Introduction

Klebsiella pneumoniae (Kp) are Gram-negative bacteria that can cause various infections. There are mainly two pathotypes that pose a threat to our health: hypervirulent (hvKp) and classical (cKp). The most common subtype of the K. pneumoniae strains is classic K. pneumoniae (cKp) notorious for their resistance to common antibiotics [1,2,3]. An emerging subtype, termed hypervirulent K. pneumoniae (hvKp), was first described in 1986 [4]. The hvKp strains exhibit unique features compared to cKp. The hvKp strains exhibit hypermucoviscosity to cause various severe infections in immunocompetent and young healthy individuals in addition to diseased patients [5,6,7,8,9], liking pyogenic liver abscesses (PLA) [4, 10]. However, the definition of hvKp is controversial. Host, pathogen, and host–pathogen interactions should be considered comprehensively for defining hvKp. However, most published studies have focused on the bacteria alone. A previous study concluded that major histocompatibility complex (MHC) variants, eating habits, nutritional status, and gut microbiota composition are essential host factors to investigate to enhance our understanding of the hypervirulence phenomenon [11]. Moreover, some controversies exist about the relationship between the virulent and morphological phenotype (hypermucoviscosity) [12, 13]. Using in vitro and in vivo assays, various studies showed that few hypermucoviscous K. pneumonia (hmvKp) strains are associated with high virulence [12, 13]. In animal models, hypermucoviscous K. pneumonia did not cause more severe infections and a higher mortality rate than non-hypermucoviscous K. pneumonia. In vitro and in vivo experiments showed that a few (1/5) hypermucoviscous K. pneumoniae isolates had a high virulence. Thus, identifying hvKp by the string test alone is not sufficient [11, 14].

Recently, aerobactin has been regarded as a critical virulence factor for hvKp [14,15,16], which is often concomitant with the mucoid phenotype. Based on this finding, a multi-centre research in China first stated the clinical and molecular characteristics of hvKp (defined as aerobactin-positive) isolate [14]. The results showed that invasive infections (especially PLA), hypermucoviscosity and most of virulence factors (K1, K2, K20, rmpA) genes are highly associated with aerobactin-positive Kp. In addition, some studies have reported that iron acquisition factors and the genes encoding the hypermucoviscous phenotype are located on the same virulence plasmid, which is not frequently present in cKp strains [5, 17,18,19]. Therefore, aerobactin combined with hypermucoviscosity may be a defining hvKp trait. Additionally, the elderly often has various underlying diseases, poor nutritional status and atypical manifestations.

To date, no data about antimicrobial susceptibility, epidemiology and risk factor of hvKp in the elderly has been described. Thus, we conducted a comparison of hvKp (hypermucoviscous- and aerobactin-positive) and cKp considering the host nutritional status, pathogen and host–pathogen interactions.

Methods

Patients

A retrospective study was conducted on K. pneumoniae culture-positive patients diagnosed at Chinese PLA General Hospital between January 2008 and January 2014. Duplicate isolates from the same patient were excluded. The basic demographics and clinical characteristics (underlying diseases, invasive procedures, nutritional status, and survival) of patients infected by K. pneumoniae were collected. Sequential Organ Failure Assessment (SOFA) scores were evaluated within the first 24 h after admission. To further assess the host response and nutritional status between the two pathotypes, we monitored white blood cell count (WBC), percentage of neutrophils (NEU%), total protein (TP) and albumin (ALB) as biomarkers. The study was approved by the Chinese PLA General Hospital Ethics Committee and the Guidelines for Human Experimentation (PR. China) were followed throughout. The main inclusion criteria were (1) the definition of the elderly has being 65 years old or older (≥ 65 year); (2) at least one K. pneumoniae positive culture; (3) Patients with all the indicators(WBC, NEU %, TP, ALB, SOFA score) were recruited in this study when their clinical specimens were identified as Kp. The exclusion criterions were (1) insufficient clinical data (lacking one of these above indicators) or bacterial strain sample storage and (2) co-infection cases. Infections were considered to be community-acquired infections if K. pneumoniae-positive culture was obtained from a sample isolated upon admission to the study center within 24 h. Cases without these conditions were defined as nosocomial infections.

Clinical K. pneumonia isolates

These specimens were from sputum, urine, blood and drainage fluid. The standardized isolation, culture and identification were conducted in the Department of Clinical Microbiology. All strains were stored at − 80 °C. All the strains were identified by the API 20 NE system and the Vitek II system. Moreover, species identification was further confirmed by 16S rRNA gene sequencing. The definition of hvKp required that both hypermucoviscosity and aerobactin were positive. Hypermucoviscosity was confirmed by the positive string test as previously described [20].

Antimicrobial susceptibility testing and phenotypic confirmation of extended spectrum beta lactamases (ESBL)

Antimicrobial susceptibility testing was conducted using the microbroth dilution method as previously described [6]. The following antibiotic agents were included: Amikacin, Gentamicin, Ampicillin/Sulbactam, Aztreonam, Cefazolin, Cefepime, Ceftriaxone, Ceftazidime, Ciprofloxacin, Levofloxacin, Piperacillin/Tazobactam, Trimethoprim/Sulfamethoxazole, Imipenem, Meropenem and Tobramycin. The results were interpreted using the 2017 Clinical and Laboratory Standards Institute (CLSI) guidelines. ESBL was confirmed by agar dilution test using ceftazidime and cefotaxime combined with clavulanate [14]. Multidrug-resistant isolate was defined as resistant to three or more antimicrobial classes [21].

Detection of virulence-associated gene and capsular serotype-specific (cps) genes

Genomic DNA was extracted from all K. pneumoniae isolates. Polymerase Chain Reaction (PCR) for virulence-associated genes (such as rmpA, rmpA2, magA and aerobactin) were conducted as previously described [14, 22, 23]. Capsular serotype-specific genes (K1, K2, K5, K20, K54, and K57) were amplified by PCR [14, 24]. The primers used are listed in Additional file 1: Table S1.

Multilocus sequence typing

The primers and reaction conditions of seven housekeeping genes (gapA, mdh, phoE, tonB, infB, pgi, and rpoB) were utilized according to the K. pneumoniae MLST website (http://bigsdb.pasteur.fr.html) (Additional file 1: Table S1). Allelic profiling and sequence types (STs) determination were also confirmed using the above website. In addition, for further analyses the relationship among different STs, phylogenetic analysis of housekeeping genes was performed. The concatenation of the seven housekeeping genes of K. pneumonia was conducted. A dendrogram was constructed from the concatenated sequences using the neighbour-joining method (MEGA 6.05).

Statistical analysis

SPSS software (version 20.0) was used for data analysis. Measurement data were reported as the mean ± standard deviation (SD), and count data were analysed as percentages. Student’s t-tests and the Wilcoxon rank-sum tests were performed for the analysis of continuous variables. The χ2 or Fisher’s exact test was used for categorical variables. All tests were 2-tailed. The P-value < 0.05 was considered statistically significant. To determine the risk factors for hvKp, univariate logistic regression analyses were performed. All variables with a P value < 0.05 were included in the multivariate model.

Results

Patient Characteristics

Between January 2008 and January 2014, 175 cases are appropriate for this study. Aerobactin-positive and hypermucoviscous strains were defined as hvKp, which was determined by PCR and string test. Eighty of 175 (45.7%) isolates were hvKp. The distribution of the main infection types in the hospital was hospital acquired pneumonia (130, 72.3%), urinary infection (28, 16.0%), abdominal infection (24, 13.7%) and bacteraemia (9, 5.14%). Overall, 170 (97.1%) patients were males and five (2.9%) were females; the mean age was 84.84 ± 8.48 years.

Clinical characteristics (including host response and nutritional status) of hvKp infection

The basic clinical characteristics, host response and nutritional status of patients with hvKp infections are shown in Table 1. The mean age of patients infected with hvKp is significantly younger than the cKp group (83.2 ± 8.75 years vs 86.2 ± 8.04 years, P = 0.020). A significantly higher number of patients with hvKp had diabetes (76.3% versus 54.7%; P = 0.003) as their underlying diseases. Compared with the cKp group, more patients with hvKp infections presented with invasive infections (28.8% versus 6.3%; P = 0.000), liver abscess (10.0% vs 1.1%; P = 0.008), other abscesses (16.3% vs 3.2%; P = 0.035), sepsis shock (11.3% versus 3.2%; P = 0.035) and abdominal infection (22.5% vs 6.3%; P = 0.035). However, the rate of urinary infection in the hvKp group is lower (10.0% vs 21.1%, P = 0.047). In addition, stomach tube is also less common in the hvKp group (56.3% vs 74.7%, P = 0.01). With regard to the host response, both WBC (12.87 ± 4.24 vs 10.34 ± 2.95, P = 0.000) and NEU % (78.87 ± 7.60 vs 74.23 ± 7.83, P = 0.000) are higher in patients with hvKp than the cKp group. However, patients infected with hvKp are more likely to have a lower TP (65.14 ± 4.93 vs 62.96 ± 4.71, P = 0.003) and ALB (35.54 ± 2.75 vs 34.45 ± 3.44, P = 0.021). It was also noted that although the SOFA score in the hvKp group is higher (6.84 ± 2.81 vs 4.93 ± 2.59, P = 0.000), the mortality at 28 days (17.5% vs 17.9%, P = 0.946) was not significantly different between the two groups (Table 1).

Table 1 Clinical and microbiological characteristics, host response and nutritional status of hvKp
Full size table

Genetic characteristics of hvKp vs cKp

Previous reports showed that the virulence-associated genes rmpA, rmpA2, magA and (K1, K2, K5, K20, K54, and K57) genes for capsular K antigens are associated with hvKp [25,26,27]. All isolated strains were tested for the above genes by PCR. K1, K2, rmpA, rmpA2 and magA were highly associated with hvKp (P = 0.000, 0.001, 0.000, 0.000, and 0.012, respectively). However, K5, K20, K54, and K57 were not associated with hvKp (P = 0.276, 0.354, 0.795, and 0.974, respectively). There is no strain in cKp group with K5 (Table 1).

Antimicrobial resistance and prevalence of ESBL genes among K. pneumoniae isolates

The resistance rate of almost all antibiotic agents for cKp was significantly higher than that of the hvKp group, with the exception of ampicillin, imipenem, and meropenem (Additional file 1: Table S2). All hvKp strains were resistant to ampicillin. Two hvKp isolates were resistant to carbapenems. Among hvKp strains, 16 strains (20.0%) were identified as multi-drug resistant bacteria (MDR). Fifty-one strains were identified as ESBL-producing, which was more common in the cKp group (40.0% vs 16.3%, P = 0.001). In the hvKp group, 16.3% (13/80) samples were ESBL-producing isolates, and 2 of them presented with carbapenems resistance. The detailed information about the 13 ESBL-producing hvKp strains is shown in Table 2.

Table 2 Clinical and microbiological characteristics of ESBL-producing hvKp isolates
Full size table

The distribution time and the rate of multi-drug resistance of hvKp were investigated. During the periods from January 2008 to January 2010, February 2010 to January 2012, February 2012 to January 2014, 12, 30, and 48 hvKp isolates were detected, respectively. At the three time points, 2, 6, and 5 ESBL-hvKp strains and 2, 8, and 6 MDR-hvKp strains were detected, respectively. Furthermore, an increase in the number of ESBL-hvKp isolates was detected during the periods from January 2008 to January 2010 (n = 2), February 2010 to January 2012 (n = 6), and February 2012 to January 2014 (n = 5). Additionally, 2, 8 and 6 MDR-hvKp stains were observed in the above three time points, respectively (Fig. 1).

Fig. 1
figure 1

Number of hypervirulent Klebsiella pneumoniae (hvKp), ESBL-hvKp and MDR-hvKp strains detected between January 2008 and January 2014. (black, hvKp; red, ESBL-hvKp; blue, MDR-hvKp)

Full size image

Risk factors: hvKp vs cKp

In this study, univariate regression analysis showed that diabetes (odds ratio [OR] = 2.655) and digestive diseases (OR = 2.152) were statistically significant risk factors associated with hvKp infections (Table 2). Indwelling stomach tube (OR = 0.435) is a protective factor for hvKp infection. Moreover, multivariate analysis revealed that diabetes (OR = 2.548) and digestive diseases (OR = 2.196) were independent risk factors for hvKp infections (Table 3).

Table 3 Risk factor for hvKp vs cKp
Full size table

Risk factors: ESBL-hvKp vs Non-ESBL-hvKp

Patients infected in the ICU department (OR = 5.826) and indwelling stomach tube (OR = 6.421) are significant independent risk factors for ESBL-producing hvKp infections by regression analysis (Table 4).

Table 4 Risk factor for ESBL-hvKp vs Non-ESBL-hvKp
Full size table

MLST genotypic analysis

Among the 175 K. pneumoniae isolates, 119 STs were identified by MLST analysis, including 37 novel STs (ST2868–2869, ST2871–2878, ST2882–2884, ST2887–2892, ST2894–2901, ST2905–2906, ST2908–2909, ST2911, ST2914, ST2916–2918, ST2920). The most prevalent ST in this study was ST23 (n = 22;18.5%), followed by ST37 (n = 6;5.0%), ST11 (n = 5;4.2%), and ST412 (n = 5;4.2%). These 4 STs accounted for 27.7% (33/119) of the total strains. Moreover, 97 isolates identified another 97 distinct STs. ST23, ST412, ST218, ST375, and ST65 were strongly associated with hvKp, while ST11, ST37, and ST461 were more common in the cKp group. The most common clone complex (CC) of the ESBL-hvKp strains were CC1 (N = 2), CC23 (N = 2) and C292 (N = 2), followed by CC412, CC101, CC17, CC34, CC11 and two singletons. The phylogenetic tree showed that the ST347 isolate produced a serious infection (SOFA = 8), and the other STs (ST595, ST2906, ST1469) resulted in death (Fig. 2).

Fig. 2
figure 2

Neighbour-joining dendrogram of concatenated sequences of seven housekeeping genes from the MLST database

Full size image

Discussion

To our knowledge, our study is the first systematic study of hvKp defined as hypermucoviscosity and aerobactin positive and provides a comprehensive assessment of this definition regarding the host nutritional status, pathogen and host–pathogen interactions in the elderly. In the present study, nearly half of K. pneumonia (45.7%) accounted for infection in the elderly. Additionally, it is noted that, in the elderly, the detection of hvKp among the K. pneumoniae isolates increased from 2008 to 2014, indicating an elevated risk for hvKp infection, which is consisted with a previous study focusing on adults in China [20]. In our study, 45.7% of K. pneumonia were identified as hypermucoviscous through a positive string test, which is higher than a previous retrospective study conducted at a single centre in China, with a prevalence of 33% in Beijing [20]. HvKp is emerging in the elderly and may be a potential “superbug” for further clinical practice. However, the hypermucoviscous phenotype may not the unique key trait of hvKp. Moreover, patients with WBC, NEU%, TB, ALB can be included into this study. Therefore, the prevalence of hvKp in the elderly may be incorrectly estimated due to the lack of objective diagnostic methods and small sample size.

The resistance rate to common antibiotics (except carbapenems) in hvKp strains was still significantly lower than that in the cKp group in this study, particularly with regard to ESBLs. In addition, 16.3% of ESBL production was found among hvKp strains in our study, which is higher than previous article [14]. It is widely recognized that carbapenemase-producing hvKp (CR-hvKp) strains have cause various fatal infections, especially an outbreak in critical patients [17, 28, 29]. It was confirmed that the carbapenemase-producing plasmid could be successfully transferred into hvKp strains, leading to a large burden of disease for the public health [30]. In this study, MDR-hvKp is increasing and 2 hvKp isolates show high resistant to carbapenems in the elderly. It is alarming that CR-hvKp isolates are emerging, and it is a big challenge for medical workers to put forward new clinical intervention and prevention. Taken together, these data revealed that antimicrobial resistance is increasing among hvKp strains, which is consisted with a previous study [20]. However, the conclusion requires further investigation at multi-centres with a larger cohort of individuals to be confirmed. Moreover, the results show that the ESBL-hvKp is highly associated with magA in the study. The genetic characteristics and outer genetic environment of the two genes need to be further studied by whole genomic sequencing.

With regard to virulence factors, various types of K-antigens have been reported by now [24, 31, 32]. The most important elements are K1 and K2, which frequently result in serious infection [33, 34]. In our study, K1 and K2 are significantly higher in hvKp group than cKp group. RmpA/RmpA2 and MagA responsible for hypermucovicosity phenotype was proposed as another virulent factor in addition to cps K1/K2 [19, 23, 35, 36]. Our results showed that rmpA, rmpA2 and magA were closely related to hvKp group. These results revealed that most of the virulence factors are highly associated with this new definition of hvKp in the elderly.

Previous studies showed that hmvKp are frequently cause of invasive severe infection [37] in young people without underlying disease, such as PLA [2], suppurative endophthalmitis [38], and meningitis [39, 40]. In this study, the results show that the mean age of hvKp group is slightly younger than cKp. Invasive infection, especially liver abscess and other abscesses, occurred significantly more often with the new definition of hvKp group. In addition, the nutritional status (TP and ALB), host reaction (WBC and NEU %) and SOFA score of the hvKp group are significantly higher than cKp group. Moreover, the above results may also reveal that from the host, pathogen, and host–pathogen interactions, the new definition for hvKp may be highly associated with the real hypervirulence. Thus, focusing only on STs, serotypes, and other pathogen genomic data may not be sufficient to define hvKp. Host, pathogen and host–pathogen interactions should be taken into consideration when defining hvKp. The inflammatory factors (such as interleukin, C-reactive protein, tumour necrosis factor) and nutritional status (prealbumin, thickness of subcutaneous fat) may be more comprehensively considered in future studies.

It is essential for clinicians to respond immediately to hvKp infections, which could cause serious infections and a more severe inflammatory reaction than cKp, especially in the elderly, children and immunocompromised patients. Thus, developing a better understanding of the risk factors for hvKp is urgent and essential. Our results demonstrate that patients with diabetes and digestive diseases are more likely to be infected with hvKp, which is consistent with a previous study in China [14, 20]. Additionally, infections in the ICU and patients with indwelling stomach tube are risk factors for ESBL-hvKp, which may be related with potentially prolonged hospitalized course and antibiotic exposure. Clinicians should pay close attention to these risk factors in clinical practice to reduce emergence of MDR isolates. Previous study [28] suggested that wards previously infected with CR-hvKp should be left unoccupied for more than 2 weeks after disinfection and before the admission of new patients. However, it may be difficult to be implemented in China, a populous and developing country. Thus, it is urgent to make a cluster strategy from the host nutritional status, pathogen invasiveness and host–pathogen reaction to prevent MDR-hvKp, especially CR-hvKp.

There were some limitations in our study. First, it was a retrospective study at a single centre over 6 years. More inflammatory factors and nutrition indicators were not measured. Second, in vitro and in vivo experiments, such as galleria mellonella model, mouse models and a human neutrophil assay, may be further needed for identifying this new definition of hvKp. Third, to further explore the pathogen genomic characteristics, whole genome sequencing may be needed for further study. A prospective multi-centre study that includes more isolates, focusing on host, pathogen and host–pathogen interactions, is needed to better define the hvKp strains.

Conclusions

The hvKp strains defined as hypermucoviscous and aerobactin positive are more likely to cause more severe inflammatory reaction in host and invasive infection, such as PLA and sepsis shock. To further understand hvKp, the host, pathogen and host–pathogen interactions may be the key element. At present, the prevalence of hvKp in the elderly, especially ESBL-hvKp and MDR-hvKp is increasing. It is essential to enhance the clinical awareness and management of hvKp infections.