Genotypic characterization of neonatal and pediatric CRE isolates
A total of 163 CRE isolates were recovered from clinical samples of 155 infant patients under the age of 1 year admitted in both neonatal and pediatric wards of our hospital during the period from April 2013 to May 2018. CRE prevalence especially occurred in newborns under three days, accounting for 71.78% (117). KP (95.71%, 156) was the most prevalent organism, followed by E.coli (3.07%, 5) (Table 1). The most frequently identified carbapenemase gene was NDM-1 (63.19%, 103), followed by KPC-2 (24.54, 40), and IMP-4 (12.27%, 20).
Of the 1,650 patients who were prospectively screened for CRE colonization during the two study periods (948 in the first period and 702 in the second period), 257 patients were identified to be colonized by carbapenem-resistant organisms (CROs) in which 244 cases were carried with NDM-1 producing Enterobacteriaceae strains. Among these colonized patients, 30 (12.3%) had developed pulmonary infections with these organisms. Co-carriage of two NDM-1 strains was found in 43 (17.12%) (26 (15.57%) of 167 in the first period vs 17 (22.08%) of 77 in the second period, P = 0.215) of the 244 CRE patients. CRE rectal prevalence rate in the 29 DO-1 YO groups were significantly decreased from 14.29% (15/105) in the first period to 4.07 % (7/172) in the second period (P = 0.002). The rectal prevalence rates of NDM-1 strains in both the 0-3 DO and the 4-28 DO groups were decreased, but no statistical significances were reached (15.53%, 100/644 vs 11.78%, 51/433, P = 0.082 and 27.37%, 52/190 vs 19.59%, 19/97, P = 0.148, respectively).
In the first period, 193 CRE were isolated, including 182 NDM-1 CPKP of 7 STs, with ST20 (40.11%, 73), ST2068 (36.26%, 66), and ST36 (19.23%, 35) being the 1st, 2nd and 3rd most common types, 8 NDM-1 E.coli of 6 STs and 3 NDM-1 Enterobacter cloacae strains. Although transmission events continued to occur during the second period, isolates of ST types or species were completely different from those in the first period, and all isolates were NDM-1-producers consisting of 65 CPKP of 8 STs with a predominance of ST17 (73.85%, 48) and 29 E.coli of 4 STs with ST325 (89.62%, 26) as the major type (Table 1).
Neonatal outbreak evolution and infection control measures
In April 2013, a sporadic epidemic of NDM-1 ST39 KP occurred in neonatal ward, involving 13 cases less than 3 days old. The strain was initially isolated in March 2013 from a 5-month-old child with community-acquired pneumonia in the pediatric ward and might subsequently be transmitted to the neonatal ward via rotating medical staff. In July 2014, KPC-2 ST11 KP and IMP-4 ST307 KP clones were introduced into the ward and then spread rapidly, affecting 40 and 17 neonatal cases, respectively. Despite the implementation of basic infection control practices, especially enhanced environmental disinfection and reinforced hand hygiene before and after patient contact, new cases were still emerging. Routine environmental screening demonstrated a heavy environmental contamination with CROs (Supplementary Table 1 and Supplementary Figure 1). ERIC analysis of patients’ isolates and environmental isolates recovered from both the neonatal ward and adult ICU showed that some Aci.baumannii strains and ST11 KP clones were, respectively, closely related (Supplementary Figure 1), suggesting an important route of ‘non-patient transfer’ transmission between wards. So a new ward was opened in another building at the end of 2014 to receive new admissions. Those who had been affected remained in the original ward until they were discharged. Extensive environmental screening cultures were performed one month before and after the opening of the new ward, but did not yield any CROs.
In November 2015, the new unit was partially closed to external admissions due to NDM-1 ST20 and ST2068 KP . To determine the probable sources of the outbreak strains, 342 stool or perianal samples and 286 amniotic fluid samples collected from 342 pregnant women with suspected intrauterine infection in the maternity ward were screened for CROs, but no isolates were identified to be carbapenem-resistant. Therefore, the likely route of CRO through vertical transmission from the mothers at birth was excluded. The fact that there were no CROs circulating in the maternity ward and that no CROs had been isolated from routine obstetric environmental screenings did not support the transmission from maternity to neonatal ward. Next, we speculated that asymptomatic carriers might bring these organisms into the unit, so active rectal screening cultures were initiated for infant patients under one year old.
In October 2016, a new NDM-1 ST36 KP clone emerged and spread, affecting 33 cases. In March 2017, the initial results of rectal cultures were fed back to the ward, emphasizing that the frequent community importations of asymptomatically colonised patients and subsequent transmissions should be considered as key elements for intervention. A series of intensified measures were then instituted and implemented, including enhanced environmental disinfection with chlorine-based compound in frequency and extent, especially the carriers’ surroundings, the reduction of the total beds of the room from 14 to 10, increase of staff numbers, and reinforcement of hand hygiene compliance. Afterward, the number of clinical CRE isolates decreased drastically, and the CROs ceased to be detected from the ward environment.
In September and October 2017, two new NDM-1 strains of ST325 E.coli and ST17 KP were separately introduced and silently spread, leading to colonization in 23 and 40 cases, respectively. This suggested that these measures were far from sufficient to avoid silent transmissions. We then separately performed prospective matched case-control studies to determine neonatal risk factors associated with rectal colonization of NDM-1-producing Enterobacteriaceae in the 0-3 DO group and the 4-28 DO group. In the multivariate analysis, we found that gastric lavage and enema were independent risk factors for colonization in 0-3 DO group, whereas gastric lavage was a risk factor independently associated with colonization in 4-28 DO group (Table 2). These results strongly suggested that improperly sterilized procedures had probably direct causal role in neonatal colonization and transmission. So strict aseptic procedures were emphasized and monitored to insure stringent compliance in caring for high-risk patients. With the addition of these interventions, sporadic new CRE cases were still identified in the following years, but no outbreaks occurred.
Neonatal outbreak tracking revealed rivers contaminated with NDM genes
A detailed analysis of patient admission data combined with cultures and molecular typing revealed that multiple or repeated independent introductions of diverse NDM-1 producers into the neonatal ward direct from the community through colonized/infected infant patients, especially patients older than 3 days, resulted in continued clone transmission events (Figure 2). None of these colonized cases and their family members had any prior history of foreign travel to, or previous exposure to, known areas of high NDM endemicity. Multivariate analysis showed that residence in rural area was an independent risk factor, so we tried to put patients’ home addresses on a regional map to determine the scope of CRE endemicity in the community. We found that most patients lived in the vicinity of Hutuo River or tributaries where people frequently visited in summer (Supplementary Figure 2). Significantly increased incidences of carriage with community-acquired NDM-1strains were observed during summers (river contact seasons) compared to winters (no river contact seasons) (16 (24.62%) of 65 vs 15 (12%) of 125, P = 0.026). The obvious seasonality in NDM carriage rates might be, in large part, associated with increased chances of exposure to the rivers.
To determinate the rivers as the likely sources, extensive water samplings were carried out. A total of 83 CROs were recovered and screened for the presence of common carbapenemase genes (Supplementary Table 2 and Table 3). NDM producers could only be isolated at sampling sites C, D, and E (Figure 1). NDM-1 isolates (Aci.baumannii, Aci.lwoffiii, KP of ST37 and ST1306), NDM-5 E.coli of ST167, IMI-3 E.cloacae, OXA-24 Aci.lwoffiii and blaOXA-24, and blaOXA-51 co-producing Aci.baumannii isolates were detected in the rivers. The antimicrobial susceptibility profiles of water NDM isolates were similar to those of clinical isolates, showing high-level resistance to penicillins and cephalosporins but susceptibility to aminoglycosides and quinolones. ERIC analysis showed that water isolates of Aci.lwoffiii were unrelated to each other, suggesting horizontal plasmid transmission (Figure 3A), while ST37 KP and ST167 E.coli isolates recovered from river were genetically related to those isolated from patients (Figure 3B and 3C). Water ST37 isolate harbored extended-spectrum beta-lactamase genes (ESBLs) (SHV-31/CTX-15) and IncX3 replicon, whereas neonatal ST37 isolate possessed TEM-1/SHV-11/CTX-M-14 and A/C replicon. Both water and clinical isolates of ST167 E.coli had the same ESBLs (TEM-1/CTX-15) and IncFII replicon, but colistin resistance gene MCR-1 was detected only in the latter, suggesting that this strain likely originated from the river and had underwent a complex evolutionary process before transmission to humans.
Transferability of water isolates via conjugation
Water isolates of ST1306 NDM-1 KP and ST167 NDM-5 E.coli, and NDM-1Aci.lwoffii isolates were able to transfer, while the ST37 NDM-1 KP strain and one Aci.lwoffii isolate that co-produced both NDM-1 and OXA-24 failed to transfer.