Clinical and microbiological characteristics of nosocomial, healthcare-associated, and community-acquired Klebsiella pneumoniae infections in Guangzhou, China

Background Klebsiella pneumoniae (K. pneumoniae) is a common pathogen associated with hospital and community-onset infections. This study aimed to compare the clinical and microbiological characteristics of nosocomial, healthcare-associated (HCA), and community-acquired (CA) K. pneumoniae infections. Methods Clinical data were extracted from electronic medical records and analyzed retrospectively. Antimicrobial susceptibility and extended-spectrum beta-lactamase (ESBL) production were determined for all identified strains. Carbapenemase and ESBL genes were amplified by PCR. Genotyping of carbapenem-resistant K. pneumoniae (CRKP) and ESBL-producing strains was performed by pulsed-field gel electrophoresis (PFGE). Results Of 379 K. pneumoniae infections, 98 (25.9%) were nosocomial, 195 (51.5%) were healthcare-associated, and 86 (22.6%) were community-acquired. Hematological malignancy (OR = 4.467), and hypertension (OR = 2.08) and cerebral vascular disease (OR = 2.486) were associated with nosocomial and HCA infections respectively, when compared to CA infections. Overall, the incidence of antimicrobial resistance for the majority of agents tested was similar between nosocomial and HCA infections (P > 0.05) and both groups had a higher incidence than CA infections (P < 0.05). Moreover, 95.1% (78/82) of CRKP strains were isolated from the nosocomial and HCA groups. The blaKPC was the most prevalent carbapenemase gene among CRKP strains (80.5%, 66/82). ESBL-producing strains were prevalent among nosocomial (40.8%), HCA (35.9%) and CA groups (24.4%). The blaCTX-M-9-group and blaCTX-M-1-group genes were predominant in nosocomial (65.0%) and CA strains (66.7%), respectively. PFGE results showed ESBL-producing and CRKP strains were genetically diverse. Identical PFGE profiles were observed among HCA and nosocomial strains. Conclusions Nosocomial and HCA K. pneumoniae infections presented similar clinical features and antimicrobial resistance, and both two types of infections were different to CA infections. CRKP and ESBL-producing strains were disseminated mainly in HCA and nosocomial groups, and showed a clonal diversity. The cross transmission of CRKP was existed among HCA and nosocomial patients. This finding suggests that similar empirical therapy should be considered for patients with nosocomial and HCA K. pneumoniae infections and bacterial resistance surveillance of these infections is necessary.


Background
Klebsiella pneumoniae is a threatening human pathogen that causes various diseases, including respiratory tract and urinary tract infections, blood infections, and liver abscesses [1]. Traditionally, bacterial infections have been classified into community-acquired and nosocomial infections according to the route of infection. As the healthcare-associated effects on community-onset infections cannot be overlooked, these infections have been divided into community-onset healthcare-associated (HCA) infections for patients with a recent history of medical care, and communityacquired infections for patients without [2]. Previous studies have described that HCA infections are different from CA infections and share some similarities with nosocomial infections regarding clinical characteristics, outcome and microbiological features [2,3]. Although there have been studies on HCA infections, the data about antimicrobial resistance and molecular features of HCA K. pneumoniae infections are limited [4,5]. Inadequate understanding of these infections could lead to inappropriate empirical treatment and, in consequence, higher morbidity and mortality.
Antimicrobial resistance in K. pneumoniae is a significant concern to public health. The CRKP and ESBLpositive K. pneumoniae strains are the most important and challenging drug-resistant bacteria, leading to limited treatment options and high mortality [6][7][8]. Thus, understanding the distribution of such highly resistant strains in both hospitals and communities is vital to planning interventions.
The production of carbapenemase is the main carbapenem-resistance mechanism in K. pneumoniae [9]. Carbapenemase encoding genes reported worldwide mainly include bla KPC , bla NDM , and bla OXA-48-like [10][11][12]. Bla SHV , bla TEM , bla CTX-M , and bla OXA are the major ESBL genes identified to date [13,14]. These antibiotic resistance genes are normally located on mobile genetic elements, such as plasmids and transposons, thus, they can be widely spread in both nosocomial and community environments [15,16].
Strain typing is of great significance for epidemiological surveillance. PFGE is a gold standard for the genotyping of strains. The molecular typing of K. pneumoniae isolates contributes to the identification of specific clone dissemination and the source of infections, on which the prevention approach depends.
Currently, there are limited data comparing the clinical and microbiological characteristics of nosocomial, HCA, and CA K. pneumoniae infections. And the study data on antimicrobial resistance of HCA K. pneumoniae infections is scarce. Furthermore, strains isolated from different regions present their own distinct clinical and microbiological features. Therefore, this study aimed to characterize the clinical features, antimicrobial resistance and the distribution of CRKP and ESBL-positive strains among nosocomial, HCA, and CA infections caused by K. pneumoniae, and to characterize the genetic relatedness of CRKP and ESBL-positive isolates identified from the three different groups. The results from this study will be helpful in developing effective preventive and treatment strategies against K. pneumoniae infections both in hospitals and in communities.

Study design and population
This retrospective study was conducted at a 2300-bed tertiary hospital in Guangzhou, China. All patients presenting K. pneumoniae infections from January 2019 to December 2019 were included in the study. K. pneumoniae strains were consecutively collected from these patients. For patients with multiple K. pneumoniae isolates, only the first isolate was analysed. During the study period, the number of admittances and outpatient treatments were about 103,630 and 1,610,618, respectively.
The definitions of nosocomial, HCA, and CA K. pneumoniae infection was based on previously described criteria [2]. A nosocomial K. pneumoniae infection was defined as an infection that occurred 48 h after the patient's admission. A HCA K. pneumoniae infection was defined as an infection developing within 48 h of admission in patients presenting any of the following risk factors [2]: a history of intravenous therapy or renal dialysis in the 30 days before the K. pneumoniae infection; a history of hospitalization for 2 or more days in the three months before the K. pneumoniae infection; or residence in a nursing home or long-term care facility. A CA K. pneumoniae infection was defined as an infection occurring in patients who did not fulfill the definitions of nosocomial or HCA infections.

Clinical data collection
We extracted the following clinical data from electronic records for all patients: demographic characteristics, underlying disease, site of infection, immunosuppression, Keywords: Klebsiella pneumoniae infections, Nosocomial, Healthcare-associated, Community-acquired, Antimicrobial resistance and mortality. Exposure to prior antimicrobial treatment was defined as any treatment received for at least 48 h in the 30 days before the K. pneumoniae infection.
Multidrug resistant (MDR) strains were defined as strains with no susceptibility to at least three different classes of antimicrobials [17]. Carbapenem-resistant K. pneumoniae (CRKP) isolates were defined as isolates that displayed resistance to one or more carbapenem agents such as meropenem, ertapenem, or imipenem [18].

Detection of resistance genes
Polymerase chain reaction (PCR) was performed to detect carbapenemase genes (bla KPC , bla NDM , bla GIM , bla SPM , bla IMP , bla SIM , bla VIM , bla IMI and bla OXA-48-like ) in the CRKP strains and ESBLs genes (bla TEM , bla SHV , bla CTX-M-group , bla OXA-1-like , bla CTX-M-1-group, bla CTX-M-2group , bla CTX-M-9-group ) in the ESBL-producing strains [17,[19][20][21][22]. The commercial kits Green Taq Mix (Vazyme, Nanjing, China) and specific primers were used to detect resistance genes. Amplification conditions are as follows: initial denaturation at 95℃ for 3 min, 30 cycles of 95℃ for 40 s, primer annealing temperature for 40 s, 72℃ for 50 s, and 72℃ for 7 min. Bacterial strains with resistance genes detected by PCR and DNA sequencing were used as positive controls for the subsequent PCR runs. The distilled water was used as a negative control. Amplification products were analyzed by agarose gel electrophoresis. The specific primers are presented in Additional file 1: Table S1. The PCR products were sequenced and the sequencing results were analyzed on the BLAST website (http://blast .ncbi.nlm.nih.gov).

Pulsed-field gel electrophoresis (PFGE)
All 186 CRKP or ESBL-producing strains were subjected to PFGE following the PulseNet standardized procedure (http://www.cdc.gov/pulse net/proto cols.htm). Briefly, the strain DNA were digested with Xbal for 2 h at 37 ℃. Electrophoresis was performed using the CHEF-DRIII system (120 angle, 6 V/cm), with a running time of 19 h and a switch time of 2.16-54 s. The PFGE fingerprint patterns were uploaded to the K. pneumoniae database of the Chinese Pathogen Identification Net (http://139.9.117.189/CPIN/). A Dice coefficient-based PFGE dendrogram was constructed using the unweighted pair-group mean Analysis. Isolates were grouped into the same PFGE cluster if they shared ≥ 80% similarity.

Statistical analyses
Categorical variables were analyzed by Chi-square or Fisher's exact tests (SPSS software, version 20.0). P-values < 0.05 (two-tailed) were considered statistically significant. Multivariable logistic regression analysis was conducted to identify variables associated with nosocomial, HCA, and CA infections. Variables regarding age, underlying disease, and immunosuppression with P < 0.05 in the univariate analysis were included in the multivariable logistic regression model.

Clinical features of K. pneumoniae infection
In total, 379 patients with K. pneumoniae infection were included (98 nosocomial, 195 HCA and 86 CA infections) in the analysis. The incidence of K. pneumoniae infections was approximately 3.66 per 1000 admissions in the hospital. Male patients accounted for 64.12% of K. pneumoniae infections. The overall 30-day crude mortality rate was 12.4% (47/379).
The clinical features of K. pneumoniae infections are presented in Table 1. When compared to HCA (6.2%) and CA infections (5.8%), nosocomial infections (21.4%) occurred more frequently in patients aged ≤ 1 years old. Multivariate analysis further showed that patients aged ≤ 1 years old (nosocomial vs. HCA: OR = 5.665, nosocomial vs. CA: OR = 4.456) was an independent variable associated with nosocomial infections ( Table 2).
There were differences between the three groups regarding the infection type, with bloodstream infections being dominant in nosocomial patients (45.9%), respiratory tract infections being frequent in HCA patients (33.9%), and skin and soft tissue infections being predominant in CA patients (25.6%).
Nosocomial infections and HCA infections shared similar features in terms of underlying diseases. However, CA infections differed from both nosocomial and HCA  infections. In multivariate analysis, hematological malignancy (OR = 4.467) and brain tumor (OR = 7.865) were independent variables associated with nosocomial cases compared to CA cases, whereas diabetes mellitus was more frequently associated with CA cases (OR = 0.342). Cerebral vascular disease (OR = 2.486) and hypertension (OR = 2.080) were independent variables associated with HCA infections compared to CA infections. Both patients with nosocomial and HCA infections presented a higher prevalence of prior antibiotic exposure, accounting for 54.1% and 59.0%, respectively. No prior antibiotic exposure was observed in CA infection patients. Patients with nosocomial (19.4%) and HCA (12.3%) infections exhibited higher rates of 30-day mortality than patients with CA infections (4.7%).

Antimicrobial resistance of nosocomial, HCA, and CA K. pneumoniae isolates
The distribution of the percentage of antimicrobialresistant K. pneumoniae strains isolated from the three groups is summarized in Table 3. Overall, the resistance rates to most of the antibiotics tested were similar for nosocomial isolates and HCA isolates (P > 0.05), and both the strains were significantly more resistant than the CA strains (P < 0.05). The percentages of CRKP strains, ESBL-producing strains and MDR strains were similar between the nosocomial and HCA cases and were higher in both cases than in CA cases: (

Distribution of carbapenem-resistance genes and ESBL genes
A total of 82 CRKP isolates were identified from the three groups. The distribution of carbapenem resistance genes of the CRKP isolates among the three groups is shown in Table 4 (Table 5). The bla CTX-M-2-group gene was not detected in ESBL-producing strains. No significant differences were found in the distribution of bla SHV , bla TEM , and bla OXA-1like genes among the nosocomial, HCA, and CA infection groups. However, the bla CTX-M-1-group and bla CTX-M-9-group genes were disproportionately distributed between nosocomial, HCA, and CA isolates (Table 5). Fifty-one of the 131 ESBL-positive strains were observed to co-harbor bla SHV and bla CTX-M-1-group , with 74.5% (38/51) having  co-existence of bla SHV , bla TEM and bla CTX-M-1-group, and these strains were more commonly identified in the CA group than in the nosocomial group (61.9% vs. 20.0%, P = 0.001; 42.9% vs. 17.5%, P = 0.032, respectively). Conversely, strains possessing both bla SHV and bla CTX-M-9-group were more likely to be found in the nosocomial infection group than in the CA group (55.0% vs. 23.8%, P = 0.02).

Pulsed-field gel electrophoresis
All CRKP or ESBL-producing strains were subjected to PFGE for assessing the clonality of the strains. Three ESBL-producing isolates failed to be genotyped by PFGE. Fifty-six nosocomial strains were divided into 36 clusters at a cut-off of 80% similarity, with the major cluster C13 accounting for 14 ESBL-producing strains (Fig. 1). One hundred and three HCA strains were separated into 65 clusters, with two major clusters C82 and C83, accounting for 17 and 10 CRKP strains respectively. Twenty-four CA strains distributed among 22 clusters. Despite the high genetic diversity, several sets of identical profiles were found among the three groups. In PFGE cluster C13, three sets of identical profiles were observed, and 9 of which were nosocomial strains. In PFGE cluster C82, two sets of identical PFGE profiles were detected among HCA isolates. In cluster C82-II, four HCA isolates and one nosocomial strain showed 100% similarity. Two HCA isolates and one nosocomial strain in cluster C82-III also shared complete similarity. In addition, we observed that some strains with the same PFGE profiles but carried different resistance gene profiles. For instance, isolate 268, 182, 255, and 166 in cluster C13 presented identical profiles, but carried different bla genes. Isolate 98 carrying bla KPC showed 100% similarity to isolate 49 carrying bla NDM .

Discussion
In the current study, we used the definition proposed by Friedman et al. [2] to define healthcare-associated K. pneumoniae infections, and we found that HCA infections accounted for over half of the K. pneumoniae infections. HCA infections showed similar clinical features with nosocomial infections. CA infections presented different characteristic from both HCA and nosocomial infections. Hematological malignancy was more commonly related to nosocomial infections, whereas diabetes mellitus was more frequently associated with CA infections. These findings were also observed in previous reports from Taiwan and Korea [23,24]. It has been reported that cancer was more commonly associated with HCA K. pneumoniae bacteremia patients than CA bacteremia patients [4,5,25]. In the present study,  however, cerebral vascular disease and hypertension were more associated with HCA infections compared to CA infections. This discrepancy may be due to the regional differences in underlying diseases. Indeed, cerebral vascular disease and hypertension are epidemics in Chinese communities [26,27]. Patients with these diseases have a greater demand for medical care services. In this study, the 30-day mortality rates were similar for patients with nosocomial and HCA infections and both patients exhibited significantly higher 30-day mortality rates than CA infection patients. These results were similar to other studies [4,5] and may largely be due to the variation in underlying diseases between the three groups, which are classical risk factors for prognosis of infectious diseases.
An important finding in this study was that HCA K. pneumoniae strains exhibited similar antimicrobial resistance patterns with nosocomial strains, and both the nosocomial and HCA strains were significantly more resistant than the CA strains. This observation indicates that clinicians should consider similar empirical therapy for patients with nosocomial and HCA K. pneumoniae infections. A recent study in Taiwan reported that the rate of resistance in the group of HCA K. pneumoniae bacteremia was between those groups of nosocomial and CA bacteremia [28]. This discrepancy may be due to the different infection types of the population studied, or the regional differences of resistance characteristics in K. pneumoniae. In our study, 29.7% of the HCA isolates were carbapenem resistant, 35.9% expressed ESBL, and 52.3% were MDR, indicating the severe drug resistance in isolates from the HCA infections. The high degree of antibiotic resistance in HCA strains may be related to the frequent contact with healthcare facilities among HCA patients, where the drug-resistant bacteria could spread. In addition, other factors such as overuse of antibiotics might also responsible for developing antimicrobial resistance. Our results suggest that increased attention should be paid to the HCA K. pneumoniae infections.
The spread of CRKP strains has become a new public health crisis worldwide. Published studies on the distribution of CRKP isolates among nosocomial, HCA, and CA infections are scarce. The present study showed that CRKP strains were rare in the CA group, but more prevalent in both HCA and nosocomial groups. Thus, it should be of caution when use carbapenems to treat patients with HCA and nosocomial infections. In agreement with other studies, the bla KPC was the dominant carbapenemase gene carried by CRKP strains [18,29]. In this investigation, only 12 and 3 CRKP strains harbored bla NDM and bla OXA-48-like genes respectively. Indeed, the bla OXA-48-like producers were more prevalent in western countries such as France, Spain and Germany, but were