In Vitro Activity of Neomycin, Streptomycin, Paromomycin and Apramycin against Carbapenem-Resistant Enterobacteriaceae Clinical Strains

We determined the in vitro susceptibility of four aminoglycosides, which are not of the 4,6-disubstituted deoxystreptamine (DOS) subclass against a collection of carbapenem-resistant Enterobacteriaceae (CRE). CRE clinical strains (n = 134) were collected from multiple hospitals in China and carried blaNDM (blaNDM−1, blaNDM−5 or blaNDM−7; n = 66), blaKPC−2 (n = 62) or blaIMP−4 (n = 7; including one carrying blaNDM−1 and blaIMP−4). MICs of neomycin, paromomycin, streptomycin and apramycin as well as three 4,6-disubstituted DOS aminoglycosides (amikacin, gentamicin and tobramycin) were determined using the broth microdilution with breakpoints defined by the Clinical Laboratory Standards Institute (for amikacin, gentamicin and tobramycin), US Food and Drug Administration (streptomycin), the National Antimicrobial Resistance Monitoring System (apramycin) or la Société Française de Microbiologie (neomycin and paromomycin). Apramycin-resistant strains were subjected to whole genome sequencing using Illumina X10 platform. Among CRE strains, 65.7, 64.9, 79.1, and 95.5% were susceptible to neomycin (MIC50/MIC90, 8/256 μg/ml), paromomycin (4/>256 μg/ml), streptomycin (16/256 μg/ml) and apramycin (4/8 μg/ml), respectively, while only 55.2, 28.4, and 35.1% were susceptible to amikacin (32/>256 μg/ml), gentamicin (128/>256 μg/ml) and tobramycin (64/>256 μg/ml), respectively. Six CRE strains including five Escherichia coli of different sequence types and one Klebsiella pneumoniae were resistant to apramycin and the apramycin-resistant gene aac(3)-IVa was detected in all of these strains. In conclusion, neomycin, paromomycin, streptomycin and apramycin retain activity against most CRE strains. Although none of these non-4,6-disubstituted DOS aminoglycosides are suitable for intravenous use in human at present, these agents warrant further investigations to be used against CRE infections.

The antimicrobial options against CRE are very limited (Van Duin et al., 2013). Colistin is the last resort antimicrobial agent commonly used to treat CRE infections but colistin resistance among CRE has also emerged worldwide (Poirel et al., 2017). Therefore, it is crucial to find antimicrobial agents as alternative choices. As very few new antimicrobial agents will most likely become available in the near future, repurposing currently available agents is another yet more realistic option. Aminoglycosides that are commonly used to treat clinical infections caused by bacteria other than Mycobacterium spp. include amikacin, gentamicin and tobramycin, all of which belong to the 4,6-disubstituted deoxystreptamine (DOS) subclass (Mingeot-Leclercq et al., 1999). These aminoglycosides retain activities against certain CRE strains (Livermore et al., 2011b) and their combination with other agents has been successful for treating some CRE infection cases (Hirsch and Tam, 2010;Falagas et al., 2014;Rafailidis and Falagas, 2014;Shields et al., 2016). Nonetheless, resistance to these 4,6-disubstituted DOS aminoglycosides among CRE is extensive (Livermore et al., 2011b;Almaghrabi et al., 2014;Smith and Kirby, 2016). In addition, the emergence of acquired 16S rRNA methylases including ArmA, RmtA to RmtH and NmpA, which confer highlevel resistance to the 4,6-disubstituted DOS aminoglycosides, imposes another serious challenge for clinical management (Doi et al., 2016). Many CRE, in particular those carrying bla NDM carbapenemase gene, have 16S rRNA methylase genes (Livermore et al., 2011b). Neomycin and paromomycin (also called neomycin E) belong to the 4,5-disubstituted DOS subclass (Mingeot-Leclercq et al., 1999), while apramycin is of the 4-monosubstituted DOS subclass and streptomycin is a non-DOS aminoglycoside as it contains an aminocyclitol other than DOS (Mingeot-Leclercq et al., 1999). Streptomycin is not affected by all 16S rRNA methylases, while the 4,5-disubstituted or 4-monosubstituted DOS aminoglycosides are only affected by NpmA (Doi et al., 2016), which is not common in the Enterobacteriaceae. Therefore, the non-4,6-disubstituted DOS aminoglycosides, which are usually not included into the panel for susceptibility tests or even are not approval for human use, may be potent against CRE. Here we present susceptibility results of four non-4,6-disubstituted DOS aminoglycosides (apramycin, streptomycin, neomycin and paromomycin) against 134 CRE clinical strains that were collected at multiple hospitals in China.

Bacterial Strains
Non-duplicate CRE clinical strains (n = 134) from discrete patients were consecutively collected at 17 hospitals in 11 cities of Sichuan Province, western China since June 2016 to April 2017. Species identification was performed using the Vitek II system (bioMerieux, Durham, NC, USA). Acquired carbapenemaseencoding genes bla GES , bla KPC , bla IMP , bla NDM , bla OXA−48 -like, and bla VIM were screened for CRE strains using PCR as described previously (Poirel et al., 2000;Bradford et al., 2004;Mendes et al., 2007;Zong and Zhang, 2013). The specific allelic variants of the carbapenemase genes were obtained using PCR with additional primers able to amplify the whole encoding sequence (Zhang et al., 2012;Zong and Zhang, 2013) and Sanger sequencing.

Genome Sequencing and Analysis
Genomic DNA of apramycin-resistant CRE strains were prepared using the QIAamp DNA Mini Kit (Qiagen, Hilden, Germany) and was subjected to whole genome sequencing with 200 × coverage using the HiSeq X10 Sequencer (Illumina, San Diego, CA). Reads were trimmed using Trimmomatic (Bolger et al., 2014) and were then assembled to contigs using the SPAdes program (Bankevich et al., 2012) with careful mode turned on. Antimicrobial resistance genes were identified from genome sequences using the ResFinder tool at the Center for Genomic Epidemiology (CGE, http://genomicepidemiology. org/) program. Sequence types were determined using the genomic sequence to query the multi-locus sequence typing (MLST) database using the MLST tool available at CGE.
Among aminoglycosides tested, apramycin appears to be the most promising as its MICs against almost all (95.5%, 128/134) of the CRE strains were 8 or less (MIC 50 and MIC 90 was 4 and 8 µg/ml, respectively). Nonetheless, six CRE strains including five E. coli and one K. pneumoniae were resistant to apramycin, all of which were high-level resistance (MIC of apramycin, >256 µg/ml). Draft whole genomic sequences of these strains were obtained. A total of 4,453,184-7,526,752 clean reads and 1.34-2.26 Gb clean bases were generated for the six strains, which were then assembled to 106-229 contigs (79-187 were ≥1,000 bp in length) with a 50.25-50.57% GC content for E. coli and 57.03% for K. pneumoniae, respectively ( Table 3). The five apramycin-resistant E. coli belonged to five sequence types (ST101, ST167, ST206, ST6388, and ST6823), suggesting that the apramycin-resistant strains were not clonal. The apramycin-resistant K. pneumoniae belonged to ST340. All of apramycin-resistant E. coli and K. pneumoniae had aac(3)-IVa, which encodes an aminoglycoside-modifying enzyme conferring resistant to apramycin (Shaw et al., 1993).

DISCUSSION
Neomycin is not given via injection due to its nephrotoxicity, while paromomycin is on the List of Essential Medicines of World Health Organization (World Health Organization, 2015) and is used to a number of parasite infections such as amebiasis and giardiasis by oral or intramuscular injection. Both neomycin and paromomycin are poorly absorbed when taken orally. Nonetheless, oral administration of neomycin has been used for inhibiting the overgrowth of gut microflora (Clark, 1977) and for decolonizing the intestinal carriage of extended-spectrum β-lactamase(ESBL)-producing Enterobacteriaceae (Huttner et al., 2013). Paromomycin has also been applied for decolonizing ESBL-producing Enterobacteriaceae (Buehlmann et al., 2011;Rieg et al., 2015). As both agents are not given by intravenous injection, they are unlikely to be used for treating systematic infections caused by CRE. However, our results suggest that neomycin and paromomycin deserve further investigations    to be used enterally and to be modified to reduce their nephrotoxicity.
The fact that close to 80% of CRE strains were susceptible to streptomycin suggests that streptomycin may have a potential role in treating CRE infections. However, streptomycin is only available for intramuscular injection and therefore may not be appropriate for treating patients with critical illness. Nonetheless, the efficacy of streptomycin, probably in combination with other agents such as β-lactams, in treating CRE infections warrants to be explored.
The vast majority of CRE strains were susceptible to apramycin, suggesting the excellent in vitro activity against CRE. A previous study has found that 70.8% of 71 CRE clinical strains collected in USA were susceptible to apramycin (Smith and Kirby, 2016), while in another study 80 of 82 (97.6%) CRE clinical isolates, most of which were collected in UK, were susceptible to apramycin (Livermore et al., 2011a). These findings suggest that the excellent activity of apramycin against CRE is not geographically restricted. Unfortunately, apramycin is a veterinary agent and has not been approved for clinical use, which is likely due to its narrow therapeutic index (Livermore et al., 2011a). The use of aminoglycosides has been limited by its toxicity, particularly nephrotoxicity and ototoxicity. However, apramycin has low ototoxicity (Matt et al., 2012) and fewer nephrotoxic side effects (Kostrub et al., 2009;Kang et al., 2017) as it appears to have higher affinity to bacterial over mitochrondrial ribosomes (Perzynski et al., 1979;Kang et al., 2017). In clinical settings, aminoglycosides are commonly used in combination with other antimicrobial agents, particularly β-lactams and sometimes fluoroquinolones. There are no data about the combination of these non-4,6-disubstituted DOS aminoglycosides with other agents against CRE at present. Nonetheless, in animal models the combination of apramycin and a fluoroquinolone (enrofloxacin) shows synergic effect to increase the efficacy against Gram-negative bacilli such as E. coli and Salmonella and is also able to decrease the emergence of mutational resistance to fluoroquinolones (Randall et al., 2016). Apramycin therefore warrants further investigations as a repurposed agent against CRE. The unusual structure of apramycin also provides a scaffold for further modification to generate new potent and safe compounds for treating CRE infections (Livermore et al., 2011a).
Although our results suggest that the non-4,6-disubstituted DOS aminoglycosides had good in vitro activities against CRE, the results should be interpreted with cautions. First, unlike 4,6-disubstituted DOS aminoglycosides, the susceptibility interpretation (breakpoints of MICs) of the four non-4,6disubstituted DOS aminoglycosides against Enterobacteriaceae has not been well established and validated. In particular, apramycin is a veterinary agent and its susceptibility/resistance breakpoints may not be suitable for human medicine. Second, there are toxicity concerns, particularly nephrotoxicity, related to the four non-4,6-disubstituted DOS aminoglycosides and none of the four agents are suitable for intravenous use in human at present. Additional studies must be carried out to address any safety risks. Third, development of resistance to the four aminoglycosides also poses a huge challenge for their potential use against CRE. For instance, high level resistance to streptomycin can be rapidly developed (Sinha, 1986).
In conclusion, most (64.9-95.5%) CRE strains in this collection were susceptible to neomycin, paromomycin, streptomycin or apramycin, while less strains (28.4-55.2%) were susceptible to the mainstream clinically-available aminoglycosides amikacin, gentamicin and tobramycin. In particular, almost all CRE strains were susceptible to apramycin, suggesting that apramycin may be an excellent candidate for modification to generate new potent and safe aminoglycosides. Although none of these non-4,6-disubstituted DOS aminoglycosides are suitable for intravenous use in human at present, they warrant further investigation for treating CRE infections.

AUTHOR CONTRIBUTIONS
ZZ designed the experiments, analyzed the data and wrote the manuscript. YH and LL performed the experiments and analyzed the data. XZ and YF contributed to analyzing the data and co-wrote the manuscript.