CRISPR-Cas influences the acquisition of antibiotic resistance in Klebsiella pneumoniae

In the US Carbapenem resistance in Klebsiella pneumoniae (Kp) is primarily attributed to the presence of the genes blaKPC-2 and blaKPC-3, which are transmitted via plasmids. Carbapenem-resistant Kp (CR-Kp) infections are associated with hospital outbreaks. They are difficult to treat, and associated with high mortality rates prompting studies of how resistance is obtained. In this study, we determined the presence of CRISPR-Cas in 304 clinical Kp strains. The CRISPR-Cas system has been found to prevent the spread of plasmids and bacteriophages, and therefore limits the horizontal gene transfer mediated by these mobile genetic elements. Here, we hypothesized that only those Kp strains that lack CRISPR-Cas can acquire CR plasmids, while those strains that have CRISPR-Cas are protected from gaining these plasmids and thus maintain sensitivity to antimicrobials. Our results show that CRISPR-Cas is absent in most clinical Kp strains including the clinically important ST258 clone. ST258 strains that continue to be sensitive to carbapenems also lack CRISPR-Cas. Interestingly, CRISPR-Cas positive strains, all non-ST258, exhibit lower resistance rates to antimicrobials than CRISPR-Cas negative strains. Importantly, we demonstrate that the presence of CRISPR-Cas appears to inhibit the acquisition of blaKPC plasmids in 7 Kp strains. Furthermore, we show that strains that are unable to acquire blaKPC plasmids contain CRISPR spacer sequences highly identical to those found in previously published multidrug-resistance-containing plasmids. Lastly, to our knowledge this is the first paper demonstrating that resistance to blaKPC plasmid invasion in a CRISPR-containing Kp strain can be reversed by deleting the CRISPR-cas cassette.


Introduction
Klebsiella pneumoniae (Kp) is a gram-negative bacterium that predominately causes nosocomial infections including sepsis, urinary tract infections, and pneumonia. With increasing spread they now also cause community-acquired infectons. Carbapenem pneumoniae (CR-Kp) has increased in prevalence in the United States from 1.6% in 2001 to 10.4% in 2011 [1]. Treating CR-Kp is challenging. The combination of increased vulnerability of hospitalized patients with serious comorbidities, and inadequate treatment options explain overall 30-day mortality rates that are as high as 39-50% in bacteremic CR-Kp infected patients [2][3][4][5][6][7]. In the United States, resistance to carbapenem antibiotics in Kp strains is predominantly conferred by the two antibiotic resistance genes, bla KPC-2 and bla  , which encode class A carbapenem-hydrolyzing β-lactamases, the Klebsiella pneumoniae carbapenemases (KPC) are transfered through plasmids. In these KPC ß-lactamases convey resistance also against clavulanic acid, monobactams, and cephalosporins [8,9]. The majority of clinical isolates of CR-Kp in the US belong to the multi-locus sequence type (MLST)-defined clone ST258 [4]. This clone emerged in the US in the 2000s and is now prevalent worldwide [3,10]. Finally, numerous hospital outbreaks have been associated with the ST258 clone and its related variants making its investigation clinically imperative [11,12].
While the presence of either bla KPC-2 or bla KPC-3 is likely not the only factor contributing to the success of ST258, their acquisition provides the bacteria with a significant advantage against antibiotic therapeutics. It is noteworthy that not all ST258 strains exhibit resistance to carbapenem. As transmission of bla KPC genes is facilitated by plasmids, we analyzed the presence and function of CRISPR-Cas in carbapenem-sensitive (CS)-Kp and CR-Kp hospital strains. The CRISPR-Cas system is composed of cas genes followed by groupings of brief direct repeats and variable spacer regions that match infective agents of bacteria and archea, such as bacteriophages and plasmids [13][14][15][16]. This system has been found to prevent the spread of plasmids and bacteriophages [17][18][19], and therefore limit the horizontal gene transfer by these mobile genetic elements. For example, it was recently determined that hospital acquired, multidrug resistant strains of Enterococcus faecium lack CRISPR while community-acquired, drug sensitive strains of E. faecium contained CRISPR [20]. Additional studies demonstrated a significant association between antibiotic resistance in E. faecium strains and the absence of a CRISPR-Cas system, and correlation between decreased CRISPR-Cas and reduced defense against plasmid acquisition [21,22]. CRISPR-Cas has therefore been compared in its function to an acquired immune system in bacteria.
We explored presence of the CRISPR-Cas systems in clinical Kp strains including those of the ST258 clonal background. Previous analysis on Kp genomes have identified presence of CRISPR-Cas, though its role in this species is yet to be defined [23,24]. Until this study, little was known about the prevalence of CS-Kp strains and the predisposition of Klebsiella strains to become resistant to carbapenem antibiotics. While investigation of the overall prevalence of CRISPR-Cas in Kp has begun, the presence of CRISPR-Cas in CR versus CS-Kp and in the clinically important ST258 clone has not yet been studied. This project aimed to further elucidate the function of CRISPR-Cas in Kp immunity and examine the ST258 clonal background that has spread globally and has been highly associated with recent hospital outbreaks.

Clinical isolates
Specimen collection was done under the approved IRB#648612 from Stony Brook University. No informed consent was obtained as samples were laboratory leftovers and deidentified. 304 hospital Kp isolates derived from bodily fluids from 265 patients were randomly collected from the microbiology laboratory at Stony Brook University Hospital (SBUH) between 2015-2016 (over a 14-month period) (S1 Table). The SBUH Microbiology laboratory regardless of the site of isolation kept isolates and our laboratory collect them. 83.3% were isolated from urine samples; 9.8% from respiratory samples, 4.6% from wounds and 2.2% from blood. This unbiased approach was chosen to permit identification of carbapenem sensitive ST258 strains. Sensitivity to all antibiotics was determined by the SBUH microbiology laboratory as part of standard diagnostic work up and obtained by us by review of medical records.

Sequence typing of ST258 clone
Screening of ST258 clonal background was performed by PCR and sequencing of the gapA and tonB genes (S1 Fig). PCR was performed on colony DNA, analyzed by gel electrophoresis, and PCR products were purified using the Qiagen PCR purification kit. Purified PCR products were sent for Sanger sequencing to Genewiz™. If gapA and tonB results were gapA3 and tonB79, ST258 clonal background was confirmed by regular MLST typing of the rest of the gene array [25].

CRISPR-Cas detection
Primers designed to detect cas1, cse1 and cas3 genes by PCR are listed in Table 1. Primers were designed using sequences provided by the webtool CRISPRFinder, which contained 111 CRISPR positive K. pneumoniae genomes and were representative of different CRISPR-Cas cas genes consistently seen in Kp CRISPR-Cas systems [16]. PCR was performed on colony DNA and results were analyzed by gel electrophoresis.

Transformation assays
bla KPC-2 and bla KPC-3 plasmids were extracted by classic alkaline lysis extraction from previously reported CR strains M#34 and M#36 [4]. pPROBEKT plasmid (Km R , pVS1/p15a vector, with GFP-reporter gene) was used as a control for transformation [26]. 40 CS-Kp strains, 20 with and 20 without CRISPR-Cas, were randomly chosen for these studies and belong to different ST types as indicated by the combination of gapA and tonB alleles. CS-Kp strains were rendered electrocompetent and 100 ng of each of the plasmids was used in each of the transformation assays.

Spacer analysis
Based on previously published CRISPR-Cas architectures, primers were designed to amplify the CRISPR region in the two different CRISPR-Cas systems, type I-E and type I-E � , differentiated by flanking 3' and 5' genes ( Fig 1) [22]. Primers used are listed in Table 1.
Genewiz™ was used to sequence PCR products. PCR sequences were analyzed with the CRISPRfinder tool to identify direct repeats and spacer sequences. Spacer sequences were then BLASTed using the NCBI non-redundant nucleotide database than includes all deposited genomes, plasmids or phages. BLAST search identified sequence homologies to previously reported DNA sequences.

CRISPR deletion isogenic strains
To produce isogenic strains lacking CRISPR spacer regions, a Red/ET recombination system was used following manufacturer's protocol (Quick and Easy Escherichia coli gene deletion kit; Gene Bridges, Heidelberg, Germany), but with modifications. Briefly, a linear DNA fragment with a chloranphenicol resistance and 70bp arms homologous to DNA upstream and downstream of I-E � CRISPR2 was amplified by PCR from the plasmid pCLF3 [27] with the primers listed in Table 1. The linear fragment was used to replace I-E � CRISPR2 regions in the 7 CRISPR-positive strains analyzed using the pRed/ET Tetracycline resistant plasmid. Positive transformants were confirmed by PCR using I-E � CRISPR2 specific primers.

ST258 is significantly more carbapenem-resistant than other Kp clonal backgrounds
A PCR-based approach was designed that would allow identification of ST258 strains including those with maintained sensitivity to carbapenem antibiotics. We randomly collected 304 Kp strains (resistance unknown at collection time) that were identified by the microbiology laboratory of Stony Brook University Hospital (SBUH). PCR sequence typing for 2 genes (gapA, tonB) was performed to identify strains that strains belong to the ST258 clonal background. Those strains bearing gapA3 and tonB79 alleles were further typed by standard MLST (see Materials and methods section and S1 Fig). This approach identified a total of 45 ST258 strains (14.8%), 40 CR (89.2%) and 5 CS (11.1%) (Fig 2A). The ST258 clonal background accounted for 85.1% of all CR samples identified ( Fig 2B). 257 Kp strains were identified as CS  (84.5%), 5 of which were ST258. As expected Kp strains of the ST258 clonal background were significantly more likely to be CR compared with Kp strains of the non-ST258 clonal backgrounds (p < 0.0001, two-sided T test).

CRISPR-Cas inhibits transformation of bla KPC plasmids in CS-Kp
Next, transformation studies were performed to explore whether the presence of CRISPR-Cas protects against the acquisition of bla KPC plasmids in CS-Kp strains. These experiments showed successful transformation with a control plasmid in 11 of 20 CS-Kp strains with CRISPR-Cas (Fig 3). However, of these 11 strains, only 2 (18%) could be transformed with either the bla KPC-2 (SBU#63) or the bla KPC-3 (SBU#20) plasmid (but not both) (Fig 3). In contrast, 15 of 20 CS-Kp isolates that lacked CRISPR-Cas sequences were successfully transformed with the control plasmid and 11 of those (73%) were successfully transformed with both bla KPC-2 and bla KPC-3 . Hence, these data indicate higher transformation success of bla KPC plasmids in strains lacking CRISPR-Cas compared with strains containing CRISPR-Cas (Fisher's exact test, p-value = 0.0017 for both plasmids). No significant difference in control plasmid transformation was observed between groups. We also attempted to transform all CS ST258 strains (n = 5) with a control plasmid and bla KPC . Since kanamycin resistance was required for the selection of transformed bacteria with this control plasmid, only 2 of 5 isolates could be studied in this experiment (SBU#53 and SBU#181) as the rest were kanamycin resistant. Both strains were able to be transformed with bla KPC-2 , but not with bla KPC-3 . Table 2. Antibiogram analysis of CS-Kp strains. Numbers and percentages represent non-susceptibilities for that specific antibiotic.

CRISPR spacer sequences in strains protected from plasmid acquisition match other multidrug-resistant plasmids
The CRISPR spacer sequences in strains that could not be transformed with bla KPC-2 and bla KPC-3 plasmids were further analyzed. Specifically, we explored whether spacer sequences in the CRISPR array of these strains contained genomic sequences homologous to known multidrug resistant plasmids. CRISPR in these strains was amplified and then sequenced via a PCR primer set generated from a previously published CRISPR-Cas database of 55 Kp strains (18). These PCR primers amplify all described Kp-CRISPR-Cas architectures [24] (Fig 1). PCR was performed on 7 of the CRISPR-Cas positive strains (SBU #28, 31, 61, 63, 70, 83, and 142), all of which were resistant to transformation with bla KPC plasmids. In each of these strains, sequence data identified the CRISPR-Cas system as I-E � CRISPR type [24] (Fig 1). Strains SBU #28, 70, and 142 had I-E � CRISPR2 and I-E � CRISPR3, whereas strains SBU #31, 61, 63 and 83 contained only I-E � CRISPR2. The I-E � CRISPR2 and I-E � CRISPR3 direct repeat sequences were found to be identical to the previously reported consensus sequences [24]. Each CRISPR-Cas positive strain contained between 3 and 11 spacer sequences. Spacer sequences exhibited an 87-100% match to exogenous DNA from either known plasmids or phages [28]. Notably, at least one spacer sequence in each of the clinical isolates analyzed either partially or fully matched plasmids (89-100% identity) that were reported to confer resistance to one of the following: metals, non-carbapenem antibiotics, carbapenem antibiotics via bla KPC genes plasmids, or carbapenem antibiotics via bla NDM genes (Table 3).

With deletion of CRISPR spacer sequences, transformation was enabled in Kp strain #61
Attempts were made to delete the I-E � CRISPR2 spacer sequences in the 7 above described CRISPR-Cas positive strains (using homologous recombination, as this CRISPR2 was present in both I-E and I-E � strains). Inherent resistance of some strains to the pRedET plasmid permitted successful deletion of the I-E � CRISPR2 spacer in only strains SBU#28, SBU#61, and SBU#70. Once the I-E � CRISPR2 spacer was deleted, only SBU#61 was successfully transformed with bla KPC-2 and bla KPC-3 plasmids. Notably, SBU#61 only carries I-E � CRISPR2, whereas SBU#28 and SBU#70 also carry I-E � CRISPR3.

Discussion
This study determined the prevalence of CS-Kp in ST258 colonized patients in one major hospital in New York State. CR-Kp was responsible for 15.5% of Kp infections in our patient  [29]. ST258 was responsible for 85.1% of the CR-Kp in our patient population, which is higher than previously reported [4,7]. Additionally, this study determined the prevalence of CRISPR-Cas in CR-Kp and CS-Kp for both ST258 and all other strains tested. CRISPR-Cas was only identified in CS-Kp strains that were not of the ST258 clonal background, which has been reported by others [30][31][32]. Overall the proportion of strains with CRISPR-Cas systems in our isolate population was only 11.5%, consistent with data of other analyzed Kp genomes, but much lower than data derived from Escherichia coli isolates [33]. This difference could be related to the species itself or to technical differences in our studies, as our approach detected only the presence of Cas proteins (full CRISPR-Cas systems). As the CRISPR-Cas system appears not to be highly prevalent in Kp species, it remains to be determined how the absence or presence of the system could be contributing to the evolution of Kp strains. Our data describe CS strains that belong to the ST258 clonal background that lacked CRISPR-Cas systems an. Of these strains 80% strains exhibit other drug resistance. Future in depth characterization of the sole pan-sensitive CS ST258 strain identified in this study, and others if identified, might help explain why some strains do not acquire antibiotic resistance. Ultimately, it would be interesting to know what, if any, fitness advantage there is for pan-sensitive strains.
Most of the Kp strains containing CRISPR-Cas were pan-sensitive whereas resistance to multiple antibiotics was observed in the strains lacking CRISPR. In line with this, our transformation studies demonstrated successful transformation of the bla KPC-2 and bla KPC-3 plasmids into a significantly larger proportion of Kp strains lacking CRISPR-Cas. Although some strains could not be transformed with the control plasmid, successful control plasmid transformation was comparable among strains with and without CRISPR-Cas systems. The impedance of transformation may have been due to transformation competency preparation restrictionmodification systems inherent to the strains or incompatibility (Inc) groups if the strains harbored other plasmids.
Notably, transformation of the bla KPC-2 plasmid was successful in only 1 CS-Kp strain containing CRISPR-Cas (SBU#63), as was transformation of the bla KPC-3 plasmid (SBU#20). Interestingly, the former strain had a CRISPR-Cas spacer sequence that matched a published bla KPC-3 plasmid. The ability to be transformed with one plasmid but not the other could be related to the CRISPR-associated spacer sequences that match either bla KPC-2 or bla KPC-3 plasmids, but not both, as is the case in SBU#63. Sequence analysis of spacers in 7 CRISPR-Cas positive strains indicated that they contained sequences that were more than 90% identical to sequences found in multidrug-resistance plasmids carrying either carbapenemase genes or resistance genes against heavy metals. CRISPR-Cas and its associated spacer sequences could therefore help explain why some Kp strains are less susceptible to the acquisition of antibioticresistance plasmids.
Furthermore, after successful deletion of the I-E � CRISPR2 region in strain SBU#61, which previously could not be transformed with any bla KPC plasmids, transformation of this strain with bla KPC-2 and bla KPC-3 plasmid was successful. This supports the notion that CRISPR sequences initially shielded this strain from acquiring carbapenem resistance by preventing integration of the bla KPC-2 and/or bla KPC-3 plasmid. At this point it remains unclear why transformation with bla KPC-2 and or bla  was not successful in other tested strains SBU#28 or SBU#70, even though the CRISPR cassette was also deleted. It is important to note, however, these strains contained CRISPR3 spacers, and our deletion strategy only targeted the CRISPR2 locus. This CRISPR3 sequences may still target sequences in the plasmids, preventing the transformation.
One major limitation of our data is that we do not have the complete sequences of the bla KPC-2 and bla KPC-3 plasmids of our Kp strains, and, in addition, we did not do transformation studies using other plasmids identified to have high identity to the spacer regions. As a result, we cannot conclusively demonstrate that CRISPR-Cas spacers convey immunity due to similarity of the CRISPR spacers to plasmid sequences. Lack of transformation success could also be due to the presence of other plasmids that share the same incompatibility system with the bla KPC plasmids, although 6 of the 7-sequenced CRISPR-Cas positive were pan-sensitive to all antibiotics pointing against that line of thought. Finally, these results, while based on limited data from 7 CRISPR-Cas positive strains, do differ from a previous analysis of CRISPR-Cas in E. coli isolates where the proportion of spacers matching plasmid genomes was found to be low (20%), and there was no association between CRISPR-Cas and antibiotic resistance [34].
PCRs done to amplify CRISPRs sequences assigned all of our strains to the I-E � CRISPR type, a subtype of the Kp I-E CRISPR type [24]. It would be interesting to investigate whether phenotypic or geographic differences exist between strains harboring different CRISPR types.
A limitation of this study is that this study only analyzed clinical isolates from SBUH and results may therefore not reflect the transmission or characteristics of Kp strains and plasmids present in other geographical areas. Future studies are needed to investigate why some ST258 strains remain sensitive to carbapenems and whether these strains compensate by having other fitness advantages or virulence genes. It is possible that the ST258 clone evolved rapidly in the past in a selective pressure environment (i.e. the clinical setting) without the CRISPR--Cas system as a barrier against acquisition of exogenous DNA. Although whole genome sequence studies suggested that ST258 spread via local expansion rather than by repeated uptake of multiple antibiotic-resistance plasmids, acquisition of novel virulence genes has been described in ST258. This acquisition remains a concern for other Kp clones as well because it may contribute to the evolution of hypervirulent, multidrug-resistant strains [35,36]. While further studies need to be done, decreased prevalence of the CRISPR-Cas system in the ST258 clonal background may be a major contributor to its clonal success, acquisition of antibiotic resistance plasmids and virulence genes, hospital outbreaks, and global spread.