K1 and K2 capsular serotypes

The capsule is an important virulence factor of encapsulated pathogens, including K. pneumoniae. Capsular polysaccharides of K. pneumoniae are divided into at least 78 serotypes (Pan et al., 2008; Wyres et al., 2016). Notably, many reports have shown that K1 and K2 serotypes are strongly associated with hvKP (Paczosa and Mecsas, 2016; Catalán-Nájera et al., 2017) (Table ​(Table11 and Table S1). In China, among the hypermucoviscous K. pneumoniae clinical isolates associated with various types of invasive infections, including pyogenic liver abscess, 42.9% were the K2 serotype and 23.8% were the K1 serotype (Guo et al., 2017). Another study in China also showed that 68.75% of the hypermucoviscosity-positive K. pneumoniae isolates have the K2 serotype (Zhao et al., 2016). A South Korea study searching K. pneumoniae strains isolated from the urine of hospitalized patients showed that K1 and K2 capsular serotypes were more common in hypermucoviscosity-positive strains than in hypermucoviscosity-negative strains (Kim et al., 2017). Most hypermucoviscous K. pneumoniae isolates from patients with bacteremia also had the K1 or K2 serotype (78%) (Jung et al., 2013). Many hvKP isolates reported in Europe and America also had the K1 or K2 serotype (Table S1).

Why are the K1 and K2 serotypes so prevalent in hvKP? Two studies tried to resolve this question (Yeh et al., 2007; Lin et al., 2014). Serotype K1 and K2 K. pneumoniae isolates demonstrated significantly more virulence in the mouse model than did non-K1/K2 isolates (survival rates of 0 vs. 79.2%, respectively) (Yeh et al., 2007). Typically, K1 and K2 strains are more resistant to phagocytosis and intracellular killing by macrophages and neutrophils than strains with other serotypes, regardless of their hypermucoviscosities (Kabha et al., 1995; Lin et al., 2004; Lee et al., 2014; Paczosa and Mecsas, 2016). Several reasons were suggested for enhanced virulence of the K1 and K2 strains, relative to other strains. First, K1 and K2 strains lack the specific mannose residue repeats that are recognized by the host factors, such as the mannose-binding receptor on macrophage and the lung-secreted surfactant protein A (SP-A) (Kabha et al., 1995; Sahly et al., 2008). Second, K1 and K2 strains have a host-specific monosaccharide sialic acid on their surfaces (Lee et al., 2014), that is known to mimic host cells, thereby allowing evasion of the host immune cells. Third, K1 and K2 strains may induce a smaller release of reactive oxygen species by neutrophils than strains with other serotypes, allowing better survival in human tissues (Sahly et al., 2004; Paczosa and Mecsas, 2016). Fourth, K1 and K2 strains have more diverse O serotypes, compared with strains with other K serotypes (Follador et al., 2016), which may help K1 and K2 strains evade host immune systems. The O serotyping is based on the recognition of distinct variations of O-antigens, the outermost part of the lipopolysaccharide (LPS) (Trautmann et al., 1997).

Although the K1 and K2 serotypes are typically found in hvKP, several strains of cKP also possess these serotypes. At the same time, some hvKP strains have a non-K1 or -K2 serotype (Fang et al., 2007; Shon et al., 2013; Guo et al., 2017). In addition to K1 and K2 capsular serotypes, other capsular serotypes, including K5, K16, K20, K28, K54, K57, K63, and KN1, have also been found in hvKP (Table ​(Table11 and Table S1). A study compared the prevalence and importance of virulence factors for K. pneumoniae liver abscesses between isolates with capsular K1/K2 serotypes and isolates with non-K1/K2 serotypes (Yu et al., 2008). When injected into mice intraperitoneally, K. pneumoniae isolates with the hypermucoviscosity phenotype as well as presence of rmpA and iucABCD-iutA genes exhibited high virulence for mouse lethality, regardless of capsular type (Yu et al., 2008). However, in other reports, K1/K2 cKP strains were not significantly less virulent than K1/K2 hvKP strains (Lin et al., 2004), and serotype K1 and K2 isolates demonstrated significantly more phagocytic resistance and virulence than non-K1/K2 strains, regardless of the presence of the rmpA gene (Yeh et al., 2007), suggesting that the presence of the K1/K2 serotypes themselves is responsible for increased virulence. Therefore, these studies indicate that the presence of the K1/K2 serotypes themselves is partially responsible for the hypervirulence of the hvKP strains and the combination with other hypervirulence-associated factors may enhance virulence of K. pneumoniae.

ST23 and CC23 sequence types

Recent studies using a 694-gene core genome multilocus sequence typing (MLST) analysis or whole genome-sequencing revealed that K1 hvKP clonal complex 23 (CC23) isolates form a unique clonal lineage, whereas K2 hvKP isolates are genetically more diverse (Bialek-Davenet et al., 2014; Struve et al., 2015). These results indicate that the clonal lineage CC23 has a specific genetic background conferring hypervirulence and fitness. Another study has also revealed that the sequence type 23 (ST23) clone is strongly associated with the K1 capsular serotype in hvKP strains that were isolated specifically from liver abscesses (Siu et al., 2011). Among 47 serotype K1 isolates, 85.1% belonged to ST23, one isolate belonged to ST163 (a single-locus variant of ST23), and two isolates were ST249 (a three-locus variant of ST23) (Siu et al., 2011). In China, ST23 was found to be the most prevalent ST among 69 hypermucoviscous K. pneumoniae isolates and was found only among K1 isolates (Guo et al., 2017). Another study in China showed that 57.8% of K. pneumoniae strains from liver abscess patients belonged to ST23, and 96.2% of ST23 K. pneumoniae isolates were the K1 serotype (Qu et al., 2015). In South Korea, ST23 strains accounted for 97.3% of hvKP strains with the K1 serotype collected from patients with community-acquired liver abscesses (Chung et al., 2008). Similar results were found in many Western countries (Table S1). Reasons for the association of the K1 serotype with ST23 are unclear. Meanwhile, hvKP strains of the K2 serotype are genetically diverse (Shon et al., 2013). A recent study analyzing serotype K2 isolates from three different regions in Asia showed that eight different sequence types were found among serotype K2 isolates (ST65, ST66, ST86, ST373, ST374, ST375, ST380, and ST434) (Lin et al., 2014). There were two major groups, ST65-like (42%) and ST86-like (46%) (Lin et al., 2014). Notably, ST23 hvKP clones with the K1 serotype were associated with pyogenic liver abscess, whereas ST65 hvKP clones with the K2 serotype were correlated with various invasive infections (Guo et al., 2017).

Whole genome-sequencing of hvKP isolates from different geographic origins revealed the epidemiology of the hvKP CC23 expansion. Only minimal geographical grouping of the isolates was observed in the CC23 phylogeny, suggesting that CC23 may spread worldwide through multiple international transmission events rather than by local expansions (Struve et al., 2015). Indeed, infections of hvKP linked to recent travels to Asia were reported in several cases from Western countries (Fang et al., 2005; Gunnarsson et al., 2009; Rivero et al., 2010). However, because there were many cases not linked to travel or close contact with individuals from epidemic regions (Nadasy et al., 2007; Doud et al., 2009; Frazee et al., 2009; Pinsky et al., 2009; Decré et al., 2011; Vila et al., 2011; Merlet et al., 2012; Pomakova et al., 2012; Bilal et al., 2014; Coutinho et al., 2014; Gundestrup et al., 2014; Holmås et al., 2014; Melot et al., 2016; Surgers et al., 2016), an independent local establishment of the CC23 lineage outside of Asia is also possible.

The virulence plasmid pLVPK and KPHP1208 pathogenicity Island

Notably, a large virulence plasmid (pLVPK, a pK2044-like plasmid) was detected in all whole genome-sequenced hvKP clonal lineages (Struve et al., 2015). This plasmid encodes aerobactin, salmochelin, and RmpA, which were found to be restricted to hvKP isolates (Struve et al., 2015). A CC23 strain lacking pLVPK was found to have significantly reduced virulence (Lin et al., 2011, 2012), indicating an important role of this plasmid in the hypervirulence of K. pneumoniae.

Additionally, a genomic comparison revealed a distinct variant of a pathogenicity island (KPHP1208) specifically associated with CC23 (Struve et al., 2015). Most CC23 isolates contain genes encoding yersiniabactin, colibactin, and microcin E492, in the KPHP1208 pathogenicity island (Struve et al., 2015). Colibactin is a polyketide-peptide genotoxin that causes DNA damage in eukaryotic cells (Lai et al., 2014). A recent study using a meningitis mouse model showed that colibactin was necessary for the meningeal tropism of K1 CC23 K. pneumoniae (Lu et al., 2017), indicating the importance of the KPHP1208 genomic island in the hypervirulence of K. pneumoniae.

RmpA and MagA

RmpA (regulator of the mucoid phenotype A gene) activates capsule production, resulting in the hypermucoviscosity phenotype and increase in virulence (Cheng et al., 2010). Three rmpA genes were reported in hvKP strains: two large plasmid-carried genes (p-rmpA and p-rmpA2) located in a virulence plasmid (pLVPK) and one chromosomal gene (c-rmpA) (Wacharotayankun et al., 1993; Hsu et al., 2011). In K. pneumoniae CG43, both p-rmpA and p-rmpA2 contributed to the enhancement of capsular production (Lai et al., 2003; Cheng et al., 2010), whereas in NTUH-K2044 (ST23, K1 serotype), the first genome-sequenced hvKP, only p-rmpA, but not p-rmpA2 or c-rmpA, contributed to the increased expression of capsule genes (Hsu et al., 2011). Therefore, additional studies in other strains are needed to clarify the role of each RmpA in hypervirulence.

Many results have revealed a strong association between RmpA and hypervirulence (Table ​(Table11 and Table S1). Whole-genome sequencing of 30 hvKP isolates from different geographic origins revealed that all hvKP were rmpA positive (Struve et al., 2015). Several studies from China also showed that 92–100% of hvKP isolates were positive for rmpA (Liu et al., 2014; Sun et al., 2016; Wu et al., 2017). Regardless of the investigated infection site, most hvKP isolates appear to have the rmpA gene (Ku et al., 2008; Yu et al., 2008; Lee et al., 2010; Decré et al., 2011; Qu et al., 2015; Guo et al., 2016; Zhang et al., 2016; Wu et al., 2017; Zhan et al., 2017). Despite the strong association between hvKP and RmpA, some rmpA-positive isolates did not exhibit the hypermucoviscosity phenotype (Yu et al., 2006). However, recent studies have shown that the lack of hypermucoviscosity and low virulence in some rmpA-positive K. pneumoniae strains were caused by a concurrent mutation of rmpA and rmpA2 genes in the absence of c-rmpA (Yu et al., 2015a,b).

The magA gene was discovered in 2004 as a gene required for the hypercapsular phenotype from K. pneumoniae strains isolated from invasive liver abscesses (Fang et al., 2004). Subsequent works using bioinformatic analysis and genetic experiments determined that magA is the serotype K1 allele of the polymerase gene wzy in the cps gene cluster (Fang et al., 2007; Yeh et al., 2010). Consequentially, magA was renamed wzy_K1 (Fang et al., 2010). A study in Taiwan showed that 83% of K. pneumoniae strains from patients with pyogenic liver abscess were magA positive, whereas only 3% of strains isolated from patients with other invasive infections were magA positive (Chuang et al., 2006). Additionally, all 36 magA-positive strains were identified as serotype K1, whereas the 38 magA-negative strains were not serotype K1 (Chuang et al., 2006). Another study also reported the association of magA with the capsule serotype K1 (Yeh et al., 2006). A recent study showed that magA was found more prevalently in hvKP strains with the K1 serotype than in non-K1 strains (Guo et al., 2017). These results suggest that the magA gene is responsible for the K. pneumoniae capsular serotype K1.

Antibiotic-resistant hvKP in Asia

Except for an intrinsic resistance to ampicillin, most hvKP strains are rarely resistant to antibiotics and are generally susceptible to commonly used antimicrobial drugs (Paczosa and Mecsas, 2016; Zhang et al., 2016). However, with the global dissemination of mobile genetic elements harboring various antibiotic resistance genes, including the K. pneumoniae carbapenemase (KPC), NDM, and oxacillinases-48 (OXA-48) types of carbapenemases (Lee et al., 2016), antibiotic-resistant hvKP isolates have begun to emerge in the past few years (Zhang et al., 2015b). In 2010, a study from Taiwan analyzed various features of isolates with and without the hypermucoviscosity phenotype (Lee et al., 2010). The prevalence rates of extended-spectrum β-lactamase (ESBL) were significantly higher in K. pneumoniae isolates without the hypermucoviscosity phenotype (34%) than in isolates with the hypermucoviscosity phenotype (4%), and only one ESBL-producing strain among the 24 isolates with the hypermucoviscosity phenotype was detected (Lee et al., 2010). Several ESBL-producing hvKP strains causing bloodstream infections were found in China (Liu et al., 2014; Yan et al., 2016). Another report in China showed that 17% of hvKP strains expressed ESBL and several hvKP isolates showed resistance to all the tested antimicrobials, except carbapenems and amikacin (Li et al., 2014b). Furthermore, antimicrobial resistance in hvKP strains increased over time (Li et al., 2014b). Like the report in Taiwan, the resistance to antibiotics in China was found to be significantly greater in cKP strains than in hvKP strains (Li et al., 2014b). Similarly, a report in 2016 showed that 12.6% of hvKP isolates from several invasive infections produced ESBLs, and most of them carried bla_CTX−M genes (Zhang et al., 2016).

Recently, in China, the prevalence of hvKP in invasive infections was studied in several reports. In a study in 2016, 90.9% of the pathogens causing pyogenic liver abscess were hvKP (Ye et al., 2016). A study analyzing 369 K. pneumoniae isolates associated with various invasive infections revealed that 22.8% (84/369) of the isolates were identified as hvKP (Guo et al., 2017). With a national prevalence of hvKP, several studies in China have revealed the emergence of hvKP that have acquired extensive antibiotic resistances, including carbapenem resistance. In 2015, among 33 patients with carbapenem-resistant K. pneumoniae, carbapenem-resistant hvKP isolates were found in four patients (Yao et al., 2015). From these four patients, seven carbapenem-resistant hvKP isolates were identified and the bla_KPC−2 gene was detected in six of the seven hvKP isolates (Yao et al., 2015). Notably, carbapenem-resistant hvKP strains belonged to three sequence types, [ST65 (five isolates), ST25 (one isolate), and ST11 (one isolate)], and two capsular serotypes, [K2 (six isolates) and non-typeable (one isolate)] (Yao et al., 2015). A retrospective study analyzing 28 cases of carbapenem-resistant K. pneumoniae infections from nine cities in China showed a similar result. Among the 28 carbapenem-resistant K. pneumoniae isolates, five carbapenem-resistant hvKP isolates were detected and three isolates had the bla_KPC−2 gene (Zhang et al., 2015b). Interestingly, the sequence types of carbapenem-resistant hvKP strains were ST11 (three isolates), ST65 (one isolate), and ST1700 (one isolate). The serotypes were K2 (one isolate) and non-typeable (four isolates) (Zhang et al., 2015b). A recent study analyzed a total of 1,838 K. pneumoniae isolates collected from various patient specimens of patients between 2013 and 2015. Results were similar, identifying 21 carbapenem-resistant hvKP isolates, with all 21 isolates positive for bla_KPC−2 (Zhan et al., 2017). Additionally, all the carbapenem-resistant hvKP isolates except one harbored bla_SHV−11. Notably, the carbapenem-resistant hvKP strains belonged to five sequence types, [ST11 (16 isolates), ST65 (one isolate), ST268 (two isolates), ST595 (one isolate), and ST692 (one isolate)], and three capsular serotypes, [K20 (six isolates), K2 (one isolate), and non-typeable (14 isolates)] (Zhan et al., 2017). In a 2016 case report in China, the carbapenem-resistant ST11 hvKP strain was detected (Wei et al., 2016). This strain also harbored bla_KPC−2. In conclusion, the prevalence of hvKP among carbapenem-resistant K. pneumoniae isolates in China have significantly been high, ranging from 7.4 to 15% (Zhang et al., 2015a; Zhan et al., 2017).

Carbapenem-resistant ST23 hvKP strains with the K1 serotype have not often been detected in China. Most of the identified carbapenem-resistant hvKP strains have the bla_KPC−2 gene. However, a study in 2016 revealed the detection of carbapenem-resistant ST23 hvKP strains with the K1 serotype. A retrospective study in 2016 was conducted on 148 meropenem-resistant K. pneumoniae strains collected from two hospitals in Zhejiang Province in 2013, and among these isolates, seven K1, one K2, and five non-K1/K2 hvKP strains were identified by the loop string test (Zhang et al., 2015a). Among them, seven carbapenem-resistant hvKP strains with the K1 serotype belonged to only two sequence types, ST23 (four isolate) and ST1797 (three isolates); five strains harbored bla_KPC−2 (Zhang et al., 2015a). In this study, the bla_KPC−2 gene also was detected in most carbapenem-resistant hvKP strains (Zhang et al., 2015a). In China, the KPC types of carbapenemases are more prevalent than other global carbapenemases, such as NDM and OXA-48 types (Lee et al., 2016). KPC-2 was the most frequently identified carbapenemase in K. pneumoniae in many countries, including China (Gupta et al., 2011; Munoz-Price et al., 2013; Li et al., 2014a; Lee et al., 2016), and ST11, which is rarely found in hvKP strains (Liu et al., 2014; Zhang et al., 2016; Table ​Table1),1), is the prevalent clone associated with the spread of KPC-producing K. pneumoniae in Asia (particularly in China and Taiwan) (Qi et al., 2011; Lee et al., 2016). Therefore, these epidemiological features of carbapenemase-producing K. pneumoniae in China imply the emergence of carbapenem-resistant hvKP strains by interaction between cKP and hvKP strains.

In China, the predominant clone of carbapenem-resistant K. pneumoniae is KPC-2-producing ST11 (Qi et al., 2011; Lee et al., 2016) and ST23 is one of the dominant clones of hvKP (Table ​(Table1).1). Among 36 carbapenem-resistant hvKPs identified in China at the present time (Yao et al., 2015; Zhang et al., 2015a,b; Wei et al., 2016; Wu et al., 2017; Zhan et al., 2017), the rates of KPC-2-producing ST11 and KPC-2-producing ST23 clones are 50% (18/36) and 8.3% (3/36), respectively. Notably, KPC-2 is found in 94.4% (34/36) of carbapenem-resistant hvKP isolates. These results imply that the exchange of mobile genetic elements between KPC-2-producing ST11 cKP and ST23 hvKP has occurred. Additionally, because the rate (50%) of KPC-2-producing ST11 hvKP isolates is far higher than that of KPC-2-producing ST23 hvKP (8.3%), the transmission of hypervirulence-related mobile genetic elements, such as the pLVPK plasmid, from ST23 hvKP strains to KPC-2-producing ST11 cKP strains seems to be a prevalent event. Of course, this assumption should be verified by many additional carbapenem-resistant clinical isolates of hvKP, because the number of currently identified isolates is too small (36 isolates). A study in 2014 showed the in vitro transmission and retainment of a plasmid carrying bla_KPC−2 from a parental non-serotype ST258 K. pneumoniae to a recipient invasive virulent K2 ST65 K. pneumoniae strain (Siu et al., 2014). However, the in vitro transmission of plasmids with hypervirulence-related genes, such as pLVPK, from the parental serotype K1 ST23 K. pneumoniae to the recipient carbapenem-resistant ST11 K. pneumoniae strain has not been reported yet. Notably, a recent study in China showed that five carbapenem-resistant ST11 hvKP strains had the pLVPK-like plasmid (Gu et al., 2017). Genomic analyses showed that the emergence of these ST11 carbapenem-resistant hvKP strains was due to the acquisition of the pLVPK-like plasmid by ST11 carbapenem-resistant cKP strains (Gu et al., 2017). Therefore, additional studies are needed to verify the transmission of mobile genetic elements between cKP and hvKP strains.

94.4% of carbapenemases found in carbapenem-resistant hvKP isolates in China are KPC-2. KPC-2 was the most prevalent carbapenemase in K. pneumoniae in China (Li et al., 2014a; Lee et al., 2016). The global epidemiology of carbapenem-resistant K. pneumoniae showed that the most prevalent carbapenemase of each country varies (Lee et al., 2016). For example, the most prevalent carbapenemases of Pakistan and Turkey is NDM and OXA-48, respectively (Lee et al., 2016). If the most prevalent carbapenemase of carbapenem-resistant hvKPs in Pakistan and Turkey are identified as NDM and OXA-48, respectively, then these results suggest the necessity of investigating the possibility for the direct transmission of mobile genetic elements between cKP and hvKP. In any case, results from China caution researchers and clinicians of the potential for the emergence and global dissemination of hvKP strains with acquired carbapenem resistance.

Colistin (polymyxin E) is regarded as a key component for the combination antimicrobial therapy used for the treatment of severe carbapenem-resistant K. pneumoniae infections (Lee et al., 2016). However, many recent studies have shown that some strains acquire colistin resistance through LPS modification (Cannatelli et al., 2013, 2014), and the frequency of these colistin-resistant isolates is gradually increasing (Giani et al., 2015; Lee et al., 2016). Particularly, the emergence of a plasmid carrying the mcr-1 gene, which mediates colistin resistance by modifying LPS, was reported in China (Liu et al., 2016). A recent study in China reported the detection of an hvKP isolate harboring the mcr-1 gene (Gu et al., 2016). This isolate was identified from a stool specimen of an infant patient with diarrhea in China and was identified as the ST661 clone with the K1 serotype (Gu et al., 2016). This particular case indicates that hvKP can acquire various types of antibiotic resistances.

A recent study showed that the acquisition of colistin resistance of ST23 hvKP strains was accompanied by reduced capsule production, impaired virulence, and a significant fitness cost (Choi and Ko, 2015). Similar results regarding the effects of colistin resistance on virulence and fitness costs have been observed in other Gram-negative pathogens, including Acinetobacter baumannii and Salmonella enterica (Sun et al., 2009; López-Rojas et al., 2011; Hraiech et al., 2013; Beceiro et al., 2014). Although the effect of antibiotic resistance on virulence and fitness costs in hvKP strains has not been assessed for other antibiotics, including carbapenems, high fitness burdens in antibiotic-resistant hvKP strains may explain partially the low prevalence and restricted spread of antibiotic-resistant hvKP strains.

Antibiotic-resistant hvKP in Europe and America

A few cases of antibiotic-resistant hvKP were reported in Europe and America (Table ​(Table1).1). In France, an ESBL (bla_CTX−M−3)-producing hvKP strain was identified in a patient born in Algeria, who alternately resided in France and Algeria for several years without travel to any other countries (Surgers et al., 2016). The strain was the ST86 clone with the K2 serotype. In Brazil in 2014, a polymyxin B-resistant and KPC-2-producing hvKP was detected (Andrade et al., 2014); the strain was the ST11 clone. KPC-2 is the most prevalent carbapenemase in Brazil (Lee et al., 2016), and the correlation is consistent with cases in China. Recently, a study in Italy showed the detection of a KPC-3-producing hvKP isolate from a 52 year old Caucasian patient with liver abscess (Arena et al., 2017). The sequence type for this strain was ST512 (a single-locus variant of ST258). KPC-type enzymes were the most common carbapenemase in Italy (89.5% of carbapenemase producers), followed by VIM-1 (9.2%) and OXA-48 (1.3%) (Giani et al., 2013; Lee et al., 2016). These results imply the possibility for the direct transmission of mobile genetic elements between cKP and hvKP strains.