Introduction

The emergence of resistant Enterobacteriaceae that produce carbapenemases globally has led to the development of serious infections associated with significant morbidity and mortality [1••, 2]. In 2014, the World Health Organization (WHO) released a report that stated that carbapenem-resistant Klebsiella pneumoniae has spread to all regions of the world [3••]. The CDC currently reports that all states in the United States, with the exception of Maine and Idaho, have reported cases of KPC-producing organisms [4•]. There are multiple elements that contributed to the rapid spread of KPCs, such as plasmid borne genes and widespread international travel [5•].

Infections caused by these organisms are often difficult to treat due to resistance. Agents that are used include carbapenem combinations, polymyxins, fosfomycin, tigecycline, aminoglycosides, ceftazidime-avibactam, and meropenem-vaborbactam, however, trials supporting their use are uncommon. New treatments with activity against KPCs are currently being studied to help mitigate the threat of resistance.

This paper will review the evolution and epidemiology of KPCs, current treatment options, and pipeline antibiotics for infections caused by KPC-producing organisms.

Evolution and Epidemology

KPC enzymes are class A carbapenemases that contain a serine residue in the active site and are comprised of 265–269 amino acids. The term “KPC” arose from “Klebsiella pneumoniae carbapenemase,” as they were first identified in this organism [6]. To date, at least nineteen variants of KPC have been reported although KPC-2 and KPC-3 appear to be most prevalent [7].

KPC-producing organisms typically exhibit high levels of penicillin and cephalosporin resistance and variable carbapenem resistance. However, unlike other Ambler Class A beta-lactamase producers, KPCs are insufficiently inhibited by the beta-lactam-based beta-lactamase inhibitors clavulanate, tazobactam, and sulbactam [8]. KPC-producing organisms also tend to be multi-drug resistant as bla KPC is carried on large plasmids with accompanying resistance determinants, including those responsible for resistance to aminoglycosides, quinolones, trimethoprim, sulfonamides, and tetracyclines [9, 10].

The first organism harboring bla KPC was a K. pneumoniae isolate recovered in 1996 from a patient in North Carolina [11]. In 1998, bla KPC was identified in an isolate of Salmonella cubana that was recovered from the stool of a 4-year-old child with gastroenteritis in Maryland [12]. In the same year, KPC was detected in K. pneumoniae isolated from four adult patients in a Baltimore hospital and K. oxytoca in a patient from New York [12, 13]. In 2000–2001, a New York City hospital reported the first known outbreak of hospital-acquired KPC-producing organisms, which involved 14 infected patients, eight of whom died as a result of their infection [14]. By the mid-2000s, some hospitals in New York City were reporting that nearly half of their K. pneumoniae isolates carried bla KPC genes [15•]. The CDC currently reports that all states in the United States with the exception of Maine and Idaho have reported infections with bacteria containing bla KPC [4•]. Interestingly, epidemiologists in New York City recently reported significant declines in prevalence of KPC-producers and suggested that reduction of indwelling device utilization (e.g., urinary catheters), improved techniques for sterilization of endoscopes, and application of infection prevention strategies may have contributed to this decline [16•]. Additional surveillance data will be required to determine if this trend occurs in other parts of the USA and to see if it continues.

The first report of a KPC-producing organism outside of the United States was a patient with prostatic carcinoma who had a bilateral nephrostomy performed in December 2004 at a New York City hospital. He was admitted to a hospital in Paris in February 2005 for acute urine retention. K. pneumoniae with bla KPC was isolated from urine and blood cultures [17]. Additional early cases were reported in Brazil, the United Kingdom, France, and China. Subsequently, KPCs have disseminated rapidly throughout the globe and have been detected in virtually all clinically relevant Enterobacteriaceae [18].

A number of factors contributed to the rapid spread of KPC enzymes. First, bla KPC is plasmid-borne. Plasmids with bla KPC undergo horizontal transfer through conjugation with other bacterial cells. Thus, rapid movement of bla KPC from cell-to-cell is a major contributor to its general spread and accompanying resistance determinants. Second, by the late 2000s, multilocus sequencing typing (MLST) revealed that while bla KPC is present in over 100 different strains of K pneumoniae, ST258 predominates in the United States and around the world. The reason underlying the success of ST258 is unclear but its remarkable ability to disseminate suggests a possible selective or fitness advantage over other clones [5•]. Finally, widespread international travel most certainly facilitated intercontinental transmission of organisms harboring bla KPC .

Outbreaks involving KPC-producing organisms have been described primarily in healthcare settings [19••]. Risk factors for acquisition of KPC-producing organisms are not specific, but include exposure to acute care hospitals or long-term acute care facilities, higher degrees of comorbidity, prior antibiotic use, elevated colonization pressure in endemic settings, and the prolonged use of indwelling and central venous catheters [14, 19••, 20•, 21]. Transfer of patients between long-term acute care hospitals (LTACs) with high KPC levels and other healthcare facilities has been identified as an important mechanism of hospital-to-hospital transmission [22]. Lübbert et al. showed that following an outbreak involving KPC-producing organisms, nearly two-thirds of the 84 colonized study patients spontaneously cleared and tested negative repeatedly by KPC PCR after 6 months, but that colonization for > 3 years was observed in some patients. This underscored the importance of following the colonization status of these cases and consistently applying isolation precautions when they access healthcare [23]. Finally, environmental contamination is thought to be an important mechanism of transmission of KPC-producing organisms. Leitner et al. described a prolonged outbreak involving K. oxytoca with bla KPC in an Austrian hematology ward. Identical ST4 clones of bla KPC-2 were recovered from the sinks in the rooms where cases were admitted in addition to the sink of a medication room. The authors hypothesized that water hitting the mesh in the sink drains aerosolized organisms with KPC when patients were using the sinks for personal hygiene. Renewed enforcement of contact precautions, hand hygiene, and cleaning of the environment (including sinks and equipment) eventually controlled the outbreak [24•].

The use of bundled interventions that include 2% chlorhexidine baths, contact precautions, enhanced environmental cleaning, surveillance cultures, serial point prevalence surveillance, and personnel training has been reported to mitigate patient-to-patient transmission [24•, 25, 26•]. The use of fecal microbiota transplantation (FMT) to displace colonizing multi-drug resistant organisms is a new and novel approach that is under investigation [27••].

Treatment

Treatment options for infections caused by KPC-producing organisms are limited and outcomes for these infections are generally poor. It has been found that patients with bloodstream infections caused by carbapenem-resistant organisms are more likely to have a longer hospital stay, be admitted to an intensive care unit, and suffer increased mortality compared to patients with either ESBL-producing K. pneumoniae or susceptible K. pneumoniae [1••, 28]. Delays in administration of active therapy to patients with KPC-producing infections are common and may affect outcomes, including mortality [29,26,31]. Appropriate treatment is a modifiable predictor of infection outcome.

Current antimicrobial options for infections caused by KPC-producing organisms are limited by their pharmacokinetics, toxicities, and lack of strong clinical evidence supporting their efficacy. Two new beta-lactamase inhibitor combinations have been made available with in the last few years that have been shown to be safer and possibly more efficacious for the treatment of infections caused by KPC-producing organisms compared to some of the older agents, however, supportive data is limited. In addition, KPCs continue to evolve and have developed resistance against some of the current treatment options. Below are currently available treatment options and evidence supporting (or not supporting) their use.

Carbapenem Combinations

Carbapenems are counterintuitive choices of therapy for KPC-producing organisms, but they have been described as useful in some studies, particularly when administered as prolonged or continuous infusions or in combination with other agents, particularly for organisms with only moderately elevated minimum inhibitory concentrations (MICs). However, KPC-producing K. pneumoniae exhibit a marked inoculum effect and MICs may not predict the efficacy of carbapenems against KPC-producing infections [32]. Combination of a prolonged carbapenem infusion with another agent with activity against carbapenemases is a better option than carbapenems alone. A multi-centered, retrospective, cohort study with 661 adult patients illustrated that combination regimens that included meropenem were associated with significantly higher survival rates when the KPC-producing isolate had a meropenem MIC ≤ 8 μg/mL [33].

Dual carbapenem therapy has also been studied for the treatment of infections caused by KPC-producing organisms. This combination is active due to the higher affinity of KPCs for ertapenem, allowing it to function as a suicide substrate and enhance the activity of the concomitant carbapenem [34]. A few case reports and series support the use of dual carbapenem therapy for treatment of infections caused by KPC-producing organisms [35•, 36]. A recent case-control study investigated the clinical impact of regimens including two carbapenems in 144 patients with infections caused by carbapenem-resistant K. pneumoniae (90% producing KPCs) and found that 28-day mortality was lower in patients who received a double-carbapenem regimen compared to those who received treatment with dual or triple agent combinations of colistin, tigecycline, and gentamicin (p = 0.04). Likewise, clinical cure and microbiological eradication were significantly higher when double-carbapenem therapy was used in patients infected with carbapenem-resistant K. pneumoniae resistant to colistin (13/20 (65%) versus 10/32 (31.3%), p = 0.03 and 11/19 (57.9%) versus 7/27 (25.9%), p = 0.04, respectively) [37••].

Although the use of carbapenem-based combination regimens for the treatment of infections caused by KPC-producing organisms has some utility, evidence is limited and testing for synergy is difficult. There are pros and cons with using these regimens (Table 1). If one of them is used, it is important that the pharmacodynamics of the carbapenem used be maximized through prolonged or continuous infusions.

Table 1 Pros and cons of treatments for infections caused by Klebsiella pneumoniae carbapenemase-producing organisms

Polymyxins

The polymyxins are a class of polypeptide antibiotics that have been revitalized for the treatment of infections caused by resistant Gram-negative organisms, including those producing KPCs. Polymyxins have in vitro activity against most K. pneumoniae, however, heteroresistance is common [38••, 39, 40]. When the isolate is exposed to a polymyxin, killing of the susceptible population amplifies resistant subpopulations. In vitro studies suggest that polymyxin monotherapy may lead to emergence of resistance and that they should be administered in conjunction with other agents [41, 42••]. Carbapenems have been studied in combination with polymyxins and showed in vitro synergy with polymyxins against carbapenem-resistant K. pneumoniae, including KPC-producing K. pneumoniae [42••, 43].

When using polymyxins, appropriate dosing and potential adverse effects are concerns. Modern pharmacokinetic and pharmacodynamic studies show that dosing recommendations in both polymyxin B and colistin package inserts are inaccurate [44, 45]. The prodrug colistimethate sodium (CMS) is cleared renally while the active drug colistin is eliminated through non-renal means. However, since colistin is administered as CMS, the amount of formed active drug is dependent on renal function. In fact, these studies show difficulty in achieving adequate plasma concentrations of formed colistin in patients with a creatinine clearance > 80 mL/min/1.73m2 [46••]. The ability to reach therapeutically useful lung concentrations is also uncertain [47••]. In contrast, pharmacokinetic studies investigating polymyxin B showed that doses should not be based on renal function as recommended in the package insert because polymyxin B is not renally cleared [48]. Accordingly, colistin would be the better choice of the polymyxins for urinary tract infections (UTIs). Although both agents have the potential to cause nephrotoxicity and neurotoxicity, several comparisons have shown a significantly higher risk of acute kidney injury in patients who received colistin compared to polymyxin B [49••, 50, 51].

In practice, these agents are limited by their toxicity, the unfavorable pharmacokinetics of colistin, and a lack of availability of polymyxin B in some parts of the world. Though polymyxin-based regimens have been commonly used for KPC-producing infections, recent small studies of new beta-lactamase inhibitor combinations have shown favorable results compared to these polymyxin-containing regimens.

Fosfomycin

Fosfomycin is frequently active against KPC-producing organisms [52]. However, resistance develops rapidly when fosfomycin is used as monotherapy, therefore, combination therapies are recommended [53]. In vitro studies have evaluated fosfomycin in combination with imipenem, ertapenem, tigecycline, colistin, meropenem, gentamicin, and amikacin and showed additive effects against KPC-producing strains with fosfomycin and amikacin showing the most potent activity even when the causative bacteria were resistant to amikacin [54, 55•, 56].

The utility of fosfomycin is limited in the United States since it is only available as an oral formulation. Due to its moderate absorption, fosfomycin is limited to the treatment of urinary tract infections. While this is a limitation, UTIs are commonly caused by KPC-producing organisms. Only two case reports showing that it has been clinically effective against infections caused by KPC-producing organisms have been published [57]. A study of IV fosfomycin for UTIs for approval in the United States is currently being conducted (NCT02753946) so it may be a treatment option in combination with another agent in the future in the United States.

Aminoglycosides

Aminoglycosides are well-established agents that have potent bactericidal activity against Gram-negative bacteria [55•]. It is important to note that aminoglycosides do not have equal utility against KPC-producing isolates. When tested in vitro against 25 KPC-producing isolates, susceptibility rates were 48% to amikacin (CLSI breakpoint ≤ 16 mg/mL, 44% to gentamicin (at ≤ 4 mg/mL) and 8% to tobramycin (at ≤ 4 mg/mL) [56]. Notably, CLSI breakpoints are higher than those suggested by some other organizations, and if lower breakpoints are used many less isolates would be considered susceptible [58].

Based on limited data available, aminoglycosides appear to be a suitable option for the treatment of UTIs caused by KPC-producing isolates [55•, 59, 60•]. Patients in one small study of UTIs caused by carbapenem-resistant K. pneumoniae had significantly higher clearance rates when treated with an aminoglycoside (88%) compared to those receiving either a polymyxin (64%) or tigecycline (43%) [61]. For other infections, evidence supporting aminoglycosides as monotherapy is lacking, and studies have shown a mortality rate from 6.3–80% in patients that received them alone [62]. This may be due to their limiting pharmacokinetic profile, however, they may have a place in therapy when given in combination with another antibiotic. A systematic review of 20 clinical studies evaluating patients being treated for infections caused by carbapenemase-producing K. pnemuoniae including KPC-type enzymes showed that the combination of aminoglycosides with a carbapenem had the lowest mortality rates [62].

The role in therapy for aminoglycosides as monotherapy should be limited to UTIs. They also have a role in combination regimens with other active agents, though the need for this may be somewhat mitigated by the introduction of the new beta-lactamase inhibitor combinations. Though evidence is lacking, they may also have an adjunctive role in as aerosolized agents in patients with pneumonia caused by KPC-producing organisms [63]. The utility of aminoglycosides must always be weighed against their well-known toxicities.

Tigecycline

Tigecycline, a glycyclcycline tetracycline, has broad spectrum activity including against KPC-producing Enterobacteriaceae. Some in vitro studies show all KPC-producing isolates to be susceptible to tigecycline [64]. However, clinicians should be cautious when using tigecycline since it has limitations. Resistance development during therapy has been seen in the setting of carbapenem-resistant K. pneumoniae bacteruria [65]. Pharmacokinetic limitations include bloodstream concentrations that average below the CLSI Enterobacteriaceae breakpoint of 2mcg/mL due to its extensive tissue penetration [66]. Also, since tigecycline is eliminated primarily through biliary excretion, minimal concentrations in the bladder lead to questionable utility for UTIs. Finally, a clinical study of tigecycline compared with imipenem for ventilator-associated pneumonia was statistically inferior to imipenem, leading to the concern that higher-than-approved doses may be necessary. Several small studies have evaluated the use of tigecycline in KPC-producing infections, in combination with other agents and often in higher doses than recommended in the labeling (200 mg loading dose, than 100 mg IV q12 hours) [67,64,65,70]. One multicenter retrospective cohort study in 661 patients found that triple-combination therapy with colistin, tigecycline, and meropenem was associated with lower mortality [OR: 0.11;95%CI:0.02–.69]; p = 0.01, in bloodstream infections caused by KPC-producing isolates compared to single-drug regimens of colistin, tigecycline, or gentamicin [33].

The results of evaluations of tigecycline-contained studies have been variable and difficult to interpret due to the multiple regimens used in the studies, but some have shown a mortality benefit when using combination regimens that include tigecycline. It is important to note that these studies were conducted prior to the availability of the newer beta-lactamase-inhibitor-containing regimens and comparisons between these regimens are limited. In addition, any discussion of tigecycline warrants mention of the boxed warning of a higher risk of death compared to other antibacterial drugs [70]. This warrants caution when considering tigecycline when other highly active agents are available but should not prevent its use when they are not.

Ceftazidime-Avibactam

Ceftazidime-avibactam is a newly approved agent combining an antipseudomonal cephalosporin and novel non-beta-lactam beta-lactamase inhibitor. Avibactam has a diazabicyclooctane structure that is not based on the beta-lactam ring and does not function as a suicide substrate like beta-lactam-based inhibitors. It is also stable against KPC enzymes [71]. One surveillance study showed ceftazidime-avibactam active against 97.5% (117/120) of KPC-producing clinical isolates [72]. Prospective studies support the in vivo efficacy of ceftazidime-avibactam against ceftazidime-resistant isolates, with one study showing it to be superior to best available therapy (BAT), mostly carbapenems, in UTIs caused by these organisms [73•]. There are limited but growing clinical data illustrating ceftazidime-avibactam efficacy in treating infections in humans caused by KPC-producing pathogens [74,71,76].

Two small studies have shown clinical efficacy of monotherapy ceftazidime-avibactam in treating a variety of infections caused by CRE [77••, 78, 79]. Two studies have been performed comparing ceftazidime-avibactam-based regimens with other regimens for CRE infections, predominantly caused by KPC-producing pathogens. The first retrospectively evaluated carbapenem-resistant K. pneumoniae (97% KPC-producing strains) bloodstream infections and found that ceftazidime-avibactam treatment in these patients was associated with significantly higher clinical success 11/13 (85%) compared to those that received another treatment regimen which included a carbapenem plus aminoglycoside, (12/25(48%); p = 0.04) or colistin (12/30(40%); p = 0.009) and other regimens (15/41(37%); p = 0.004) [80••]. The second study was a multicenter observational study comparing ceftazidime-avibactam (n = 38) to colistin (n = 99) for the treatment of multiple types of infections caused by KPC-producing CRE. Most of these infections were bloodstream (n = 63, 46%) and respiratory (n = 30, 22%), and most patients received an additional anti-CRE agent. From a multivariate analysis, among patients treated with ceftazidime-avibactam versus colistin, all cause in-hospital mortality 30 days after the start of treatment was 3/38 (8%) versus 33/99 (33%), respectively. After adjustment, the 30-day mortality after starting treatment was 9% vs. 32% (p = 0.0012), for ceftazidime-avibactam versus colistin, respectively [81••].

Resistance to ceftazidime-avibactam has been reported to emerge during therapy. In a retrospective study in one medical center, emergent ceftazidime-avibactam resistance was detected in 3/10 (30%) of microbiological failures with carbapenem-resistant K. pneumoniae in patients that were treated with it. Most mutations have occurred within the KPC Ω-loop position, resulting in enhanced ceftazidime binding and restricted avibactam binding [82••, 83]. It is important that clinicians are aware of the possibility of emergent resistance leading to ceftazidime-avibactam treatment failure, retesting susceptibilities if KPC-producing organisms are persistent in cultures.

Meropenem-Vaborbactam

Meropenem-vaborbactam received FDA approval in August 2017. Vaborbactam is a novel boron-containing serine-beta lactamase inhibitor that works by creating a covalent bond between its boron moiety and the serine side chain of beta-lactamases, preventing them from destroying beta-lactams. When tested in vitro against 133 clinical KPC-producing Enterobacteriaceae strains, 131 (98.5%) were inhibited by meropenem-vaborbactam [84•].

The clinical data supporting meropenem-vaborbactam for KPC-producing infections is limited to a single randomized trial (TANGO-2). It investigated the efficacy of meropenem-vaborbactam compared with investigator-chosen BAT in patients with serious infections caused by confirmed or suspected carbapenem-resistant Enterobacteriaceae. Randomization in the trial was stopped early after the results of an interim analysis showed statistically significant differences in favor of meropenem-vaborbactam over BAT for clinical cure at the test of cure visit (meropenem-vaborbactam 57.1% (16/28) versus BAT 26.7% (4/15); 30.5% [95% CI: 1.5%–59.5%], p = 0.04. Mortality at 28 days in the pooled patient population with bacteremia, hospital-acquired bacterial pneumonia, or ventilator-associated bacterial pneumonia was 17.9% (5/28) for meropenem-vaborbactam versus 33.3% (5/15) for BAT [85•].

Treatments in Late Stage Development

After years of stagnation, governmental attention and incentives have encouraged the development of new antibiotics. Several drugs in late-stage clinical development are active against KPC-producing organisms.

Cefiderocol is a cephalosporin with a novel mechanism of action. Unlike other beta-lactams, it contains a catechol moiety that acts like a siderophore, taking advantage of active iron transport mechanisms in bacteria instead of relying solely on porin channels to penetrate organisms. Cefiderocol has been shown to have potent in vitro activity against CRE, including both KPC-producing Enterobacteriaceae and those producing metallo-beta-lactamases [86]. Cefiderocol is currently under clinical development and recently a registration study was completed in patients with complicated urinary tract infection (cUTI) (APEKS*-cUTI) (NCT02321800) (Table 2). Cefiderocol met the primary efficacy endpoint and was superior to imipenem-cilastatin with a weighted difference in primary outcome of 18.58% (95% CI 8.23%–28.92%). Additional Phase 3 trials are being conducted (Table 2).

Table 2 Treatment options in late stage development

Plazomicin is an aminoglycoside with activity against KPC-producing isolates. Two phase 3 trials have been completed looking at plazomicin for treatment of cUTI (NCT01970371) and serious infections due to CRE (NCT02486627) (Table 2). For the treatment of complicated urinary tract infections including pyelonephritis, a statistically significant difference favoring plazomicin over meropenem was demonstrated at test of cure (TOC) visit in clinical cure and microbiological failure, 81.7% versus 70.1%, respectively, 11.6% (95% confidence interval (CI): 2.7–20.3%) driven primarily by a higher microbiological eradication rate at TOC in the plazomicin group. In addition, at the late follow-up visit, the composite cure rate was significantly higher in the plazomicin group compared with the meropenem group. For the treatment of serious carbapenem infections, when used as a part of a combination regimen (tigecycline or meropenem), plazomicin (n = 17) compared to colistin (n = 20), plazomicin was associated with reduced all-cause mortality at day 28, 11.8% vs. 40%; 28.2% (95%CI: 0.7–52.5%). In addition to lower mortality rates, it was associated with a favorable safety profile compared with colistin, including a lower incidence and magnitude of serum creatinine elevations [87].

Relebactam is a bicyclic diazabicyclooctane, non-beta-lactam beta-lactamase inhibitor of class A (ESBLs and KPCs) and class C beta-lactamases that is similar in structure to avibactam. It is being studied in combination with imipenem-cilastatin. An amount of 4 μg/ml relebactam was required for imipenem-resistant K. pneumoniae isolates expressing KPC carbapenemases to become susceptible to imipenem and reduced MICs of imipenem against Enterobacter spp. isolates, including imipenem-resistant strains producing KPC enzymes [88•, 89]. The efficacy and safety of imipenem-cilastatin-relebactam has been studied compared to the combination of colistin and imipenem-cilastatin in a phase III trial of imipenem-resistant infections (NCT02452047). Results are pending (Table 2).

Eravacycline is a flurocycline tetracycline with a similar mechanism of action to other tetracyclines. This drug is different from others in its class because its antibacterial activity is barely affected by active drug efflux systems and ribosomal protection proteins [90]. Eravacycline has potent in vitro activity against KPC-producing E. coli and K. pneumoniae, with MIC90s of 0.5 μg/ml and 2 μg/ml, respectively [91, 92•]. Two clinical trials investigating the use of eravacycline for cIAI (NCT01844856) and cUTI (NCT01978938) have been completed. For the treatment of cIAI, the rates of clinical cure at the test-of-cure visit were eravacycline 86.8% (191/220) versus ertapenem 87.6% (198/226) for the microbiological intent-to-treat population. The difference in clinical cure rates between the groups was −0.80% (95%CI,−7.1–5.5%) [93••]. However, when compared to levofloxacin for the treatment of complicated UTIs, eravacycline did not achieve the primary endpoint.

Conclusion

KPC-producing organisms cause infections with high morbidity and mortality. Limited treatment options are available, though new therapies have been developed. It is likely that the reported poor outcomes in patients with KPC-producing infections are partially due to both delays to active therapy and the historical need to use antibiotics with suboptimal pharmacokinetics and tolerability. Increased use of rapid diagnostics can help to address therapy delays, while new agents are beginning to address the latter.

Emerging data suggests that new beta-lactamase inhibitor combinations are superior choices of therapy for most patients, though more supportive data is needed. It should be noted that studies showing better outcomes with combination therapies for KPC-producing infections were conducted prior to the availability of new beta-lactamase inhibitor combinations. Pipeline agents are likely to have a place in therapy for the treatment of infections caused by KPC-producing isolates. However, KPCs will continue to evolve, and new agents or regimens should continue to be studied to mitigate the never-ending threat of antibiotic resistance.