Hypervirulent Klebsiella pneumoniae

The first clinical report that brought hvKp to the forefront was a 1986 publication by Liu et al., who reported 7 cases of invasive K. pneumoniae infection in individuals from the community who presented with hepatic abscess in the absence of biliary tract disease and septic endophthalmitis (31). Some individuals had additional infectious syndromes, such as meningitis, pneumonia, and prostatic abscess. Several features of these patients were distinctive and characteristic for hvKp. First, those infected were healthy members of the community, although 4/7 were diabetic. Second at presentation, patients either had multiple sites of infection or had experienced subsequent metastatic spread. At that time most infections due to K. pneumoniae were occurring the health care environment, and unlike the case for selected Gram-positive pathogens (e.g., Staphylococcus aureus), it was unusual for infections due to Enterobacteriaceae to involve multiple sites or undergo metastatic spread. However, the moniker of hvKp was not yet assigned to these strains.

Interestingly, a 1986 nonclinical report by Nassif and Sansonetti described seven strains of K. pneumoniae (K1 and K2 serotypes) that were highly virulent in mice as demonstrated by a 50% lethal dose (LD₅₀) of <10³ CFU (32). A more detailed analysis of 4 of these strains demonstrated the presence of a large (180-kb) plasmid that contained genes for the production of aerobactin and its cognate receptor. This plasmid was absent in avirulent strains as defined by an LD₅₀ of >10⁶ CFU. Subsequent studies demonstrated that this plasmid also contained genes that conferred a hypermucoid phenotype, which proved to be mediated by the capsular polysaccharide regulator RmpA (33, 34). Details on the clinical syndromes caused by these strains were not reported. However, based on our present understanding of genes and phenotypes that define hvKp (17), these isolates would be predicted to be hvKp.

In 2004, Fang et al. reported that K. pneumoniae strains that caused hepatic abscesses in patients from Taiwan were more likely to possess a hypermucoviscous phenotype than noninvasive strains (15). Hypermucoviscosity was defined by the formation of viscous strings >5 mm in length when a loop was used to stretch the colony on agar plate, also known as a positive string test (15). A subsequent report further supported this association (35). As a result, for a period hvKp strains were sometimes designated in the literature as hypermucoviscous. However, eventually the designation hypervirulent K. pneumoniae was more commonly utilized (36–39). The report by Pomakova et al. also designated the pathotype responsible for the majority of health care associated infections as classical K. pneumoniae (39). This distinction between cKp and hvKp frames the genotypic and phenotypic differences between these pathotypes. As discussed, the use of the term hypermucoviscous has proven to be problematic, since this phenotype is not optimally sensitive or specific for hvKp strains: not all hvKp strains are hypermucoviscous, and some cKp demonstrate this phenotype (16, 17). Some studies used solely a positive string test to define hvKp strains, which has resulted in some strains of K. pneumoniae being misclassified as hvKp and consequently created some confusion in the literature.

The clinical syndrome Friedlander’s pneumonia was eventually recognized to be due to K. pneumoniae (40–42). This entity and the offending agent were first described in 1882 by Friedlander (43), hence the initial designation as Friedlander’s bacillus (Bacillus friedlanderi). A subsequent and now antiquated designation was Bacillus mucosus capsulatus (44). The acute syndrome has a number of distinctive clinical features which are consistent with some, if not all, of the offending strains being either hvKp or at least K. pneumoniae isolates that had acquired a portion of the hvKp virulence factor repertoire.

In the preantibiotic era, Friedlander’s pneumonia had a mortality rate of approximately 80%, which was 3- to 4-fold greater than pneumonia caused by Streptococcus pneumoniae (45, 46). Presentations were usually acute, and death could ensue within 24 to 48 h and on average occurred 5.5 days after presentation, compared to 9 days for pneumonia due to S. pneumoniae (46, 47). Bacteremia was noted, on average, in 60% of cases (45, 46). Radiographically, findings were indistinguishable from pneumococcal pneumonia, with bronchopneumonic, lobular, and lobar manifestations observed, which were often multifocal. However, in contrast to pneumococcal pneumonia, tissue destruction leading to overt cavitation and/or necrosis observed on histology was far more likely to develop (48). Although not diagnostic, a bulging fissure and/or cavitation increased the likelihood that the pneumonia was due to Friedlander’s bacillus (40). After cavitation, empyema was the next most common pulmonary complication; purulent pericarditis also could develop (48, 49). If the patient survived the acute episode, progression to chronic cavitary disease that mimicked tuberculosis and persisted for months could develop (50). Nearly all cases of Friedlander’s pneumonia occurred in ambulatory patients. Although chronic alcoholism was touted as a critical risk factor, many patients were healthy hosts (45, 48, 49, 51). Perhaps the increased risk of infection in alcoholics was due not solely to compromised host defense factors but also to the increased likelihood of macroaspiration. Males were more commonly infected, and although infections were reported in all age groups, the fifth and sixth decades of life were most common (47–49, 52). Likewise, in most contemporaneous studies of hvKp, men are more commonly infected than women (9, 53, 54). Mercifully, Friedlander’s pneumonia accounted for only 0.5 to 5% of community-acquired pneumonias (42, 48, 49, 52).

Friedlander’s bacillus has also been implicated in a variety of extrapulmonary infections in the presence or absence of pneumonia. These include renal abscess, hepatic abscess, osteomyelitis, cavernous sinus thrombosis, abscess of the jugular bulb, meningitis, brain abscess, splenic infection, spontaneous bacterial peritonitis, and soft tissue abscesses in the neck and arm (47–49, 55–61). Similar to the case with individuals infected with hvKp, multiple sites of infection were noted in a number of patients. Extrapulmonary sites of infection were undoubtedly underestimated in an era in which advanced imaging modalities were nonexistent. Interestingly, septic endophthalmitis was not noted, which could be easily diagnosed clinically.

Additional features suggested that Friedlander’s bacillus was most consistent with hvKp isolates. A phenotypic feature of many hvKp strains is hypermucoviscosity, i.e., an inoculation loop can generate a viscous string >5 mm in length from the bacterial colony; this trait is due to increased capsular polysaccharide production mediated by RmpA and/or RmpA2 (62). Although hypermucoviscosity is not pathognomonic for hvKp since this phenotype also can be observed in cKp strains (17), it is suggestive. Numerous reports remark on Friedlander’s bacillus possessing this characteristic. In cases of meningitis the spinal fluid is often referred to being “gelatinous” and “the drawing out of a filament from the stylet of the spinal needle” (56). Likewise, “[o]n agar plates the colonies appear…as round, raised, slimy, gray colonies, which string out when drawn up with a wire loop” (48). Sputum was commonly described as “tenacious” or gelatinous (41, 49), and cut lung sections were described as “covered by a characteristic viscid, abundant, mucinous exudate which sticks to the knife” (48). Another paper states that “[t]he name Bacillus mucosus capsulatus (Friedlander’s bacillus) draws attention to the two most prominent and distinctive features of the organism, namely, the marked degree of capsular development and its power of producing large amounts of mucoid material in its growth both on artificial media and in the human body” (44). A serotyping schema was developed for Friedlander’s bacillus with types A (equates to K1), B (equates to K2), and C (equates to K. pneumoniae rhinoscleroma), and group X (other) (63, 64). The majority of strains were types A and B (45, 46, 48, 49, 57), with type A being most common, similar to the observation for hvKp strains (2, 17, 65), although cKp strains also can produce K1 and K2 capsules (66).

In one series that reviewed cases of meningitis due to K. pneumoniae, patients were more likely to have diabetes mellitus (56), also similar to what has been observed in most studies on hvKp-infected individuals (67–75). Additionally, in a recent study authored by Lam et al., hvKp sequence type 23 (ST23) was calculated to have evolved around 1878, lending further credibility to Friedlander’s bacillus being the first description of an hvKp strain (76).

Lastly, in experimental reports studying Friedlander’s bacillus, “[s]ubcutaneous or intraperitoneal injections of 1:1 million or 1: 1 billion dilutions of young cultures often kill mice in 1–3 days” (49); likewise, another report presented data that 10⁻⁵, 10⁻⁶, and 10⁻⁷ dilutions of a culture grown “4-6 h” resulted in 100% mortality of mice challenged intraperitoneally with 0.5 ml of diluted bacteria over 15 to 48 h (63). Although some guesswork is required, if one assumes a “young culture” and a “4-6 h” culture maximally consist of 1 × 10⁹ CFU/ml, then lethal doses would be in the range of 1.0 to 5,000 CFU. Another study reported that 9.0 × 10⁴ CFU killed mice within 18 h, “a result not obtainable with coliform organisms or A. aerogenes” (59). A lethal effect from these low challenge inocula would clearly identify such strains as hvKp and not cKp (17).

Clinical challenges with Friedlander’s pneumonia included early recognition and appropriate treatment. S. pneumoniae was responsible for the overwhelming majority of cases of pneumonia, and at that time treatment with penicillin was efficacious but was ineffective for Friedlander’s bacillus, for which tetracyclines and/or streptomycin were the preferred antimicrobials. Given the fulminant course and high mortality seen with untreated Friedlander’s pneumonia, a lack of recognition was problematic. A similar scenario occurred with meningitis due to Neisseria meningitidis versus Friedlander’s bacillus (58). This scenario echoes a different form of diagnostic issues that occur with hvKp today: K. pneumoniae can be readily identified, but differentiating cKp from kvKp is more challenging. As discussed below, hvKp presents different management challenges and if this pathotype is unrecognized, the consequences could be significant, especially for XDR hvKp strains (22).

Taken together, the described features of infection due to Friedlander’s bacillus, namely, the ability to cause life-threatening disease in healthy patients from the community, multiple sites of infection or subsequent metastatic spread (including meningitis and brain abscess), hypermucoviscosity, and capsule type, as well as experimental mouse data are consistent with at least some of these strains being hvKp. Of course, it would be interesting and informative if properly stored isolates of Friedlander’s bacillus were available for sequencing and in vivo assessment in appropriate infection models. These data would also generate insights into the evolution of present-day hvKp strains. For example, was a virulence factor(s) which enables present-day hvKp to cause endophthalmitis absent from the Friedlander’s bacillus? Did Friedlander’s bacillus have a greater tropism for the lung than hvKp, or were nonpulmonary sites underrecognized due to the lack of modern-day imaging technologies? Further, it is intriguing to speculate that cKp evolved from hvKp by loss of the virulence plasmid and perhaps other genetic material once introduced into the health care environment.

K. pneumoniae organisms can be members of normal animal and human microbiotas and/or the microbiotas of various environmental habitats (110, 111). Acquisition of and colonization with K. pneumoniae appear to be requisite for, but do not necessarily lead to, infection (112–117). Otherwise healthy individuals from the community are at risk for developing hvKp infection, whereas it is uncommon for cKp infection to develop in this population. Healthy people can be colonized with cKp, but in the absence of some form of host compromise, infection rarely occurs. By contrast, healthy individuals colonized with hvKp are at much greater risk for developing infection. However, the frequency with which infection develops after colonization with hvKp and the factors that modulate this risk are not well understood. The relative importance of colonization versus that of the quantity of the colonizing hvKp strain, host factors, and degrees of hvKp virulence is an important issue that requires active investigation.

In healthy humans from the community in Western countries, the prevalences of K. pneumoniae colonic colonization ranged from 5 to 35% (112, 117, 118). In Asian countries, K. pneumoniae colonization rates in stool from health adults were 87.7%, 61.1%, 75%, 58.8%, 57.9%, 18.8%, 52.9%, and 41.3% for Malaysia, Singapore, Taiwan, Hong Kong, China, Japan, Thailand, and Vietnam, respectively (115). Another study from Korea reported a K. pneumoniae colonization rate in stool of 21.1% (248/1,175) (113).

In Western countries, 1 to 5% of healthy humans from the community are nasopharyngeally colonized with K. pneumoniae. In children <10 years of age from Brazil and Vietnam, 1.4% (17/1,192) and 1.6% had nasopharyngeal colonization with K. pneumoniae, respectively (119, 120). By contrast, 7% of Indonesian children (16/243) were colonized with K. pneumoniae (121). Nasopharyngeal colonization rates increase with age; in Indonesia 15% of adults (38/253) were found to be colonized with K. pneumoniae in the nasopharynx (121) and in Vietnam the overall colonization rate was 14.7% but exceeded 20% in those >40 years of age (120). Increased nasopharyngeal colonization rates were associated with poorer states of sanitation and increased contamination of food and water (121), age, smoking, alcohol use, and living in a rural community (120). In Malaysia, 32% of samples from street food were contaminated with K. pneumoniae (122). Interestingly, in a study of community-acquired pneumonia from Indonesia, K. pneumoniae was the most common bacterial agent identified, causing 14% of 148 cases; by contrast, S. pneumoniae caused 13% of cases (123). Data such as these support the concept that increased colonization has the potential for increasing the incidence of infection.

The dogma is that skin colonization with K. pneumoniae is uncommon and transient. However, in one study from the United States, axillary colonization was relatively common, occurring in up to 50% of individuals, and colonization of other dermal sites, albeit uncommon, increases in warmer months (124).

Obtaining accurate data on hvKp colonization rates of individuals from the community is more challenging since hvKp-specific markers have not always been used to determine their relative proportions compared to cKp. A colonic colonization study of healthy Koreans demonstrated a colonization rate of 4.6% for hvKp (based on less sensitive K1 capsule and ST23 sequence typing) (113). Another report demonstrated colonization rates of 14.1%, 14.9%, 11.3%, 12%, 11.7%, 16.7%, 2.7%, and 0% for healthy individuals from Malaysia, Singapore, Taiwan, Hong Kong, China, Japan, Thailand, and Vietnam, respectively, for putative hvKp strains (based on K1 and K2 capsule types, phenotypes not optimally sensitive or specific) (115). For Australia, 1.3% (1/80) of K. pneumoniae isolates from rectal swabs were hvKp strains (28). In a study that performed nasopharygeal cultures on adults seen at an outpatient otorhinolaryngology clinic for sinusitis and rhinitis in Taiwan, 11.5% (39/340) of isolates were K. pneumoniae and 77.5% of K. pneumoniae isolates tested were predicted to be hvKp based on being rmpA positive (125). We were unable to identify data on dermal colonization with hvKp. Despite the lack of optimal data, it is clear that a significant minority of Asians are colonized with hvKp. More data on hvKp colonization rates from Western countries, in which there is a lower incidence of hvKp infection, would be welcomed. These data may generate insights into the relative risk of acquisition versus genetic factors (e.g., ethnic background) for subsequent infection. Likewise, data on skin colonization could be insightful, since this represents a potential source of entry.

A point of concern is that a variety of Gram-negative bacilli, including cKp, emerge as the dominant colonizers of both mucosal and skin surfaces in the health care setting, particularly in association with antimicrobial use, indwelling devices, severe illness, and extended length of stay (117, 126, 127). Hospitals and long-term-care facilities have been identified as important reservoirs for XDR cKp (128, 129). In these settings, transmission from health care workers to patients, especially with lax hand hygiene, and transmission via instrumentation are important mechanisms that could be minimized via appropriate infection control measures (130, 131). Although this venue was once the realm of cKp, that reality is changing as of late. In part, this is due to XDR cKp strains that acquired the hvKp virulence plasmid, thereby evolving into XDR hvKp strains (22). Additional cases of health care acquisition of hvKp also have been described (132, 133). If hvKp even partially replaces cKp as a colonizer in the health care setting, which will undoubtedly lead to infection in a proportion of these patients (134), then the incidence of hvKp infections, morbidity, mortality, and costs are predicted to increase, given the vulnerability of this patient cohort and the virulence of hvKp, particularly if XDR hvKp is the offending pathogen (135, 136).

One of the initial defining features of hvKp infection is acquisition in the community (31, 35, 70, 75, 137). The mechanism for acquisition of hvKp within the community is presently undefined, but based on data from cKp and other Enterobacteriaceae, some combination of contaminated food or water, person-to-person transmission (e.g., close contacts such as family members or sexual partners), and perhaps zoonotic transmission is possible. Support for food as a possible source comes from a report that identified two probable hvKp strains (ST23, K1, positive string test) that harbored bla_KPC-2 from cucumber (138).

Although the majority of hvKp infections are community acquired, there are increasing numbers of reports that describe infections developing in health care settings (22, 132, 133). This is due to a combination of increasing recognition of antimicrobial-sensitive hvKp strains causing nosocomial infection, an increasing prevalence of hvKp strains that have acquired antimicrobial resistance determinants and therefore are more likely to survive and be transmitted in this setting, and lastly, the fact that XDR cKp strains that were entrenched in the health care environment have evolved into hvKp strains by virtue of acquiring the hvKp virulence plasmid (22).

The predominance of infections to date has been reported from the Asian Pacific Rim. However, infections are increasingly being reported worldwide (8, 9, 139). With an increasing awareness and the identification of validated biomarkers (140), hopefully an accurate assessment of the incidence of hvKp infection across the globe will be achieved. Table 3 summarizes some of the data presently available.

The epidemiology of acquisition and colonization by hvKp has been discussed. A number of bacterial factors are required to overcome the combination of host factors and competing microbes, which may be site specific. It is important to delineate the factors and define the mechanisms that enable hvKp to successfully colonize various epidermal and mucosal surfaces since this represents a potential point of intervention for decreasing the incidence of infection. To date, the bulk of studies have focused on gastrointestinal colonization, and the majority of genes identified are also variably present in cKp strains.

Colibactin is a peptide-polyketide that is produced by nonribosomal synthesis. Its biosynthetic genes (pks) are located within a mobile genetic element (ICEKp10) in hvKp strains which also usually contains genes for yersiniabactin and microcin E492 synthesis (76, 90). This element is present in most CG23 (K1 capsule type) hvKp strains, but less commonly in other hvKp strains (76, 94, 103). The acquisition of ICEKp10 by the sublineage CG23-I was calculated to have occurred in 1928, with subsequent global dissemination (76). This suggests that colibactin may have been an evolutionary asset. The pks gene cluster also can be present in cKp strains (90), and these genes are highly homologous to those reported for Escherichia coli (152) and other Enterobacteriaecae (153). Colibactin has been shown to promote colonization in E. coli (154) and in hvKp strain 1084 (97). Microcin E492 is an 8-kDa bacteriocin that is active against Enterobacteriaceae (155). Activity requires attachment of salmochelin, which enables the uptake of microcin by the target bacteria (156). Therefore, hvKp strains that produce the combination of colibactin, microcin E492, and salmochelin would be predicted to have a significant colonization advantage in the competitive colonic environment.

Several genes have been identified by signature-tagged mutagenesis that appear to play a role in some combination of intestinal colonization and/or invasion across the mucosal barrier after intragastric (i.g.) challenge in mice (157). hvKp strain CG43 (ST86, K2 serotype) was used in this study. After i.g. instillation, the recovery of 28 mutant derivatives was less from liver and spleen homogenates than that of their wild-type parent. Gastric challenge with mutant derivatives with single disruptions of genes that encoded a LuxR family transcriptional regulator (kva15), a putative type III fimbrial usher protein (mrkC), a monamine regulon positive regulator (moaR), a two-component regulator system (kvgA-kvgS, which has been shown to contribute to capsule production) (158), a uracil permease (kva28), or 2 hypothetical proteins (kva7 and kva21) resulted in 0% mortality, compared to 100% mortality observed for their parent, CG43. However, after intraperitoneal (i.p.) challenge, these mutants were as lethal as CG43. These data are consistent with a role for these factors in intestinal colonization and/or invasion across the mucosa. However, all of these genes are also present in cKp strains and therefore are not hvKp specific.

A mediator of ferric iron uptake, Kfu, is more prevalent in hvKp stains than in cKp strains. Kfu was shown to contribute to virulence after i.g., but not i.p., challenge in mice (159, 160). These data support a role for intestinal colonization and/or invasion. However, given its known role in free iron acquisition, which is available in the gastrointestinal tract, a contribution to colonization seems more likely. Likewise, genes that metabolize allantoin are more prevalent in hvKp strains with a K1 capsule type than in cKp strains, but not in hvKp strains with non-K1 capsule type serotypes (65, 66, 161). Similarly, these genes were requisite for maximal virulence after i.g., but not i.p., challenge in mice, thereby supporting a role for intestinal colonization and/or invasion (161). The products of these genes enable nitrogen assimilation from either exogenous allantoin or the catabolism of purines, substrates present in the gastrointestinal tract. Therefore, although a role in mucosal invasion cannot be excluded, it is biologically more plausible that the products of these genes play a role in colonization. It should be noted that the absence of these genes in non-K1 hvKp strains suggests that the ability to metabolize allantoin is not requisite for the hvKp hypervirulent phenotype but may increase the pathogenic potential of K1 strains by increasing their ability to colonize the gastrointestinal tract.

The sensitivity to antimicrobial peptides (SAP) transporter was shown to increase colonic colonization of the undefined K. pneumoniae strain Ca0437 in a mouse i.g. challenge model (162). The SAP transporter also enhanced adherence to intestinal epithelial cells in vitro. Interestingly, and unpredictably, the SAP transporter affected transcriptional levels of other genes, including those that encoded type I fimbriae. Therefore, it is unclear whether the effect was direct or indirect.

Lastly, in the hvKp strain NTUH-K2044, the disruption of treC, whose product enables trehalose utilization, resulted in decreased intestinal colonization in mice when the strain was in competition with its wild-type parent. Additional effects observed were decreased capsule production and biofilm formation, suggesting potential mechanisms (163). Similarly, the result of the loss of celB, whose product is needed for the transport of cellobiose into the cytoplasm, led to decreased biofilm formation, intestinal colonization, and lethality in mice challenged i.g. (164). Although the experimental design is not discriminatory for which step(s) in the pathogenic process was affected, these data are consistent with capsule and/or biofilm formation as important factors in mediating intestinal colonization, a first and requisite step in Klebsiella pathogenesis (163, 164). However, it has also been shown that capsule promotes colonic colonization for cKp strains as well (165). It is unclear whether the ability of NTUH-K2044 and other hvKp strains to produce more capsule and biofilm than cKp strains (163) enhances their ability to colonize compared to that of cKp and perhaps other Enterobacteriaceae strains.

For some hvKp infections, the route of entry seems straightforward. Oropharyngeal colonization could lead to pneumonia via micro- or macroaspiration. Colonic/perineal colonization could lead to urinary tract infection via the ascending route, although this appears to be uncommon for hvKp (2). For some patients, cryptogenic liver abscesses due to K. pneumoniae have been associated with colonic carcinoma (166); although the K. pneumoniae strains were undefined, it is likely that a number of them were hvKp. In the health care setting, disruptions of mucosal or epithelial barriers (e.g., endotracheal tubes, surgical incisions, and catheters) may enable entry (22).

However, for most patients that develop hvKp infection, the initial site of entry is unclear. To date, most hvKp infections are community acquired, and they often occur in otherwise healthy hosts for whom there is no overt mucosal or epithelial barrier disruption. Since hepatic abscess is the most commonly reported infectious syndrome, it seems logical that hvKp is able to cross the intestinal mucosa and seed the liver through a portal bacteremia. A gut-vascular barrier exists which protects the host from indiscriminate entry of pathogens from the intestinal microbiota (167). Certain pathogens, such as Salmonella enterica serovar Typhi, have developed the ability to translocate across the normal intestinal mucosa, mediated in part by type III secretion systems (168). But for hvKp strains in humans, this does not routinely appear to be the case, since colonization does not necessarily result in infection. The mechanism by which hvKp is able to cross mucosal and/or epithelial barriers in humans is unclear. Considerations include entry through occult disruptions in the skin, with resultant bacteremia and subsequent spread to distant sites, a mechanism employed by Staphylococcus aureus. Another possibility is that the quantity of hvKp organisms colonizing the intestinal tract is a critical factor for entry. In support of this hypothesis, oral ampicillin or amoxicillin treatment, agents inactive against Klebsiella, was associated with an increased risk of subsequent hvKp infection (169). If the magnitude of the intestinal colonizing hvKp titer proves to be an important factor for facilitating entry, then as hvKp becomes increasingly antimicrobial resistant, it is predicted that the risk of infection will increase with antimicrobial use, especially in the health care setting. Of course, yet-to-be-defined host genetic differences that enable increased intestinal binding and/or translocation may be contributory. Needless to say, an improved understanding of the mechanism that enables hvKp to enter the host would facilitate preventative strategies.

Experimentally, hvKp strains have been shown to be capable of translocating across human intestinal epithelial cell lines and mouse colonic epithelial cells (170). A number of bacterial factors have been identified that could facilitate entry through the intestinal barrier. Based on experimental design, the factors identified by Tu et al. could contribute to entry and/or colonization (please see above for details) (157). The SAP transporter was shown to increase translocation of Ca0437 across intestinal epithelial cell monolayers in vitro in a mouse i.g. challenge model (162). But, the models used to identify these factors may not optimally translate. The identified factors are not hvKp specific and are present also in cKp and, in some instances, other Enterobacteriaceae. In contrast to hvKp, these pathogens can colonize the intestinal tract but require an overt mucosal disruption for entry. Therefore, it would appear that these factors alone cannot be responsible for hvKp entry leading to infection.

Paradoxically, studies using hvKp strain 52145 demonstrated that capsule impeded entry into A549 epithelial cells (171), and those using undefined K. pneumoniae strains demonstrated that capsule impeded invasion of an ileocecal epithelial cell line, at least in part by decreasing adherence (172). Further studies on the effect of increased capsule production that occurs with hvKp strains on adherence and cellular invasion would be of interest.

hvKp strains have been shown to be more resistant to phagocytosis, neutrophil- and complement-mediated activity, and neutrophil extracellular traps (NETs) than cKp strains (15, 39, 99, 173). hvKp strains also demonstrate enhanced growth in human body fluids ex vivo and increased virulence in a variety of infection models compared to cKp strains (39, 174). The hvKp-specific factors identified to date that mediate these phenotypes and clinical manifestations are discussed in the following sections. However, undoubtedly additional factors will be defined in the future.

In humans, multiple sites of infection and/or metastatic spread are more common with hvKp than cKp strains (17, 174). cKp and other members of the Enterobacteriaceae family rarely establish infection in secondary sites as a result of bacteremia, except in the setting of an immunocompromised host (e.g., neutropenia). The primary means by which hvKp strains infect multiple sites is via the bloodstream; whether this occurs at the time of entry and/or when a primary site of infection (e.g., hepatic abscess) is established (which, in turn, serves as the source for subsequent bacteremias and distant spread) is unclear. Likely both mechanisms are operational. Regardless, the ability to invade the bloodstream and survive the resident host defense factors is the first requisite step. Resistance of hvKp strains to the bactericidal activity of complement, which is mediated in part by capsule, is needed to accomplish this and contributes to this step in the process (99). However, most pathogenic members of the Enterobacteriaceae family are also capable of causing bacteremia and are resistant to the bactericidal activity of complement as well as the other host defense factors within the bloodstream. This suggests that hvKp is more efficient in invading tissue from the bloodstream. The mechanism by which this occurs is incompletely defined. Although the animal models used presently to study hvKp can measure bacteremia and subsequent spread to various organs or sites, models that directly measure hvKp tissue invasion from the bloodstream at the cellular level could assist in identifying potential factors that enable systemic tissue invasion.

Thrombophlebitis of the hepatic and portal veins has been described in association with hepatic abscess (199–201), but this complication could only account for spread to the lungs unless a patent ductus arteriosus also was present. Further, thrombophlebitis of the hepatic and portal veins and metastatic spread has been described in association with hepatic abscess due to other bacterial pathogens, albeit less frequently (201).

A “Trojan horse” mechanism has been postulated with neutrophils implicated as a possible vehicle (202). hvKp strains were shown to be able to survive within neutrophils (202, 203) and delay apoptosis up to 24 h (203), and i.p. injection of infected neutrophils resulted in disseminated infection; however, it was unclear if the integrity of the injected neutrophils was maintained postinjection (202). However, a cKp strain was also shown to be able to survive within macrophages and trigger apoptosis (176). Further, this concept, at least for professional phagocytes such as neutrophils or macrophages, seems counterintuitive, since these cells are attracted to the site of infection and would not be expected depart this inflammatory environment after acquiring bacterial cargo.

In most infections complicated by bacteremia, titers of 1 to 10² CFU/ml commonly are seen (204). Titers of >10² CFU/ml for S. aureus, H. influenzae, S. pneumoniae, and N. meningitidis have been associated with more severe disease and metastatic spread (204–207). Therefore, a quantitatively high titer of hvKp during bacteremia could be responsible or contributory. There is some limited support for this hypothesis. Albeit in subcutaneously infected outbred CD1 mice, 10⁴ to 10⁵ CFU/ml were measured in blood 48 h after bacterial challenge (191) and titers of 10⁵ CFU/ml were measured after intranasal or intravenous challenge (97). Another possibility is that spread could be driven hvKp’s increased capsule production. Although speculative, perhaps this phenotype results in greater in vivo clumping of bacterial cells, which, in turn, enhances survival with hematogenous dissemination.

Colibactin has been suggested to contribute to meningeal spread (97). Whether this is directly mediated by colibactin or due to the magnitude of bacteremia is unclear. It should be noted that the hvKp strain NTUH-K2044 was isolated from a patient with meningitis, but this isolate does not produce colibactin (93).

hvKp infection often results in abscess formation. However, to date there is limited insight into the responsible bacterial or host factors. The best-defined factor is colibactin, which has been shown to be genotoxic, causing DNA damage and cell death (94, 97). However, non-colibactin-producing hvKp strains (e.g., non-CG23-K1 capsule type) also have caused abscesses in multiple sites. This suggests the possibility that unrecognized hvKp factors also may be contributory. Undoubtedly an unregulated host response resulting in collateral damage is contributory to some degree as well.

The incidence of colonic cancer increased over the first decade in the new millennium in Taiwan. Several studies have identified an increased risk of colonic cancer in patients with hepatic abscess due to K. pneumoniae compared to patients without a hepatic abscess or with hepatic abscesses due to non-K. pneumoniae strains (208, 209). Since some K. pneumoniae strains, including hvKp, can produce the genotoxin colibactin, it has been hypothesized that such strains are the causative agent. An alternative hypothesis would be that hvKp-mediated hepatic abscess is a marker for cryptogenic colonic carcinoma, similar to Streptococcus bovis bacteremia/endocarditis (210) or infection due to Clostridium septicum (211). Further, in another epidemiologic study from Taiwan, an increased risk of hepatocellular carcinoma was observed in patients with a history of liver abscess (212). Given the well-established risk of hepatocellular carcinoma due to hepatitis B and C virus infection, although those infections are more chronic, there is a biologic plausibility for this association. More data are clearly needed to define cause and effect and the relative risks. However, presently it is reasonable to consider a screening colonoscopy for patients diagnosed with a hepatic abscess due to hvKp, especially for those not recently studied or with risk factors (e.g., age or family history). Whether long-term follow-up should be performed for the possible development of hepatocellular carcinoma in patients who have suffered a hepatic abscess due to hvKp is less clear.

Although hvKp infections occur in all ethnic groups, even those acquired in Western countries commonly involve Asians (37, 39, 213–221) and (to a lesser degree) Pacific Islanders and Hispanics (9, 215, 216). One explanation could be that hvKp infection occurs more frequently in Asians from geographically defined pathogen exposure, acquisition, and increased colonization (113), a requisite step for subsequent infection. In some cases, there is a history of Asians infected in the West that travelled to or were exposed to individuals who had recently been in the Asian Pacific Rim (39). In fact, a case in a Caucasian from Denmark who traveled to Shanghai, China, suggests this very scenario (220). Alternatively, an underlying genetic abnormality that is more commonly, but not exclusively, present in Asians, Pacific Islanders, and Hispanics could also be contributory. This concept of genetic susceptibility is consistent with a case report involving a Japanese father and son who developed infection due to hvKp, but the mother was only colonized (114). Further, another report from Singapore suggests that genetic background may contribute to the development of hvKp infection. In 70 patients with liver abscess, among whom 67 cases were probably caused by hvKp, only 1.4% of these infections were in Indians, despite Indians comprising 7.4% of the population and being more likely to be diabetic. However, differences in diet and colonization were potential factors that were not accounted for (81). Conversely, in a study from France, of 14 patients with hepatic abscess due to hvKp, 5 were Caucasian, 6 were African, and only 3 were Asian (8).

Data are needed to address whether a genetic variation is contributory to hvKp infection. These data would have important implications for the development of preventative strategies and/or infection control processes (see “Infection Control and Prevention” below). Potential mechanisms remain speculative. However, considerations include genetic variants that enhance of binding of hvKp within the gastrointestinal tract or enable increased survival within the host. A nongenetic mechanism that negatively modulates host defenses is the development of autoantibodies directed against a host factor; although this has not been described for hvKp, it has been implicated in infections mediated by other pathogens (222, 223).

Diabetes has been implicated as a risk factor for a variety of infections, including those due to hvKp. The majority of studies support this contention (67–75), although some reports concluded that diabetes mellitus (DM) was not predictive for developing hvKp infection (99, 132, 174). Poor glycemic control has been associated with an increase in metastatic complications (224). For the syndromes of endophthalmitis and necrotizing fasciitis due to hvKp the association is stronger in most (54, 225–229) but not all (230) studies. An unsubstantiated but biologically plausible mechanism for this observation is metastatic seeding by hvKp during bacteremia as a result of DM-related loss of vascular patency or integrity. Equally important as the association of hvKp infection with DM is the fact the hvKp infection is frequently seen in younger, otherwise healthy individuals. The clinician needs to be cognizant of the fact that this potentially devastating disease can develop in the absence of DM or any other comorbidities (99).

Males might be slightly more likely to be infected than females. Siu et al. reviewed several studies of hepatic abscess due to hvKp (53). Overall, males were more likely to be infected than females (55% [483/871]); however, geographic differences were noted. Males were infected in 68% (26/38), 42% (136/321), and 63% (321/512) of cases in the United States, South Korea, and Taiwan, respectively. Among 61 patients with hvKp infection complicated by endophthalmitis, 80.3% (49/61) were male (231). By contrast, in another study, men and women were similarly infected by hvKp strains causing liver abscesses: 52.5% (21/40) were male and 47.5% (19/40) were female (78).

Recurrent hvKp bacteremia and low levels of IgG2 have been reported (232). Additional studies that examined immunoglobulin levels and hvKp infection would be of interest.

A retrospective population study reported that individuals who used ampicillin or amoxicillin in the prior 30 days had an increased risk (odds ratio [OR], 3.5) of developing a cryptogenic hepatic abscess due to K. pneumoniae (it is predicted that most of these cases would be due to hvKp) (169). Mice challenged orally with hvKp and then treated orally with ampicillin for 5 days starting 2 days after bacterial challenge had increased mortality compared to mice treated with water alone (169). Similarly, individuals that used proton pump inhibitors had an increased risk (OR, 4.7) of developing a cryptogenic hepatic abscess due to K. pneumoniae compared to controls (233). Although these are single reports, both are biologically plausible. Proton pump inhibitors could lead to increased acquisition as has been described for enteric pathogens (234). Treatment with ampicillin or amoxicillin could increase the titer of colonizing hvKp (given the intrinsic resistance of K. pneumoniae to these agents) by killing bacteria that mediate colonization resistance, which, in turn, could lead to increased entry. But, this remains speculative since the mechanism for entry by hvKp is presently unknown.

There have been several reports from Taiwan in which various treatments for esophageal varices were complicated by central nervous system (CNS) infection or endophthalmitis due to K. pneumoniae (235–237). Although no studies were performed to confirm that these infections were due to hvKp, the clinical sites of infection and geographic location are highly suggestive.

A significant issue that is presently impeding optimal care of hvKp-infected patients is the inability of clinical microbiology laboratories to distinguish cKp from hvKp. To resolve this shortcoming, the availability of a commercial test that has been approved by appropriate regulatory bodies is needed. Identification of hvKp as the infecting agent would be important. Specifically, if infection due to hvKp was not previously considered, this would suggest to the treating physician the need to obtain additional imaging (computed tomography [CT] or magnetic resonance imaging [MRI]) in search of additional sites of infection, which may be unrecognized and could require source control (e.g., drainage) (295). Further, the identification of certain occult sites of infection, such as in cases of meningitis (which may be mistaken for sepsis-related changes in mental status), brain or prostatic abscesses, or endophthalmitis, would be important since site-specific antimicrobial regimens that achieve adequate drug concentrations are needed for an optimal outcome (31). Of particular note is endophthalmitis, a devastating complication of hvKp infection. Ocular infection may not be present or apparent at presentation. Due to the rapidity of tissue damage at this site, immediate treatment is required to maximize the odds of maintaining visual acuity. Therefore, established or suspected hvKp infection dictates involvement of an ophthalmologist, who has the capability of performing an appropriate evaluation and immediately initiating treatment specific for endophthalmitis, such as vitrectomy and intravitreal antibiotics if needed (229, 231). The hypermucoviscous phenotype of hvKp can be problematic in the management of abscesses. Its increased viscosity can impede percutaneous drainage (PD) and increase the likelihood of the catheter becoming clogged (39, 310). If it is known at the time of catheter placement that hvKp is the infecting strain, then the clinician can consider the use of the largest bore practical. Once the drainage catheter is placed, it would be equally important to perform more frequent irrigations than usual to try to maintain drainage, which, in turn, will decrease the need for subsequent open surgical drainage (SD) if percutaneous drainage fails. Anecdotally, hvKp infection has been associated with relapse (69, 114, 245, 250). In the absence of controlled data, when hvKp is identified as the infecting agent, a more prolonged treatment course may be needed to maximize cure rates and minimize relapse.

Presently, the suspicion of an astute clinician is the only means by which hvKp would be considered (Table 1). However, clinical criteria are becoming increasingly problematic. The clinical definition that requires the occurrence of a tissue-invasive, community-acquired infection in a healthy host precludes recognition of hvKp infection in patients who have comorbidities, are immunocompromised, or are in a health care setting; the last scenario is becoming increasingly common.

Fortunately, peg-344, iroB, iucA, _prmpA, _prmpA2, and siderophore production greater than 30 μg/ml have been shown to accurately differentiate hvKp from cKp strains (17). These biomarkers could enable the development of a diagnostic test which, if FDA approved, could be used by clinical laboratories for the identification of hvKp strains. A sensitive and specific test is needed not only for patient care but also for use in surveillance and research studies (140). This study and others have demonstrated that the accuracy of the string test for defining hvKp was not optimal, especially in low-prevalence regions, and that this diagnostic modality should no longer be used as the sole means for identifying hvKp (16, 17). However, it is important to note that genetic variation is commonplace with K. pneumoniae. Therefore, it would therefore seem logical that the best marker for hvKp strains would be one that clearly contributes to the hypervirulence phenotype. Total siderophore production has been shown to strongly correlate with in vivo virulence (17, 86, 87), and aerobactin has been shown to be the dominant siderophore produced by hvKp strains and the critical siderophore that enhances virulence ex vivo and in vivo (87). Further, increased production of capsule, which is mediated by rmpA or rmpA2, also has been shown to contribute to the hypervirulent phenotype (33, 62, 178, 179). Therefore, total siderophore production, or iuc and/or either rmpA or rmpA2, would be predicted to be the most accurate and durable marker. This concept is supported by the report by Gu et al. in which an XDR cKp strain became hypervirulent despite iro, peg-344, and rmpA, but not iuc and rmpA2, being deleted in the acquired hvKp plasmid (22). Of course, as more hvKp-specific genes are identified, additional markers may prove to be viable alternatives or even more accurate.

A facile assay to measure quantitative siderophore production would require development for routine use in clinical laboratories. However, identification of genetic biomarkers could be easily achieved via a PCR assay (311). Identification of hvKp via matrix-assisted laser desorption ionization–time of flight (MALDI-TOF) mass spectrometry (MS) has been explored but requires further refinement (312–314).

hvKp infection is characterized by often having multisite involvement upon presentation or subsequent metastatic spread. Infected sites may declare themselves by some combination of signs or symptoms or may be occult (292, 295). Defining the full extent of hvKp infection is critical since selected sites may require a modification of the antimicrobial regimen (e.g., CNS, prostate) and/or dictate the need for source control. Therefore, the clinician should have a low threshold for performing appropriate radiographic studies with the goal of defining the full extent of infection.

hvKp infection is often manifested by abscess development. The liver is the most common location, but nonhepatic abscesses have been described for virtually every organ and anatomic location. There is a limited body of data that addresses the management of hvKp mediated abscess. Further, extrapolation of data from non-hvKp strains may not be applicable due to important differences between hvKp and other pathogens, namely, its hypermucoviscous phenotype, which can result in an extremely viscous fluid content within the abscess. Lastly, management may be driven by the location of the abscess and acuity of illness.

A retrospective study from Singapore compared percutaneous drainiage with open surgical drainage (SD) for the management of hepatic abscess larger than 5 cm; 80% were multiloculated (310). K. pneumoniae was the offending pathogen in 67% (24/36) and 61% (27/44) of the PD and SD groups, respectively. Although the pathotypes were undefined, it is likely that a high proportion of these K. pneumoniae strains were hvKp given the location of the study and that the majority of abscesses were cryptogenic (8). Overall, there was no difference in mortality between the groups, but the SD group had a shorter length of stay, were less likely to fail therapy, and required fewer subsequent procedures; no subgroup analysis for the K. pneumoniae infections was performed. A retrospective study from Taiwan compared PD alone (n = 46), PD followed by hepatic resection (PD-HR) for patients that failed PD (n = 19), and HR as the initial treatment for patients (n = 16) with liver abscess and an APACHE II score of ≥15 (315). The mortality rate was lower in the HR group than in the PD group, but the rates were similar between the HR and PD-HR groups. The HR group was reported to consist of patients with peritonitis, biliary tract infection, or suspected malignancy. No microbiologic data were reported, and the majority of abscesses were described as cryptogenic. Another study from Singapore reported that hepatic abscesses due to K. pneumoniae (unclear what proportion of organisms were hvKp) that were >5 cm (71/109) were more likely to have a delayed response to therapy but not an increased risk of death or readmission; the presence of loculations was not predictive of a delayed response. Of these 109 patients, 7 were treated with open or laparoscopic surgical drainage, 65 with percutaneous drainage, and 37 with antimicrobials alone (26 of which had abscess <5 cm) (316).

Given the limited published data set, which primarily addresses hepatic abscesses, a reasonable approach for hvKp-mediated abscesses in any location should be an individualized assessment of risk and benefit. Abscesses in critical sites (e.g., brain and epidural space) or ruptured abscesses would warrant consideration for more immediate surgical intervention. Large abscesses in noncritical sites may be best managed by percutaneous drainage pending accessibility, with a surgical intervention reserved for failures. Since abscess fluid with hvKp infection is commonly viscous, the largest-bore drainage catheter feasible should be used, with frequent flushing to try to prevent catheter blockage, a reported complication (310). Smaller abscesses (<5 cm) have the potential to be cured with antimicrobial therapy alone. Imaging can be used to assess the response to therapy and define duration of therapy. With the availability of biomarkers to accurately identify hvKp (17), controlled trials that address various issues related to source control would be welcomed.

hvKp infection also can result in necrotizing fasciitis (296). Although there are no controlled data that addresses the management of this manifestation of hvKp infection, the standard of care for this potentially lethal entity is rapid surgical intervention (fasciotomy, debridement, and amputation) in addition to medical treatment (228, 296).

Antimicrobial resistance can be mediated by several mechanisms (317), all of which are variably occurring. The first is acquisition by an hvKp strain of a conjugal plasmid(s) that contains antimicrobial resistance determinants (19). Since hvKp strains have not acquired antimicrobial resistance determinants as rapidly as cKp strains, there has been speculation that plasmid incompatibilities, a physical barrier due to overexpression of capsule, and CRISPR systems could be contributory (76). The second mechanism consists of acquisition and integration of an ICE containing antimicrobial resistance determinants into either the chromosome or virulence plasmid of an hvKp strain (20, 318). The third mechanism is disruption or mutations in chromosomal genes (e.g., genes for outer membrane proteins [OMPs]) (319). The fourth is acquisition of the hvKp virulence plasmid by MDR or XDR cKp strains (22, 319). The hvKp virulence plasmid appears to be nonconjugal. However, a potential mechanism for mediating transfer has been postulated. Dong et al. demonstrated that there is a common 11.2-kb region between at least some hvKp virulence plasmids and a conjugal plasmid that encodes the KPC carbapenemase. This suggests the possibility of initial integration of these extrachromosomal elements, subsequent transfer, and resolution. However, this hypothesis requires experimental confirmation (319).

The literature is conflicting on the effect of antimicrobial resistance genes on hvKp biofitness. This may due to the mechanism (autonomous resistance plasmid versus integrated resistance determinants) and the nature (e.g., ESBL versus carbapenemase) of resistance. Reports in the clinical literature of lethal outbreaks due to XDR hvKp strains are highly concerning but uncontrolled (22). Some studies in which hvKp strains were unable to maintain antimicrobial resistance plasmids interpreted this finding as a potential negative effect on biofitness (336). By contrast, an hvKp strain that contained a conjugal plasmid that produced an AmpC β-lactamase demonstrated plasmid stability in the absence of selection and high lethality, similar to the case with the prototypical strain NTUH-K2044 in a mouse infection model (323). A 2014 report described the introduction of a KPC-containing plasmid into an hvKp strain without a loss in resistance to complement-mediated bactericidal activity or lethality in mice (339). A clinical hvKp isolate (strain 3) that was carbapenem resistant due to the combination of ESBL production and decreased OMPK35/36 expression was highly virulent in a mouse lethality assay, but two other hvKp strains that expressed KPC were not (328).

There is a lack of data to guide hvKp infection prevention and control practices beyond standard precautions. The reservoirs and mechanism of spread for hvKp strains have not been established. Pending the generation of hvKp-specific data, utilizing cKp-based data seems reasonable. Potential reservoirs include both the environment and colonized patients. Neonatal ICU incubator water (347) and waste disposal hoppers (348) were recently identified as a potential environmental sources for NDM-1- and KPC-producing K. pneumoniae, respectively. However, it is not known whether these findings are applicable to hvKp. Further, it remains unclear which of these reservoirs, if either, is the most important source for transmission. Data for cKp from a multicenter ICU study demonstrated that cephalosporin-resistant K. pneumoniae appeared to be more easily transmissible than E. coli; this finding heightens concern for spread regardless of the source (349).

Regardless of the mechanism, it is clear that hvKp strains can undergo nosocomial spread. Gu et al. described a lethal outbreak in an ICU due to an XDR hvKP strain (22). Likewise, additional studies are suggestive of nosocomial dissemination (331, 332). In addition, data published by Harada et al. (114) and unpublished data from our group have demonstrated that healthy, close contacts of hvKp-infected patients may be colonized and/or infected with the index hvKp strain. However, presently the relative roles of environmental and person-to-person acquisition are unclear.

Whether enhanced infection control is beneficial for antimicrobial-resistant hvKp is a critical question that is presently unanswered. It is further complicated by the fact that many hvKp infections may not be recognized, which would preclude institution of appropriate infection control measures if deemed appropriate. Until clinical microbiology laboratories implement testing to identify hvKp, a presumptive or potential diagnosis of hvKp infection relies on an astute physician (Table 1). This task is feasible when a patient from the community presents with typical epidemiologic and clinical features, but it presently is difficult to impossible with health care-associated hvKp infection.

Nonetheless, if a presumptive diagnosis of hvKp infection is made, it seems reasonable to assume that hospitalized patients would be at a greater risk for developing infection once colonized. The frequency remains undefined. Data on which patient groups (e.g., those of particular ethnic backgrounds) or settings (e.g., ICUs), if any, are at greater risk for developing or fostering infection also would be informative, and perhaps these data could be used for targeted interventions. Regardless, when considered from a risk-benefit perspective, the consequences of hvKp infection due to antimicrobial-sensitive strains can be severe even in an otherwise healthy host. It is predicted that consequences would be more significant in patients with comorbidities or various degrees of immunocompromise. Therefore, contact precautions are worthy of consideration. At this time, there is a lack of data to determine if healthy contacts or hospitalized patients exposed to hvKp-infected persons should be empirically treated to decrease the chances of subsequent infection or whether these individuals should be screened for hvKp colonization and, if colonized, preemptively treated.

The need for point-of-care testing to diagnose infection due to XDR hvKp would appear to be less pressing since infection control measures for patients infected with antimicrobial-resistant hvKp strains should be equivalent to those employed for other highly drug-resistant pathogens independent of virulence potential. For patients infected with carbapenemase-producing hvKp strains, contact precautions, informing receiving facilities that the patient being transferred is colonized or infected with a carbapenem-resistant isolate, and cleaning of areas that the patient is in contact with on a daily basis are recommended (350). Consideration should also be given to active surveillance for extensively drug-resistant strains and contact screening.

There is an ongoing debate as to whether ESBL-producing Enterobacteriaceae warrant contact precautions, at least in the nonoutbreak setting (351, 352). Similar to considerations for antimicrobial sensitive hvKP, the consequences of infection with MDR-hvKp is increased relative to other Enterobacteriaceae and likely antimicrobial-sensitive hvKp due to decreased treatment options. Therefore, until data become available, it seems reasonable to strongly consider contact precautions and, if feasible, screening of contacts and active surveillance for MDR hvKp, particularly in an outbreak setting.