and codon-optimized for chromosomal expression, unless otherwise indicated. Strain YZ4 also contains a chromosomally integrated Cas9 gene.

We next explored constitutive expression of PtNTT2(66-575) from a low-copy plasmid or a chromosomal locus, which we anticipated would not only eliminate the need to use T7 RNAP but would also impart the SSO with greater autonomy (eliminating the need to induce transporter production), and, importantly, would result in more homogeneous transporter expression and triphosphate uptake across a population of cells, which we reasoned might improve UBP retention. We explored expression of PtNTT2(66-575) in E. coli BL21(DE3) with the promoters P lacI , P bla , and P lac from a pSC plasmid, and with P bla , P lac , P lacUV5 , P H207 , P λ , P tac , and P N25 from the chromosomal lacZYA locus (Dataset S1). We also explored the use of a codon-optimized variant of the truncated transporter (see Dataset S1). Although increasing expression of PtNTT2(66-575) (as measured by uptake of [α-32P]-dATP) was correlated with increasing doubling time, indicating that expression of PtNTT2(66-575) still exhibited some toxicity (uptake of [α-32P]-dATP is itself not toxic), each strain exhibited an improved ratio of uptake to fitness compared with DM1 (Fig. 1B). Strain YZ3, which expresses the codon-optimized, chromosomally integrated PtNTT2(66-575) from the P lacUV5 promoter, exhibited an optimal compromise of robust growth (<20% increased doubling time relative to the isogenic strain without the transporter), and [α-32P]-dATP uptake, and was thus selected for further characterization.

To determine whether the optimized transporter system of YZ3 facilitates high UBP retention, we constructed three plasmids that position the UBP within the 75-nt TK1 sequence (14) [with a local sequence context of d(A-NaM-T)]. These include two high-copy pUC19-derived plasmids, pUCX1 [referred to in previous work as pINF (14)] and pUCX2, as well as one low-copy pBR322-derived plasmid, pBRX2 (Fig. S2). In addition to allowing us to examine the effect of copy number on UBP retention, these plasmids position the UBP at proximal (pUCX1) and distal (pUCX2 and pBRX2) positions relative to the origin of replication, which we previously speculated might be important (14). Strains YZ3 and DM1 were transformed with pUCX1, pUCX2, or pBRX2 and directly cultured in liquid growth media supplemented with dNaMTP and d5SICSTP [and isopropyl β-D-1-thiogalactopyranoside (IPTG) for DM1 to induce the transporter], and growth and UBP retention were characterized at an OD 600 of ∼1 (Fig. 2A and Fig. S3A; see also Materials and Methods). Although DM1 showed variable levels of retention and reduced growth, especially with the high-copy plasmids, YZ3 showed uniformly high levels of UBP retention and robust growth with all three UBP-containing plasmids (Fig. 2B and Fig. S3A).

Fig. 2. UBP retention assay and the effects of transporter and UBP optimization. (A) Schematic representation of the biotin shift assay used to determine UBP retention. The plasmid DNA to be analyzed is first amplified in a PCR supplemented with the unnatural triphosphates, and the resulting products are then incubated with streptavidin and subjected to PAGE analysis. X = dNaM, or, in the PCR, its biotinylated analog dMMO2biotin. Y = d5SICS in the PCR, whereas Y = dTPT3 or d5SICS in the plasmid DNA, depending on the experimental conditions. Lane 1 is the product from the oligonucleotide analogous to that used to introduce the UBP during plasmid assembly, but with the UBP replaced by a natural base pair (negative control). This band serves as a marker for DNA that has lost the UBP. Lane 2 is the product from the synthetic oligonucleotide containing the UBP that was used for plasmid assembly. The shift of this band serves as a marker for the shift of DNA containing the UBP. Lane 3 is the product from the in vitro-assembled plasmid before SSO transformation (positive control). The unshifted band results from DNA that has lost the UBP during in vitro plasmid assembly. Lane 4 is the product from an in vivo replication experiment. (B) UBP retentions of plasmids pUCX1, pUCX2, and pBRX2 in strains DM1 and YZ3. Error bars represent SD of the mean, n = 4 transformations for pUCX1 and pUCX2, n = 3 for DM1 pBRX2, and n = 5 for YZ3 pBRX2. (C) UBP retentions of pUCX2 variants, wherein the UBP is flanked by all possible combinations of natural nucleotides (NXN, where N = d(G, C, A, or T) and X = dNaM), in strain YZ3 grown in media supplemented with either dNaMTP and d5SICSTP (gray bars) or dNaMTP and dTPT3TP (black bars).

Fig. S2. Plasmid maps. Promoters and terminators are denoted by white and gray features, respectively; pMB1* denotes the derivative of the pMB1 origin from pUC19, which contains a mutation that increases its copy number (41). Plasmids that contain a UBP are generally indicated with the TK1 sequence (orange), but, as described in Results and Discussion and indicated above, pUCX2 and pAIO variants with other UBP-containing sequences also position the UBP in the approximate locus shown with TK1 above; sgRNA (N) denotes the guide RNA that recognizes a natural substitution mutation of the UBP, with N being the nucleotide present in the guide RNA; sgRNA (∆) denotes the guide RNA that recognizes a single nucleotide deletion of the UBP; this sgRNA and its associated promoter and terminator (indicated by ‡) are only present in certain experiments. serT and gfp do not have promoters.

Fig. S3. Additional characterization of UBP propagation. (A) Growth curves for the experiments shown in Fig. 2B. YZ3 and DM1 (induced with 1 mM IPTG) were transformed with the indicated UBP-containing plasmids, or their corresponding fully natural controls, and grown in media containing dNaMTP and d5SICSTP. Each line represents one transformation and subsequent growth in liquid culture. The x axis represents time spent in liquid culture, excluding the 1 h of recovery following electroporation (Materials and Methods). Growth curves terminate at the OD 600 at which cells were collected for plasmid isolation and analysis of UBP retention. Staggering of the curves along the x axis for replicates within a given strain and plasmid combination is likely due to minor variability in transformation frequencies between transformations (and thus differences in the number of cells inoculated into each culture), whereas differences in slope between curves indicate differences in fitness. Growth of YZ3 is comparable between all three UBP-containing plasmids (and between each UBP-containing plasmid and its respective natural control), whereas growth of DM1 is impaired by the UBP-containing plasmids, especially for pUCX1 and pUCX2. (B) Retentions of gfp pUCX2 variants propagated in YZ3 by transformation, plating on solid media, isolation of single colonies, and subsequent inoculation and growth in liquid media, in comparison with retentions from plasmids propagated by transformation and growth of YZ3 in liquid media only. Cells were plated from the same transformations used in the experiments for Fig. 2C. Solid and liquid media both contained dNaMTP and dTPT3TP. Cells were harvested at OD 600 of ∼1. Five colonies were inoculated for each of the pUCX2 variants indicated, but some colonies failed to grow (indicated by a blank space in the table). Retentions for samples isolated from transformants grown solely in liquid media were assayed from the same samples shown in Fig. 2C, but were assayed and normalized to an oligonucleotide control in parallel with the plated transformant samples to facilitate comparisons in retention. See Materials and Methods for additional details regarding UBP retention normalization. For samples with near-zero shift, we cannot determine whether the UBP was completely lost in vivo or if the sample came from a colony that was transformed with a fully natural plasmid (some of which arises during plasmid assembly, specifically during the PCR used to generate the UBP-containing insert).

Given that no plasmid locus or copy number biases on UBP retention were observed in YZ3, we chose pUCX2 as a representative UBP-containing plasmid to explore the effect of local sequence context on UBP retention, and we constructed 16 pUCX2 variants in which the UBP was flanked by each possible combination of natural base pairs within a fragment of the GFP gene (see Dataset S1). Under the same growth conditions as above, we observed a wide range of UBP retentions, with some sequence contexts showing complete loss of the UBP (Fig. 2C). However, since the development of DM1 with the dNaM-d5SICS UBP, we have determined that ring contraction and sulfur derivatization of d5SICS, yielding the dNaM-dTPT3 UBP (Fig. 1A), results in more efficient replication in vitro (17). To explore the in vivo use of dNaM-dTPT3, we repeated the experiments with YZ3 and each of the 16 pUCX2 plasmids, but with growth in media supplemented with dNaMTP and dTPT3TP. UBP retentions were clearly higher with dNaM-dTPT3 than with dNaM-d5SICS (Fig. 2C).

Although dNaM-dTPT3 is clearly a more optimal UBP for the SSO than dNaM-d5SICS, its retention is still moderate to poor in some sequence contexts (Fig. 2C). Moreover, several sequences that show good retention in YZ3 cultured in liquid media show poor retention when growth includes culturing on solid media (Fig. S3B). To further increase UBP retention with even these challenging sequences and/or growth conditions, we sought to selectively eliminate plasmids that lose the UBP. In prokaryotes, the clustered regularly interspaced short palindromic repeats (CRISPR)-CRISPR-associated (Cas) system provides adaptive immunity against viruses and foreign plasmids (18⇓–20). In type II CRISPR-Cas systems, such as that from Streptococcus pyogenes (21, 22), the endonuclease Cas9 uses encoded RNAs [or their artificial mimics known as single-guide RNAs (sgRNAs) (23)] to introduce double-strand breaks in cDNA upstream of a 5′-NGG-3′ protospacer adjacent motif (PAM) (24), which then results in DNA degradation by exonucleases (25) (Fig. 3A). In vitro, we found that the presence of a UBP in the target DNA generally reduces Cas9-mediated cleavage relative to sequences that are fully complementary to the provided sgRNA (Fig. S4). We thus hypothesized that, within a cell, Cas9 programmed with sgRNA(s) complementary to natural sequences that arise from UBP loss would enforce retention in a population of plasmids (by eliminating those that lose the UBP), which we refer to as immunity to UBP loss. To test this, we used a p15A plasmid to construct pCas9, which expresses Cas9 via an IPTG-inducible lacO promoter, as well as an 18-nt sgRNA that is complementary to the TK1 sequence containing the most common dNaM-dTPT3 mutation (dT-dA) via the constitutive proK promoter (Figs. S2 and S5A). Initial exploratory experiments were carried out with strain YZ2 (a forerunner of YZ3 with slightly less optimal transporter performance, but which was available before strain YZ3; Fig. 1B and Fig. S1 D and E). Strain YZ2 was transformed with corresponding pairs of pCas9 and pUCX2 plasmids (i.e., the pUCX2 variant with the UBP embedded within the TK1 sequence such that loss of the UBP produces a sequence targeted by the sgRNA encoded on pCas9), grown to an OD 600 of ∼4, diluted 250-fold, and regrown to the same OD 600 . UBP retention in control experiments with a nontarget sgRNA dropped to 17% after the second outgrowth; in contrast, UBP retention in the presence of the correct sgRNA was 70% after the second outgrowth (Fig. 3B). Sequencing revealed that the majority of plasmids lacking a UBP when the correct sgRNA was provided contained a single nucleotide deletion in its place, which was not observed with the nontarget sgRNA (Fig. S5 B and C). With a pCas9 plasmid that expresses two sgRNAs, one targeting the most common substitution mutation and one targeting the single nucleotide deletion mutation (Fig. S5A), and the same growth and regrowth assay, loss of the UBP was undetectable (Fig. 3B).

Fig. 3. Cas9-based editing system. (A) Model for Cas9-mediated immunity to UBP loss. (B) UBP retention for pUCX2 TK1 is enhanced by targeting Cas9 cleavage to plasmids that have lost the UBP. The bars labeled hEGFP correspond to UBP retention with growth (black) and regrowth (gray) with an sgRNA that has a sequence taken from the hEGFP gene and thus does not target DNA containing the TK1 sequence (negative control). The bars labeled TK1-A or TK1-A/∆ correspond, respectively, to growth and regrowth with an sgRNA that targets the dNaM to dT mutation or two sgRNAs that individually target the dNaM to dT mutation and a single nucleotide deletion of the UBP.

Fig. S4. Effect of dNaM-dTPT3 on Cas9-mediated cleavage of DNA in vitro. Cas9-mediated in vitro cleavage was assessed for six DNA substrates, wherein the third nucleotide upstream of the PAM is one of the four natural nucleotides, dTPT3, or dNaM. The four sgRNAs that are complementary to each natural template were prepared by in vitro transcription with T7 RNAP. To account for differences in sgRNA activity and/or minor variations in preparation, a relative percent maximal cleavage for each sgRNA vs. all six DNA substrates is shown in parentheses (SI Materials and Methods). Values represent means ±1 SD (n = 3 technical replicates). In several cases, the presence of an unnatural nucleotide significantly reduced cleavage compared with DNA complementary to the sgRNA. These data suggest that Cas9 programmed with sgRNA(s) complementary to one or more of the natural sequences would preferentially cut DNA that had lost the UBP.

Fig. S5. (A) The sgRNA sequences used to enhance retention of the UBP in Fig. 3B (red denotes guide RNA nucleotides mismatched with the DNA target; the position of dTPT3 is denoted by Y and shown in green); hEGFP is a nontarget sgRNA. (B) Sanger sequencing chromatogram illustrating mutation of dNaM to dT in the absence of an sgRNA to target Cas9 nuclease activity. (C) Sanger sequencing chromatogram illustrating that, in the presence of Cas9 and a targeting sgRNA (TK1-A), sequences containing the dNaM to dT mutation are likely depleted by Cas9 cleavage, thus resulting in the accumulation of other mutations that are either not targeted by the TK1-A sgRNA (∆, a single nucleotide deletion of dNaM) or targeted by the TK1-A sgRNA, but less efficiently because of a mismatch between the guide and the mutation sequence (dNaM to dG). UBP-containing species were depleted before sequencing (SI Materials and Methods). The position of the mutation in the chromatograms shown in B and C is indicated by an arrow.

To more broadly explore Cas9-mediated immunity to UBP loss, we examined retention using 16 pUCX2 variants with sequences that flank the UBP with each possible combination of natural base pairs but also vary its position relative to the PAM, and vary which unnatural nucleotide is present in the strand recognized by the sgRNAs (Table S1). We also constructed a corresponding set of 16 pCas9 plasmids that express two sgRNAs, one targeting a substitution mutation and one targeting the single nucleotide deletion mutation, for each pUCX2 variant. Strain YZ2 carrying a pCas9 plasmid was transformed with its corresponding pUCX2 variant and grown in the presence of the unnatural triphosphates and IPTG (to induce Cas9), and UBP retention was assessed after cells reached an OD 600 of ∼1. As a control, the 16 pUCX2 plasmids were also propagated in YZ2 carrying a pCas9 plasmid with a nontarget sgRNA. For 4 of the 16 sequences explored, UBP loss was already minimal without immunity (nontarget sgRNA), but was undetectable with expression of the correct sgRNA (Fig. 4A). The remaining sequences showed moderate to no retention without immunity, and significantly higher retention with it, including at positions up to 15 nts from the PAM.

Fig. 4. Variation in Cas9-mediated immunity with sequence context. (A) UBP retentions of pUCX2 variants in strain YZ2 with a pCas9 plasmid that expresses a nontarget sgRNA (gray) or an on-target sgRNA (black). Error bars represent SD of the mean, and n = 3 transformations for all sequences except on-target CXA and CXG, where n = 5. (B) UBP retentions of pAIO plasmids in strain YZ3 (gray), which does not express Cas9, or in strain YZ4 (black) with expression of Cas9. In A and B, the nucleotides immediately flanking X = dNaM are indicated, as is distance to the PAM. “(N)” denotes the nucleotide N in the sgRNA that targets a substitution mutation of the UBP; all pCas9 and pAIO plasmids also express an sgRNA that targets a single nucleotide deletion of the UBP. Error bars represent SD of the mean, and n ≥ 3 colonies; Table S1 for exact values of n, sequences, and IPTG concentrations used to induced Cas9 in YZ4. Materials and Methods for additional experimental details.

Table S1. Cas9 NXN sequences

To further simplify and streamline the SSO, we next constructed strain YZ4 by integrating an IPTG-inducible Cas9 gene at the arsB locus of the YZ3 chromosome, which allows for the use of a single plasmid that both carries a UBP and expresses the sgRNAs that enforce its retention. Sixteen such “all-in-one” plasmids (pAIO) were constructed by replacing the Cas9 gene in each of the pCas9 variants with a UBP sequence from the corresponding pUCX2 variant (Fig. S2 and Table S1). YZ4 and YZ3 (included as a no-Cas9 control due to leaky expression of Cas9 in YZ4) were transformed with a single pAIO plasmid and cultured on solid growth media supplemented with the unnatural triphosphates and with or without IPTG to induce Cas9. Single colonies were used to inoculate liquid media of the same composition, and UBP retention was assessed after cells reached an OD 600 of ∼1 to 2 (Fig. 4B). Despite variable levels of retention in the absence of Cas9 (YZ3), with induction of Cas9 expression in YZ4, loss was minimal to undetectable in 13 of the 16 sequences. Although retention with the three problematic sequences—d(C-NaM-C), d(C-NaM-A), and d(C-NaM-G)—might be optimized, for example, through alterations in Cas9 or sgRNA expression, the undetectable loss of the UBP with the majority of the sequences after a regimen that included growth both on solid and in liquid media, which was not possible with our previous SSO DM1, attests to the vitality of YZ4.

Finally, we constructed a pAIO plasmid, pAIO2X, containing two UBPs: dNaM paired opposite dTPT3 at position 453 of the sense strand of the GFP gene and dTPT3 paired opposite dNaM at position 36 of the sense strand of the SerT tRNA gene, as well as encoding the sgRNAs targeting the most common substitution mutation expected in each sequence (Fig. S2). YZ4 and YZ3 (again used as a control) were transformed with pAIO2X and subjected to the challenging growth regime depicted in Fig. 5, which included extensive high-density growth on solid and in liquid growth media. Plasmids were recovered and analyzed for UBP retention (Fig. S6) when the OD 600 reached ∼1 to 2 during each liquid outgrowth. In YZ3, which does not express Cas9, or in the absence of Cas9 induction (no IPTG) in YZ4, UBP retention steadily declined with extended growth (Fig. 5). With induction of immunity (20 or 40 μM IPTG), we observed only a marginal reduction in growth rate (less than 17% increase in doubling time; Fig. S7), and, remarkably, virtually 100% UBP retention (no detectable loss) in both genes.

Fig. 5. Simultaneous retention of two UBPs during extended growth. Strains YZ3 and YZ4 were transformed with pAIO2X and plated on solid media containing dNaMTP and dTPT3TP, with or without IPTG to induce Cas9. Single colonies were inoculated into liquid media of the same composition, and cultures were grown to an OD 600 of ∼2 (point 1). Cultures were subsequently diluted 30,000-fold and regrown to an OD 600 of ∼2 (point 2), and this dilution−regrowth process was then repeated two more times (points 3 and 4). As a no-immunity control, strain YZ3 was grown in the absence of IPTG, and two representative cultures are indicated in gray. Strain YZ4 was grown in the presence of varying amounts of IPTG, and averages of cultures are indicated in green (0 µM, n = 5), blue (20 µM, n = 5), and orange (40 µM, n = 4). Retentions of the UBP in gfp and serT are indicated with solid or dotted lines, respectively. After the fourth outgrowth, two of the YZ4 cultures grown with 20 µM IPTG were subcultured on solid media of the same composition. Three randomly selected colonies from each plate (n = 6 total) were inoculated into liquid media of the same composition, and each of the six cultures was grown to an OD 600 of ∼1 (point 5), diluted 300,000-fold into media containing 0, 20, and 40 µM IPTG, and regrown to an OD 600 of ∼1 (point 6). This dilution−regrowth process was subsequently repeated (point 7). The pAIO2X plasmids were isolated at each of the numbered points and analyzed for UBP retention (Fig. S6). Cell doublings are estimated from OD 600 (Materials and Methods) and did not account for growth on solid media, thus making them an underestimate of actual cell doublings. Error bars represent SD of the mean.

Fig. S6. Representative biotin shift assay gels for Fig. 5. Each lane (excluding the oligonucleotide controls) corresponds to a pAIO2X plasmid sample isolated from a clonally derived YZ4 culture, grown with the IPTG concentration indicated, after an estimated 108 cell doublings in liquid culture (point 7 of Fig. 5). Each plasmid sample is split and analyzed in parallel biotin shift reactions that assay the UBP content at the gfp and serT loci (red and blue primers, respectively).