Hypervirulent Klebsiella pneumoniae: a Call for Consensus Definition and International Collaboration

TEXT

Klebsiella pneumoniae infections have mainly been observed among hospitalized or immunocompromised patients, but clinicians in Taiwan first recognized an increasing number of pyogenic K. pneumoniae infections in the 1980s occurring among community-residing individuals without apparent immunodeficiency other than diabetes mellitus or anatomical abnormalities in biliary tracts (1, 2). The most common presentation of such infections was liver abscess sometimes complicated by endophthalmitis and meningitis, but other invasive infections were also reported, including osteomyelitis and necrotizing fasciitis (3, 4). Fang and colleagues observed that most strains causing these pyogenic infections showed hypermucoviscosity when grown on agar plates and proposed the generation of viscous strings exceeding 5 mm between a colony and a loop as the diagnostic criterion, which is commonly referred to as the string test (5). These strains were much more resistant than classic K. pneumoniae to in vitro killing by serum or to phagocytosis by neutrophils and macrophages and caused liver abscess and meningitis in mice; they have thus been referred to as hypervirulent K. pneumoniae (hvKp) strains.

The first gene identified as contributing to the hypermucoviscous phenotype in the representative strain NTUH-K2044 was magA (mucoviscosity-associated gene A), which also conferred increased virulence in mouse models (5). magA was later found to be identical to the wzy gene in K1 (wzy_K1), which is part of the capsule biosynthesis operon of strains belonging to capsule serotype K1; therefore, serotype K1 was recognized as the primary serogroup causing hvKp infections in Taiwan (1, 6). These serotype K1 strains are relatively monophyletic, belong to clonal group 23 (CG23) predominated by sequence type 23 (ST23), and have since been reported frequently from East Asian countries besides Taiwan, including South Korea, China, and Japan (7–9).

In addition to wzy_K1, several virulence factors have been identified that characterize the hypermucoviscous and hypervirulent phenotype in serotype K1 strains. They include the rmpA (regulator of mucoid phenotype A) genes and various siderophore genes. The prototypical hypermucoviscous K1 strain NTUH-K2044 carries two rmpA plasmid-borne genes (_prmpA and _prmpA2) on the same large plasmid (pK2044) and another copy on the chromosomal integrative conjugative element (ICE) (10). Epidemiological comparison of the mucoviscous phenotype and rmpA carriage status among K. pneumoniae bacteremia strains in Taiwan has demonstrated significant correlation between them and additionally identified hypermucoviscous strains with rmpA in serotypes other than K1 (11). On the other hand, rmpA- and rmpA2-positive, nonhypermucoviscous strains and rmpA- and rmpA2-negative, hypermucoviscous strains have also been found, suggesting the presence of yet other factors involved in the expression of the hypermucoviscous phenotype (12). Siderophores are important virulence factors that scavenge scarce ferric iron from the environment. In K. pneumoniae, aerobactin (encoded by iuc genes), salmochelin (encoded by iro genes), and yersiniabactin (encoded by irp genes) are the representative siderophores. In the serotype K1 NTUH-K2044 strain, the iuc and iro genes are located on the same plasmid pK2044 as the _prmpA and _prmpA2 genes (13). Among them, the iuc genes appear to be critical in conferring accentuated virulence (14, 15). More recently, some non-K1 K. pneumoniae strains, such as K2-ST86 and K2-ST380, have also been reported to cause community-onset pyogenic infections (16, 17). These non-K1 hypervirulent strains are more clonally diverse than K1-ST23 strains but often carry pK2044-like plasmids encoding rmpA and iuc genes (13).

Until recently, hvKp strains had been relatively susceptible to commonly used antimicrobial agents, with only a few exceptions, including several strains producing extended-spectrum beta-lactamase (ESBL) (7). However, a hospital-based outbreak caused by a carbapenem-resistant, KPC-producing K. pneumoniae ST11 strain which had acquired a pK2044-like plasmid was recently reported from China (18). The strain was hypervirulent in vivo and in fact caused high mortality among the infected patients. ST11 is the most common sequence type among KPC-producing K. pneumoniae strains in China and is considered to be a classical opportunistic K. pneumoniae clonal lineage in terms of virulence. Conversely, hypervirulent, antimicrobial-susceptible ST23 strains have also acquired plasmids carrying bla_KPC and become carbapenem resistant (19). Thus, convergence of hypervirulence and multidrug resistance in K. pneumoniae, through both acquisition of resistance genes by hypervirulent strains and through acquisition of virulence genes by multidrug-resistant strains, poses a consequential public health challenge (20).

In this issue of the Journal of Clinical Microbiology, T. A. Russo and colleagues set out to evaluate biomarkers that can accurately differentiate hvKp and classical K. pneumoniae (cKp) strains (21). To do so, they first identified two sets of strains. One set consisted of an hvKp-rich cohort of 85 clinical strains that caused pyogenic infection in otherwise healthy outpatients in Taiwan and the United States; 45 of these strains represented disseminated infections, including endophthalmitis and necrotizing fasciitis. The second set consisted of a cKp-rich cohort of 90 randomly selected bacteremia strains collected in the United States, Canada, and the United Kingdom, under the assumption that hvKp strains would be extremely rare if existent at all in these countries. As a result, 5 of the 10 virulence genes investigated (peg-344, a putative inner membrane transporter gene located on the same virulence plasmids [22], iroB, iucA, _prmpA, and _prmpA2) showed a diagnostic accuracy of >0.95 in predicting strains in the hvKp-rich cohort. In addition, siderophore levels of ≥30 μg/ml as measured by a quantitative assay had a diagnostic accuracy of 0.96. The hypervirulence of the strains meeting these criteria was corroborated extensively in a murine sepsis model. On the other hand, the other attributes which have been associated with hvKp, including the typical capsule types (K1, K2, K5, K20, K54, and K57) and the string test, had a lower diagnostic accuracy of 0.90 in predicting hvKp over cKp.

The findings of this study have multiple implications for investigations of hvKp. Owing to the lack of a consensus definition of hvKp, studies have so far used various criteria (e.g., specific capsule types or a positive string test, alone or in combination) to define these strains. Because of this, whether a strain defined as hvKp in one study would also qualify as such in another study has been unclear, which has made interpretation and comparison of data challenging. If we can reach a consensus definition of hvKp by incorporating genotypic or phenotypic biomarkers validated to achieve the high diagnostic accuracy in the present study, there will be multiple benefits to the research and clinical communities. First of all, the microbiological and clinical features of hvKp can be accumulated and compared across different geographical areas. Second, early identification of hvKp infections may inform clinicians, allowing them to conduct early and vigorous searches for disseminated infection sites and potentially consider longer antimicrobial therapy than with cKp infections. Clinical studies can be designed to determine whether such strategies can improve patient outcome. Third, improved definition of hvKp at the strain level will also facilitate investigation of the potential impact of these strains on hospitalized and immunocompromised patients. If hvKp is found to cause severe infections in these special populations, then infection control practices may need to be evaluated to possibly implement contact precautions of carriers based on virulence rather than resistance. Finally, departure from the string test, which is intuitive and has served us well but has now been shown to have less than optimal accuracy, will allow for more systematic investigations and identification of novel serotypes or clones of hvKp which may be circulating or may emerge in the future.

The study by Russo and colleagues also identifies areas for further research. While comprehensive detection of peg-344, iroB, iucA, _prmpA, and _prmpA2 is ideal, a typical clinical microbiology laboratory is not yet equipped to routinely perform such PCR-based tests. It is also not yet clear if detection of one of the five genes with high diagnostic accuracy suffices. These genes often exist on the same virulence plasmids, and thus detection of multiple genes may become redundant; but genetic changes like deletion may alter the utility of each gene as a biomarker, depending on which specific strains are prevalent locally. For example, the KPC-producing hvKp strain that caused the fatal outbreak in China possessed _prmpA2 but not _prmpA (18), and strains from China were not included in this study. Detection of more than one gene may be a prudent approach for this reason. The siderophore assay proposed in this work may be a reasonable substitute if the process can be simplified enough to be feasible in clinical laboratories, especially in areas of endemicity. Also, whether emerging hvKp serotypes and clones exert their virulent phenotypes through the same sets of virulence factors remains unknown. Similar studies in different epidemiological contexts would be highly useful to see if these results can be replicated.

In clinical practice, microbiologists and infectious disease physicians have always inferred virulence of a microorganism based on the species identification and acted accordingly. However, if the approach to therapy can be individualized based on the predicted virulence of the specific infecting strain, in addition to the species and antimicrobial susceptibility data, to optimize clinical outcome, there could be a significant impact on how we approach treatment of bacterial infections. The distinct clinical behavior of hvKp and cKp strains implies that K. pneumoniae may be an ideal pathogen to explore such an approach once we can come to a consensus on the definition of hypervirulence in K. pneumoniae. The present study sets the framework for this exciting possibility. Furthermore, having such a common definition would facilitate designing and conducting international studies aimed at elucidating the global epidemiology and clinical features of this emerging pathogen.