Abstract
Genetic tools are critical to dissecting the mechanisms governing cellular processes, from fundamental physiology to pathogenesis. Species of Burkholderia have emerged over recent decades as both important opportunistic pathogens in immunocompromised patients and as a rich source of secondary metabolites for biotechnology. With large, high GC content genomes, there are limited tools available for genetic manipulation of Burkholderia species. To address this, we have developed CRISPR interference (CRISPRi) technology for gene silencing in Burkholderia. Tunable expression was provided by placing a codon-optimized dcas9 from Streptococcus pyogenes under control of a rhamnose-inducible promoter, which was validated by immunoblot. Expression from a multicopy plasmid caused a growth defect; therefore, the mini-CTX single-copy insertion system was used to deliver dcas9 to the chromosome, resulting in wild-type growth. As a proof of concept, the paaABCDE operon controlling genes necessary for phenylacetic acid degradation was targeted by plasmid-borne sgRNAs, resulting in near complete inhibition of growth on phenylacetic acid as a sole carbon source. The utility of CRISPRi to probe other functions at the single cell level was demonstrated knocking down phbC and fliF, which dramatically reduces polyhydroxybutyrate granule accumulation and motility, respectively. As a hallmark of the mini-CTX system is the broad host-range of integration, we putatively identified 67 genera of Proteobacteria that might be amenable to modification with our CRISPRi toolkit.
Significance statement Members of the genus Burkholderia have potential for biotechnological applications but can also cause disease in humans with a debilitated immune system. The lack of suitable genetic tools to edit Burkholderia GC-rich genomes has hampered the exploration of useful capacities and the understanding of pathogenic features. Our CRISPRi tool kit provides a simple and rapid way to silence gene expression to produce an observable phenotype. Linking genes to functions with CRISPRi will facilitate genome editing with the goal of enhancing biotechnological capabilities while reducing Burkholderia’s pathogenic arsenal.
Introduction
The genus Burkholderia comprises a diverse group of Gram-negative bacteria characterized by a remarkable biotechnological potential, with species that can be exploited for bioremediation purposes, production of bioactive compounds, and agricultural growth promotion. Perhaps due to the phylogenetic diversity of the genus, some species of Burkholderia can also pose a threat to human health (Eberl and Vandamme, 2016). Within the multiple deep-branching groups of the Burkholderia phylogenetic tree (Depoorter et al., 2016), one branch contains the clades known as the Burkholderia cepacia complex (Bcc) and the B. pseudomallei group. The Bcc group comprises species that cause severe infections in people with the genetic disease cystic fibrosis and other patients with a compromised immune system (Mahenthiralingam et al., 2005). B. cenocepacia and B. multivorans are two relevant species of this group, which are prevalent in cystic fibrosis patients (Kenna et al., 2017). The B. pseudomallei group contains the risk group 3 members B. mallei, B. pseudomallei, and its close relative B. thailandensis, the last one being a frequent surrogate used in B. pseudomallei and B. mallei research (Howe et al., 1971). Other branches, such as the B. xenovorans group comprise species isolated from diverse environmental sources, like polluted soils, and plant rhizospheres (Depoorter et al., 2016). While species with and without pathogenic potential tend to cluster separately in most phylogenetic trees, the genetic characteristics that define the pathogenic potential of Burkholderia are poorly understood. Environmental species can cause serious infections (Nally et al., 2018), calling for caution in the use of Burkholderia strains for biotechnological applications. The incomplete understanding of Burkholderia pathogenic potential may be related to the limited tools available to link gene to function. Most genome editing methods designed for Gram-negative bacteria are inefficient in Burkholderia due in part to their high resistance to the antibiotics used as genetic markers (Rhodes and Schweizer, 2016), and the high GC content of their large genomes.
Tools that facilitate controlled gene expression are necessary for interrogation of gene function. Programmable control of gene expression by promoter replacement is a valuable tool to link gene to phenotype (Judson and Mekalanos, 2000). Yet, promoter replacement implies that the natural regulatory circuitry of the target gene is interrupted. Instead, clustered regularly interspaced short palindromic repeats interference (CRISPRi) (Qi et al., 2013) is a method of silencing native gene expression, which is based on a dead Cas9 (dCas9), a nuclease-inactive version of the RNA-guided endonuclease Cas9 (Cho et al., 2013). In CRISPRi, a single guide RNA (sgRNA) designed towards the 5’ end of the target gene and the dCas9 protein form an RNA-protein complex that recognizes the target region by base-pairing, and sterically blocks transcription initiation, if targeting the promoter, or elongation if targeting downstream of the promoter, by the RNA polymerase (Figure 1A).
Initially developed in Escherichia coli, CRISPRi technology has been applied to functionally characterize genomes of a handful of bacteria (Lee et al., 2019; Liu et al., 2017; Peters et al., 2016). To achieve genome-wide control of gene expression a wide range of dCas9 expression levels are necessary, which can be provided by expressing dcas9 with strong inducible promoters (Peters et al., 2016) or by providing multiple copies of dcas9 engineered in a plasmid (Lee et al., 2019). Limitations to the success of these genomic efforts are the proteotoxicity of dCas9 when expressed at high levels (Cui et al., 2018) and the necessity of customizing CRISPRi delivery tools across bacteria. Recently, the use of Tn7 transposon mutagenesis, which specifically delivers genetic constructs close to a single glmS site (Choi and Schweizer, 2006; Choi et al., 2005) was applied to deliver a mobile CRISPRi system across bacteria (Peters et al., 2019). However, the Tn7 system is less suitable for Burkholderia species as their genomes contain multiple copies of glmS, requiring additional steps to confirm the site of chromosomal integration (Choi et al., 2006).
In this work, we employ a mini-CTX-derived mutagenesis system (Hoang et al., 2000) to achieve specific chromosomal delivery of dcas9 in three species of Burkholderia, B. cenocepacia, B. thailandensis and B. multivorans. Placing the chromosomal copy of dcas9 under the control of the E. coli rhamnose-inducible promoter (Cardona and Valvano, 2005), we demonstrate durable and tunable control of endogenous gene expression, which affects cellular function producing observable phenotypes. We extend the usability of our CRISPRi tool kit by exploring other bacterial genomes for putative mini-CTX insertion sites.
Results
Construction of CRISPRi system
The dCas9 from Streptococcus pyogenes has been shown to provide robust gene repression in diverse bacteria (Peters et al., 2016, 2019; Qi et al., 2013); we therefore selected it as our first approach. To function, the dCas9 binds a sgRNA and, by complementary base pairing, is guided to target and silence a gene of interest by sterically blocking the RNA polymerase (Figure 1A) (Qi et al., 2013). However, the genome of S. pyogenes has low GC-content (∼40%), and from inspection of the dcas9 gene, we expected poor codon usage in high GC-content organisms such as Burkholderia (67%) and subsequently low levels of expression. Indeed, we did not observe detectable levels of dCas9 by immunoblot upon expression of the native gene from S. pyogenes from a single copy in the B. cenocepacia K56-2 chromosome (Supplemental Figure 1). Upon codon optimization for B. cenocepacia, we first introduced the gene into a multicopy plasmid under the control of the rhamnose-inducible promoter (Cardona and Valvano, 2005); however, we observed a severe growth defect upon induction, except at minute concentrations of rhamnose (Supplemental Figure 2A). Growth inhibition was not observed in the vector control (Supplemental Figure 2B) and it remains unclear if the inhibitory effect of dCas9 expression on growth was caused by metabolic load from expression of a large protein in multicopy, or from proteotoxicity.
We reasoned a single chromosomal copy of dcas9 may provide sufficient levels of expression as observed previously (Choudhary et al., 2015; Peters et al., 2016, 2019). Using the mini-CTX system, we introduced a single copy of dcas9 under control of the rhamnose-inducible promoter into B. cenocepacia K56-2 (Supplemental Figure 3A and B). We observed titratable dCas9 expression at various levels of rhamnose by immunoblot (Figure 1B and Supplemental Figure 4A). At rhamnose concentrations up to 0.2% there was no growth defect (Figure 1C) compared to the vector control mutant (Supplemental Figure 4B) Furthermore, dCas9 expression in the presence of a sgRNA with a randomized binding region (sgRNA-non-target) had no effect on growth (Supplemental Figure 4C), demonstrating the current system is well-tolerated in B. cenocepacia K56-2.
Tunable and durable CRISPRi silencing of paaA strongly suppresses growth on phenylacetic acid
To evaluate the utility of the CRISPRi system for gene repression in B. cenocepacia K56-2, we first chose to target the paaA gene, which encodes phenylacetate-CoA oxygenase subunit PaaA (Teufel et al., 2010). This gene, and the rest of the paaABCDE operon, enable growth with PA as a sole carbon source in B. cenocepacia K56-2 (Pribytkova et al., 2014), with the lack of growth being a clearly observable phenotype when the paaA gene is disrupted. In addition, as this is the first characterization of CRISPRi in Burkholderia, we also wished to assess the effect on repression efficiency when targeting the non-template (NT) and template (T) strands, as previous studies have demonstrated profound differences (Bikard et al., 2013; Qi et al., 2013). We therefore designed five sgRNAs: three sgRNAs targeted the promoter elements and adjacent to the transcription start site (TSS) on the NT strand (sgRNA 1, 2 and 3), one sgRNA targeted near the start codon of paaA on the NT strand (sgRNA 4), and one sgRNA targeted near the start codon of paaA on the T strand (sgRNA 5) (Figure 2A). For phenotypic characterization of the paaA mutants, we used M9 minimal medium with PA (M9+PA) as the sole carbon source. Upon induction of dCas9, the growth of all mutants (except the controls) was suppressed approximately 30-fold (Figure 2B) to the same level of a ΔpaaABCDE mutant, which is unable to utilize PA as a sole carbon source (Pribytkova et al., 2014). Furthermore, in the absence of rhamnose all of the mutants grew at or near wild-type levels, suggesting that dcas9 expression is tightly repressed in non-inducing conditions. Interestingly, we did not observe a differential effect of placement of the sgRNA-binding site, as the growth of all mutants was suppressed equally. We also found there to be no difference in control strains; mutants expressing dCas9 and either a guideless or non-targeting sgRNA displayed the same levels of growth.
Next, we sought to determine the tunability of our CRISPRi system in Burkholderia. Tuning is useful to control the level of transcription inhibition when precision is required, such as in the study of essential genes. To that end, we examined the growth of the dCas9 mutant of B. cenocepacia K56-2 with the sgRNA1 (targeting paaA), using PA as a sole carbon source in the presence of a 2-fold serial dilution of rhamnose. The results showed that our CRISPRi system is tunable, exhibiting growth reduction in a dose-dependent manner with variable repression across sub-saturating rhamnose concentrations (Figure 2C). Intermediate levels of growth were observed in concentrations of rhamnose between 0.00625% and 0.05%. We observed a nearly 30-fold repression in OD600nm at concentrations well below maximum induction as identified by immunoblot, suggesting our system produces more dCas9 than is required for maximum repression. Rhamnose levels beyond 0.2% were intrinsically growth inhibitory precluding their use for inducing gene expression (Figure 2C).
Although once rhamnose is removed from the culture, expression of dCas9 is no longer induced, it remained possible that the dCas9 already synthesized may persist and cause long-term, or durable, silencing. To address this, the mutant strains harbouring sgRNAs targeting paaA, were grown overnight in rich medium with 0.2% rhamnose, effectively priming the cells with dCas9. When grown in M9+PA with and without rhamnose, we again observed strong repression of growth (∼30-fold) in all conditions regardless of the presence of rhamnose in the M9+PA, and at levels similar to the ΔpaaABCDE mutant (Figure 2D). This suggests that after the inducer is removed, dCas9 is either slowly degraded in K56-2 or is present at high enough levels in the cells for durable repression.
Single-cell analysis reveals an ‘all or none’ effect in B. cenocepacia K56-2
While at the culture level the effect of CRISPRi is tunable, we further explored the effect of maximum dCas9 induction at the single-cell level. We therefore targeted fliF, a gene encoding a transmembrane protein that forms the M ring of the basal body complex of the flagellum. We targeted fliF by introducing four sgRNAs designed to bind near the putative promoter and start codon (Supplemental Figure 5A). Our goal was to assess individual cell flagellation as it compares to overall culture swimming motility, and indeed, insertional inactivation of fliF completely ablates swimming motility (Supplemental Figure 5B), supporting previous findings (Tomich et al., 2002). While we observed effective inhibition of swimming motility compared to controls in a plate-based assay (approximately 4-fold reduction) (Supplemental Figure 5B and C), we were unable to observe flagella in any of the mutants harbouring gRNAs targeting fliF (Supplemental Figure 5D). It is possible that interfering with fliF expression rendered fragile flagella that could not remained attached to the cell during the staining process, while still being functional when grown in culture (Komatsu et al., 2016).
As a second option, we targeted phbC, a gene encoding poly-β-hydroxybutyrate polymerase, an enzyme required for PHB synthesis. To repress phbC expression, we designed three sgRNAs to target the region up to 50 bp before the start codon, at the putative promoter site (Figure 3A). Polyhydroxyalkanoate (PHA) granule accumulation was assessed by fluorescence microscopy with Nile red staining after overnight induction with rhamnose. For comparison to a null phenotype, we created a phbC insertional mutant (phbC::pAH27) which was unable to produce PHA granules (Figure 3B and D). Compared to the wild-type and non-target controls, markedly few cells harbouring sgRNAs targeting phbC contained PHA granules, ranging from 17-70% depending on the sgRNA (Figure 3B). sgRNA6 rendered the strongest level of repression, with only 17.1% of cells containing PHA granules in contrast to 86.9% for the wild-type (Figure 3B). Interestingly, although few cells possessed PHA granules, the granules were of identical size to those in wild-type cells, averaging 0.65 µm in diameter (Figure 3C and D). As shown by the insertional mutant, inactivation of phbC ablates PHA granule accumulation; therefore, the presence of granules of the same size in the dCas9 mutants as in the wild-type is indicative of an ‘all-or-none’ effect in B. cenocepacia K56-2, where most cells display the silenced phenotype, but some manage to escape the effect of CRISPRi.
Broad host range of the mini-CTX system extends applicability to other species
Having shown that the S. pyogenes dCas9 renders strong gene repression in B. cenocepacia K56-2, we next turned our attention to other important species of Burkholderia. We introduced the codon-optimized dcas9 gene into the chromosome of B. multivorans ATCC 17616 and targeted the paaA and phbC genes by designing three sgRNAs for each centered around the putative promoter (Figures 4A and C). Targeted repression of paaA by each of the three sgRNAs strongly suppressed growth in M9+PA by approximately 25-fold (Figure 4B), very similar to the activity observed in B. cenocepacia K56-2. However, we observed poor expression of dCas9 by immunoblot (Supplemental Figure 6A). Given that the observed phenotype is indicative of effective gene silencing, it is unlikely that the actual expression of dCas9 is as low as detected by immunoblot; however, the basis for this discrepancy is unknown. Expression of dCas9 in the presence of the non-targeting sgRNA did not affect growth in B. multivorans ATCC 17616 (Supplemental Figure 4C). Following this, targeting phbC rendered relatively poor inhibition of PHA granule accumulation with between 35-50% of cells containing PHA granules, depending on the sgRNA (Figure C and D). In contrast to what was seen in B. cenocepacia K56-2, we observed granules of reduced size (Figure 4E and F), suggesting there might be compensatory PHA synthesis pathways in B. multivorans ATCC 17616.
Members of the B. pseudomallei group are phylogenetically distinct from the Bcc (Eberl and Vandamme, 2016), but remain of interest due to the ability to cause infection (such as melioidosis and glanders) and for their remarkable capacity for secondary metabolite production (Mao et al., 2017). B. thailandensis is a commonly used model for the pathogenic members of the B. pseudomallei group, we therefore also introduced the codon-optimized dcas9 gene into the chromosome of B. thailandensis strain E264. When induced with rhamnose, dCas9 was highly expressed (Supplemental Figure 6) and did not impair growth in the presence of a non-targeting sgRNA (Supplemental Figure 4C). Similarly, as for B. cenocepacia K56-2 and B. multivorans ATCC 17616, we designed three sgRNAs targeting the putative promoters of the paaA and phbC genes (Figure 5A and C). Upon induction of dCas9 with 0.2% rhamnose, growth of the mutants harbouring the paaA-targeting sgRNAs was suppressed in M9+PA to varying levels ranging from 3-fold (sgRNA19) to 25-fold (sgRNA20) (Figure 5B). The fact that some sgRNAs yielded only modest repression was not unexpected as the work characterizing the paaA promoter done in K56-2 (Pribytkova et al., 2014) would not likely apply as well to E264 as to the more closely related ATCC 17616. Lastly, for the sgRNAs targeting phbC, the results mirror those seen in ATCC 17616, as depending on the sgRNA there was variation in the percent of cells with PHA granules (10-40%) and the diameter of the granules (overall decrease in size) (Figure 5C and F). Though the exact contributions to PHA synthesis in E264 remain unclear, it has been previously suggested that PhbC is not the only polyhydroxyalkanoate polymerase in E264 (Funston et al., 2017).
The integration vector we implemented to deliver dcas9 to the chromosome of various species of Burkholderia relies on the expression of the φCTX integrase and recombination using the plasmid-borne attP site with the chromosomal attB site. φCTX is a Pseudomonas-infecting phage, and the mini-CTX integration system was originally designed for use in P. aeruginosa (Hoang et al., 2000). The mini-CTX integration system has been used successfully in other species; however, this has been mostly limited to members of Burkholderia (Chapalain et al., 2017; Le Guillouzer et al., 2017). The utility in Burkholderia has been comparable to P. aeruginosa, in part owing to efficient integration. In the species used in this study, we observed the integration efficiencies to be 6×10−7 in K56-2, 6×10−8 in E264, and 5×10−9 in ATCC 17616 (Supplemental Figure 7A), compared to 10−7 to 10−8 observed previously in P. aeruginosa (Hoang et al., 2000). To further broaden the scope of the applicability of our CRISPRi system, we used NCBI BLAST to search all published genomes for putative attB sites. While the full-length attB site is 30 nt, integration is known to occur if only the 5’ 19 nt are completely complementary, such as for many species of Burkholderia. This shorter attB site was therefore used as a BLAST query, resulting in 1760 hits with 100% alignment (Supplementary Table 4). Enterobacteria were excluded from the search as the pMB1 oriR in the mini-CTX system is functional is these species; therefore, integrants cannot be easily isolated. Furthermore, the search parameters were modified to only include species of Proteobacteria. Overall, there were 168 unique species from 67 genera. As expected, the most abundant hit corresponded to species and strains of Pseudomonas (480 hits), then followed by Acinetobacter (443 hits), Burkholderia (276 hits), Neisseria (170 hits), and Ralstonia (146 hits), with members of the other 62 genera comprising the remaining 245 hits. A summary of the major hits and species of interest (pathogenic, environmental, biotechnological, etc) can be found along with the genomic context in Table 1. We note that the hit table comprises species with both high and low GC-content genomes, and while the GC-rich codon-optimized dcas9 (in pAH-CTX1-rhadCas9) may be better suited for species with high GC-content genomes, such as those in the families Pseudomonadaceae and Alcaligenaceae, the native dcas9 (in pAH-CTX1-rhadCas9-native) may have better functionality in species with low GC-content genomes, such as those in the families Moraxellaceae and Neisseriaceae.
A further consideration for use of our CRISPRi system is that the pFLPe recombinase-expressing plasmids might not be functional in all species. In which case, the int and tet genes along with the oriT and pBM1 oriR would not be excised and the mutant would remain resistant to tetracycline. To explore if experiments could be performed in the absence of tetracycline selection, we assessed the stability of the mini-CTX integration in K56-2, ATCC 17616, and E264 in serial passages over four days without tetracycline. For all species over the entire experiment we determined CFU counts on agar with and without tetracycline selection and found equal recovery of tetracycline-resistant colonies as total colonies (Supplemental Figure 7B-D), suggesting the integration is stable.
Discussion
Genetic tools are necessary to dissect the molecular mechanisms governing cellular processes. Here, we report the development of a CRISPRi system for efficient repression of gene expression in Burkholderia. By mobilizing the dcas9 gene that was codon-optimized for the GC-rich Burkholderia on a broad host range mini-CTX1 integration vector, we demonstrate robust, tunable, and durable repression of endogenous genes. While others have shown effective repression using the native dcas9 gene in E. coli (Qi et al., 2013), B. subtilis (Peters et al., 2016), Staphylococcus aureus, and Acinetobacter baumannii (Peters et al., 2019), codon-optimization was a necessary step for K56-2, as expression of the native dcas9 gene was not detectable in K56-2. Indeed, for expression in species with high GC-content, codon optimization appears to be necessary. In P. aeruginosa (66.3% GC) both the S. pyogenes dCas9 (Peters et al., 2019) and the S. pasteurianus dCas9 (Tan et al., 2018) were not expressed unless first codon-optimized, albeit for Mycobacterium (∼67% GC) or Homo sapiens (optimized at 50.2% GC). Additionally, the S. thermophilus dCas9 was codon-optimized for Mycobacterium for efficient expression in M. tuberculosis and M. smegmatis (Rock et al., 2017).
Despite codon-optimizing the dcas9 gene, we observe a severe growth defect when expressed from a multicopy pBBR1 origin plasmid. At this time, we are unsure if this was caused by a metabolic burden of expressing a large protein from a multicopy plasmid or from direct proteotoxicity. Previous studies have demonstrated that expression of the canonical S. pyogenes dCas9 causes toxicity in M. smegmatis, M. tuberculosis (Rock et al., 2017), and E. coli (Cho et al., 2018), which provides rationale for developing a system with low-enough levels of dCas9 expression to maintain cell viability without sacrificing repression activity. While this effort has spurred the exploration of alternative dCas9 orthologues (Rock et al., 2017), we found that introducing dcas9 in single copy in the chromosome provided a balance of repression activity without affecting growth. Furthermore, while other systems display up to 3-fold repression in the absence of inducer (Peters et al., 2016; Tan et al., 2018), our application of the tightly regulated rhamnose-inducible promoter from E. coli does not display such a phenomenon in any of the three species tested. Consequently, we suggest this attribute is suited for the investigation of essential genes, as we would predict wild-type growth in the absence of inducer. However, at the cost of apparently negligible uninduced expression, the rhamnose-inducible promoter is incapable of attaining the high levels of maximal expression seen in other systems, such as the xylose-inducible promoter used in the B. subtilis CRISPRi library (Peters et al., 2016) (Supplementary Figure 4A). While we did observe increased expression at rhamnose concentrations of 0.5% and 1.0%, it was accompanied by a substantial growth defect. This is similar to the use of the arabinose-inducible promoter in K56-2, in which the 2-3% arabinose necessary to induce maximal expression caused marked changes in cell morphology reminiscent of osmotic stress (Cardona and Valvano, 2005). As a result of the lower levels of maximal expression, it remains possible that our CRISPRi system may not be capable of effectively repressing highly transcribed genes, but this is under investigation. While work by Vigouroux et al. (Vigouroux et al., 2018) suggests that when the sgRNA binding site is saturated with the dCas9:sgRNA complex, fold change in repression is independent of promoter strength; however, the absolute repression of highly active promoters is lower and it is unclear what constitutes saturating levels of dCas9. Furthermore, multiplexing CRISPRi with simultaneous use of more than one sgRNA might reduce the repression at each locus by titrating away dCas9, and this effect may be even more apparent if lower levels of dCas9 are present.
Critical to fine manipulation of gene expression in precision experiments is tunability. We show that using the rhamnose-inducible promoter can provide intermediate levels of repression and a resulting phenotype of gene depletion rather than complete silencing. Combined with the lack of repression in the uninduced state, we propose that our CRISPRi system will be useful in providing dynamic gene repression. Moreover, the repression caused by dCas9 is durable, as even a full 24 hours after removal of the inducer, we observed the same phenotype as in the cultures grown with inducer. This is in line with previous findings in Giardia lamblia (McInally et al., 2019) and in vitro (Jones et al., 2017; Sternberg et al., 2014). Interestingly, binding of the dCas9:sgRNA complex to a target is nearly irreversible once the sgRNA has bound with complete complementarity to its target site (Sternberg et al., 2014). Furthermore, this means that the dCas9:sgRNA complex does not follow Michaelis-Menton enzyme kinetics, with each dCas9 acting like a single-turnover enzyme that, in vitro, can only be removed from the target by strong detergents (Sternberg et al., 2014). In vivo however, although the RNA polymerase is largely incapable of displacing the dCas9:sgRNA complex, the DNA replication machinery is not blocked by dCas9. In fact, it has been observed that the dissociation of the dCas9:sgRNA complex correlates well with the generation time in E. coli (Jones et al., 2017). This is an important consideration for understanding the durability of growth suppression we observed when using PA as a sole carbon source. When priming the cells by pre-incubation with rhamnose and then placing the cells in M9+PA, they were unable to grow, that is paaA became a conditionally essential gene, thereby shutting down cell division and DNA replication. Hence, the DNA replication machinery would not displace the dCas9:sgRNA from the paaA gene, resulting in long-term growth suppression. By extension, we predict that strong knockdown of any (conditionally) essential gene that causes a halt in DNA replication would be durable in this manner. This is a useful aspect of CRISPRi that we are investigating further.
Although it has been reported that the CRISPRi system is less effective in silencing the gene expression when the template strand (T) is targeted compared to the non-template (NT) strand, our results demonstrated nearly 30-fold repression of the growth irrespective of the paaA target strand (Figure 2B). This suggests that the efficiency of the CRISPRi system might not strand specific at all loci, supporting the findings from Howe et al. (Howe et al., 2017). However, this effect might have been masked by the strong repression we observed, as it is difficult to compare across null phenotypes. Additionally, sgRNAs targeting different regions of the 5’ region of paaA produced a strong 25-30-fold repression in all three species studied here, suggesting that knockdown efficiency is largely unaffected by sgRNA placement on targeting regions overlapping or adjacent to the −35 and −10 promoter elements. possibly resulted from the proximity of the large dCas9:sgRNA complex to the promoter sterically interfering with the RNA polymerase binding. Indeed, the ∼160 kDa dCas9 enzyme has an average DNA footprint of 78.1 bp, much larger than most promoter regions (Josephs et al., 2015).
Single cell level analysis of our system demonstrated a unimodal ‘all or none’ effect in K56-2 with ∼17% cells escaping the silenced phenotype when targeting phbC compared to an insertion mutant, which did not possess any granules. The PHB granules in these cells were of identical diameter to wild type, averaging 0.65µm in diameter. The ‘all or none’ effect might be attributed to an uneven distribution of the membrane transporter involved in transporting rhamnose into the cell, essentially escaping the rhamnose mediated dCas9 induction to silence the target gene (Siegele and Hu, 1997). Interestingly, targeting phbC in B. multivorans ATCC17616 and B. thailandensis E264 resulted in a mixed phenotype with a reduced number of cells with granules as well as an overall decrease in granule size. Even though some of the cells seemed to have escaped the repression (up to 35% in ATCC17616 and 6% in E264), more than 90% of the cells contained granules of reduced size. The residual presence of PHA granule suggests the existence of an unelucidated PHA synthesis pathways in B. thailandensis E264 and B. multivorans ATCC 17616, supporting the finding from Funston et al. (Funston et al., 2017).
One of the hallmarks of CRISPRi is the broad-range amenability in diverse bacteria, enabling synthetic biology and mechanistic investigations into many dozens of species in innovative ways. We wished to apply our CRISPRi system in this manner and therefore mobilized both the native dcas9 gene, suitable for low/medium GC-content organisms, and the dcas9 gene, codon-optimized for GC-rich Burkholderia, on the mini-CTX1 integration vector. Our analysis of putative hosts (with attB sites near the 3’ end of the serine tRNA) identified 168 unique species in 67 genera, mostly from the β- and γ-Proteobacteria. Previous works have mobilized dcas9 on broad host-range integrative plasmids (Peters et al., 2019; Tan et al., 2018); however, both systems use the mini-Tn7 system, which has multiple insertion locations in certain genomes, such as many species of Burkholderia. Together, our work contributes to the available genetic toolkit for rapid functional analysis of bacteria.
Methods
Strains, selective antibiotics, and growth conditions
All strains and plasmids are found in Supplemental Table 1. All strains were grown in LB-Lennox medium (Difco). B. cenocepacia K56-2 and strains of E. coli were grown at 37°C, while B. thailandensis E264 and B. multivorans ATCC 17616 were grown at 30°C. The following selective antibiotics were used: chloramphenicol (Sigma; 100 µg/mL for B. cenocepacia, 20 µg/mL for E. coli), trimethoprim (Sigma; 100 µg/mL for strains of Burkholderia, 50 µg/mL for E. coli), tetracycline (Sigma; 50 µg/mL for all strains of Burkholderia, 20 µg/mL for E. coli), kanamycin (250 µg/mL for B. thailandensis, 150 µg/mL for B. multivorans, 40 µg/mL for E. coli), ampicillin (100 µg/mL for E. coli), gentamicin (50 µg/mL for all strains of Burkholderia).
Construction of pSC-rhadCas9, pAH-CTX1-rhadCas9, pAH-CTX1-rhadCas9-native, and dCas9 insertional mutants
The endogenous cas9 gene from S. pyogenes has low GC content (averaging 34.1%) and subsequently poor codon usage for GC-rich organisms (http://www.kazusa.or.jp/codon/). The nuclease-inactive variant (dcas9) was therefore codon optimized for B. cenocepacia by purchasing the optimized gene in two fragments from IDT (2462 bp and 1849 bp, Supplemental table 2), each with 38 bp overlapping regions. Strong, rho-independent terminators were added following the gene. The full-length gene (with terminal NdeI and HindIII cut sites) was synthesized by overlap-extension PCR using Q5 polymerase with high GC buffer (NEB) and primers 979 and 987 (Supplemental Table 2). The first ten rounds of PCR were performed without primers to synthesize the full-length product using the overlap regions; primers were added for the following 25 cycles. The cycle parameters are as follows: 98°C for 30 sec, (98°C for 10 sec, 67.5°C for 20 sec, 72°C for 2.5 min)x10 cycles, 72°C for 5 min, 98°C for 30 sec, (98°C for 10 sec, 62.5°C for 20 sec,72 °C for 2.5 min)x25 cycles, 72°C for 10 min. The 4267 bp product was gel-purified (Qiagen) and introduced into pSCrhaB2 by restriction cloning using NdeI and HindIII (NEB). The resulting plasmid, pSC-rhadCas9, was transformed into E. coli DH5α, and trimethoprim-resistant colonies were screened by colony PCR with primers 954 and 955. Triparental mating with E. coli MM290/pRK2013 as a helper was performed as previously described (Hogan et al., 2018).
Serial restriction cloning was used to introduce the optimized dcas9 into the mini-CTX1 insertion plasmid (Hoang et al., 2000). The rhamnose-inducible promoter from pSC201 (Ortega et al., 2007) was first PCR amplified with Q5 polymerase and primers 976 and 1071, containing HindIII and SpeI restriction sites, respectively. This fragment was introduced into mini-CTX1, to create pAH-CTX1-rha, and tetracycline-resistant E. coli DH5α were screened by colony PCR using primers 957 and 1074. The fragment containing the dCas9 gene was PCR amplified as above, but instead using primers 1072 and 1073, introducing SpeI and NotI restriction sites, respectively. This fragment was introduced into pAH-CTX1-rha, to create pAH-CTX1-rhadCas9, and tetracycline-resistant E. coli DH5α colonies were screened by colony PCR using primers 954 and 955.
The native (non codon-optimized) dcas9 was also introduced into pAH-CTX1-rha. The native dcas9 and transcriptional terminators were PCR amplified from pdCas9-bacteria (Addgene plasmid # 44249) with Q5 polymerase (NEB) using primers 1216 and 1217. The PCR product was cloned into pAH-CTX1-rha (creating pAH-CTX1-rhadcas9-native) using NotI and SpeI restriction sites then transformed into E. coli DH5α. Tetracycline-resistant colonies were screened by PCR using primers 954 and 1218.
pAH-CTX1-rha, pAH-CTX1-rhadCas9, and pAH-CTX1-rhadCas9-native were introduced into Burkholderia species by triparental mating using E. coli MM290/pRK2013 as a helper as above. Tetracycline-resistant colonies were screened by colony PCR using primer 954 (for pAH-CTX1-rha) or 1075 (for pAH-CTX1-rhadCas9) or 1219 (for pAH-CTX1-rhadcas9-native) and 1008 (for B. cenocepacia), 1167 (for B. multivorans), or 1168 (for B. thailandensis).
While the pFLPe plasmids have been used to remove the tetracycline resistance and integrase genes in other species (Choi et al., 2008), sensitive colonies could not be obtained in B. cenocepacia. Therefore, the clean deletion method of Flannagan et al was used as previously described (Flannagan et al., 2008). To do this, a fragment with 475 bp overlapping the upstream and downstream regions of the FRT sites when pAH-CTX1-rhadCas9 was inserted into the genome (Supplemental Figure 2), was designed and synthesized (IDT) with KpnI and EcoRI restriction sites, respectively. The fragment was ligated into pGPI-SceI (Flannagan et al., 2008) via the KpnI and EcoRI restriction sites, creating pAH18, and transformed into E. coli SY327. Trimethoprim-resistant colonies were screened for the insertion of the fragment with primer 153 and 154. pAH18 was introduced into the mutant backgrounds via triparental mating, as described above. Trimethoprim-resistant K56-2 were screened by PCR for both possible integration orientations using primers 154 and 1126, or 153 and 1133. To initiate the second recombination, an SceI-expressing plasmid is required; however, the conventional plasmid, pDAI-SceI, confers tetracycline resistance and could not be selected for in the mutant background. Therefore, the tetracycline resistance cassette was removed by digestion with AgeI and XhoI. The chloramphenicol resistance gene cat was PCR amplified from pKD3 (Datsenko and Wanner, 2000) using primers 1084 and 1150, then ligated into the pDAI-SceI backbone and transformed into E. coli DH5α, creating pAH25-SceI. Chloramphenicol-resistant colonies were screened with primers 1091 and 1150. pAH35-SceI was introduced into the mutant backgrounds by triparental mating as described above. Chloramphenicol-resistant colonies were screened for sensitivity to trimethoprim (indicating excision of pAH18) and tetracycline (indicating excision of the genes between the FRT sites), and then screened by PCR with primers 1126 and 1133, which bridge the excision.
The pFLPe system was used to remove the tetracycline resistance and integrase genes in the dCas9 mutants in B. multivorans ATCC 17616 and B. thailandensis E264. Triparental mating to introduce pFLPe4 into the strains was performed as for K56-2 above, except 0.2% rhamnose was added to the mating and antibiotic selection plates. Tetracycline-sensitive colonies were screened by PCR using primers 957 and 1194 (for B. multivorans) or 1195 (for B. thailandensis). pFLPe4 has a temperature-sensitive origin of replication; therefore, mutants were grown overnight in LB without antibiotics at 37°C. Single colonies were then tested for kanamycin sensitivity and then by colony PCR for pFLPe4 using primers 1128 and 1129.
Design and construction of the sgRNA-expressing plasmids
PAM sequences closest to the 5’ end of the transcription start site (TSS) were first identified on both the non-template and template strands. We extracted 20-23 nucleotides adjacent to the PAM sequence to design the base-pairing region of the sgRNAs in the following format: 5’-CCN-N(20-23)-3’ for targeting the non-template strand and 5’N(20-23)-NGG-3’ for the template strand. To score the specificity and identify off-target binding sites, the 5’ end of the 20-23nt variable base-pairing sequences were trimmed one base at a time and the remaining base-pairing region was searched against the appropriate organism’s reference genome. This was repeated until only 10 nt was used as a search query. Potential sgRNAs were discarded if off-target sites were discovered in this manner.
The expression vector pSCrhaB2 (Cardona and Valvano, 2005) was chosen as the method of sgRNA expression due to the broad host range of the pBBR1 origin of replication. The sgRNA cassette from pgRNA-bacteria (Qi et al., 2013) (Addgene plasmid # 44251) was introduced into pSCrhaB2 by restriction cloning with EcoRI and HindIII (NEB) to create pSCrhaB2-sgRNA. To remove rhaS and rhaR, inverse PCR was performed using Q5 polymerase (NEB) and primers 847 and 1025. The resulting fragment was ligated by blunt-end ligation using 1µL of PCR product incubated with 0.5 µL DpnI, 0.5 µL T4 polynucleotide kinase, and 0.5 µL T4 ligase (NEB) with quick ligation buffer (NEB) at 37°C for 30 minutes. The resulting plasmid, pSCB2-sgRNA, was screened using primers 781 and 848, which span the ligated junction. Individual sgRNAs were introduced into pSCB2-sgRNA using inverse PCR as previously described (Qi et al., 2013) (Supplemental Table 3).
Construction of the fliF insertional mutant
Inactivation of fliF was performed with the mutagenesis system of Flannagan et al. (Flannagan et al., 2007). Briefly, a 322 bp internal fragment of fliF was PCR amplified from the K56-2 genome using primers 1156 and 1157 and Q5 polymerase (NEB). The fragment and pGPΩ-Tp were double digested with KpnI and EcoRI (NEB) and ligated with T4 ligase (NEB). The resulting plasmid, pAH26, was electroporated into E. coli SY327, and trimethoprim-resistant colonies were screened by colony PCR for the fliF fragment. Triparental matings were performed as above. Trimethoprim-resistant exconjugants, were screened by motility assay (below).
Construction of the phbC insertional mutant
Inactivation of phbC (WQ49_RS30385) was performed as for fliF. Briefly, a 328 bp internal fragment of phbC was PCR amplified from the K56-2 genome using primers 1196 and 1197 and Q5 polymerase (NEB). The plasmid, pAH27, created from ligating the fragment into pGPΩ-Tp using KpnI and EcoRI (NEB) restriction sites, was electroporated into E. coli SY327 and trimethoprim-resistant colonies were screened by colony PCR for the phbC fragment. Triparental matings were performed as above, and trimethoprim-resistant exconjugants were screened by staining for polyhydroxyalkanoate granule accumulation (below).
Assays for integration efficiency and stability of the mini-CTX1-based system
To assess integration efficiency, triparental matings were started as above. However, after the mating on LB agar, the pellicles were serially diluted and plated for CFU/mL on LB agar with 50 µg/mL gentamicin and LB agar with 50 µg/mL tetracycline and 50 µg/mL gentamicin.
To assess stability of the integration, cultures of the dCas9 mutants (containing the tetracycline resistance cassette) were serial passaged over 4 days without antibiotics. Each day, a fresh culture was started by 1:2500 dilution of the previous day’s stationary phase culture. In addition, the cultures were serially diluted and plated for CFU/mL on LB agar without antibiotics and LB agar with 50 µg/mL tetracycline.
Growth assay with phenylacetic acid as the sole carbon source
Overnight cultures, started from isolated colonies, of the appropriate strains were washed at 4000x g for 4 minutes and resuspended in PBS (2.7 mM KCl, 136.9 mM NaCl, 1.5 mM KH2PO4, 8.9 mM Na2HPO4, pH 7.4) to remove growth medium. The OD600nm of the cultures was normalized to 0.01 in M9 medium supplemented with 5 mM phenylacetic acid, 100 µg/mL trimethoprim, and 0.2% rhamnose as required. The culture was added to wells of a 96-well plate and incubated with continuous shaking at 37°C (for B. cenocepacia K56-2) or 30°C (for B. multivorans ATCC 17616 and B. thailandensis E264). The OD600nm of the cultures was measured after 24 hours for B. cenocepacia K56-2 and B. multivorans ATCC 17616, or 48 hours for B. thailandensis E264.
Fluorescent microscopy and polyhydroxyalkanoate granule detection
Overnight cultures of the appropriate strains with or without rhamnose were first washed to remove growth medium and resuspended in PBS. Cells were fixed in 3.7% formaldehyde + 1% methanol at room temperature for 10 minutes (B. cenocepacia K56-2) or 20 minutes (B. multivorans ATCC 17616 and B. thailandensis E264) then quenched by the addition of an equal volume of 0.5 M glycine. The cells were washed and resuspended in PBS with 0.5 µg/mL Nile Red (Carbosynth) and stained at room temperature in the dark for 20 minutes, after which the cells were washed to remove excess stain and resuspended in PBS. The cells were mounted on 1.5% agarose pads and imaged by fluorescence microscopy at 1000x total magnification on an upright AxioImager Z1 (Zeiss). Nile red was excited at 546/12 nm and detected at 607/33 nm.
Plate-based motility assay
Assays were performed as previously described (Kumar and Cardona, 2016), with some modifications. Briefly, strains were grown on LB agar with the appropriate antibiotics and single colonies were stab-inoculated into motility medium consisting of nutrient broth (Difco) with 0.3% agar. Medium was supplemented with rhamnose (Sigma) as appropriate. Plates were incubated right-side up for 22 hours at 37°C.
Flagellum staining
Staining was performed as previously described (Kumar and Cardona, 2016). Briefly, an overnight culture was rested statically at room temperature for 20 minutes. Gently, a 1 in 10 dilution was prepared in water and rested statically for a further 20 minutes. A small drop of the diluted culture was placed on a clean glass slide and rested for 20 minutes. A coverglass was gently applied and one side was flooded with Ryu flagellum stain (Remel), then allowed to dry for 2 hours at room temperature. Slides were observed by light microscopy at 1000x total magnification on an upright AxioImager Z1 (Zeiss).
SDS-PAGE and Immunoblotting
Cells from an overnight culture were subcultured into fresh medium and grown at 37°C (30°C for B. thailandensis) to an OD600nm of 0.4, then exposed to various concentrations of rhamnose for 3 hours. Soluble protein was isolated first by sonicating the cells in TNG Buffer (100 mM Tris-HCl, 150 mM NaCl, 10% glycerol, pH 7.4) then by centrifugation at 15 000g for 20 minutes. Following boiling denaturation in SDS loading buffer (50 mM Tris-HCl, 2% SDS, 0.2% bromophenol blue, 20% glycerol, 100 mM DTT, pH 6.8), samples were run on an 8% Tris/glycine gel. To ensure equal loading, 20 µg protein was loaded per well (as determined by NanoDrop) and gels were run in duplicate (one for immunoblot, and another for Coomassie staining). Protein was transferred by iBlot to a PVDF membrane, blocked in 5% skim milk-TBST (150 mM NaCl, 10 mM Tris-HCl, 0.5% Tween-20, pH 7.5) at room temperature for 1 hour, then probed with a 1:2 000 dilution of primary α-Cas9 antibody (ThermoFisher 10C11-A12) in 5% skim milk-TBST overnight at 4°C. Following washes, the blot was probed with a 1:20 000 dilution of secondary antibody linked to alkaline phosphatase (ThermoFisher G-21060) in 5% skim milk-TBST for 1 hour at room temperature. Protein was detected by incubation with a solution of NBT/BCIP (Roche) as per the manufacturer’s protocols.
Author contributions
STC conceived the idea and design of the research; AMH designed and cloned the dCas9 constructs; AMH and ASMZ designed the sgRNAs, assessed knockdown phenotypes, processed data, and wrote and edited the manuscript; TJL performed qRT-PCR analysis and edited the manuscript; STC supervised the work and provided financial support.
Conflict of Interest Statement
The authors declare no conflict of interest.
Acknowledgements
This work was financially supported by grants from the Cystic Fibrosis Foundation, Cystic Fibrosis Canada, and the Natural Sciences and Engineering Research Council of Canada (NSERC) to STC; AMH was supported by grants from the Canadian Institutes of Health Research (CIHR) and Cystic Fibrosis Canada.
The authors are grateful to Eric Déziel from the Institut National de la Recherche Scientifique – Institut Armand-Frappier for providing the miniCTX1 integration vector.