Nationwide Surveillance and Molecular Characterization of Critically Drug-Resistant Gram-Negative Bacteria: Results of the Research University Network Thailand Study

ABSTRACT A large-scale surveillance is an important measure to monitor the regional spread of antimicrobial resistance. We prospectively studied the prevalence and molecular characteristics of clinically important Gram-negative bacilli, including Escherichia coli, Klebsiella pneumoniae, Acinetobacter baumannii complex (ABC), and Pseudomonas aeruginosa, from blood, respiratory tract, urine, and sterile sites at 47 hospitals across Thailand. Among 187,619 isolates, 93,810 isolates (50.0%) were critically drug resistant, of which 12,915 isolates (13.8%) were randomly selected for molecular characterization. E. coli was most commonly isolated from all specimens, except the respiratory tract, in which ABC was predominant. Prevalence of extended-spectrum cephalosporin resistance (ESCR) was higher in E. coli (42.5%) than K. pneumoniae (32.0%), but carbapenem-resistant (CR)-K. pneumoniae (17.2%) was 4.5-fold higher than CR-E. coli (3.8%). The majority of ESCR/CR-E. coli and K. pneumoniae isolates carried blaCTX-M (64.6% to 82.1%). blaNDM and blaOXA-48-like were the most prevalent carbapenemase genes in CR-E. coli/CR-K. pneumoniae (74.9%/52.9% and 22.4%/54.1%, respectively). In addition, 12.9%/23.0% of CR-E. coli/CR-K. pneumoniae cocarried blaNDM and blaOXA-48-like. Among ABC isolates, 41.9% were extensively drug resistant (XDR) and 35.7% were multidrug resistant (MDR), while P. aeruginosa showed XDR/MDR at 6.3%/16.5%. A. baumannii was the most common species among ABC isolates. The major carbapenemase gene in MDR-A. baumannii/XDR-A. baumannii was blaOXA-23-like (85.8%/93.0%), which had much higher rates than other ABC species. blaIMP, blaVIM, blaOXA-40-like, and blaOXA-58-like were also detected in ABC at lower rates. The most common carbapenemase gene in MDR/XDR-P. aeruginosa was blaIMP (29.0%/30.6%), followed by blaVIM (9.5%/25.3%). The findings reiterate an alarming situation of drug resistance that requires serious control measures.

C ritically drug-resistant Gram-negative bacilli (CDRGNB) constitute a significant cause of high morbidity and mortality worldwide. The most common CDRGNB in clinical practice include Escherichia coli, Klebsiella pneumoniae, Acinetobacter baumannii complex (ABC), and Pseudomonas aeruginosa. Among E. coli and K. pneumoniae, serious concerns are specifically for extended-spectrum cephalosporin-resistant (ESCR) and carbapenemresistant (CR) isolates. The most important species of ABC is A. baumannii. Some isolates of ABC and P. aeruginosa demonstrate high resistance to several antimicrobial classes, defined as multidrug-resistant (MDR) and extensively drug-resistant (XDR) bacteria (1). Infections due to these CDRGNB likely contribute to significant long-term stays and high mortality rates (2). The World Health Organization announced that these bacteria were critical priorities for research and new antimicrobial agent development (3). A hospital surveillance system is a proactive approach to taking action on antimicrobial resistance and nosocomial infection. At the international level, surveillance data have been employed to monitor the emergence of resistant bacteria and prevent global spread.
Due to antimicrobial resistance, the mortality rate has strikingly exceeded 54 per 100,000 population in Thailand (4). The availability of over-the-counter drugs and overprescribing in hospitals of many antibacterial agents have provoked the situation. The survey revealed that 29.7% of Thai people were self-medicated with antibacterial agents, and most of them had symptoms resembling viral infection (5). The significant correlation between carbapenem consumption and the increase of CR-E. coli and CR-K. pneumoniae has also been demonstrated (6). While CDRGNB infection is on the rise in most Southeast Asian countries, large-scale surveillance of clinically significant isolates with molecular analysis remains very limited (7). This study aimed to conduct a nationwide prospective survey of CDRGNB prevalence in Thailand and investigate the molecular characteristics of beta-lactamase genes. CDRGNB isolates were classified into critically drug-resistant (CDR) groups, including CDR-E. coli (ESCR-E. coli and CR-E. coli), CDR-K. pneumoniae (ESCR-K. pneumoniae and CR-K. pneumoniae), CDR-ABC (MDR-ABC and XDR-ABC), and CDR-P. aeruginosa (MDR-P. aeruginosa and XDR-P. aeruginosa). The project was a part of the Research University Network Thailand's collaborative effort in partnership with government, military, and private hospitals to address the threatening health issues.

RESULTS
Prevalence of CDRGNB from clinical specimens. There were 47 hospitals across all Thailand regions that participated in this study, including university, government, military, and private hospitals of various sizes and types of services, as shown in Table 1. A total of 187,619 isolates of four clinically significant Gram-negative bacteria (E. coli, K. pneumoniae, ABC, and P. aeruginosa) from blood, respiratory tract, urine, and sterile site samples were reported during the study period. ABC isolates were collected as a complex according to biochemical-based identification commonly used in hospital laboratories with limited species differentiation. These Gram-negative bacteria were mainly isolated from the respiratory tract (48.7%), followed by urine (32.2%), blood (13.7%), and sterile sites (5.4%). E. coli was the most predominant organism, followed by K. pneumoniae, ABC, and P. aeruginosa (Table 2). E. coli was most commonly isolated from blood, urine, and sterile sites, while ABC was most commonly isolated from the respiratory tract. K. pneumoniae was the second most common pathogen from all specimen types. Isolates in eight CDR groups comprised half of the study isolates, indicating the overall prevalence of CDRGNB at 50% from clinical specimens (Table 3). CDRGNB had high rates in all specimens, especially for the respiratory tract, urine, and sterile site samples, which were shown in over 50% of isolates. ESCR-E. coli was the most prevalent CDRGNB in overall specimens and was found at the highest rates from urine and blood, respectively, while ESCR-K. pneumoniae was most detected from the respiratory tract samples. XDR-ABC had a higher ratio over MDR-ABC and were most predominant in respiratory tract samples, followed by sterile sites. P. aeruginosa had less prevalence of MDR and XDR isolates than ABC and was also commonly found in the respiratory tract samples. The distribution of each CDR group in clinical specimens is shown in Fig. 1. Overall, the prevalence of ESCR-K. pneumoniae was lower than ESCR-E. coli, but the CR-K. pneumoniae rate was 4.5-fold higher than that for CR-E. coli, resulting in a CDR-K. pneumoniae rate (49.2%) slightly higher than the CDR-E. coli rate (46.3%). A combination of ESCR-E. coli and CR-E. coli constituted more than half of E. coli isolates from the respiratory tract (54.7%) and sterile site (51.6%) samples and at high rates in urine (47.4%) and blood (38.0%) samples. ESCR-K. pneumoniae and CR-K. pneumoniae were detected in more than half of K. pneumoniae isolates from urine (59.6%) and sterile site (53.0%) samples and also at high rates in the respiratory tract (47.0%) and blood (37.1%) samples. CR-K. pneumoniae was at a higher rate than CR-E. coli in all specimen types.  CDR-ABC were predominant among ABC isolates in all specimens, ranging from 66.4% to 78.7%. XDR-ABC had the highest rate in respiratory tract samples and higher rates than MDR-ABC in all specimen types except urine. MDR-P. aeruginosa and XDR-P. aeruginosa were less prevalent overall than other CDRGNB. CDR-P. aeruginosa was exceptionally high in urine samples, with a larger ratio of XDR-P. aeruginosa than in other specimens. Molecular characteristics of resistance genes among CDRGNB. A total of 12,915 isolates, or approximately 13.8% of CDRGNB isolates, were randomly selected for resistance gene detection. The results of bla genes detected among these isolates are shown in Table  4. bla CTX-M was highly prevalent among CDR-E. coli (65.9%) and CDR-K. pneumoniae (80.6%) at nearly similar rates between ESCR and CR groups of each organism. Among CR-E. coli isolates, bla NDM was the predominant carbapenemase gene, followed by bla OXA-48-like . However, both bla NDM and bla OXA-48-like were common among CR-K. pneumoniae, with a slightly higher prevalence for bla OXA-48-like ; 12.9% and 23.0% of CR-E. coli and CR-K. pneumoniae, respectively, carried both bla NDM and bla OXA-48 . Lesser numbers of CR-E. coli and CR-K. pneumoniae carried bla VIM , bla IMP , and bla KPC , which were slightly more common in CR-E. coli than in CR-K. pneumoniae. Based on the detection of bla OXA-51-like , 86.5% and 89.5% of ABC isolates were determined to be A. baumannii, and the rest were other species of ABC. bla OXA-23-like was the most prevalent carbapenemase gene in both MDR-A. baumannii and XDR-A. baumannii isolates and was shown at a much higher rate than in other ABC species. Metallo-b-lactamase (MBL) genes (bla IMP and bla VIM ) and other OXA genes (bla OXA-40-like and bla OXA-58-like ) were also found with much less prevalence. Notably, bla IMP and bla OXA-58-like were more common among other ABC species than in A. baumannii. In addition to bla OXA-51-like , 67 (5.8%) and 38 (2.8%) isolates of MDR-A. baumannii and XDR-A. baumannii, respectively, carried both bla OXA-23-like and bla OXA-58-like . Three MDR-A. baumannii isolates (0.3%), but not XDR-A. baumannii, had four bla OXA genes (bla OXA-23-like , bla OXA-40-like , bla OXA-51-like , and bla OXA-58-like ). However, all MDR-ABC and XDR-ABC isolates with an MBL gene carried either bla IMP or bla VIM , i.e., none of them had both genes. bla IMP was prominent at a similar rate in both MDR-P. aeruginosa and XDR-P. aeruginosa, followed by bla VIM , which was more pronounced in XDR-P. aeruginosa than in MDR-P. aeruginosa. Eight MDR-P. aeruginosa (0.9%) and 21 XDR-P. aeruginosa (3.9%) isolates had both bla IMP and bla VIM .
Antimicrobial susceptibility of CDRGNB. Table 5 shows susceptibility rates of CDRGNB groups to alternative antimicrobial agents. XDR-ABC and XDR-P. aeruginosa are not shown due to their resistance to all tested drugs. Carbapenems and amikacin were highly effective against ESCR-E. coli and ESCR-K. pneumoniae, and amikacin remained moderate against CR-E. coli and CR-K. pneumoniae. Piperacillin-tazobactam had good activity for ESCR-E. coli but not ESCR-K. pneumoniae. ESCR-E. coli and ESCR-K. pneumoniae isolates were moderately susceptible to netilmicin and cefoxitin, but most were resistant to ciprofloxacin, trimethoprim-sulfamethoxazole, and tetracycline. CR-E. coli and CR-K. pneumoniae were resistant to most drugs, except aminoglycosides may have moderate activity against them. A. baumannii isolates were slightly less susceptible than other ABC species, except for trimethoprim-sulfamethoxazole. All XDR-A. baumannii and 98.3% of MDR-A. baumannii Species of Acinetobacter baumannii complex other than A. baumannii based on the absence of bla OXA-51 .
isolates were resistant to at least a carbapenem, resulting in the rate of carbapenem resistance among A. baumannii of 77.0%. Among P. aeruginosa, 96.8% of MDR-P. aeruginosa and all XDR-P. aeruginosa isolates were resistant to at least a carbapenem, given P. aeruginosa's carbapenem resistance rate of 22.3%. Amikacin showed moderate activity against MDR-ABC and MDR-P. aeruginosa. Trimethoprim-sulfamethoxazole and netilmicin had partial action against MDR-A. baumannii and MDR-P. aeruginosa, respectively.

DISCUSSION
Infection due to CDRGNB has raised a serious concern worldwide and requires urgent solutions to reduce the wide spread of these organisms. Systematic surveillance of drug-resistant bacteria plays a vital role in monitoring and controlling the situation (8). This study conducted large-scale surveillance of four clinically significant Gram-negative bacteria in eight CDR groups. CDR-E. coli and CDR-K. pneumoniae were common causes of various infections, especially in the urinary tract, intra-abdominal area, and bloodstream, with high rates of extended-spectrum b-lactamase production and low susceptibility to fluoroquinolones, indicating multiple resistance mechanisms (9)(10)(11). The prevalence of CR-E. coli and CR-K. pneumoniae in Thailand has rapidly increased in the past decade. Studies in the late 2000s and early 2010s showed that E. coli and K. pneumoniae isolates from Thai patients were less than 1% and 5% resistant to carbapenems, respectively (12,13). In this study, the CR-E. coli prevalence has risen to 3.8% and varied in different specimen types, ranging from 1.9% to 6.0%. An average proportion of CR-K. pneumoniae was 4.5-fold (ranging from 2.4-to 7.7-fold) higher than CR-E. coli. Therefore, within less than a decade, CR-E. coli and CR-K. pneumoniae prevalences in Thailand have risen more than three times. Inappropriate consumption of carbapenems was shown to contribute to an increased prevalence of CDRGNB, including CR-E. coli and CR-K. pneumoniae (6,14). The increasing trends of CR-E. coli and CR-K. pneumoniae have also been reported from many countries in Southeast Asia, in which CR-K. pneumoniae often showed higher prevalence (15). Malaysia, Myanmar, Indonesia, and Vietnam were countries with a high prevalence of CR-E. coli (1.9% to 10.0%) and CR-K. pneumoniae (7.9% to 9.5%). Our study was among the most extensive and recent studies in Thailand. It revealed significantly higher CR-E. coli and CR-K. pneumoniae rates than what was shown in previous reports from Thailand and neighboring countries, raising a critical need for infection control actions. Aminoglycosides seemed to be the only class that is active for CR isolates but with modest activities. With a rise of CR isolates among Enterobacteriaceae, polymyxins, tigecycline, and fosfomycin were often choices for the alternative or last-line drug. However, these agents still have limitations due to toxicity or availability in different regions (16).
The most common carbapenemase genes in CR-E. coli and CR-K. pneumoniae detected in this study were bla NDM and bla OXA-48-like. In contrast to North America,  Latin America, some European countries, and China, where bla KPC appears commonly, bla KPC has rarely been detected in Thailand and Southeast Asia (17)(18)(19). bla NDM has been endemic to South Asia and later widespread in Southeast Asia (15,17). A study in Myanmar also revealed that all carbapenemase gene-carrying Enterobacteriaceae clinical isolates had bla NDM , of which bla NDM-5 was the most predominant (20). bla NDM was first reported in Thailand in 2012 (21), and this study, conducted approximately 6 years later, showed a very high prevalence of bla NDM in both CR-E. coli and CR-K. pneumoniae, suggesting a successful spreading of the gene in this country. Various mobile genetic elements mediate its rapid dissemination, such as transposons and conjugative plasmids. bla NDM was also found in CR-E. coli and CR-K. pneumoniae isolates from the environment, likely related to contracting these organisms through traveling in the region of endemicity (18). While bla IMP and bla VIM were found at a low prevalence, bla OXA-48-like , a class D carbapenemase, was strikingly high among CR-E. coli and CR-K. pneumoniae. Originating in and primarily endemic to European and Middle East countries, bla OXA-48-like and its derivatives have rapidly spread to Africa, South America, and Asia (17,18). OXA-48 and OXA-48-like carbapenemases display unique characteristics of a modest hydrolytic activity for penicillins and carbapenems but not for cephalosporins, making it problematic for in vitro detection (22). Although OXA-48 usually confers weak resistance to carbapenems in Enterobacteriaceae, it may be cocarried with an MBL gene, e.g., bla NDM , as shown in this study, resulting in highlevel carbapenem resistance (23). CDR-ABC was shown to be the main portion of ABC isolates in all specimen types. The most common gene that mediates carbapenem resistance in ABC is bla OXA , which produces a class D OXA b-lactamase. The majority species of ABC was A. baumannii, based on the presence of intrinsic bla OXA-51-like (24). Although bla OXA-51-like may not be the most accurate marker for species confirmation, it is practical to screen for A. baumannii in an extensive survey. Besides chromosomally encoded bla OXA-51-like , we investigated three mainly plasmid-borne bla OXA carbapenemase-related genes, bla OXA-23-like , bla OXA-40-like (formerly known as bla OXA-24 ), and bla OXA-58-like , commonly present among ABC in Asian countries (25)(26)(27)(28). All four OXA carbapenemases and their derivatives are transferable by a mobile genetic element. Either OXA-23 or OXA-51 is sufficient to confer carbapenem resistance in ABC, but generally, other carbapenemases and other mechanisms, e.g., efflux pump, often collectively coproduce high-level resistance. We showed that bla OXA-23-like was most common among MDR-A. baumannii and XDR-A. baumannii but was less prevalent among other species of ABC. bla OXA-40-like and bla OXA-58-like are less common and usually confer low activity against carbapenems. This study reported that ABC isolates carry up to four bla OXA genes, making them highly resistant to carbapenems. Although other resistance mechanisms were not investigated, it is likely that a combination of resistance determinants attributed to the extreme resistance phenotype in ABC.
The worldwide SENTRY surveillance program demonstrated that CDR-P. aeruginosa had a declining rate and was most susceptible to colistin, followed by amikacin (29). A previous study in Thailand among hospitalized patients during 2014 showed a higher rate of XDR-P. aeruginosa but a lower rate of MDR-P. aeruginosa compared to this study, suggesting a decreasing trend of XDR-P. aeruginosa but not MDR-P. aeruginosa (30). Amikacin was shown in the SENTRY report and this study to be most active against MDR-P. aeruginosa, yet with a modest susceptibility rate. Most CDR-P. aeruginosa isolates were resistant to carbapenem. The carbapenem resistance rate among P. aeruginosa isolates tends to be stable or decreasing over time in many regions (29,31). P. aeruginosa often intrinsically harbors several b-lactam resistance mechanisms, including membrane modification, efflux pump, and AmpC enzyme hyperproduction (32). MBL enzymes were sporadically studied among carbapenem-resistant P. aeruginosa, of which IMP and VIM were the most common carbapenemases (33,34). We reported that bla IMP was detected at approximately similar rates in both MDR-P. aeruginosa and XDR-P. aeruginosa, while bla VIM was detected 2.7-fold more often in XDR-P.
aeruginosa than MDR-P. aeruginosa. In addition, XDR-P. aeruginosa isolates that carried both bla IMP and bla VIM genes were 4.3-fold more common than MDR-P. aeruginosa. Therefore, bla VIM could contribute to extended resistance to multiple drugs among P. aeruginosa isolates, especially when cocarrying bla IMP . Despite a lesser proportion of resistant isolates than other CDRGNB, both MDR-P. aeruginosa and XDR-P. aeruginosa remain a severe threat because of their high resistance to commonly used antipseudomonal drugs and their critical role in nosocomial infection. Carbapenem resistance in P. aeruginosa is often mediated by MBL genes acquirable through a mobile plasmid and is significantly associated with high mortality (34).
This study has limitations in exploring specific and additional resistance genes and other mechanisms that could contribute to resistance due to a large number of study isolates. In addition, susceptibility to some alternative drugs, e.g., colistin, was not tested because the recommended testing methods were not available at all participating hospitals. Although A. baumannii is the most common species of ABC, this study used bla OXA-51-like as a marker for A. baumannii but did not specifically differentiate all species. Therefore, phenotypic and genotypic resistance profiles of each species of ABC were not analyzed separately. These limitations would be further studied in selected groups to understand the evolving resistance among CDRGNB and their possible relatedness of resistance clones in various geographic regions. We emphasize in this report a high prevalence of CDRGNB in Thailand in eight CDR groups, raising concern for a severe threat in the Asia-Pacific region. These CDRGNB isolates are usually encountered in various infections, mainly associated with hospitalized patients. The threat of bacterial resistance should be critically addressed in Thailand and in the region for the necessity of appropriate antimicrobial prescription in clinical practice. In addition, the public policy regarding the self-consumption of antibacterial agents should be emphasized to reduce the risk of acquiring drugresistant infections. Continuously prospective surveillance is an active measure to monitor and control the spread of resistance and to support the importance of antimicrobial stewardship programs. Further studies are recommended to investigate the impact of CDRGNB on various clinical aspects, such as community-acquired infections, nosocomial infections, and treatment outcomes. Multiple organizations and professionals are needed to build a scholarly society to prevent spreading resistance.

MATERIALS AND METHODS
Study design and study sites. Forty-seven hospitals participated as study sites were located across all regions of Thailand (Fig. 2). The Faculty of Medicine, Siriraj Hospital, Mahidol University (Bangkok), served as the study center, and an additional six major university hospitals served as the study hubs. The other 40 hospitals were designated satellite hospitals. Isolates and data from four clinically significant Gram-negative bacteria (E. coli, K. pneumoniae, ABC, and P. aeruginosa) were collected from blood, respiratory, urine, and sterile site (including cerebrospinal fluid, pleural fluid, abdominal fluid, joint fluid, and other fluids from sterile body parts) samples from patients between October 2017 and January 2019. Isolate collection was inclusive for all patient groups at each hospital without stratification according to disease onset or illness history. Isolates were evaluated for prevalence, antimicrobial susceptibility, and molecular characteristics. Each hospital performed bacterial identification and antimicrobial susceptibility testing, including quality control testing, at its laboratory based on standardized microbiological procedures, and additional external quality assessment was provided. Repeated isolates from the same patient were excluded. Random CDRGNB isolates were sent to a study hub or selected study centers for molecular study. The molecular study included the first three isolates of the month, if available, for each CDRGNB group in every clinical specimen from all hospitals for up to 12 months during the collection period. This procedure ensured that isolates were enrolled without selection bias of patient demography, geographic area, or time of collection. This study was approved by the Institution Review Board or Ethical Committee of all participating hospitals.
Resistance definitions. CDRGNB isolates were divided into eight CDR groups, including ESCR-E. coli, CR-E. coli, ESCR-K. pneumoniae, CR-K. pneumoniae, MDR-ABC, XDR-ABC, MDR-P. aeruginosa, and XDR-P. aeruginosa. As defined by the Centers for Disease Control and Prevention, ESCR isolates of E. coli and K. pneumoniae were resistant to at least one of these cephalosporin agents: ceftriaxone, cefotaxime, ceftazidime, and cefepime. CR isolates were resistant to at least one of these carbapenem agents: imipenem, meropenem, doripenem, and ertapenem (35). A. baumannii and P. aeruginosa were classified according to their resistance as MDR and XDR. MDR isolates were nonsusceptible to at least one agent in at least three antibacterial agent categories, and XDR isolates were nonsusceptible to at least one agent in all but two or fewer antimicrobial categories (1).
Molecular characterization of bla genes. Eight groups of CDRGNB from various specimens were collected for further molecular characterization. Each CDRGNB group was randomized monthly to ensure suitable distribution and that isolates collected each month during the study period were included. Bacterial DNA was extracted from colonies using the boiling method. PCR was used for the detection of beta-lactamase (bla) genes, including bla CTX-M , bla KPC , bla NDM , bla IMP , bla VIM , bla OXA-23-like , bla OXA-40-like , bla OXA-48-like , bla OXA-51-like , and bla OXA-58-like . The multiplex PCR primer sets (M1, M2, M3, and M4) and simplex PCR primer (S) were used to detect various specific genes, as detailed in Table 6. According to the corresponding reference, PCR was performed under the relevant conditions except where indicated, and the amplified targets were verified by DNA sequencing if necessary.
Antimicrobial susceptibility testing. Susceptibility testing was performed by using a standard disk diffusion assay or an automated system, either Sensititre (Thermo Fisher Scientific, OH, USA), Vitek2 (bioMérieux, Marcy-L'Etoile, France), or Phoenix (Becton, Dickinson, MD, USA), available at each study site. Results were interpreted according to the Clinical and Laboratory Standards Institute (CLSI) guidelines (36). All bacterial identification and antimicrobial susceptibility test results were submitted electronically and verified according to the resistance definitions.