A highly effective and self-transmissible CRISPR antimicrobial for elimination of target plasmids without antibiotic selection

Bacterial strains and plasmids

Donor host cells in all experiments were E. coli EcGT2 provided by Dr. Harris H. Wang, University of Columbia, USA (Ronda et al., 2019). Recipient host cells in all experiments were E.coli SAR08 provided by Dr. Ellen Zechner, University of Graz, Austria (Reisner, Wolinski & Zechner, 2012). pgNDM1 is identical to plasmid pMM441 (#61271; Addgene, Watertown, MA, USA) provided by Dr. Timothy K. Lu, MIT, USA (Citorik, Mimee & Lu, 2014). pEMPTY was modified from pMM441 by replacing its CRISPR cassette with a gRNA gene whose spacer region was flanked by BsaI restriction sites. pgGFP was constructed by cloning a spacer sequence targeting a green fluorescent protein (GFP) gene into BsaI sites of pEMPTY. ptNDM1 and ptGFP were built by cloning an NDM1 fragment and a GFP gene, respectively, into the backbone of pSEVA231-CRISPR, provided by Dr. Victor de Lorenzo, Nacional de Biotecnología (CSIC) in Madrid, Spain (Aparicio, de Lorenzo & Martínez-García, 2018). The NDM1 fragment was amplified from Acinetobacter baumannii AB377 provided by Dr. Sutthirat Sitthisak, Naresuan University, Thailand. pHELP is identical to plasmid pRL443 provided by C. Peter Wolk, Michigan State University, USA (Elhai et al., 1997). Additional details about plasmid sequences and bacterial genotypes were provided in the Supplementary Materials (Table S2 and Supplementary Files).

Constructions of donor and recipient cells

pHELP was delivered to donor cells via conjugation (see below). pgNDM1, pgGFP, ptNDM1 and ptGFP were introduced to donor cells via TSS chemical transformation (Chung, Niemela & Miller, 1989). All plasmids were introduced to recipient cells via conjugation from E.coli EcGT2 donor cells and selected on Luria-Bertani (LB) agar with appropriate antibiotic and without diaminopimelic acid (DAP).

Media and culture conditions

Unless otherwise noted, E. coli were grown in LB broth or agar plates (LB with 15 g agar/L) at 37 °C. Where indicated, antibiotics were added to the selection medium to the following final concentrations: 25 μg/mL of chloramphenicol (Cm) 25 μg/ml of kanamycin (Km) 30 μg/mL of gentamicin (Gm) and 10 μg/mL of tetracycline (Tc). E. coli EcGT2 were grown in LB with 50 μM of DAP.

Conjugation assay

Donors and recipients were each inoculated from a single colony into LB broth with appropriate antibiotics and DAP (for EcGT2 donor cells) and cultured overnight in a 37 °C 200 rpm shaker. Saturated overnight cultures were regrown at 1% dilution ratio in fresh LB broth with appropriate antibiotics and DAP (for EcGT2 donor cells) until OD₆₀₀ reached 0.3–0.7. Cells were spun down at 5,000 g for 5 min and resuspended with 1 mL of LB broth three times in order to wash off residue antibiotics. The volumes of LB broth used in the final resuspension were adjusted so that both donor and recipient cells had OD₆₀₀ ~ 0.3l. Unless otherwise noted, donor cells and recipient cells were mixed at 1:100 ratio by volume. For EcGT2 donor and SAR08 recipient, OD₆₀₀ of 1 is equivalent to 1.14E + 09 and 0.63E + 09 CFU/m, respectively (see Table S4). Thus, donor-to-recipient ratio of 1:100 by volume (at the same OD₆₀₀) is approximately equivalent to donor-to-recipient ratio of 1:180 by cell number. 10 μL of mixed cells were dropped on 1 × 1 cm nitrocellulose membrane on LB agar with DAP. Unless otherwise noted, conjugation between donor and recipient on membranes were allowed to proceed for 18 h at 37 °C. Following conjugation, each membrane was placed in a 1.5 mL tube containing 1 mL of phosphate buffer saline and vortexed for 10 s three times to remove the bacteria from the membrane. The supernatant was serially diluted and plated on LB agar with appropriate antibiotic selection. Agar plates were incubated overnight at 37 °C for 16–24 h. Colonies were counted manually. Conjugation frequencies were measured as the number of transconjugants per recipient. Unless otherwise noted, each experiment was performed in triplicate.

Data visualization and statistics

Data visualization and statistical analysis were performed using R statistical computing software (https://www.r-project.org/). Scatter plots were generated using ggplot2 package. Student’s t-test of the difference between two data sets were conducted using t.test function in R. For multiple comparisons, we performed ANOVA with Tukey’s test for post hoc analysis using aov and TukeyHSD function in R.

Testing conjugatively delivered CRISPR/Cas for target plasmid elimination

A system was set up for conjugative delivery of CRISPR/Cas plasmid (pCRISPR) from donor cells and testing functionality of CRISPR/Cas in recipient cells (Fig. 1A, Fig. S1). pCRISPR was a plasmid encoding a R1162 broad-host-range origin of replication/transfer, a chloramphenicol resistance marker, a cas9 endonuclease gene, a tracrRNA gene and a CRISPR cassette or a gRNA. pTarget was a plasmid encoding a pBBR1 origin of replication, RK2 origin of transfer, a kanamycin resistance marker and a target gene for CRISPR/Cas. Once in a recipient cell, crRNA, tracrRNA and Cas9 expressed from pCRISPR can together cut the target gene and eliminate pTarget (Fig. 1B).

The spacer sequence of CRISPR cassette or gRNA could be re-designed for different target genes. In the first experiment, we had two versions of pCRISPR, pgNDM1 and pgGFP, with a spacer targeting New Delhi metallo-beta-lactamase 1 (NDM1) gene and Green Fluorescent Protein (GFP) gene, respectively. A total of two versions of pTarget, ptNDM1 and ptGFP, had a fragment of NDM1 and a GFP gene, respectively. We delivered pCRISPR from a donor host cell, EcGT2, which had an RK2 conjugation machinery integrated to the genome. The donor host cell was also auxotrophic for an essential cell-wall component diaminopimelic acid (DAP), thus requiring DAP supplementation in the growth media. This allowed us to select against donors simply by not adding DAP to growth media. Four mating experiments were conducted between donor cells (with pgNDM1 or pgGFP) and recipient cells (with ptNDM1 or ptGFP) (Fig. 1C). We showed that the number of pCRISPR transconjugants that still had pTarget (Fig. 1C, triangle) is less than the total number of transconjugant (Fig. 1C, circle) when the spacer sequence on pCRISPR matched the target gene on pTarget. Therefore, we concluded that CRISPR/Cas system expressed from pCRISPR could cut and eliminate specific pTarget in the recipient cell.

Next, we attempted to show that our mobilizable CRISPR/Cas can eliminate the target gene in a mixed recipient cell population (Fig. 1D). We conducted two mating experiments using EcGT2 donors with pgNDM1 or pgGFP. For both experiments, a recipient population consisted of cells with ptNDM1 or with ptGFP mixed together at 1:1 ratio. The percentage of green (GFP expressing) cells indicated the percentage of cells with ptGFP in the recipient population. Among transconjugants that maintained pTarget, the percentage of green cells increased to near 100 percent when pgNDM1 was delivered but decreased when pgGFP was delivered (Fig. 1E, triangle). This result implies that the delivery of pgNDM1 completely eliminated ptNDM1 in transconjugants. All surviving transconjugants with pTarget were green, i.e., having ptGFP. On the contrary, the delivery of pgGFP eliminated most of ptGFP in transconjugants. In this case, most surviving transconjugants with pTarget were white, i.e., having ptNDM1. Therefore, mobilizable CRISPR/Cas can specifically knockdown subpopulation, thereby altering the composition of the mixed population.

Among all transconjugants, the percentages of green cells in both experiments were only slightly different (Fig. 1E, rhombus). This could be explained by the fact that CRISPR/Cas mediated elimination of pTarget did not lead to cell death. The delivery of pgNDM1 may eliminate ptNDM1 in a recipient cell but the transconjugant still survived and remained as part of white cell population. The delivery of pgGFP led to elimination of ptGFP. Even if the cell survives, GFP signal is lost, resulting in reduction of green colonies (Fig. 1E triangle and Fig. S2). Without transconjugant selection, the percentages of green cells in both experiments were similar (Fig. 1E, circle). This is because most recipient cells did not receive pCRISPR and thus were not affected by CRISPR/Cas (Table S3). These results imply that the efficiency of mobilizable CRISPR/Cas in altering cell populations was limited by efficiency of DNA delivery.

Self-transmissible CRISPR/cas delivery systems

Since conjugative transfer of pCRISPR relied on conjugation machinery in the donor (EcGT2) genome, pCRISPR can only be mobilized to recipients immediately adjacent to donors. In order to deliver pCRISPR further, we introduced pHELP, a self-transmissible plasmid that can also mobilize pCRISPR from a transconjugant to the next recipients (Fig. 2A). pHELP is an RK2 conjugative plasmid with a tetracycline resistance marker, RK2 origin of transfer and conjugation machinery that can mobilize both RK2 and R1162 origin of transfer. By having both pHELP and pCRISPR in donor cells, we expected that pCRISPR can spread to a larger proportion of recipients even in the absence of transconjugant selection.

We conducted mating experiments using EcGT2 donors with pCRISPR alone or with both pCRISPR and pHELP (Fig. 2B). Note that for all following experiments, we used pgNDM1 as our only pCRISPR. Thus, we referred to this plasmid simply as pCRISPR (and its targeted plasmid, ptNDM1, as pTarget). Donors and recipients were mixed together at 1:18 or 1:180 donor-to-recipient ratio. At both donor-to-recipient ratios, mating experiment with pHELP yielded significantly higher pCRISPR transconjugant per recipient than mating experiment without pHELP (Fig. 2B, rhombus vs. triangle). For mating experiment with pHELP, pCRISPR transconjugant per recipient did not significantly decreased as we increased donor-to-recipient ratio from 1:18 to 1:180. Therefore, the use of pHELP could allow us to deliver pCRISPR widely in the population even at low donor-to-recipient ratio and in the absence of transconjugant selection.

We performed further experiments to explore how pHELP and pCRISPR affect transfer efficiencies of each other (Fig. 2C). The mating experiments were conducted at 1:180 donor-to-recipient ratio for 18 h on LB agar. We found that pHELP by itself can spread from donors to nearly 100% of recipients. pCRISPR by itself was only delivered, using conjugation machinery in the genome of donor cells, to only 0.3% of all recipients on average. In the presence of pHELP, the pCRISPR transconjugants increased to 31.9% of all recipients on average. On the contrary, the presence of pCRISPR decreased pHELP transconjugants down to 36% on average. In other words, pHELP boosted pCRISPR spreading while pCRISPR hindered pHELP spreading. About 9% of recipients had both pCRISPR and pHELP. To explore the dynamics of pCRISPR and pHELP transfer, we measured the number of transconjugants with pCRISPR, pHELP or both per recipient over time. The experiment started at 1:180 donor-to-recipient ratio; the donor had both pCRISPR and pHELP (Fig. 2D). Early on, pCRISPR appeared to spread faster but were later caught up and surpassed by pHELP. We found that after 2 days (2880 min) over 50% of recipient cells received pCRISPR. Thus, since the percentage of recipients with pCRISPR continued to increase over time, it should be possible to deliver mobilizable CRISPR/Cas plasmid to the entire bacteria population without transconjugant selection.

However, a self-transmissible CRISPR delivery system could also mobilize undesirable plasmids (e.g., antibiotic resistance or virulence plasmids) in the population. In the following experiment, we started with a recipient population in which only 1% of recipient cells had pTarget (representing an undesirable antibiotic resistance plasmid) which had a target site for pCRISPR but can also be mobilized by pHELP. We mated this recipient population to donor cells with only pCRISPR (no self-transmissible helper plasmid), to donor cells with only pHELP or to donor cells with both pHELP and pCRISPR. For the experiment with pCRISPR alone, the percentage of recipient cells with pTarget decreased by over a hundred folds relative to the control experiment (no plasmid in donors). On the contrary, when the donor cells had only pHELP, the percentage of recipient cells with pTarget increased by nearly a hundred folds relative to the control experiment. When both pCRISPR and pHELP were present, the percentage of recipient cells with pTarget remained almost equal to the control experiment. In other words, while pHELP increased overall delivering efficiency of CRISPR/Cas cassette (on pCRISPR in this case), the target plasmid was eliminated less efficiently. Our simulation results also confirmed the possibility of such scenarios in which the presence of a self-transmissible CRISPR delivery system could hinder rather than help the elimination of the target plasmid (Figs. S5, S6).

Elimination and replacement of the target plasmid with an incompatible plasmid

To fully utilize a self-transmissible CRISPR antimicrobial system, one has to prevent undesirable spread of the target plasmid by the helper plasmid (pHELP). We reasoned that the existence of a plasmid in a population could slow down or stop the entry of another plasmid from the same incompatibility group. While plasmid incompatibility does not directly blocking plasmid transfer, an incoming plasmid would be unlikely to establish in the population in the absence of selective advantage. In all following experiments, we used ptGFP as “pSubst”, a plasmid for replacing ptNDM1 (from now referred to as “pTarget”). pSubst and pTarget belong to the same incompatibility group and should not co-exist stably in the same host cell. We used pgNDM1 as “pCRISPR” which can express CRISPR/Cas specifically for cleaving pTarget but not pSubst.

We demonstrated a multi-plasmid system that can both eliminate an undesirable plasmid (pTarget) and replace it with another plasmid (pSubst) from the same incompatibility group. EcGT2 donors with both pCRISPR and pSubst were mated to recipients that had pTarget. pSubst by itself is unlikely to be able to enter, compete and replace pTarget that has already established in recipient population. However, when the donor has both pCRISPR and pSubst, pCRISPR can first be mobilized into recipients and mediate the degradation of pTarget. Once pTarget is eliminated, pSubst can then be mobilized into the recipient. We measured the fractions of recipients in each stage of plasmid replacement (Fig. 3A). We found that over 96% of recipients did not receive any plasmid (Fig. 3A, cell type X). Among transconjugants (cell type Y, Z and W), the majority of cells had both pCRISPR and pSubst (cell type W). Only a tiny fraction of transconjugants had both pCRISPR and pTarget (cell type Y). Overall, although recipients were unlikely to receive any plasmid, most recipients that received pCRISPR also received pSubst.

Next, we attempted to confirmed that a new plasmid cannot establish in the population that is already saturated with another plasmid from the same incompatibility group. We conducted two mating experiments using (1) EcGT2 donors with pCRISPR and pSubst or (2) EcGT2 donors with pEMPTY and pSubst. pEMPTY was similar to pCRISPR but without a spacer for targeting pTarget. All recipients in both experiments already had pTarget before mating. Donors were mated with recipients at 1:180 donor-to-recipient ratio. We confirmed that transfer efficiency of pEMPTY and pCRISPR were not significantly different (Fig. 3B). When the donors had pCRISPR, we found that 75% of transconjugants received pSubst and turned green (Fig. 3C, rhombus). With pEMPTY in donors instead of pCRISPR, all transconjugants remained white. This experiment confirmed that the replacement plasmid by itself could not enter, compete and replace the target plasmid that had already established in the population. The use of CRISPR plasmid to eliminate pTarget in recipients was critical for the replacement. Moreover, the majority of recipients that received pCRISPR also received pSubst. In other words, these two plasmids were likely to be transferred together. Nonetheless, like in the previous experiment (Fig. 3A), only a small fraction of recipients received any plasmid from the donors in the absence of transconjugant selection. For this reason, only a small percentage of all recipients received pSubst and turned green (Fig. 3C, circle).

To demonstrate generalizability of a plasmid substitution system, we performed an experiment similar to that shown in Fig. 3 but using a different pair of target and substitution plasmids (Fig. S8). These new pairs of plasmids have RK2 origin of replication (rather than pBBR1) and different backbone sequences from pTarget and pSubst. We confirmed that functional CRISPR/Cas was still critical for elimination and replacement of the target plasmid. Additionally, we performed conjugation experiments of EcGT2 donors with pTarget and pEMPTY to recipients without or with pSubst. Our results indicated that the presence of pSubst could prevent the delivery of pTarget to recipients albeit incompletely (Fig. S9).

Self-transmissible CRISPR/cas delivery and target plasmid replacement

The utilization of CRISPR antimicrobial requires high DNA transfer frequency. Thus far, we have shown that the use of self-transmissible helper plasmid (pHELP) can greatly increase transfer efficiency of a CRISPR plasmid (e.g., pCRISPR) even at low donor-to-recipient ratio and without antibiotic selection (Figs. 2B and 2C). Nonetheless, pHELP could also promote the dissemination of undesirable mobilizable plasmid we would like to eliminate (Fig. 2E). We have also shown that the use of CRISPR plasmid can facilitate the replacement of the target plasmid with a replacement plasmid from the same incompatibility group (Fig. 3). Here, we would like to explore whether self-transmissible CRISPR delivery system together with the use of plasmid incompatibility could drive complete elimination of undesirable mobilizable plasmid.

Our experimental system consisted of the following plasmids and E. coli strains (Fig. 4A). pTarget represented, an undesirable target plasmid (e.g., resistance or virulence plasmid) we would like to eliminate from the bacteria population. pSubst represented a replacement plasmid from the same incompatibility group as pTarget (but without any harmful genetic element). pCRISPR represented a CRISPR plasmid targeting pTarget. pHELP was a self-transmissible helper plasmid that can mobilize pCRISPR, pTarget and pSubst. An E. coli cell in the population can be classified into different “types” according to the type(s) of plasmid(s) the cell possessed. For example, a cell type A has no plasmid, a cell type B has a pTarget and a cell type G has both pTarget and pHELP. Any E. coli strain in the population can receive pHELP, pCRISPR, pTarget and pSubst thereby converting into another strain (blue, black, red and green arrow, respectively in Fig. 4A). For example, an E. coli strain-E that receives pCRISPR becomes an E. coli strain-J. pTarget and pSubst belong to the same incompatibility group. We previously showed that pSubst cannot enter and establish in the recipient that already had pTarget. Similarly, it is likely pSubst could also hinder the establishment of pTarget. Additionally, if pCRISPR and pTarget are in the same host cell, pTarget will be eliminated (dashed red arrow in Fig. 4A). If all four plasmids are present in the population, all recipients in the population are expected to end up receiving three plasmids (e.g., pCRISPR, pSubst and pHELP) while pTarget are eliminated completely from the population. We conducted two mating experiments using (1) EcGT2 donors with only pCRISPR or (2) EcGT2 donors with pHELP and pCRISPR. For both experiments, the recipient population consisted of cells with pTarget and cells with pSubst mixed at 1:1 ratio.

The EcGT2 donors were mated with mixed recipients at 1:180 donor-to-recipient ratio. We experimented on three different selection schemes: (1) selecting for pCRISPR alone, (2) selecting for pTarget or pSubst, (3) selecting for both pCRISPR and pTarget or pSubst. Unless selecting for both pCRISPR and pSubst, mating experiment in the presence of pHELP led to significantly higher percentage of cells that had pSubst. Notably, without selecting for pCRISPR, the percentage of cells with pSubst remained near 50% unless pHELP was present. Hence, pHELP helped mobilize pCRISPR to eliminate pTarget in recipients as well as mobilizing pSubst to replace pTarget (Fig. 4B).

Next, we conducted two mating experiments using (1) EcGT2 donors with pCRISPR and pSubst or (2) EcGT2 donors with pHELP, pCRISPR and pSubst. For both experiments, the EcGT2 donors were mated at 1:180 donor-to-recipient ratio to recipients that has pTarget. We experimented on five different selection schemes: (1) non selection, (2) selecting for pTarget or pSubst, (3) selecting for pCRISPR, (4) selecting for both pCRISPR and pTarget or pSubst and (5) selecting for pHELP. We found that, for the donor with pHELP, 100% of recipients had pTarget replaced by pSubst even in the absence of antibiotic selection after 4 days of conjugation (Fig. 4C). We confirmed that all colonies that received pSubst had eventually lose pTarget (Fig. S4). In computer simulation, we also showed that the presence of both pHELP, pCRISPP and pSubst could all contribute to complete elimination of pTarget (Fig. S7).