A retrospective study on Escherichia coli bacteremia in immunocompromised patients: Microbiological features, clinical characteristics, and risk factors for shock and death

Abstract Background To evaluate clinical features, bacterial characteristics, and risk factors for shock and mortality of immunocompromised patients with Escherichia coli bacteremia. Methods A nearly 6‐year retrospective study of E coli bacteremia in 188 immunocompromised patients at Xiangya Hospital was conducted. Demographic, clinical, and laboratory data were documented. Phylogenetic background and virulence factors of E coli isolates were detected by polymerase chain reaction. Risk factors for shock and mortality were also investigated. Results Of all 188 E coli isolates, most prevalent virulence factors were fimH (91.0%), followed by traT (68.6%) and iutA (67.0%), while papG allele I, gafD, and cdtB were not detected. Phylogenetic group D was dominant (42.0%) among all isolates, and group B2 accounted for 17.6%, while group A and B1 accounted for 28.2% and 12.2%, respectively. In univariate analysis, ibeA and cnf1 were associated with mortality, which were not found in multivariate regression analysis. 22.3% of patients suffered shock, and 30‐day mortality rate was 21.3%. MDR (HR 2.956; 95% CI, 1.091‐8.012) was the only risk factor for shock, while adult (HR 0.239; 95% CI, 0.108‐0.527) was a protective factor. Multivariate analysis revealed that shock (HR 4.268; 95% CI, 2.208‐8.248; P < .001) and Charlson index > 2 (HR 2.073; 95% CI, 1.087‐3.952; P = .027) were associated with fatal outcome. Conclusions Escherichia coli bacteremia was highly lethal in immunocompromised patients, and host‐related factors played major roles in poor prognosis, while bacterial determinants had little effect on outcome. This study also provided additional information about the virulence and phylogenetic group characteristics of E coli bacteremia.


| INTRODUC TI ON
The gram-negative bacilli Escherichia coli strains form part of the normal microflora of the gastrointestinal tract, but are also common isolates of gram-negative bloodstream infection (BSI) worldwide. 1 In addition, E coli is an important cause of both community-onset and hospital-acquired bacteremia. [2][3][4] Escherichia coli is the leading cause of BSIs, with an incidence ranging from 31.9% to 81.0% among major gram-negative species in 28 European countries. 5 Moreover, the increasing incident of multidrug resistant (MDR) E coli strains is concerning. For example, in different region of China the isolation rate of ESBL E coli in BSI is significantly increased to more than 50.0%. 6,7 MDR strains are closely associated with appreciable mortality and carried with great economic loss. [8][9][10] Thus, there is an urgent need to understand factors associated with the causes and outcomes of these infections, in an attempt to reduce their incidence and severity and initiate appropriate therapies.
Virulence factors (VF), for example, hemolysins, adhesins, iron-acquisition systems, cytotoxins, and siderophores, play an important role in the infection. 11,12 They are essential for the interaction between E coli and its host, to overcome host defenses, invade host tissues, and trigger a local inflammatory response. There have been a number of studies describing the epidemiology, risk factor, and outcome of E coli BSIs in different countries. 6,13 Several VFs, such as afa, iroN, and cvaC, were reported to be associated with mortality. 14 Immunocompromised patients are more likely to develop bloodstream infection. 15 And these patients, usually suffering from immunosuppression, chemotherapy, particularly steroids, stem cell transplant, and long-term hospitalization simultaneously, have relatively poor prognosis and high mortality rate. 16 Thus, early treatment of bacteremia is clinically important to ultimate clinical outcome.
Yet, information about virulence genes and phylogenetic groups of E coli strains isolated from immunocompromised patients in mainland China remains scarce and incomplete. The objective of this retrospective study was to assess the clinical features and outcome of E coli BSIs in immunocompromised patients and to study the molecular epidemiology of E coli isolates.

| Patients and study design
This retrospective study was conducted at Xiangya Hospital Central South University, a 3,500-bed teaching hospital in Changsha, Hunan Province, China, from March 2013 to December 2018.
Immunocompromised patients diagnosed with at least one blood culture positive for E coli bacteremia were included in the study.
Outpatients and patients without complete medical records were excluded from the study. Clinical data were collected through a retrospective review of the electronic medical records, including demographic characteristic, clinical and microbiological data, underlying disease, laboratory data at the time of bacteremia onset, results of antimicrobial susceptibility testing, 30-day mortality, and other rel- The study was approved by the Ethics Committee of Xiangya Hospital, Central South University. No informed consent was taken because this study did not cause additional medical procedure.

| Definition
Bloodstream infection was defined as the isolation of organisms from at least one bottle of blood culture specimens from patients with compatible clinical signs or symptoms. The date of the first positive blood culture (index culture) was regarded as the date of bacteremia onset. If the first culture had been drawn more than 48 hours after admission to the hospital, the infection was classified as hospital-onset; otherwise, it was classified as community-acquired.
Neutropenia was defined as an absolute neutrophil count (ANC) of <500 cells/mm 3 . 17 Immunocompromised status was defined as any patients with at least one of the following factors: active malignancy or cancer receiving chemotherapy, history of stem cell transplantation or solid-organ transplantation on immunosuppressive agents, immunosuppressive therapy (including steroid therapy of prednisone ≥10 mg/d or its equivalent administered for ≥7 days), or other underlying immune deficiency. Polymicrobial infection was defined as the presence of microorganisms other than E coli identified from blood samples regardless of whether the isolates came from the same or different blood culture sets during the bacteremia period. The isolates that were resistant to three or more categories were defined as multidrug resistant (MDR). 18 The isolates were regarded as carbapenem-resistant Enterobacteriaceae (CRE) if they were resistant to any carbapenems by in vitro antibiotic sensitivity test. Prior antibiotic treatment was defined as the receipt of any systemic antibiotic >48 hours in the preceding 30 days. 19 Appropriate empirical antimicrobial therapy was defined that at least one active antimicrobial agent effective to the organism by in vitro test was administered within 48 hours after the onset of bacteremia. The impact of comorbidities was determined by the Charlson comorbidity index (CCI). 20

| Phylogenetic analysis
A multiplex PCR method was utilized to detect chuA, yjaA gene, and a DNA fragment TspE4.C2. The PCR steps were as follows: denaturation for 4 minutes at 94°C, 30 cycles of 5 seconds at 94°C and 10 seconds at 59°C, and a final extension step of 5 minutes at 72°C.
The amplification products were separated in 1% agarose gels. After electrophoresis, the gel was photographed under UV light, and the strains were assigned to the phylogenetic groups (group A, B1, B2, or D) by use of a dichotomous decision tree. 22

| Statistical analysis
For continuous variables, results were expressed as mean standard deviation (SD) or median with interquartile range (IQR) and categorical variables using percentages of the group from which they were derived. To evaluate continuous variables, Student's t test (for normally distributed variables) and the Mann-Whitney U test (for variables that did not have normal distribution) were used. Categorical variables were analyzed by the chi-squared or Fisher's exact test when appropriate. P values < .05 were considered statistically significant. All variables that were associated with shock and death in the univariate analysis (P < .05) were entered into a multivariate logistic regression analysis.
All statistical analyses were carried out using SPSS version 22.0.

| Demographic and clinical characteristics
A total of 188 patients were identified during the study period. The median age was 34.9 years old (range 1-73), and 94 were female (50.0%). Most patients had underlying disorders, in which hematological diseases were the most prevalent, including acute myeloid leukemia (35.1%), acute lymphoblastic leukemia (24.5%), and other hematological diseases (24.5%). Moreover, 21 patients suffered from solid tumors (11.2%) and 9 patients had autoimmune diseases (4.8%).
Among the total of 188 episodes, the most common comorbidity was liver damage (11.7%), followed by diabetes (7.4%) and hypertension

| Antibiotic susceptibility
Among all E coli strains, the MDR rate was 68.1% and 12 isolates, accounting for 6.4%, were identified as CRE, as shown in
The most common phylogenetic group is group D with the highest proportion of 42.0% (79/188), followed by group A with 28.2%, group B2 with 17.6%, and group B1 with 12.2% (Table S1).

Escherichia coli BSI
The characteristic of survivors and non-survivors is shown in Meanwhile, there were no significant differences in mortality among E coli BSI patients that received appropriate empirical treatment compared with inappropriate empirical treatment (87.5% vs 92.6%, P = .339).
However, these correlations were not found in multivariate analysis.
In our study, there was no significant association between phylogenetic group distribution and mortality (Figure 1).

Bloodstream infection is an important cause of death in immuno-
compromised patients. In recent years, E coli has gradually become the most common pathogen of bloodstream infection and received extensive attention because of its severe antibiotic resistance. In this study, we analyzed the demographic data, microbiological characteristics, and outcomes of immunocompromised patients with E coli BSIs to identify the risk factors of shock and mortality.
During the study period, a total of 188 E coli strains were isolated. infection. 25 The iutA and fyuA genes, present in 67% and 56.9% of the isolates, respectively, are included in iron uptake systems. The presence of fyuA gene was associated with increased mortality reported by Mora-Rillo et al. 26 The serum resistance-related virulence factor traT, detected in 68.6% of isolates, can subvert complement-mediated killing and is also essential for bacteremia. 21 The high prevalence of these different kinds of VFs all contributed to occurrence of E coli BSIs.
In the univariate analysis, ibeA and cnf1 were associated with 30-day mortality. The virulence gene cnf1 is a cytotoxic necrotizing factors, encoding a protein that affects cellular function, including inflammation and inhibition of neutrophil phagocytosis and chemotactic activities of neutrophils. 27 The prevalence rates reported in other studies were relatively low, ranging from 7.2% to 24%. 26,28,29 ibeA, present in 8.5% of isolates, was associated with central nervous system infections, particularly in newborns, and has been shown to be involved in interacting with surface proteins of the brain microvascular endothelial cells and facilitating invasion of the central nervous system. 30 Previous study also reported 10% prevalence of ibeA and was correlated with increased mortality. 31 In another study from China, it reported that siderophores iroN and iss were associated with 30-day mortality in univariate analysis. 32 In current study, the 25 VFs detected did not show any differences between shock and non-shock groups in multivariate analysis, and the same results were also found in survival and non-survival groups. It suggested that host determinants may override bacterial VFs in determining the mortality of E coli BSIs. Further researches are needed to explore the influence of VFs on the mortality of E coli BSIs.
Regarding to the distribution of phylogenetic groups, we found that group D was the most prevalent group, with 79 isolates accounting for 42.0%, while the prevalence of group B2 was only 17.6%, which is different from the results of most previous studies. Typically, E coli strains isolated from bacteremia mainly belong to group B2, followed by group D, and they had been confirmed as highly virulent strains. 33  In previous studies, the correlation between drug resistance and poor prognosis has been reported. 26 Therefore, high frequency of antibiotic treatment is sometimes inevitable for immunocompromised patients. However, it is still necessary for clinicians to strictly control the application of antibiotics in order to improve survival of patient and reduce the incidence of antibiotic resistance.
In summary, this study evaluated the host and bacterial factors affecting the prognosis of E coli BSIs. We found that host-related factors played a major role in the E coli BSIs, while VFs and phylogenetic groups of E coli showed little effect on prognosis of immunocompromised patients. Our study also showed that MDR rate of E coli was still concerning and was the only risk factor related to shock, while adult was the protective factor. The study revealed that shock and Charlson's comorbidities index >2 were two independent factors for mortality of E coli BSIs.
There are several limitations in our study. First, our data were based on the local epidemiology of a single center and cannot be generalized. Further studies of a larger number of patients from different regions are necessary to assess the influence of these determinants on the outcome of E coli BSIs. Moreover, due to the retrospective nature of this study, several patients were excluded from the study cohort because of incomplete clinical data, which may result in a slight degree of selection bias. At last, the present study failed to associate any particular bacterial determinant with the risk of shock and death. Thus, further research should be conducted on the genetic features of these pathogens, such as pathogenicity islands and plasmids associated with the virulence.

CO N FLI C T O F I NTE R E S T
No conflict of interests is declared.