On-Site Genomic Epidemiological Analysis of Antimicrobial-Resistant Bacteria in Cambodia With Portable Laboratory Equipment

The rapid emergence of carbapenemase-producing gram-negative bacteria (CPGNB) is a global threat due to the high mortality of infection and limited treatment options. Although there have been many reports of CPGNB isolated from Southeast Asian countries, to date there has been no genetic analysis of CPGNB isolated from Cambodia. Sequence-based molecular epidemiological analysis enables a better understanding of the genotypic characteristics and epidemiological significance of antimicrobial-resistant (AMR) bacteria in each country, and allows countries to enact measures related to AMR issues. In this study, we performed on-site genomic epidemiological analysis of CPGNB isolated in Cambodia using a portable laboratory equipment called Bento Lab, which combines a PCR thermal cycler, microcentrifuge, gel electrophoresis apparatus, and LED transilluminator, along with the MinION nanopore sequencer. PCR targeting of major carbapenemase genes using Bento Lab revealed that two Escherichia coli isolates and one Acinetobacter baumannii isolate harbored carbapenemase genes: blaNDM, blaOXA–48, and blaOXA–23, respectively. The results of phenotypic diagnostic tests for CPGNB, such as the carbapenem inactivation method and double-disk diffusion test using a specific inhibitor of metallo-β-lactamases, were consistent with their AMR genotypes. Whole-genome sequencing analysis using MinION revealed that blaNDM–5 gene was carried on a 93.9-kb plasmid with IncFIA/IncFIB/IncFII/IncQ1 replicons, and blaOXA–181 gene was carried on a 51.5-kb plasmid with the IncX3 replicon in E. coli isolates. blaOXA–23 was encoded in two locations on the chromosome of A. baumannii. Plasmids carrying blaNDM–5 or blaOXA–181 in E. coli were highly structurally identical to plasmids prevalent in Enterobacterales in China and other countries, suggesting that they disseminated from a common evolutionary origin. Our findings demonstrate the potential impact of portable laboratory equipment on AMR bacteria research in hospitals and research centers with limited research facilities, and provide the first glimpse into the genomic epidemiology of CPGNB in Cambodia.


INTRODUCTION
Antimicrobial-resistant (AMR) bacteria have emerged and spread all over the world. Among antimicrobials, carbapenems are one of the most reliable last-resort antimicrobials for infections caused by AMR gram-negative bacteria. One of the mechanisms of AMR is drug inactivation mediated by acquired enzymes, such as β-lactamases (Santajit and Indrawattana, 2016). Among β-lactamases, extended-spectrum β-lactamases (ESBLs) and carbapenem-hydrolyzing β-lactamase (carbapenemases) are clinically important, as ESBLs hydrolyze a broad range of β-lactams, including cephalosporins, and carbapenemases hydrolyze most β-lactams, including carbapenems. β-lactamases are classified using the Ambler scheme as follows. Ambler class A includes ESBLs, such as CTX-M, as well as carbapenemases, such as KPC; class B includes metallo-β-lactamases (MBLs), such as NDM, IMP, and VIM; class C includes AmpC β-lactamases; and class D includes carbapenem-hydrolyzing oxacillinases, of which OXA-48 is prevalent in Enterobacterales, and OXA-23, OXA-24, and OXA-58 are prevalent in Acinetobacter spp. ESBL and carbapenemase genes are predominantly encoded on conjugative plasmids and have been transferred among Enterobacterales and other gram-negative bacteria (Tzouvelekis et al., 2012). Moreover, carbapenemase-producing gram-negative bacteria (CPGNB) harboring carbapenemase genes often co-harbor clinically relevant antimicrobial resistance genes, such as aminoglycoside and fluoroquinolone resistance genes (Wu et al., 2007;Chmelnitsky et al., 2008;Poirel et al., 2011a). There is great concern about the global spread of plasmids that carry multiple AMR genes that can be transferred between homogeneous and heterogeneous species.
Although CPGNB has been detected in large numbers in Southeast Asia, the publicly available information on CPGNB is limited to a small number of countries (Malchione et al., 2019). In 2015, the World Health Organization (WHO) adopted a global action plan on AMR and launched the Global Antimicrobial Resistance Surveillance System (GLASS), the first global collaborative report to standardize AMR surveillance (WHO, 2015;Tornimbene et al., 2018). Cambodia's Laboratory Information System (CamLIS) was developed by the Ministry of Health in Cambodia with the support of WHO starting in 2011 (WHO, 2019). As of 2018, 35 national, provincial, and referral laboratories contribute to CamLIS. Cambodia has started to prepare the GLASS report for 2020, and the actual status of AMR bacteria in the country will be revealed in the near future. To date, however, only a few studies have examined AMR bacteria clinically isolated in Cambodia, although other Southeast Asian countries are increasingly reporting cases of AMR bacteria (Suwantarat and Carroll, 2016;Gandra et al., 2020). To date, there has been no report on the genomic epidemiology of CPGNB in Cambodia.
In this study, we introduced portable laboratory equipment, Bento Lab and MinION, for on-site genomic epidemiological analysis of CPGNB in Cambodia. Bento Lab (Bento Bioworks Ltd., United Kingdom) is a DNA analysis device small enough to fit in a laptop-sized bag. It contains a PCR thermal cycler, microcentrifuge, and gel electrophoresis apparatus with LED transilluminator, and has sufficient functionality for laboratory work (Bento Lab, 2016). The MinION nanopore sequencer (Oxford Nanopore Technologies, United Kingdom) is a portable long-read sequencer with the size of a large USB memory stick. MinION was utilized for on-site genomic epidemiological analysis of the Ebola virus outbreak in West Africa in 2016 (Quick et al., 2016) and the Zika virus outbreak in the Americas in 2017 (Faria et al., 2017). Because carbapenemase genes are mostly carried on plasmids, long-read sequencing is useful for assembling whole plasmid sequences and tracking horizontal transfer of AMR plasmids in hospitals, as well as local and global communities (Conlan et al., 2014).
We organized an international collaborative research group with researchers from Japan, United Kingdom, and Cambodia, and successfully performed on-site genomic epidemiological analysis of CPGNB clinical isolates in Cambodia, where access to laboratory equipment is limited. Our findings demonstrate the potential impact of portable laboratory equipment on AMR bacteria research and provide the first glimpse into the genomic epidemiology of CPGNB in Cambodia.

Ethics
Written informed consent was obtained from the individuals for the publication of any potentially identifiable images included in this article.

Subjects and Specimen Collection
The outpatient clinic of National Institute of Public Health (NIPH) in Phnom Penh, Cambodia has around 10 patients in a day and 456 bacterial strains were isolated from patient specimens, such as sputum, stool, urine, pus, body fluid, and cerebral spinal fluid, in 2017. Ethical approval of this study "Genomic epidemiological analysis of AMR bacterial isolates in Cambodia" was obtained from National Ethics Committee for Health Research (NECHR), Cambodia (approval no.: 178NECHR). Two carbapenemase-producing isolates NIPH17_0020 and NIPH17_0036 of Escherichia coli and one carbapenemase-producing isolate NIPH17_0019 of Acinetobacter baumannii analyzed in this study were obtained from abdominal pus, urine, and blood of patients, respectively, at NIPH, Cambodia in 2017.

PCR, Whole-Genome Sequencing, and Bioinformatics Analysis
Draft genome analysis of carbapenemase-producing isolates of E. coli (NIPH17_0020 and NIPH17_0036) and A. baumannii (NIPH17_0019) using Bento Lab and MinION was performed in NIPH, Cambodia in July, 2017. Bacterial genomic DNAs (gDNAs) were extracted using the MagAttract HMW DNA Kit (Qiagen) and quantified using a Qubit 2.0 fluorometer (Thermo Fisher Scientific). Library preparation for MinION sequencing using Rapid Sequencing Kits (SQK-RAD002 and SQK-RAD003) (Oxford Nanopore Technologies) were performed using the prototype model of Bento Lab (Bento Bioworks Ltd.) consisting of a thermal cycler, microcentrifuge, and gel electrophoresis apparatus with LED transilluminator (Bento Lab, 2016). The prototype is configured slightly differently from the current commercial versions, but there is no significant difference in performance (Supplementary Figure 1).
The extracted bacterial gDNAs were subsequently resequenced on a Illumina system in National Institute of Infectious Diseases, Japan for further error correction. Library for Illumina sequencing (insert size of 500-900 bp) was prepared using Nextera XT DNA Library Prep Kit (Illumina) and paired-end sequencing (2 bp× 150 bp) was performed using MiniSeq (Illumina). Illumina paired-end reads were mapped onto the on-site assembly sequences, and sequencing errors were corrected by extracting the consensus of the mapped reads five times using CLC Genomics Workbench v12.0 (Qiagen) with default parameters.
Genome sequences were annotated using the DFAST server 1 . Sequence type (ST), plasmid replicon type, AMR genes, and virulence genes were detected using MLST v2.0, PlasmidFinder v2.1, ResFinder v4.1, and VirulenceFinder v2.0, respectively, using the CGE server 2 with default parameters. Type IV secretion system (T4SS)-associated genes involved in conjugation were detected using TXSScan 3 with default parameters. Mobile gene elements (MGEs) were identified manually from CDS annotations and basically analyzed by comparing the sequences analyzed in previous studies. Linear comparisons of sequences carrying carbapenemase genes were performed using BLAST with default settings (the nucleotide collection database and the megablast program) and visualized using Easyfig v2.2.2 4 . The annotated bacterial circular chromosomes were visualized using the CGView Server 5 .

On-Site Genomic Epidemiological Analysis of Carbapenemase-Producing Gram-Negative Bacteria Isolated in Cambodia
In July 2017, we stayed for 5 days at the National Institute of Public Health (NIPH) in Phnom Penh, Cambodia. There, we set up potable laboratory equipment, including Bento Lab and MinION, in the Bacteriology laboratory, which has limited research facilities and no PCR machine (Figure 1). On the first and second days, we performed the carbapenem inactivation method (CIM) test, double-disk diffusion tests (DDDTs), minimum inhibitory concentrations (MICs) measurement, FIGURE 1 | On-site genomic epidemiological analysis of AMR bacteria in Cambodia. Bento Lab and MinION were used for genotype analysis and the carbapenem inactivation method (CIM) and double-disk diffusion tests (DDDTs) were used for phenotype analysis of AMR bacteria.
PCR, and MinION sequencing on two carbapenemaseproducing isolates of E. coli (NIPH17_0020 and NIPH17_0036) and one carbapenemase-producing isolate of A. baumannii (NIPH17_0019) (Figure 1) stored in the laboratory prior to this study. On the third and fourth days, we examined diagnostic testing data (Figure 2) and analyzed sequencing data (Figures 3-5). On the final day, we discussed the results of genotype and phenotype analysis with researchers and technicians belonging to the Bacteriology laboratory at NIPH.
To summarize the results, the detected carbapenemase genes had a few-percent mismatch that caused frameshifts of genes relative to their putative reference sequences (Supplementary Figure 2); hence, we avoided performing CDS annotation for the on-site sequences. We detected AMR genes and plasmid replicons in the sequences analyzed on-site by sequencebased detection (Figures 3-5 and Supplementary Table 1). AMR genes, such as bla TEM−1B , aadA2, aph(3 )-lb, aac(3)lld, and aph(6)-ld, and plasmid replicons, including IncFIA, IncFIB, IncFII, and IncQ1, were detected in provisional bla NDM−5 -carrying pNIPH17_0020_1 in E. coli NIPH17_0020 (Figure 3), and qnrS1 genes and the IncX3 replicon were detected in provisional bla OXA−181 -carrying pNIPH17_0036_1 in E. coli NIPH17_0036 (Figure 4). BLAST searches of E. coli pNIPH17_0020_1 and E. coli pNIPH17_0036_1 revealed that several plasmids from Asian and Western countries were highly identical to those plasmids (Figures 3, 4).

Comparison of Plasmids and Genomic Regions in Carbapenemase-Producing Gram-Negative Bacteria Isolated in Cambodia With Those in Other Countries
After the on-site analysis in Cambodia, we further performed Illumina sequencing of carbapenemase-producing isolates of E. coli (NIPH17_0020 and NIPH17_0036) and A. baumannii (NIPH17_0019), corrected the on-site de novo assembly sequences using Illumina reads, and compared the onsite and error-corrected sequences (Figures 3, 4, 5A, and  Supplementary Figure 3). The on-site sequences of E. coli pNIPH17_0020_1 (provisional 91.6-kb bla NDM−5 -carrying plasmid with IncFIA/FIB/FII/Q1 replicons) and E. coli pNIPH17_0036_1 (provisional 50.2-kb bla OXA−181 -carrying plasmid with IncX3 replicons) were highly identical to their error-corrected sequences (96.48% identity over 100% of the error-corrected 93.9-kb plasmid pNIPH17_0020_1: accession no. LC483178 and 96.70% identity over 100% of the error-corrected plasmid 51.5-kb pNIPH17_0036_1: accession no. LC483179, respectively) (Figures 3, 4). Moreover, the on-site sequences of bla OXA−23 -containing chromosomal regions of A. baumannii NIPH17_0019 were also highly identical to their error-corrected sequences (96.95% identity over 100% of 216,633-233,441 nt and 97.36% identity over 100% of 3,924,674-3,941,483 nt in their error-corrected chromosome of NIPH17_0019: accession no. AP024415) (Supplementary Figures 3A,B). Although there were differences of a few percent between the on-site and error-corrected sequences, the best match types of AMR genes and plasmid replicons detected from the reference libraries were consistent (Figures 3, 4, 5A).
Acinetobacter baumannii NIPH17_0019 belonged to ST471 according to MLST analysis and harbored the bla OXA−23 genes in two separate regions on the chromosome (accession no. AP024415) (Figure 5A). Both copies of bla OXA−23 were located in Tn2006 in AbaR4. AbaR is the resistance island containing AMR genes in A. baumannii (Bi et al., 2019). Comparison of the genetic environment around AbaR4 was performed between A. baumannii NIPH17_0019 and AbaR4-harboring A. baumannii D36 (accession no. JN107991), which was isolated from a human in 2008 in Australia (Figure 5B). The results revealed that two AbaR4 islands in A. baumannii NIPH17_0019 were highly identical to that of A. baumannii D36 (99.84-99.88% identity over 100% of the sequence) ( Figure 5B). Interestingly, AbaR4 was integrated into the comM gene in A. baumannii D36, whereas AbaR4 was integrated into the comM gene (AbaR4 of the 216,633-233,441 nt region) and mutY genes (for AbaR4 of the 3,924,674-3,941,483 nt region) in A. baumannii NIPH17_0019.

DISCUSSION
In this study, we successfully conducted genomic analysis of carbapenemase-producing gram-negative bacteria (CPGNB) in Cambodia and provided the first glimpse into the genomic epidemiology of CPGNB in that country. The main analysis was performed on-site using portable laboratory equipment, namely, Bento Lab and MinION. The nearly complete genomes of carbapenemase-producing E. coli and A. baumannii isolates, including their plasmids, were determined using the MinION nanopore sequencing data; detection of AMR genes and plasmid replicons, as well as sequence similarity searches in public databases, were performed at a Cambodian laboratory with limited research facilities. Because the nanopore sequencing technology is still evolving, and the accuracy of sequencing is imperfect (the rates of indels and substitutions are a few percent each) (Jain et al., 2016), it was necessary to combine other sequencing technologies, such as Illumina sequencing by synthesis, for further molecular typing analysis of bacteria that require more accurate sequences at the single-nucleotide level.
The cost of commercial versions of Bento Lab is starting from $1,646.99 6 , whereas MinION is free of charge because it is basically a rental from the company 7 . In this study, we did not use barcodes for MinION library preparation and performed MinION analysis using one flow cell per sample to ensure reliable data acquisition. The cost of the flow cell and library preparation reagent was $599 (for 6 reactions) and $900, respectively, resulting in MinION analysis cost of $999 per sample. If barcodes are used for library preparation, the unit cost of MinION analysis can be dramatically reduced. We usually analyze about six barcoded samples using one flow cell for bacterial species with genome sizes of around 5 Mb. If six samples are included in one analysis, the unit cost is going down to $171 8 .
The carbapenem inactivation method (CIM) and doubledisk diffusion test (DDDT) are simple, inexpensive, and useful, especially in developing countries where materials and facilities are limited. In this study, the results of both tests were reasonable, as validated by subsequent genetic analysis (Figure 2). The CIM test is routinely performed at NIPH, Cambodia, and NIPH had detected and stored carbapenemase-producing bacterial isolates prior to this study. The DDDTs with SMA and CVA disks clearly detected production of MBL in E. coli NIPH17_0020 and ESBL in E. coli NIPH17_0036 (Figure 2A). CIM and DDDT are important for screening for CPGNB because MICs of carbapenems against CPGNB are not always high, as is the case for bacteria with carbapenem-hydrolyzing oxacillinase genes, such as bla OXA−48 or its variants. The MICs of carbapenems against E. coli NIPH17_0036, which harbored bla OXA−181 on its plasmid (Figure 4), were relatively low, whereas CIM yielded a positive result ( Figure 2C). OXA-48 is capable of weakly hydrolyzing carbapenems while maintaining its activity against broad-spectrum cephalosporins. The bla OXA−181 gene is a variant of bla OXA−48 , and the hydrolytic activity of OXA-181 toward β-lactams is similar to that of OXA-48 (Castanheira et al., 2011).
We sequenced carbapenemase-producing E. coli and A. baumannii isolates on-site using MinION and Bento Lab, and revealed that E. coli NIPH17_0020 and E. coli NIPH17_0036 harbored bla NDM−5 and bla OXA−181 on plasmids pNIPH17_0020_1 and pNIPH17_0036_1, respectively (Figures 3, 4) and that A. baumannii NIPH17_0019 harbored two copies of bla OXA−23 on its chromosome (Figure 5). Although the MinION control software MinKNOW requires a constant internet connection, the company provided us with an offlinecapable version of MinKNOW. For de novo assembly using nanopore long-read data, we used the Miniasm software, which is fast and computationally inexpensive (Li, 2016). Because Miniasm assembles without error correction, MinION reads were error-corrected using the CLC Genomics Workbench pipeline for long-read sequencers prior to de novo assembly. The resultant assembly sequences obtained from on-site analysis still contained a few percent of errors that are responsible for gene frameshifts. However, the on-site de novo assembly sequences were sufficient for subsequent molecular epidemiological analysis (e.g., detection of AMR genes and genomic locations where the genes are encoded) (Figures 3-5 and Supplementary  Figures 2, 3), and the error-corrected sequences using Illumina sequencing were subsequently used to confirm the results of the on-site analysis (Figures 3-5).
Escherichia coli isolates NIPH17_0020 and NIPH17_0036 belonged to ST410 according to MLST analysis. ST410 is a high-risk clone associated with AMR and recently emerged in among humans and the environment in Southeast Asia (Nadimpalli et al., 2019). In this study, both E. coli isolates harbored carbapenemase genes, bla NDM−5 and bla OXA−181 , respectively, on their plasmids, and commonly harbored putative virulence genes, including gad (a glutamate decarboxylase gene), lpfA (a long polar fimbriae gene), and terC (a tellurium ion resistance gene) on their chromosomes (Supplementary Table 1). Furthermore, NIPH17_0020 harbored other virulence genes, fyuA (a siderophore receptor gene) and irp2 (a siderophore gene), and NIPH17_0036 harbored other virulence gene hra (a heat-resistant agglutinin gene) on their chromosomes (Supplementary Table 1).
Escherichia coli pNIPH17_0036_1 [51.5-kb IncX3 plasmid with bla OXA−181 , accession no. LC483179] was structurally nearly identical to E. coli pOXA-181-IHIT35346 from Italy (accession no. KX894452), E. coli pOXA181 from China (accession no. KP400525), and K. variicola pKS22 from Thailand/Vietnam (accession no. KT005457) (Figure 4). Our analysis revealed that bla OXA−181 -carrying IncX3 plasmids widespread among Enterobacterales worldwide were also present in Cambodia. pKS22 was detected in coriander imported from Thailand or Vietnam, and the international fresh vegetable trade is suspected to be a route for the spread of AMR bacteria (Zurfluh et al., 2015). Because Cambodia is geographically and culturally close to Thailand and Vietnam, it is possible for AMR bacteria to be transmitted through foods. However, this study was very small, so further analysis with larger numbers of bacterial isolates in Cambodia according to One Health approaches will be necessary to characterize AMR bacteria in this country.
Acinetobacter baumannii NIPH17_0019 belonging to ST571 harbored two copies of bla OXA−23 in the chromosome ( Figure 5A). According to a previous study of carbapenemresistant A. baumannii (Hamidian and Nigro, 2019), ST2 is the most prevalent genotype and recognized as a highrisk clone associated with AMR, and bla OXA−23 is the most widespread carbapenem-resistance gene in the world. ST571 belongs to clonal complex 2, and ST571 strains harboring bla OXA−23 are widespread in medical settings in Vietnam (Tada et al., 2015). In general, carbapenem-hydrolyzing oxacillinases hydrolyze carbapenems weakly and do not contribute to strong carbapenem resistance on their own. However, elevated expression of oxacillinase genes by the upstream insertion of IS, such as ISAba1 in Acinetobacter spp., which serves as a promoter for the downstream genes, leads to strong resistance Corvec et al., 2007). Two copies of bla OXA−23 in A. baumannii NIPH17_0019 were located downstream of ISAba1 ( Figure 5B). Both bla oxa−23 were present in Tn2006, a common 4.8-kb transposon in Acinetobacter app., with a central segment of 2,445-bp flanked by two reverse-oriented copies of ISAba1 (Nigro and Hall, 2016). Tn2006-containing AbaR4 is frequently inserted within the comM gene in the chromosome of international A. baumannii clones (Hamidian and Hall, 2011;Seputiene et al., 2012;Hamidian et al., 2015), and the comM gene is the mostpreferred hotspot for AbaR insertions (Bi et al., 2019). One of the insertion sites of AbaR4 (3,924,674-3,941,483 nt region) in A. baumannii NIPH17_0019 was not the comM gene but the mutY gene, encoding for an adenine DNA glycosylase ( Figure 5B). According to a previous study (Bi et al., 2019), AbaR insertions at acoA, pho, and uup genes were occasionally observed; however, mutY has not been previously reported as an insertion site.

CONCLUSION
Based on our on-site genomic epidemiological analysis of carbapenemase-producing gram-negative bacteria in Cambodia, we revealed for the first time that plasmids and MGEs carrying clinically relevant carbapenemase genes reported in other countries have also been spreading in Cambodia. Bento Lab and MinION are useful for genomic analysis and surveillance of AMR bacteria in hospitals and research centers with limited facilities.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary Material.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by National Ethics Committee for Health Research (NECHR), Cambodia. The patients/participants provided their written informed consent to participate in this study. Written informed consent was obtained from the individuals for the publication of any potentially identifiable images included in this article.

AUTHOR CONTRIBUTIONS
MS contributed to conceptualization. AH, HY, and MS contributed to methodology. MS contributed to software. AH, HY, HT, KY, and MS contributed to validation. AH and MS contributed to formal analysis, data curation, and visualization. AH, HY, HT, KY, KS, and MS contributed to investigation. PB, BW, VN, VL, MV, VA, and CD contributed to resources. AH, KS, and MS contributed to writing. CD and KS contributed to supervision. CD, KS, and MS contributed to project administration. KS and MS contributed to funding. All authors contributed to the article and approved the submitted version.