Epidemiology and Molecular Characterization of Bacteremia Due to Carbapenem-Resistant Klebsiella pneumoniae in Transplant Recipients



We conducted a retrospective study of 17 transplant recipients with carbapenem-resistant Klebsiella pneumoniae bacteremia, and described epidemiology, clinical characteristics and strain genotypes. Eighty-eight percent (15/17) of patients were liver or intestinal transplant recipients. Outcomes were death due to septic shock (18%), cure (24%) and persistent (>7 days) or recurrent bacteremia (29% each). Thirty- and 90-day mortality was 18% and 47%, respectively. Patients who were cured received at least one active antimicrobial agent and underwent source control interventions. Forty-one percent (7/17) of patients had intra-abdominal infections; all except one developed persistent/recurrent bacteremia despite drainage. Two patients tolerated persistent bacteremia for >300 days. All patients except one were infected with sequence type 258 (ST258), K. pneumoniae carbapenemase (KPC)-2-producing strains harboring a mutant ompK35 porin gene; the exception was infected with an ST37, KPC-3-producing strain. Seventy-one percent (12/17) of patients were infected with ST258 ompK36 mutant strains. In two patients, persistent bacteremia was caused by two strains with different ompK36 genotypes. Three ompK36 mutations were associated with significantly higher carbapenem minimum inhibitory concentrations than wild-type ompK36. Pulse-field gel electrophoresis identified a single ST258 lineage; serial strains from individual patients were indistinguishable. In conclusion, KPC-K. pneumoniae bacteremia exhibited highly diverse clinical courses following transplantation, and was caused by clonal ST258 strains with different ompK36 genotypes.


amino acid


premature stop codon at amino acid position 89




bronchoalveolar lavage


blood culture




Centers for Disease Control


carbapenem-resistant Klebsiella pneumoniae



Del AA84-87NNTE

aspragine–asparagine–threonine–glutamic acid deletion at AA position 84–87


deep wound


glutamic acid


endoscopic retrograde cholangiopancreatography


extended-spectrum β-lactamase






Gram-negative rod


intra-abdominal abscess


intensive care unit



Ins AA134–135 GD

glycine and aspartic acid at amino acid (AA) position 134–135

Ins nt382G

guanine insertion at nucleotide position 382


insertion sequence


Jackson–Pratt drain




Klebsiella pneumoniae carbapenemase


KPC-K. pneumoniae




minimum inhibitory concentration


multilocus sequence typing






outer membrane protein


open reading frame


polymerase chain reaction


pulse-field gel electrophoresis


quantitative reverse transcription-polymerase chain reaction




sequence type



TGA nonsense mutation

thymidine, guanine, adenine






target site duplication




urine culture


unweighted pair group mathematical average


vancomycin-resistant Enterococcus


Carbapenem-resistant Klebsiella pneumoniae (CR-Kp) have emerged as major pathogens [1]. Carbapenem resistance can arise through production of metallo-β-lactamases (MBLs) or nonmetallo-carbapenemases (such as Klebsiella pneumoniae carbapenemases (KPCs) and OXA-type carbapenemases). Alternatively, strains may express extended-spectrum β-lactamases (ESBLs) or AmpC β-lactamases in conjunction with loss or decreased expression of outer membrane porins (OMPs) [2-5]. In the last few years, K. pneumoniae sequence type 258 (ST258) international clonal strains producing KPCs have spread to hospitals throughout the world. At least nine KPC variants have been identified since the description of KPC-1 in 2001, with KPC-2 and -3 being most prevalent in ST258 strains [6, 7]. Unique KPCs may confer differing degrees of carbapenem resistance [8-11], but these effects are often confounded by the presence of additional β-lactamases and alterations of OMPs [3-5, 12]. Crude mortality rates among patients with CR-Kp infections exceed 40% in most studies [13-17].

Solid organ transplantation is an independent risk factor for CR-Kp infection [18], but only a few detailed studies have been performed among transplant recipients. Our understanding of CR-Kp infections in transplant and other high-risk patient populations is limited by a lack of data on long-term outcomes, which are important because persistent or recurrent infections are recognized [19, 20]. Moreover, clinical studies have not systematically characterized the molecular epidemiology or mechanisms of carbapenem resistance among infecting strains. The objectives of this study were to describe the epidemiology, clinical characteristics and outcomes of CR-Kp bacteremia among transplant recipients at our center, characterize carbapenem resistance mechanisms among CR-Kp strains and determine the genetic relatedness of strains.

Materials and Methods

We conducted a single-center, retrospective study of transplant recipients with CR-Kp bacteremia between August 2008 and July 2011. CR-Kp was defined according to Centers for Disease Control and Prevention (CDC) definitions as nonsusceptible to one of the carbapenems and resistant to all third-generation cephalosporins [21]. The onset of bacteremia was defined by the date of first positive blood culture. Portal of entry was defined as the primary source of CR-Kp infection that led to bacteremia, as assigned independently and agreed upon by two investigators (C.J.C. and M.-H.N.). The classifications of colonization or infection (catheter-associated bacteremia, intra-abdominal infection, pneumonia or urinary tract infections) were made by the treating physician and independently confirmed by the two investigators according to CDC definitions [22].

The initial treatment regimen was defined as the agent(s) used for the treatment of CR-Kp bacteremia for ≥3 days within the 7 days following the first positive blood culture. Therapy was defined as active if it included an antimicrobial agent to which the infecting strain was susceptible in vitro; therapy was inactive if strains were intermediately susceptible or resistant to antimicrobial agents. The primary endpoints were mortality at 30 and 90 days from onset of bacteremia; 14-day mortality was not used because many patients remained on antimicrobial therapy at this timepoint. Secondary endpoints were microbiological and clinical outcomes, which were defined as cures or failures. Cures were defined as clinical resolution of signs and symptoms of infection, sterilization of blood cultures within 7 days and absence of recurrent bacteremia within 6 months from the initial onset of bacteremia. Clinical failure was defined by: (1) signs and symptoms of infection or bacteremia that persisted despite at least 7 days of therapy, or (2) recurrence of CR-Kp infections after treatment was discontinued. Persistent bacteremia was defined by blood cultures that remained positive for >7 days, and recurrent bacteremia as the return of positive blood cultures that had initially cleared in the setting of clinical improvement. Patients with persistent infections were considered to be clinically improved if there was subjective improvement in clinical parameters (resolution of hypotension, fever, etc.), but blood cultures remained positive.

Strain characterizations

Susceptibility data were reported by the University of Pittsburgh Medical Center (UPMC) clinical microbiology laboratory. During the study period, tigecycline and colistin susceptibility testing was performed per physician's request only. Minimum inhibitory concentrations (MICs) were determined by standard broth microdilution for all agents [23], except for gentamicin and tigecycline, which were tested by Kirby–Bauer and E-test, respectively. Gentamicin, ciprofloxacin and carbapenem susceptibilities were defined according to the Clinical and Laboratory Standards Institute (CLSI) breakpoints for Enterobacteriaceae [23]. Tigecycline susceptibility was interpreted using US Food and Drug Administration recommended breakpoints since CLSI breakpoints are not defined; isolates exhibiting tigecycline MIC ≤2, 4 and >4 µg/mL were defined as susceptible, intermediately susceptible and resistant, respectively [24]. Since consensus breakpoints for colistin against Enterobacteriacae have not been established by the CLSI, we applied the breakpoints for Pseudomonas aeruginosa and Acinetobacter baumannii (susceptible ≤2 µg/mL) [23].

Strains were saved at −80°C and subcultured onto Mueller–Hinton agar at least twice prior to shipping to the Public Health Research Institute for molecular testing. Polymerase chain reaction (PCR) and/or DNA sequencing were used to detect resistance determinants and porin genes, as described previously [25-29]. Primers used to detect ompK35 and ompK36 genes are listed in Table S1. Nucleotide sequences were compared with data from GeneBank (http://www.ncbi.nlm.nih.gov/blast/). Strain typing by pulse-field gel electrophoresis (PFGE) was performed following XbaI restriction enzyme digestion [30]. The PFGE pattern was analyzed by BioNumerics (Applied Maths, Kortrijk, Belgium), using the Dice correlation coefficient and the unweighted pair group mathematical average (UPGMA) clustering algorithm.

Statistical analysis

Comparisons between two groups were performed by Wilcoxon rank sum (continuous variables). Correlation between two variables was assessed using Spearman's rank test. Significance was defined as p ≤ 0.05 (two-tailed).


Epidemiology and clinical characteristics

Seventeen transplant recipients developed CR-Kp bacteremia, representing 0.7% of 2427 organ transplants during the study period (Tables 1 and 2). Types of transplants were liver (1.3%, 8/610 of liver transplants), intestinal (5.4%, 6/112), lung (0.4%, 2/546) and liver–kidney (n = 1). Median time to onset posttransplant was 163 days (range: 6–1070 days) (Table 1). Fifty-three percent (9/17) of patients had CR-Kp recovered from another site prior to developing bacteremia: 41% (7/17) were colonized, and 6% (1/16) each had intra-abdominal (P9) and urinary tract infections (P13), respectively.

Table 1. Summary of demographics and clinical characteristics of transplant recipients with CR-Kp bacteremia
  • CR-Kp, carbapenem-resistant Klebsiella pneumoniae; ICU, intensive care unit.
  • 1One patient underwent a second liver transplant within 30 days prior to the diagnosis of bacteremia.
  • 2Drugs received within 30 days prior to the diagnosis of bacteremia.
  • 3Includes endocarditis and mycotic aneurysm (n = 1 each).
Median age (range)51 years (25–70 years)
Male sex59% (10/17)
Time to onset of bacteremia after transplant
Within 30 days29% (5/17)1
31–90 days12% (2/17)
91–180 days18% (3/17)
>180 days47% (8/17)
Residence in ICU at onset of bacteremia41% (7/17)
Calcineurin inhibitors94% (16/17)
Corticosteroid88% (15/17)
Mycophenolic acid41% (7/17)
Receipt of antimicrobial agents with activity against Gram-negative bacteria294% (16/17)
Ampicillin–clavulanate29% (5/17)
Piperacillin–tazobactam47% (8/17)
Cefepime41% (7/17)
Carbapenem24% (4/17)
Colistin12% (2/17)
Gentamicin6% (1/17)
Tigecycline6% (1/17)
Presentation at onset of bacteremia
Median APACHE II score (range)18 (4–26)
Septic shock18% (3/17)
CR-Kp infection identified at extra-blood site65% (11/17)
Intra-abdominal/biliary tract47% (8/17)
Upper urinary tract12% (2/17)
Lungs12% (2/17)
Cardiovascular12% (2/17)3
Median time (range) of follow-up for patients who survived >90 days after onset of bacteremia415 days (235–1599 days)
Table 2. Clinical characteristics of transplant recipients with CR-Kp bacteremia
PtAge, sexType of transplantTime to bacteremia1Complication(s) prior to CR-Kp bacteremiaOther site(s) of cultures positive for CR-Kp2Other bacteria (sites)Portal of entryTypes of disease
  • BAL, bronchoalveolar lavage; CR-Kp, carbapenem-resistant K. pneumoniae; d, day(s); wk, week(s); yr, year(s); ERCP, endoscopic retrograde cholangiopancreatography; GI, gastrointestinal tract; IAA, intra-abdominal abscess; JP, Jackson Pratt drain; Bronch, bronchoscopy; UTI, urinary tract infection; Pt, patient; s/p, status post; VRE, vancomycin-resistant Enterococcus.
  • 1Dated from transplant.
  • 2“Prior”: within 3 months prior to the onset of KPC-K. pneumoniae (KPC-Kp) bacteremia; “Bacteremia and after”: at the time of or within 1 month of the diagnosis of KPC-Kp bacteremia.
  • 3This patient developed bile leak leading to peritonitis and empyema following a re-do liver transplant. Cultures of peritoneal fluid and pleural effusion grew CR-Kp.
  • 4This patient had ascending cholangitis due to susceptible K. pneumoniae, treated with cefuroxime. He developed fever on Day 8 of cefuroxime; blood culture and IAA fluid grew CR-Kp.
  • 5This patient had purulent UTI due to CR-Kp pretransplant (treated with colistin × 7 days), then underwent orthotopic liver transplant. In the immediate posttransplant period, he had recurrent UTIs due to CR-Kp, treated with multiple antibiotics including colistin, gentamicin and doxycycline. The urinary isolates became gentamicin-resistant during therapy.
P148 yr, womanIntestinal6 dFulminant C. difficile colitis (diagnosed 5 d prior; active at time of bacteremia)Prior: urine (90 d)A. baumannii (blood)GI (ascribed to translocation due to C. difficile colitis)Septic shock
     Bacteremia and after: none   
P262 yr, manLiver714 dFulminant C. difficile colitis (diagnosed 2 wk prior; active at time of bacteremia)Prior: urine (7 and 89 d)NoneGI (ascribed to translocation due to C. difficile colitis)Septic shock
     Bacteremia and after: urine   
P370 yr, manLiver7 dPharyngeal injury due to feeding tube misplacement (diagnosed 1 d prior)Prior: noneNoneUrinary tractSeptic shock, disseminated infection
     Bacteremia and after: urine   
P441 yr, womanIntestinal/multivisceral254 dBleeding at ileostomy output (diagnosed 8 d prior; active at time of bacteremia)Prior: noneVRE, P. aeruginosa (abdominal fluid, blood)GI (ileocolonic anastomotic leak with focal IAA)IAA, endocarditis
     Bacteremia and after: leg abscess, surgical wound, IAA   
P543 yr, manLiver37 dNonePrior: noneNoneBiliary tree3Biliary sepsis, peritonitis
     Bacteremia and after: pleural effusion, peritoneal fluid   
P665 yr, manLiver6 dNonePrior: noneNoneAscribed to vascular catheterLine-associated bacteremia
     Bacteremia and after: sputum, BAL   
P743 yr, womanIntestinal407 dGraft enterectomy and gastrostomy placement (44 d prior); gastrostomy site leakage (diagnosed 10 d prior; active at time of bacteremia)Prior: urine (69 d)NoneGI (IAA)IAA, peritonitis
     Bacteremia and after: rectal abscess, JP, urine   
P851 yr, manLiver/kidney136 dNonePrior: urine (3 d)VRE (abdominal fluid)GI (IAA)IAA, pneumonia
     Bacteremia and after: IAA, intraoperative deep wound, urine, BAL   
P925 yr, womanIntestinal/multivisceral4.5 yrHemorrhagic and necrotizing pancreatitis, s/p ERCP (36 d prior) and multiple pancreatic debridements (first and last performed 61 and 2 d prior to bacteremia, respectively)Prior: JP (31 and 32 d), pancreas and wounds (2, 12 and 19 d)VRE, P. aeruginosa (abdominal fluid)GI (IAA and peritonitis)IAA, mycotic aneurysm
     Bacteremia and after: pancreas, abdominal fluid and JP drain   
P1042 yr, womanIntestinal163 dRefractory rejection requiring alemtuzumab then anti-thymocyte globulin within 90 d prior to bacteremiaPrior: NoneNoneAscribed to vascular catheterLine-associated bacteremia
     Bacteremia and after: none   
P1144 yr, womanIntestinal134 dGut perforation requiring repair (10 d prior); enterocutaneous fistula with bleeding (8 d prior; active at time of bacteremia)Prior: sputum (21 d)NoneGI (IAA)Peritonitis
    Multiple courses of steroid for rejection; alemtuzumab 28 d priorBacteremia and after: IAA   
P1266 yr, manLiver1070 dERCP (8 d prior)4Prior: noneA. baumannii (biliary fluid)Biliary tree3Ascending cholangitis
     Bacteremia and after: biliary fluid   
P1364 yr, manLiver409 dNonePrior: urine (18 d)NoneUrinary tractUrosepsis
     Bacteremia and after: urine, urethral pus, incisional wound, wound drainage, sputum, BAL5   
P1462 yr, manLiver3 dNonePrior: BAL (54 d), urine (41 and 54 d), BAL (2 d)NoneLungsPneumonia
     Bacteremia and after: sputum   
P1558 yr, womanLung201 dNeutropenia (7 d prior)Prior: noneNoneAscribed to vascular catheterLine-associated bacteremia
     Bacteremia and after: none   
P1630 yr, manLung235 dNonePrior: bronch wash (47 and 61 d)NoneAscribed to vascular catheterLine-associated bacteremia
     Bacteremia and after: catheter tip, sputum   
P1761 yr, manLiver10 dRe-anastomosis of blood vessel in transplanted liver and repair of incarcerated hernia (8 d prior)Prior: noneNoneAscribed to vascular catheterLine-associated bacteremia
     Bacteremia and after: sputum, BAL, urine   

All patients were receiving at least one immunosuppressive agent at the onset of bacteremia (Table 1). Within 3 months of bacteremia, 12% (2/17) received multiple steroid pulses, alemtuzumab and/or anti-thymocyte globulin for treatment of acute cellular rejection (P10–11) (Table 2). In the same time frame, 59% (10/17) of patients had a procedure or disease involving the gastrointestinal (GI) tract (some patients had multiple processes concomitantly). Procedures involving the GI or biliary tract were undertaken in 53% (9/17) of patients, including liver transplant (n = 4; P3, P6, P14, P17), intestinal transplant (n = 1; P1), repair of GI tract perforation (n = 2; P7, P11), re-anastomosis of liver vessels/hernia repair (n = 1; P17), debridement of necrotizing pancreatitis (n = 1; P9) and endoscopic retrograde cholangiopancreatography (ERCP; n = 2; P9, P12) (Table 2). Two patients had fulminant Clostridium difficile colitis at the time of transplant (P1–2), which was believed to serve as a portal of entry for CR-Kp into the bloodstream. In all, the ascribed portals of entry were GI or biliary tract (53%; 9/17), intravascular catheters (29%; 5/17), urinary tract (12%; 2/17) and lungs (6%; 1/17). In 59% (10/17) of patients, CR-Kp caused disease at a site in addition to the bloodstream (P3–5, P7–9, P11–14) (Table 2).

Treatment and outcomes

Mortality rates at 30 and 90 days after the diagnosis of bacteremia were 18% (3/17) and 47% (8/17), respectively (Table 3 and Table S2). Three patients died of septic shock within 6 days of the initial positive blood culture (P1–3) (Table 3). Clinical cures were achieved in 24% (4/17) of patients (P14–17). Clinical failures were observed in 59% (10/17), including patients who developed persistent or recurrent bacteremia (n = 5 each). In the former group, two patients died of persistent KPC-K. pneumoniae (KPC-Kp) infections within 50 days (P4–5), two patients tolerated persistent bacteremia for 375 and 305 days and spent extended periods of time as out-patients (P9 and P11, respectively) and one patient was cured after 19 days (P12). Among patients with recurrent bacteremia, three died within 42 days (P6–8) and two were ultimately cured (P10, P13).

Table 3. Treatment and outcomes of transplant recipients with CR-Kp bacteremia
PtAPACHE II scoreAntimicrobial susceptibility1Initial antimicrobial regimen (subsequent regimens for persistent infections)Active agent (n) in initial regimen?Source controlDuration of initial bacteremiaFinal outcome and follow-up time
Drug adverse effects (if any)Clinical outcome
  • CR-Kp, carbapenem-resistant Klebsiella pneumoniae; Pt, patient; n, number; N/A, not applicable; d, days; d/c, discontinued; IAA, intra-abdominal abscess; VRE, vancomycin-resistant Enterococcus.
  • Drugs—Col: colistin; Gent: gentamicin; Tig: tigecycline; Cipro: ciprofloxacin; I: intermediate; S: susceptible; R: resistant.
  • 1Results associated with reduced susceptibility (intermediate susceptibility or resistance) appear in bold. D0 (Day 0) represents the day that the blood culture that first revealed CR-Kp was collected.
  • 2Overall, renal failure occurred in 50% (2/4) and 60% (3/5) of patients receiving regimens that included gentamicin and colistin, respectively, for ≥3 days. Following initial episodes of bacteremia, renal failure developed in 50% (2/4) and 100% (2/2) of patients treated with gentamicin and colistin, respectively. Renal failure also occurred in one patient who received colistin for treatment of recurrent bacteremia (P13).
  • 3Patients with persistent or recurrent bacteremia were considered to be clinically improved if there was subjective improvement in clinical parameters despite ongoing or recurring positive blood cultures.
P121D0:Piperacillin–tazobactam (started 5 d prior to bacteremia), gentamicin and meropenem (started Day 1, until death)Yes (1)N/AN/ADied at 1 d
  Gent: S   Clinical failure (died) 
  Tig: S (1.5 µg/mL)     
  Col: S (0.25 µg/mL)     
P217D0:Piperacillin–tazobactam (started Day 0, until death)NoneN/AN/ADied at 6 d
  Gent: S   Clinical failure (died) 
  Tig: S (2 µg/mL)     
  Col: S (0.25 µg/mL)     
P326D0:Cefepime and gentamicin (started Day 1, until death)Yes (1)N/AN/ADied at 2 d
  Gent: S   Clinical failure (died) 
  Tig: S (2 µg/mL)    Autopsy: disseminated infection (mediastinitis, peritonitis, pneumonia)
  Col: S (0.125 µg/mL)     
P49D0:Piperacillin–tazobactam (started Day 1, ×10 d)NoneN/A34 d Clinical failure (persistent bacteremia)Died at 37 d
  Gent: S(Meropenem and gentamicin added on Day 10). Meropenem continued until death   Blood and wound culture 3 d before death remained positive for CR-Kp
  Tig: S (0.5 µg/mL)Gentamicin d/c after 9 d due to renal failure2    
  Col: S (0.25 µg/mL)     
  Gent: S     
  Tig and Col: not tested     
P518D0:Colistin and gentamicin (started Day 1, ×13 d)Yes (1)Multiple abdominal wash out and biliary leak repair25 dDied at 50 d
  Gent: RGentamicin d/c after 13 d due to resistance  Clinical failure (persistent bacteremia)On treatment at time of death. Blood culture was negative 19 d prior to death; wound culture remained positive for CR-Kp 5 d prior to death
  Tig: S (1.5 µg/mL)(Doripenem and tigecycline added on Day 15; rifampin added on Day 31)    
  Col: S (<0.125 µg/mL)Rifampin d/c after 3 d due to increase bilirubin    
  Gent: R     
  Tig: S (2 µg/mL)     
  Col: R (64 µg/mL)     
P619D0:Cefepime (started on Day 1, ×2 d)Yes (1)Vascular catheter removal4 d1 episode of recurrent bacteremia. Died at 42 d after recurrence
  Gent: SGentamicin (started on Day 3, ×21 d)  Clinical failure (recurrent bacteremia on Day 36)Cause of death unknown
  Tig: S (1 µg/mL)Gentamicin d/c after 21 d due to worsening renal failure requiring hemodialysis2    
  Col: S (0.5 µg/mL)(Treatment of recurrent bacteremia described in Table S2)    
  D36 (recurrence):     
  Gent: I     
  Tig: S (1 µg/mL)     
  Col: S (0.5 µg/mL)     
P74D0:Doripenem (started on Day 4, ×14 d)NoneGastrostomy closure and debridement of abdominal cavity1 dMultiple episodes of recurrent bacteremia. Died at 20 d after a 2nd recurrence
  Gent: R(Treatment of recurrent bacteremia described in Table S2)  Clinical failure (recurrent bacteremia on Day 32)On treatment for recurrent bacteremia at time of death. Blood culture was negative for 14 d prior to death
  Tig: S (2 µg/mL)     
  Col: R (64 µg/mL)     
  D32 (recurrence):     
  Gent: R     
  Tig: not done     
  Col: not done     
  D66 (2nd recurrence):     
  Gent: R     
  Tig: not done     
  Col: not done     
P820D0:Breakthrough bacteremia while on gentamicin (started on Day 37 prior to bacteremia for CR-Kp IAA); gentamicin d/c on Day 2 of bacteremiaYes (1)Open debridement (Day 17 and 22), and repair of gastrostomy tube leakage (Day 22)1 d1 episode of recurrent bacteremia. Died at 28 d after recurrence
  Gent: SColistin replaced gentamicin on Day 2, ×14 d  Clinical failure (recurrent bacteremia on Day 57)On treatment for recurrent bacteremia at time of death. Blood culture was negative 26 d prior to death. CR-Kp pneumonia diagnosed 9 d prior to death
  Tig: I (4 µg/mL)(Treatment of recurrent bacteremia described in Table S2)    
  Col: S (<0.125 µg/mL)     
  D57 (recurrence):     
  Gent: I     
  Tig: I (4 µg/mL)     
  Col: S (0.5 µg/mL)     
P921D0:Breakthrough bacteremia while on tigecycline (started 30 d prior to bacteremia for CR-Kp intra-abdominal infection)NoneMultiple open debridements. Last operation involved resection of mycotic aneurysm (Day 350)375 dClinical improvement on therapy but persistently bacteremic. Died at Day 405 due to complications of surgery to resect a mycotic aneurysm
  Gent: RColistin and doripenem started on Day 6. Developed worsening renal failure while on colistin, requiring hemodialysis2  Clinical improvement but persistent bacteremia3On therapy for CR-Kp infection at the time of death. Blood culture was positive for CR-Kp 30 d prior to death, but then became negative. Ascites fluid culture was positive 4 d prior to death
  Tig: I (4 µg/mL)     
  Col: S (<0.125–0.25 µg/mL)     
  D43 (abscess) and D100 (blood):     
  Gent: R     
  Tig: I (4 µg/mL)     
  Col: R (16-64 µg/mL)     
  D162 (blood):     
  Gent: R     
  Col: R (32 µg/mL)     
P1010D0 and D28 (recurrence):Ciprofloxacin and ertapenem (started on Day 2, ×6 d)Yes (1)Vascular catheter removal1 d3 episodes of recurrent bacteremia, each related to vascular catheter. Cured after the third episodes of recurrent bacteremia
  Cipro: S(Treatment of recurrent bacteremia described in Table S2)  Clinical failure (recurrent bacteremia on Day 51)Died at Day 415 after the last episode of recurrent bacteremia, due to transplant-related issues
  Gent: R     
  Tig: S (1.5 µg/mL)     
  Col: S (0.5 µg/mL)     
  D79 (2nd recurrence):     
  Cipro: I     
  Gent: S     
  Tig: S (0.4 µg/mL)     
  D118 (3rd recurrence):     
  Cipro: R     
  Gent: S     
  Tig: R (8 µg/mL)     
P1117D0:Piperacillin–tazobactam (started on Day 2, ×73 d)NoneMultiple debridements and multiple courses of immune-suppression to control rejection; graft removal (Day 302)305 dDied at Day 306 after the onset of bacteremia, after attempted corrective surgery (graft removal)
  Gent: R(Subsequently failed multiple combinations of carbapenems, colistin, tigecycline, piperacillin–tazobactam)  Clinical improvement but persistent bacteremia3On treatment for CR-Kp infection at the time of death. Blood culture, wound and hematoma fluid culture were positive 7 d prior to death
  Tig: R (6 µg/mL)     
  Col: R (64 µg/mL)     
P1218D0:Cefepime (started on Day 0, ×2 d)Yes (1)Percutaneous biliary drain19 dAlive at 2.9 years
  Gent: SGentamicin (started on Day 1, ×7 d, stopped due to breakthrough bacteremia)  Clinical improvement but persistent bacteremia3 
  Tig: S (2 µg/mL)Tigecycline started on Day 12, ×27 d    
  Col: R (64 µg/mL)     
  Gent: R     
P1318D0:Doripenem + doxycycline (started on Day 3, ×14 d)Yes (1)N/A1 d1 episode of recurrent bacteremia. Died at Day 340 after recurrent bacteremia, due to graft failure and staphylococcal sepsis
  Gent: R(Treatment of recurrent bacteremia described in Table S3)  Clinical failure (recurrent bacteremia on Day 165) 
  Col: S (<0.125 µg/mL)     
  D165 (recurrence):     
  Gent: R     
  Col: R (8 µg/mL)     
P1415D0:Doripenem + colistin (started on Day 1, ×14 d); gentamicin and tigecycline added on Day 3 (due to colistin-resistant CR-Kp and VRE, respectively)Yes (2)N/A1 dAlive at Day 675
  Gent: S   Microbiologic and clinical cure 
  Tig: S (1 µg/mL)     
  Col: R (8 µg/mL)     
P1513D0:Gentamicin (started on Day 3, ×14 d)Yes (1)Vascular catheter removal1 dAlive at 4 years
  Gent: S   Microbiologic and clinical cure 
  Tig: S (1 µg/mL)     
  Col: S (0.25 µg/mL)     
P1621D0:Gentamicin (started on Day 2, ×7 d)Yes (1)Vascular catheter removal1 dDied at Day 566 after the onset of bacteremia, due to chronic allograft rejection
  Gent: S   Microbiologic and clinical cure 
  Tig: S (1 µg/mL)     
  Col: R (8 µg/mL)     
P1714D0:Piperacillin–tazobactam (started on Day 0, ×3 d)Yes (1)Vascular catheter removal1 dDied at Day 235 after the onset of bacteremia, due to graft failure and P. aeruginosa pneumonia
  Gent: SDoripenem + colistin (started on Day 3, ×14 d)  Microbiologic and clinical cure 
  Tig: I (4 µg/mL)Patient developed worsening renal failure on colistin requiring hemodialysis2    
  Col: S (0.25 µg/mL)     

The three patients who died of acute septic shock either did not receive an antimicrobial regimen that was active against the infecting strain (P2) or received an active agent for only 1 day (P1 and P3); they are excluded from the following discussion of antimicrobial regimens and surgical interventions. Among the remaining patients, 29% (4/14), 50% (7/14) and 21% (3/14) were treated with an initial antimicrobial regimen that included agents that were inactive in vitro against the infecting strain, a single active agent or a carbapenem combined with an active agent, respectively (Figure 1). All 4 patients who did not receive an active agent in vitro developed persistent (n = 3, P4, P9, P11) or recurrent (n = 1, P7) bacteremia. Gentamicin as a single active agent was associated with cure (n = 2, P15–16) and persistent (n = 1, P12) or recurrent bacteremia (n = 1, P6). Colistin as a single active agent was associated with persistent (n = 1, P5) or recurrent bacteremia (n = 1, P8). Ciprofloxacin as a single active agent was associated with recurrent bacteremia (n = 1, P10). Combination regimens that included doripenem–colistin resulted in cures of both patients (P14, P17); tigecycline and gentamicin were included as part of the regimen in one of these patients (P14). Doripenem–doxycycline was used to treat the final patient, who had CR-Kp bacteremia and urosepsis (P13). His bacteremia resolved but recurred 165 days later. The four patients who were cured upon their initial presentation with bacteremia received 7–14 days of antimicrobial therapy.

Figure 1.

Summary of treatment and outcomes of transplant recipients with CR-Kp bacteremia.*

*Patients who died of septic shock within 6 days were excluded because they either received an inactive agent (piperacillin-tazobactam; P2) or received a regimen with a single active agent for only 1 day (P1 and P3). CR-Kp, carbapenem-resistant Klebsiella pneumoniae; n, number; (I), intermediate; (R), resistant.

Interventions directed at source control for bacteremia were undertaken in 86% (12/14) of patients, including all four patients who were cured upon their initial presentation. Fifty percent (7/14) of patients underwent intra-abdominal surgery or drainage for infections of the GI and/or biliary tracts. Clinical cure was achieved in a liver transplant recipient with ascending cholangitis who eventually cleared his persistent bacteremia of 19 days after undergoing biliary drainage (P12). The other six patients who underwent intra-abdominal interventions experienced persistent (n = 4) or recurrent (n = 2) bacteremia. Of note, the two intestinal transplant recipients with persistent bacteremia for >300 days ultimately died from complications of surgery to address presumed sources of infection (resection of mycotic aneurysm (P9) and removal of intestinal graft (P11)). Thirty-six percent (5/14) of patients had intra-vascular catheters removed for line-associated bacteremia. Three of these patients were cured (P15–17), and two developed recurrent bacteremia (P6, P10).

Antimicrobial regimens that were associated with the sterilization of blood cultures among the five patients with recurrent bacteremia were doripenem–colistin (n = 4, P6–8, P13) and meropenem (n = 1, P10) (Table S2). The duration of antimicrobial therapy ranged from 7 to 21 days. Four of the patients also underwent source control interventions, including debridement or drainage of intra-abdominal infections (n = 2, P7 and P8, respectively) and removal of intravascular catheters (n = 2, P6 and P10).

Renal failure occurred in 50% (2/4) and 60% (3/5) of patients receiving regimens that included gentamicin and colistin, respectively, for ≥3 days (Table 3 and Table S2). Three patients required hemodialysis due to antibiotic-induced renal failure (colistin, n = 2; gentamicin, n = 1; P6, P9, P17). One patient who received rifampin as part of a combination regimen developed hyperbilirubinemia that prompted discontinuation of the agent (P5).

Antimicrobial resistance and molecular typing of CR-Kp strains

The median doripenem MIC against initial bloodstream strains was 32 µg/mL (range: 2–256 µg/mL) (Table 4 and Table S3). At some point, 53% (9/17), 47% (8/17) and 29% (5/17) of patients had bacteremia due to a strain that was not susceptible to gentamicin, colistin and tigecycline, respectively. Thirty-five percent (6/17), 29% (5/17) and 24% (4/17) of initial bloodstream strains were not susceptible to the respective agents.

Table 4. Summary of ompK36 variants in ST258, KPC-2 strains undergoing molecular analysis
VariantBloodstream strainsNonbloodstream strains
Strains (n)Patients (n)1Strains (n)Patients (n)2
  • n, number.
  • 1In two patients, two different ompK36 variants of KPC-K. pneumoniae (KPC-Kp) were recovered from the blood.
  • 2In four patients, two different ompK36 variants of KPC-Kp were recovered from the blood and extra-blood site.
ins AA134-135 GD9652
NNTE deletion (AA84-87)2100
ins nt382 G1100

Ninety-four strains were available for molecular characterization, including at least one bloodstream strain from each patient (72 bloodstream strains from 17 patients; 22 nonbloodstream strains from 10 patients) (Table S3). All CR-Kp strains harbored blaKPC: 6% (1/17) of patients were infected with ST37, KPC-3-producing strains, and 94% (16/17) with epidemic K. pneumoniae clone ST258, harboring Tn4401a with the blaKPC-2 gene that encodes KPC-2 carbapenemase. KPC-3-producing ST37 strains harbored blaSHV-1, and had wild-type ompK35 and ompK36 genes. All KPC-2-producing ST258 strains harbored blaTEM-1 and blaSHV-12, and had guanine (G) inserted at nucleotide position 121 within ompK35, resulting in a premature stop codon at amino acid position 89 (AA89-STOP).

Sequence analysis of the ompK36 gene in ST258 strains identified five different genotypes (wild-type and four distinct mutations) (Table 4). Mutations included IS5 insertions in the ompK36 promoter, a 6-bp insertion that encodes for glycine and aspartic acid at amino acid (AA) position 134–135 (ins AA134–135 GD), a guanine insertion at nucleotide position 382 (ins nt382 G), and an aspragine–asparagine–threonine–glutamic acid (NNTE) deletion at AA position 84–87 (del AA84-87NNTE). KPC-2-producing ST258 strains with a mutant ompK36 gene were recovered from 71% (12/17) to 80% (8/10) of patients with bacteremia and persistent/recurrent bacteremia, respectively. In two patients, persistent bacteremia was caused by two KPC-2 strains with different ompK36 gene mutations (P4 and P9) (Table S3 and Figure S1); four patients were infected by strains with different ompK36 mutations at blood and extra-blood sites. Overall, carbapenem MICs were higher against strains with IS5 or ins AA134-135 GD mutations than against wild-type ompK36 strains (Figure 2). Carbapenem MICs were lowest against KPC-3-producing ST37 strains (Table S3).

Figure 2.

Distribution of doripenem MICs and ompK36 variants. Doripenem was chosen to represent the carbapenem class since it is the formulary agent at our center. MICs, minimum inhibitory concentrations.

Serial isolates recovered from blood and other sites of individual patients clustered closely in PFGE dendrograms, including ompK36 mutant and wild-type variants (Figure 3). All KPC-2-producing ST258 strains were clustered together, and shared >86.5% similarities (dice similarity coefficient). KPC-3-producing ST37 strains were highly distinct from ST258 strains (<65% similarities).

Figure 3.

PFGE of representative KPC-Kp isolates from the 17 patients with bacteremia. PFGE dendrogram showing the relatedness of 41 KPC-producing K. pneumoniae isolates. All KPC-2-producing ST258 strains were clustered in one group, sharing >86.5% similarities. KPC-3-producing ST37 strains were distinct from ST258 strains (<65% similarities). The numbers following the abbreviations for the patients' samples correspond to days before (number preceded by a minus sign) or after the first positive blood culture (number alone). For example, U-90 denotes urine culture obtained 90 days before the first positive blood culture. B0 denotes the first positive blood culture. T24 denotes tissue culture obtained 24 days after the first positive blood culture. PFGE, pulse-field gel electrophoresis; KPC-K. pneumoniae (KPC-Kp).


To our knowledge, this is the first comprehensive study of CR-Kp bacteremia that describes long-term outcomes among transplant recipients and evaluates the genetic profiles of a large number of infecting strains. There are three particularly noteworthy findings. First, 88% of CR-Kp bacteremia occurred among intestinal and liver transplant recipients, in whom infections were often associated with intra-abdominal abscesses, peritonitis or biliary infection. In a majority of patients, bacteremia was ascribed to a GI portal of entry. Second, outcomes among transplant recipients were distributed across a remarkably broad continuum. On either end of the continuum, 18% and 24% of patients died rapidly due to septic shock or were cured upon initial presentation, respectively. In between, outcomes ranged from subacute deaths over 31 to 90 days, to persistent bacteremia that was well tolerated for up to 375 days, to recurrent bacteremia. Third, all CR-Kp strains were KPC producers, and 94% (16/17) of patients were infected with strains derived from a KPC-2-producing ST258 clone that predominated at our center. Despite the fact that ST258 strains were clonal by conventional molecular epidemiologic criteria, they harbored five unique ompK36 porin genotypes. Three ompK36 mutants were associated with high-level carbapenem resistance compared to wild type. Therefore, the epidemiology and clinical manifestations of KPC-Kp bacteremia among transplant recipients were in keeping with the life cycle of K. pneumoniae as a GI tract colonizer [31], and the disease was characterized by diversity in both outcomes and porin genotypes of infecting strains.

KPC-Kp bacteremia was diagnosed in 5.4% and 1.3% of intestinal and liver transplant recipients, respectively, compared to 0.4% of lung transplant recipients. Intestinal transplants, in particular, carry risks because they involve exposure to highly contaminated environments, and patients require robust immunosuppressive therapy to temper the immunogenicity of the GI tract. Overall, 59% (10/17) of patients had a disease or procedure involving the GI tract within 30 days of bacteremia. Routine screening for CR-Kp carriage was not performed during the study period, so we are unable to establish conclusive links between GI colonization and subsequent disease [12, 31]. However, 53% (9/17) of patients had KPC-Kp isolated from a site of colonization or infection prior to bacteremia, suggesting that strain carriage is a risk factor. As in previous studies [32, 33], almost all patients received antimicrobials with broad-spectrum activity against Gram-negative bacteria (but not necessarily a carbapenem) in the month before bacteremia was diagnosed. In contrast to an earlier report in which the median time to CR-Kp infection following liver transplantation was 12 days [31], approximately half of our cases of KPC-Kp bacteremia occurred more than 6 months after transplant. Based on these data, KPC-Kp should be considered as a potential cause of bacteremia among intestinal and liver transplant recipients at any point posttransplant, particularly if they were previously colonized or infected, have recently had GI disease or undergone a GI procedure, received broad-spectrum antimicrobials and/or have a concurrent intra-abdominal infection.

Our 30-day mortality rate of 18% (3/17) was low compared to previous studies of KPC-Kp bacteremia among transplant and nontransplant patients, in which rates often exceeded 40% [13-18, 31]. Our data were tempered, however, by a 90-day mortality rate of 47% (8/17) and a low cure rate for initial episodes of bacteremia. It is plausible that 90-day rather than 30-day mortality more accurately describes KPC-Kp bacteremia among transplant recipients in the present era, since aggressive critical care management may prolong survival but not change final outcomes. Prolonged survival generally came at the cost of refractory bacteremia, as 71% of patients who lived more than 30 days had persistent or recurrent infections (36% each). The most striking cases were two intestinal transplant recipients with persistent bacteremia for over 300 days. Remarkably, these patients largely tolerated their bacteremia without signs or symptoms of systemic infection, in essence behaving as if they had bloodstream colonization. Regardless, the patients died following complications of surgery to eradicate presumed deep-tissue sources of persistent infection. The inability to eliminate KPC-Kp strains from the bloodstream in the majority of patients may be attributed to a combination of immunosuppression, suboptimal antimicrobial therapy, the presence of infections at other sites and/or adaptation of the infecting organisms to the host environment.

The size of this study limits our ability to draw definitive conclusions about treatment of KPC-Kp bacteremia. Nevertheless, several findings stand out. First, administration of agents that were inactive in vitro was ineffective. Second, the use of a single active agent was unsuccessful in 71% (5/7) of patients, and further limited by the loss of susceptibility to gentamicin (n = 2), colistin (n = 1) or ciprofloxacin (n = 1) in each instance of treatment failure. The emergence of gentamicin resistance offset the fact that two patients were rapidly cured of bacteremia while receiving this agent. Third, most successful regimens included the combination of doripenem–colistin. In addition to curing two patients upon initial presentation, these regimens sterilized the blood of four patients with recurrent bacteremia. Fourth, tigecycline was rarely used as primary treatment for bacteremia, likely because of its low serum concentrations and higher mortality rates than with comparative drugs in pooled clinical studies [34]. Finally, source control was an essential component of successful treatment, as the four patients who were cured underwent concurrent IV line removal or drainage of the billiary tree. Taken together, the data are consistent with studies showing that combination therapy, use of active antimicrobial agents and source control are associated with improved outcomes among patients with CR-Kp bacteremia [3, 14, 16, 35]. If the source of bacteremia and persistent foci of infection are eliminated, our experience suggests that ∼14 days of antimicrobial therapy are sufficient.

We identified three major obstacles to these treatment principles. First, salvage agents such as gentamicin, colistin and tigecycline were inactive against a significant minority of strains at the time that bacteremia was first diagnosed. In some cases, antimicrobial resistance likely stemmed from prior exposure in the setting of KPC-Kp colonization or nonbloodstream infections. However, possible nosocomial transmission was also implicated, as none of the patients who were initially bacteremic with colistin-resistant strains had received the agent previously. Second, renal toxicity developed in 50% and 60% of patients who received gentamicin and colistin, respectively, for ≥3 days. The rates of nephrotoxicity are higher than previous reports [36, 37], which may reflect the potentiating effects of sepsis, co-administration of other toxic agents such as calcineurin inhibitors and the need for prolonged treatment courses due to severity of infections and frequent persistence/recurrence. Third, source control was difficult to achieve, particularly at intra-abdominal sites. In fact, six of the seven patients who had intra-abdominal foci of infection suffered persistent or recurrent bacteremia, despite drainage or other surgical interventions. Given the poor outcomes and limited treatment options, it is clear that new antimicrobials with activity against CR-Kp are needed. In studying the potential roles for agents such as novel β-lactam–β-lactamase inhibitor combinations and neoglycosides [38, 39], it will be important to both establish therapeutic efficacy and define strategies that limit the further emergence of resistance.

Our PFGE and sequence data suggest a model for the emergence of KPC-Kp strains within our program. The predominant ST258, KPC-2-producing clone was introduced to our center prior to 2008, and harbors blaTEM-1 and blaSHV-12 ESBLs and a mutant ompK35 porin gene. Various ompK36 mutants have emerged within this clone, likely under the selection pressure of widespread, broad-spectrum antimicrobial usage. The fact that the same IS5 insertions in the ompK36 promoter region were identified in strains from different patients suggests that acquisition of nosocomial strains is more common than independent emergence of mutations. Indeed, highly susceptible transplant recipients may become infected by strains with different ompK36 porins. For example, 30% of patients with refractory bacteremia had positive blood or intra-abdominal abscess cultures that alternated between strains with different ompK36 genes over time. As illustrated by the intestinal transplant recipient in Figure 4, these refractory infections appeared to represent co-infections with different strains rather than the replacement of a wild-type ompK36 strain with an ompK36 mutant. There were no conclusive relationships between exposures to specific antimicrobial agents and the recovery of particular porin mutants.

Figure 4.

Serial ST258, KPC-2-producing strains with ompK36 variants isolated from an intestinal transplant recipient with persistent bacteremia (P9). Blue and red diamonds represent bloodstream and non-bloodstream strains, respectively.

The most common ompK36 mutations among our KPC-Kp strains (AA134-135 GD insertions and IS5 mutations in the promoter) have been linked to elevated carbapenem MICs, but the clinical implications of high-level carbapenem resistance are unknown. We recently showed that low levels of ompK35 and ompK36 expression by ST258, KPC-2-producing K. pneumoniae strains correlated with poor responses to combinations of colistin, doripenem and ertapenem in vitro [25]. These data suggest that porin function may be an important determinant of the efficacy of antimicrobial combinations, especially when a carbapenem is included. At the same time, porin mutations may impact the clinical course of KPC-Kp infections independently of their effects on antimicrobial susceptibility. ompK36 mutations, for example, are known to reduce microbial fitness and virulence, possibly through impairments in the handling of nutrients and toxic metabolites [40]. An intriguing possibility is that several of our patients tolerated long-term, persistent KPC-Kp bacteremia because they were infected by mutant strains of attenuated virulence. The data indicate that it should be possible to develop molecular markers to rapidly type KPC-Kp strains responsible for the majority of infections at our center, which would be valuable tools for epidemiologic investigations. Moreover, they may have future clinical utility, if specific strains are linked conclusively to diminished responses to antimicrobial regimens or altered virulence.

In addition to shortcomings mentioned earlier, this study was limited by its retrospective, single-center design. Analyses were restricted to existing data and available KPC-Kp strains. As such, we cannot exclude that other strains may have infected patients. Nevertheless, the study provides new insights into the clinical manifestations and outcomes of KPC-Kp bacteremia, and marks an initial attempt to link detailed clinical data with the molecular characterization of strains. In the latter regard, the study serves as a paradigm for future investigations of CR-Kp infections at other centers and in other patient populations. The extent to which our findings are particular to our center and transplant recipients is uncertain. We believe that immunosuppression and other factors render transplant patients more susceptible to extreme clinical outcomes following KPC-Kp infections, and more dependent upon optimal therapeutic interventions. As such, studies of transplant recipients are likely to be powerful models for understanding the full extent of clinical manifestations, treatment responses, outcomes and host–pathogen interactions in other populations.


We would like to thank Lloyd Clarke for his help in data collection. This study was partially supported by the University of Pittsburgh Department of Medicine, University of Pittsburgh Medical Center (to the XDR Pathogen Laboratory), Public Health Research Institute Director's Fund, and a grant (to B.N.K.) from the National Institutes of Health (1R01AI090155). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. R.K.S. is supported by the National Institutes of Health through grant numbers KL2RR024154 and KL2TR000146.


The authors of this manuscript have no conflicts of interest to disclose as described by the American Journal of Transplantation.