Macrocolony of NDM-1 Producing Enterobacter hormaechei subsp. oharae Generates Subpopulations with Different Features Regarding the Response of Antimicrobial Agents and Biofilm Formation

Enterobacter cloacae complex has been increasingly recognized as a nosocomial pathogen representing the third major Enterobacteriaceae species involved with infections. This study aims to evaluate virulence and antimicrobial susceptibility of subpopulations generated from macrocolonies of NDM-1 producing Enterobacter hormaechei clinical isolates. Biofilm was quantified using crystal violet method and fimbrial genes were investigated by PCR. Susceptibility of antimicrobials, alone and combined, was determined by minimum inhibitory concentration and checkerboard assays, respectively. Virulence and efficacy of antimicrobials were evaluated in Galleria mellonella larvae. Importantly, we verified that some subpopulations that originate from the same macrocolony present different biofilm production ability and distinct susceptibility to meropenem due to the loss of blaNDM-1 encoding plasmid. A more in-depth study was performed with the 798 macrocolony subpopulations. Type 3 fimbriae were straightly related with biofilm production; however, virulence in larvae was not statistically different among subpopulations. Triple combination with meropenem–rifampicin–polymyxin B showed in vitro synergistic effect against all subpopulations; while in vivo this treatment showed different efficacy rates for 798-1S and 798-4S subpopulations. The ability of multidrug resistant E. hormaechei isolates in generating bacterial subpopulations presenting different susceptible and virulence mechanisms are worrisome and may explain why these infections are hardly overcome.


Introduction
The Enterobacteriaceae family is composed by many bacteria, including Escherichia coli, Klebsiella spp., and Enterobacter spp., responsible for community-associated as well as healthcare-associated infections [1][2][3]. In the last decades, species of the Enterobacter cloacae complex (ECC) have aroused greater concern, since they are increasingly associated with carbapenemase-encoding genes acquisition, being the second or third most common carbapenemase-producing Enterobacteriaceae (CPE) [4][5][6][7]. Carbapenem resistance constitutes a global public-healthcare problem associated with a high mortality   Results of crystal violet assays showed that subpopulations were not able to produce much biofilm; however, we could still observe statistically differences in the ability to form biofilm among subpopulations of five macrocolonies (67, 798, 821, 977, and 1105) ( Figure 2). Considering that subpopulations of the group 798 showed a heterogeneous behavior in biofilm formation and 798-1S Bacteria were grown at 37 °C in microtiter plates containing Tryptone Soya Broth (TSB) for 24 h and then biofilm formation was quantified using crystal violet assay. The results are presented as the means and standard deviation. Significant differences in biofilm formation, among subpopulations of the same macrocolony, were pointed out when p < 0.05 (*), p < 0.01 (**), p < 0.0001 (****).   Figure 3A) and the amount of biofilm produced was higher in LB supplemented with glucose.

Subpopulations of 798
Since fimbriae play an important role in attachment and biofilm formation in gram-negative bacteria, we investigated the presence of fimbrial genes in order to understand the difference in biofilm formation ability among subpopulations. Genes encoding curli (csgA, csgB, and csgD), type 1 fimbriae (fimA and fimH), and P pili (papC and papD) were detected in all subpopulations. However, type 3 fimbriae encoding gene (mrkB) was detected only in two subpopulations (798-1S and 798-2S) ( Figure 3B). Interestingly, only subpopulations with mrkB encoding gene were able to produce biofilm.

Checkerboard Assay: Triple Combination (meropenem-rifampicin-polymyxin B) is Effective against All Subpopulations
Synergism with double and triple antimicrobial combination against subpopulations generated from the 798 macrocolony was evaluated using checkerboard assay. At first, MIC of meropenem, rifampicin, and polymyxin B was determined. According to these results, all subpopulations showed a similar antimicrobial profile: susceptible to polymyxin B, resistant to meropenem, and with high MIC for rifampicin (128 µg/mL) ( Table 2). In vitro combined inhibitory activities of meropenem-polymyxin B and rifampicin-meropenem achieved synergy against only one subpopulation (798-1S). Rifampicin-polymyxin B did not show synergistic activity against any subpopulation. On the other hand, triple combination with meropenem-rifampicin-polymyxin B presented synergistic effect against all subpopulations ( Table 2). Biofilm formation was determined in different media: Luria Bertani broth (LB), LB with glucose (0.02 M), and M9 minimal medium supplemented with glucose (0.02 M) and MgSO4 (0.002 M). The results showed that two (798-1S and 798-2S) of the four subpopulations were able to form biofilm ( Figure 3A) and the amount of biofilm produced was higher in LB supplemented with glucose.
fimA Bacteria were grown at 37 • C in microtiter plates containing LB broth, LB broth supplemented with glucose, M9 minimal medium or TSB for 24 h, after which biofilm formation was quantified. The results are presented as the means and standard deviation for three independent experiments. Significant differences in biofilm formation using LB broth + GLU were pointed out: p < 0.05 (*) and p < 0.0001 (****). (B) Fimbrial encoding genes: fimA, fimH (type 1 fimbriae genes), papC, papD (P pili genes), csgA, csgB, csgD (curli genes), and mrkB (type 3 fimbriae gene). Gene present: +; and gene absent: -. In order to determine the G. mellonella susceptibility to E. hormaechei subsp. oharae infection, larvae were infected with four different bacterial inocula (5.0 × 10 5 , 2.0 × 10 6 , 5.0 × 10 6 and 1.0 × 10 7 CFU/larva) of subpopulations 798-1S, 798-2S, 798-3S, and 798-4S. It was verified that increasing doses of bacteria resulted in reduced larval survival in a dose-dependent manner during 120 h of incubation ( Figure 4). No macroscopic changes or deaths were observed in the uninfected groups. Based on these data, 1.0 × 10 7 CFU/larva was selected as the optimal inoculum for the subsequent treatment experiments, In vivo antimicrobial treatments were evaluated against 798-1S and 798-4S subpopulations. These subpopulations were chosen because they presented the most distinct virulence pattern in G. mellonella. The 798-1S showed a tendency to be less virulent (30% of survival after 120 h of incubation) and the 798-4S the more virulent (16.7% of survival) when larvae were inoculated with 1.0 x 10 7 CFU.
Efficacy of monotherapy and triple antimicrobial combination in larvae infected with NDM-1 producing E. hormaechei subsp. oharae are shown in Figure 5. Meropenem alone ( Figure 5G) or combined with polymyxin B plus rifampicin ( Figure 5H) showed a significantly protective effect against 798-4S (p = 0.0010 and p < 0.0001, respectively). Conversely, none of the tested treatments was able to enhance the larvae survival when they were infected with 798-1S ( Figure 5A-D), although this subpopulation has killed fewer larvae than the 798-4S. In vivo antimicrobial treatments were evaluated against 798-1S and 798-4S subpopulations. These subpopulations were chosen because they presented the most distinct virulence pattern in G. mellonella. The 798-1S showed a tendency to be less virulent (30% of survival after 120 h of incubation) and the 798-4S the more virulent (16.7% of survival) when larvae were inoculated with 1.0 × 10 7 CFU.
Efficacy of monotherapy and triple antimicrobial combination in larvae infected with NDM-1 producing E. hormaechei subsp. oharae are shown in Figure 5. Meropenem alone ( Figure 5G) or combined with polymyxin B plus rifampicin ( Figure 5H) showed a significantly protective effect against 798-4S (p = 0.0010 and p < 0.0001, respectively). Conversely, none of the tested treatments was able to enhance the larvae survival when they were infected with 798-1S ( Figure 5A-D), although this subpopulation has killed fewer larvae than the 798-4S.

Discussion
Enterobacter hormaechei subsp. oharae biofilm macrocolonies were able to generate genetically and phenotypically distinct subpopulations (Figure 1) that presented differences in virulence mechanisms and antimicrobial response.
The amount of biofilm produced by subpopulations generated from the same macrocolony was different in five of the nine clinical isolates (Figure 2). Interestingly, the strongest biofilm producer isolate was obtained from urine (798). It is possible that this strain has additional virulence mechanisms that allow bacteria to adhere to bladder cells and cause urinary tract infection. Important virulence factors that contribute to biofilm formation are extracellular appendages called fimbriae. Genes encoding curli, type 1 fimbriae, and P pili were detected in all subpopulations of the group 798; however, the type 3 fimbriae gene was only detected in subpopulations able to produce biofilm (798-1S and 798-2S). On the other hand, subpopulations without mrkB (798-3S and 798-4S) did not produce biofilm in vitro using four different mediums ( Figure 3A). As far as we know, there is no data showing the role of type 3 fimbriae in biofilm formation of Enterobacter spp.
Type 3 fimbriae are encoded by the mrkABCDF gene cluster [32] and were initially identified in Klebsiella pneumoniae strains by Duguid [33]. Since then, type 3 fimbriae have been described in other members of the Enterobacteriaceae family, including Serratia spp., Enterobacter spp., and Escherichia coli isolates [34][35][36]. In K. pneumoniae, this gene cluster is chromosomally encoded [37], while in other species it is found to be encoded by conjugative plasmids [36,38]. Type 3 fimbriae are involved in attachment to abiotic and biotic surfaces and in biofilm formation [39,40]. There is little information about type 3 fimbriae related with Enterobacter spp. in the literature. Adegbola and Old investigated type 1 and type 3 fimbriae in Enterobacter spp. and showed that type 1 fimbriae is most frequent and that most strains produced only one of these fimbriae [34]. Similar findings were described by Hornick et al., where most of E. cloacae respiratory isolates produced type 1 fimbriae, and fewer numbers also expressed type 3 fimbriae [35]. Recently, Azevedo et al. investigated nine virulence genes, including mrkD (adhesin type 3 fimbriae) and fimH (adhesive subunit of type 1 fimbriae), in eight E. cloacae isolates; surprisingly, no isolates presented virulence genes [41].
Macrocolony is well described and accepted as a biofilm model. Usually, studies evaluate macrocolonies' structure and morphology, like the presence of wrinkle and ring patterns that are related with cellulose and curli fimbriae production, respectively [28,29,[42][43][44][45][46]. As far as we know, Richer et al. (2014) were the first to investigated different subpopulations/regions generated by macrocolony [47]. In a macrocolony biofilm, bacteria typically conjugate with their closest neighbors when physical contact occurs between a donor and recipient cell to transmit horizontal plasmid, creating subpopulations that are independent from each other [30]. In this sense, some studies observed that a plasmid-bearing population can originate clonal sectors of plasmid-free cells [48,49]. This can explain why plasmid encoding bla NDM-1 did not spread throughout all the macrocolony and some subpopulations restored meropenem susceptibility. We could also hypothesize that the same occurred with type 3 fimbriae, since these fimbriae are encoded by plasmids in most Enterobacteriaceae species.
Capability of bacteria to form biofilm on medical devices, such as catheter and prosthesis, has been proposed as one of the important mechanisms in nosocomially acquired and persistent infections, increasing resistance to antimicrobial treatment [16]. There are few studies evaluating biofilm formation in multidrug resistant strains. Recently, a study from Brazil demonstrated that the majority of repeated KPC infections are caused by the same strain that caused the previous infection/colonization, these findings illustrate the capacity of multiple clones producing biofilm to coexist in the same patient at the same time, serving as a constant reservoir of KPC in the hospital environment [50].
Regarding susceptibility profile, as expected, most subpopulations presented high antimicrobial resistance (Table 1). In vitro triple combination with meropenem-rifampicin-polymyxin B presented a synergistic effect against all tested subpopulations ( Table 2). Corroborating with our findings, Tangden et al. tested 14 antimicrobial combinations using time-kill experiments against two NDM-1-producing K. pneumoniae strains, and found that the combination of rifampicin-meropenem-colistin was the most effective regimen [51]. In another study, Urban et al. showed that combination of polymyxin B-doripenem-rifampicin achieved 100% bactericidal activity for Pseudomonas aeruginosa and E. coli, 80% for K. pneumoniae, and 60% for Acinetobacter baumannii despite resistance to the carbapenems and rifampicin alone [52]. Although rifampicin by itself is not considered for the treatment infections caused by gram-negatives due to the rapid emergence of resistance, in vitro studies suggest that rifampicin has a synergistic activity when used as part of a combination therapy regimen against CPE [51][52][53][54][55].
In vivo, meropenem alone or combined with polymyxin B and rifampicin showed a significantly protective effect in larvae infected with 798-4S (p = 0.0010 and p < 0.0001, respectively). Conversely, none of the tested treatments was able to enhance the larvae survival when they were infected with 798-1S ( Figure 5), although this subpopulation has shown a tendency to be less virulent then the 798-4S. Our hypothesis is that other bacterial factors, than biofilm, are being expressed in vivo and that the larval immune system can respond differently according to distinct antigenic stimuli evoked by bacterial subpopulations, influencing host survival.
Taken together, some discrepancies between in vitro and in vivo results were observed: (i) Polymyxin B was effective in vitro but not in vivo against 798-1S and 798-4S; (ii) triple combination presented synergistic effect in vitro but did not show significant enhancement in survival rates of larvae infected with 798-1S; and (iii) monotherapy with meropenem was able to increase the survival of larvae infected with 798-4S when compared with the control group treatment with water (53.3% × 25.6% survival), while in vitro evaluation classified 798-4S as resistant to this antimicrobial.
Supporting these results, discrepancies between in vitro and in vivo susceptibility for polymyxin B have been published in literature [56,57]. Yang et al., reported contradictory results considering in vitro and in vivo (G. mellonella model) colistin (polymyxin E) susceptibility in A. baumanni strains [56]. Moreover, Benthall et al. showed that the treatment with colistina in G. mellonella presented variable activity against K. pneumoniae, regardless of intrinsic susceptibility. In this same study, the carbapenems appeared to act better in vivo than in vitro, with meropenem able to clear infections caused by strains possessing bla NDM-1 and bla VIM carbapenemases [58], similarly as observed herein for subpopulation 798-4S.
In summary, our findings demonstrate discrepancies between in vitro and in vivo susceptibility of E. hormaechei subsp. oharae. subpopulations to antimicrobial agents; some treatments were effective in vitro but not in vivo and vice versa. Interestingly, these subpopulations also showed different response to antimicrobial agents in G. mellonella infection model. Additionally, we may hypothesize that type 3 fimbriae are encoded on the plasmid and have a very important role in biofilm formation of E. hormaechei subsp. oharae. As far as we know, this is the first study evaluating macrocolonies of E. cloacae complex strains highlighting the potential of the macrocolony model as a tool to study the physiological heterogeneity within the biofilm.
The results point out the ability of a multidrug resistant E. hormaechei subsp. oharae isolate in generating subpopulations, with distinct phenotypic and genetic features related to biofilm formation and antimicrobial profile. These findings might be associated to long term and chronic infections leading to an additional challenge in the treatment of bacterial infections and highlight the urgent need for newer antimicrobial development against biofilm related infection.

Bacterial Strains and Growth of Macrocolonies
Nine E. hormaechei subsp. oharae clinical isolates (Table 3) harboring bla NDM-1 gene (New Delhi metallo-beta-lactamase) were recovered from three hospitals and stored at −80 • C at the Laboratório de Pesquisa em Resistência Bacteriana of Hospital de Clínicas de Porto Alegre in Rio Grande do Sul, the southermost state of Brazil. These isolates were previously evaluated by Rozales et al. [59]. At first, clinical isolates were streaked on LB agar and incubated at 37 • C overnight. After checking purity of cultures, one colony was selected and dissolved in 5 mL of LB broth and incubated at same conditions. For macrocolonies' growth, a volume of 3 uL of these cultures was spotted on LB agar plates supplemented with Congo Red (40 mg/L) and Coomassie Brilliant Blue (20 mg/L). The plates were incubated at 35 • C for 5 days [42,47].
After incubation, macrocolonies presented different areas, termed subpopulations. Cells were directly taken from these subpopulations and frozen in 10% skim milk with glycerol for further experiments. Before each experiment, these subpopulations were streaked on LB agar and incubated at 37 • C overnight.

Biofilm Formation: Microtiter Plates Assay
Biofilm formation was quantified using crystal violet (CV) assay in 96-well microtiter plates [60]. After 24 h of incubation, plates were washed to remove unbound bacteria and the attached bacteria were heat-fixed at 60 • C for 1 h. The biofilm was then stained with crystal violet, quantified by dissolving CV in 96% ethanol, and the optical density was measured at 595 nm (OD595). Biofilm formation was tested using different media: TSB, LB, LB supplemented with glucose (0.02 M), and M9 minimal medium supplemented with glucose (0.02 M) and MgSO 4 (0.002 M).

Polymerase Chain Reaction (PCR):Fimbrial Genes Detection
DNA of bacterial cells was extracted using boiling method, which is based on thermal shock and lysis of components other than nucleic acids. The same PCR cycling conditions used were for all genes: initial denaturation at 95 • C for 5 min followed by 30 cycles of denaturation at 94 • C for 30 s, annealing at 58 • C for 30 s, extension at 72 • C for 1 min, and a final extension at 72 • C for 5 min. Primer sets were design for PCR detection of fimbrial genes (Table 4).

Minimum Inhibitory Concentration (MIC): Agar Dilution Method
MIC of antimicrobial agents ceftazidim, ciprofloxacin, gentamicin, and meropenem was determined for all subpopulations using the agar dilution method according to EUCAST guideline [31]. The MIC was defined as the lowest concentration of the drug that inhibited growth of the tested bacteria. All MIC experiments were performed at least three times.

Checkerboard Assay
Minimum inhibitory concentration of meropenem, rifampicin, and polymyxin B and synergy with double and triple combinations were performed using the checkerboard method in 96-well microtiter plates with Mueller-Hinton broth and bacterial suspension containing approximately 5 × 10 5 CFU/mL [61]. MIC and synergistic effect were visually determined as the lowest drug concentration (alone or combined, respectively) that inhibited bacterial growth. Susceptibility of meropenem and polymyxin B was interpreted according to EUCAST [31]. There is no standardized breakpoint value for rifampicin against Enterobacteriaceae according to guidelines. The synergism was determined by the fractional inhibitory concentration index (FICI) and interpreted as follows: FICI ≤ 0.5 indicates synergy; FICI > 0.5 ≤4 = no interaction; FICI > 4.0 = antagonism [61,62].

Galleria mellonella Model Studies
The whole cycle of G. mellonella was maintained in our laboratory at 28 • C. Insects were fed with an artificial diet consisting of honey and several flours. Larvae weighting 220-260 mg were randomly selected to comprise groups of ten larvae which were inoculated with 10 uL of bacterial suspension by injection into the haemocoel via the last right proleg, using a Hamilton syringe (Sigma-Aldrich). Four different bacterial suspensions in sterile phosphate-buffered saline (PBS) were tested for each subpopulation: 5.0 × 10 5 , 2.0 × 10 6 , 5.0 × 10 6 and 1.0 × 10 7 CFU/larva. Uninfected larvae (either uninoculated or injected with only PBS) were used as negative controls. Afterwards caterpillars were incubated in Petri dishes at 37 • C and were observed daily during 120 h. They were considered dead when they did not respond to touch.
Based on larvae survival curve with different inocula, bacterial concentration of 1.0 × 10 7 CFU/larva was selected to be used for evaluating treatment efficacy. Antimicrobial agents (10 uL) were administered as single injection into the last left proleg 30 min after bacterial inoculation.