Efficiency of Single Phage Suspensions and Phage Cocktail in the Inactivation of Escherichia coli and Salmonella Typhimurium: An In Vitro Preliminary Study

Enterobacteriaceae Escherichia coli and Salmonella enterica serovar Typhimurium strains are among the main pathogens responsible for moderate and serious infections at hospital and community environments, in part because they frequently present resistance to antibiotics. As the treatment of Enterobacteriaceae infections is empiric, using the same antibiotics to treat E. coli and Salmonella infections, the same concept can be applied with phages. The use of different phages combined in cocktails, frequently used to circumvent the development of phage-resistant mutants, also allows for the treatment of multiple pathogens, broadening the phages’ action spectrum. As such, the aim of this study was to evaluate the efficiency of a cocktail of two phages (ELY-1, produced on E. coli and phSE-5, produced on S. Typhimurium) to control E. coli and S. Typhimurium. Phages ELY-1 and phSE-5 were effective against E. coli (maximum reductions of 4.5 and 3.8 log CFU/mL, respectively), S. Typhimurium (maximum reductions of 2.2 and 2.6 log CFU/mL, respectively), and the mixture of both bacteria (maximum reductions of 2.2 and 2.0 log CFU/mL, respectively). The cocktail ELY-1/phSE-5 was more effective against S. Typhimurium and the mixture of both bacteria (maximum reduction of 3.2 log CFU/mL for both) than the single phage suspensions and as effective against E. coli as its specific phage ELY-1 (maximum reductions of 4.5 log CFU/mL). The use of both the phage cocktails, as well as the single-phage suspensions, however, did not prevent the occurrence of phage-resistant mutants. Overall, the results indicate that the application of the phages in the form of a cocktail show their potential to be used presumptively, that is, prior to the identification of the pathogens, paving its use to control E. coli or S. Typhimurium.

Phages ELY-1 and phSE-5 were prepared using E. coli and S. Typhimurium as the host, respectively. Phage suspensions were prepared from the phage stock prepared previously in SM buffer (0.1 M NaCl (Sigma-Aldrich, St. Louis, MO, USA), 8 mM MgSO 4 (Sigma), 20 mM Tris-HCl (Sigma), 2% (w/v) gelatin, pH 7.5). Three hundred microliters of the phage stock were added to 30 mL of E. coli or S. Typhimurium in the exponential growth phase. The suspension was grown overnight and incubated at 25 • C at 50 rpm. The lysates were incubated with chloroform (final volume of 1%) for 1 h at 120 rpm. After incubation, the lysate was centrifuged at 13,000 rpm for 10 min at 4 • C, to remove intact bacteria or bacterial debris. Phage suspension was stored at 4 • C and the titre was determined by the double-layer agar method [62]. Successive dilutions of the phage suspension were performed in PBS and 500 µL of each dilution, together with 200 µL of fresh bacterial culture, and were mixed with 5 mL of TSB 0.6% top agar layer (30 g/L TSB (Liofilchem), 6 g/L agar (Liofilchem), 0.05 g/L CaCl 2 (Sigma), 0.12 g/L MgSO 4 (Sigma), pH 7.4) and placed over a TSA plate. The plates were incubated at 25 • C for 8-12 h. After incubation, the number of plaques was counted and the results expressed as plaque-forming units (PFU)/mL.

Phage Host Range: Spot Test and Efficiency of Plating (EOP)
The phage host range was determine using single suspensions (ELY-1 and phSE-5) and the bacterial strains listed in Table 1. The phage host range was determined by a spot test according to   [14]. Briefly, three milliliters of TSB 0.6% agar (Merck, Darmstadt, Germany), previously inoculated with 300 µL of bacterial culture (see Table 1), were overlaid on solid TSA and spotted with 10 µL of the phage suspension. The plates were incubated at 25 • C and examined for lysis plaques after 8-12 h. Bacterial sensitivity of the phage was established by a lysis cleared zone at the spot. According to the clarity of the spot, bacteria were differentiated into two categories, clear lysis zone (+) and not lysis zone (−).
The EOP was determined for bacteria where a clear lysis zone occurred (positive spot test), using the double-layer agar method. The EOP was calculated as the ratio between the PFU average on the target bacteria and the PFU average on the host bacteria. EOP values are presented in the manuscript as the mean of three measurements followed by their standard deviation. The value obtained with the host strain was considered as EOP = 1.

Killing Curves
Bacterial inactivation was determined using single phage suspensions (ELY-1 and phSE-5) and a phage cocktail (ELY-1/phSE-5, the two phages were mixed together with each phage at the same concentration) using the bacterium E. coli or S. Typhimurium or a mixture of both bacteria with the same concentration, at a MOI of 100. Exponential bacterial cultures of E. coli, S. Typhimurium, or the mixture of the two bacteria were adjusted to a 0.8 O.D. at 600 nm (corresponding to a cell density of 10 9 CFU/mL). In order to obtain a MOI of 100, the exponential cultures of bacteria (final concentration of 10 6 CFU/mL) and phage suspension (final concentration of 10 8 PFU/mL) were inoculated in sterilized glass Erlenmeyer flasks with 30 mL of TSB medium and incubated at 25 • C without agitation (B+P). For each assay, two control samples were included, as follows: The bacterial control (BC) and the phage control (PC). The bacterial control was not inoculated with phages and the phage controls were inoculated with phages but not with bacteria. Controls and test samples were incubated in exactly the same conditions. Aliquots of the test samples and of the bacterial and phage controls were collected at time 0 and after 2, 4, 6, 8, 10, and 12 h of incubation. The phage titre was determined in duplicate through the double-agar layer method after an incubation period of 12 h at 25 • C. Bacterial concentration was determined in duplicate in TSA medium after 48 h at 25 • C. Sensitive and phage resistant colonies were picked and purified by successive sub-culturing in TSA agar in order to remove attached phage particles and were used in further experiments (as described in Section 2.6, Fitness of Phage Resistant Mutants). Three independent experiments were performed for each condition.

Determination of the Frequency of Emergence of Phage Resistant Mutants
The development of resistant mutants of E. coli and S. Typhimurium, in pure and mixed cultures of phages ELY-1, phSE-5, and phage cocktail ELY-1/phSE-5, were evaluated according to   [14]. To determinate the frequency of phage-resistant bacteria, ten isolated colonies from a plate with sensitive bacteria were selected and inoculated into ten tubes with 5 mL of TSB, grown at 25 • C for 12 h at 120 rpm stirring. Aliquots of 100 µL from the 10 −1 to 10 −3 dilutions of the bacterial culture and of the phage, from a stock solution at 10 9 PFU/mL, were inoculated in tubes with TSB 0.6% agar, plated on TSA plates and incubated at 25 • C for 24 h. Simultaneously, 100 µL aliquots of the 10 −5 to 10 −7 dilutions of the bacterial culture were plated by incorporation on TSA plates without phages and incubated at 25 • C for 24 h. The averaged colony number of mutants (obtained from the ten isolated colonies) in 1 mL of culture (prepared from the culture with phages) was divided by the averaged colony number of the control (prepared from the culture without phages) [63]. Three independent assays were performed.

Fitness of Phage Resistant Mutants
The growth of resistant and sensitive bacterial populations was quantified in the presence and absence of the phages ELY-1, phSE-5, and the phage cocktail ELY-1/phSE-5, in order to evaluate the toll bacteria suffers ("the fitness") to develop resistance to the phages. Exponential host bacterial cultures of E. coli, S. Typhimurium, and the mixture with the two bacteria (without phage contact) and mutants resistant to the phages ELY-1, phSE-5, and the phage cocktail ELY-1/phSE-5 were adjusted to 0.8 O.D. at 600 nm (corresponding to a cell density of 10 9 CFU/mL). The fitness of each bacterial population was evaluated by determining the bacterial concentration using the colony-counting method and by determining the bacterial growth curve using the optical density (OD).
Pure cultures of E. coli or S. Typhimurium or the mixture of the two bacteria resistant to phage ELY-1 were added to 6 samples to obtain a final concentration of 10 6 CFU/mL. Three of these samples were inoculated with a phage (resistant bacteria with phage ELY-1) to obtain a final concentration of 10 8 PFU/mL. To the remaining infected samples, no phages were added (resistant bacteria without phages). The same protocol was followed for the mutants resistant to phage phSE-5 and to the cocktail ELY1/phSE-5. The same procedure was done for E. coli or S. Typhimurium or the mixture with the two bacteria using sensitive bacteria. Samples were incubated at 25 • C and bacterial concentration was determined by the spread method in duplicate in TSA medium at time 0 and after 6 and 12 h. Three independent experiments were performed for each condition.

Statistical Analysis
The statistical analysis of data was performed using GraphPad Prism 7.04 software. Normal distribution was checked by a Kolmogorov-Smirnov test and the homogeneity of variance was assessed by Levene's test. Significance was accepted at p < 0.05. Tukey's multiple comparison test was used for a pairwise comparison of the means. The significance of bacterial and viral concentrations between treatments, and along the experiments, was tested using two-way ANOVA and the Bonferroni post-hoc test (Section 3.2, Section 3.3 and Section 3.4). For different treatments, the significance of differences was evaluated by comparing the result obtained in the test samples with the results obtained for the correspondent control samples, for the different times. One-way ANOVA was used to examine the differences between the frequency of bacteria spontaneous phage-resistant mutants (Section 3.5). Two-way ANOVA was used to examine differences between the concentration of resistant bacteria and sensitive bacteria in the presence/absence of the phage after 6 and 12 h of incubation (Section 3.6). Two-way ANOVA with repeated measures was used to analyze the statistical differences between growth curves of the sensitive and resistant bacteria in the presence and absence of the phages during the sampling time (Section 3.6).
The phage controls (PC) remained constant during the 12 h of the experiments (ANOVA, p > 0.05, Figure 1B). For phage ELY-1 and the phage cocktail, the survival factor increased up to 0.5 log PFU/mL (ANOVA, p < 0.05) after 12 h of incubation. When the phage phSE-5 was incubated in the presence of E. coli, a significant increase (1.7 log PFU/mL, ANOVA, p < 0.05) was observed after 10 h of incubation. BC increased by 2.8 log CFU/mL (ANOVA, p < 0.05, Figure 1A) during 12 h of incubation.
The phage controls (PC) remained constant during the 12 h of the experiments (ANOVA, p > 0.05, Figure 1B). For phage ELY-1 and the phage cocktail, the survival factor increased up to 0.5 log PFU/mL (ANOVA, p < 0.05) after 12 h of incubation. When the phage phSE-5 was incubated in the presence of E. coli, a significant increase (1.7 log PFU/mL, ANOVA, p < 0.05) was observed after 10 h of incubation.
The phage controls (PC) remained constant during the 12 h of the assay (ANOVA, p > 0.05, Figure 2B). In the case of the phages incubated in the presence of S. Typhimurium, a significant increase (ANOVA, p < 0.05) of 0.4 and 2.3 log PFU/mL was observed for ELY-1 and phSE-5, respectively, after 8 and 4 h of incubation ( Figure 2B). In the case of the phage cocktail incubated in the presence of the S. Typhimurium (BP), no significant difference was observed when compared with the phage control (ANOVA, p > 0.05, Figure 2B). 8 increased by 3.2 log CFU/mL (ANOVA, p < 0.05, Figure 2A) during 12 h of incubation.
The phage controls (PC) remained constant during the 12 h of the assay (ANOVA, p > 0.05, Figure 2B). In the case of the phages incubated in the presence of S. Typhimurium, a significant increase (ANOVA, p < 0.05) of 0.4 and 2.3 log PFU/mL was observed for ELY-1 and phSE-5, respectively, after 8 and 4 h of incubation ( Figure 2B). In the case of the phage cocktail incubated in the presence of the S. Typhimurium (BP), no significant difference was observed when compared with the phage control (ANOVA, p > 0.05, Figure 2B).

Effect of Phages ELY-1, phSE-5 and Cocktail ELY-1/phSE-5 on the Inactivation of the Mixture of E. coli/S. Typhimurium
The maximum inactivation of the mixture of E. coli/S. Typhimurium with the phages ELY-1, phSE-5, and cocktail phSE-5/ELY-1 was 2.2, 2.0, and 3.2 log CFU/mL (ANOVA, p < 0.05) achieved after 8, 10, and 10 h, respectively, when compared with those of the bacterial control (BC) ( Figure 3A). During treatment, the rate of inactivation of the phage cocktail was significantly higher than that obtained with phages ELY-1 and phSE-5. The bacterial inactivation of the phage cocktail started after 2 h and the rate of inactivation (1.2 log) was significantly higher (ANOVA, p < 0.05) than those of phages ELY-1 and phSE-5 (bacterial concentration similar to bacterial control). However, after 6 h of

Effect of Phages ELY-1, phSE-5 and Cocktail ELY-1/phSE-5 on the Inactivation of the Mixture of E. coli/S. Typhimurium
The maximum inactivation of the mixture of E. coli/S. Typhimurium with the phages ELY-1, phSE-5, and cocktail phSE-5/ELY-1 was 2.2, 2.0, and 3.2 log CFU/mL (ANOVA, p < 0.05) achieved after 8, 10, and 10 h, respectively, when compared with those of the bacterial control (BC) ( Figure 3A). During treatment, the rate of inactivation of the phage cocktail was significantly higher than that obtained with phages ELY-1 and phSE-5. The bacterial inactivation of the phage cocktail started after 2 h and the rate of inactivation (1.2 log) was significantly higher (ANOVA, p < 0.05) than those of phages ELY-1 and phSE-5 (bacterial concentration similar to bacterial control). However, after 6 h of treatment, the rate of inactivation of phage ELY-1 was significantly higher (2.1 log CFU/mL, ANOVA, p < 0.05) than that obtained with phage phSE-5 (0.8 log CFU/mL) ( Figure 3A) and significantly lower than that obtained with the phage cocktail (3.0 log CFU/mL). After 12 h of incubation, the bacterial inactivation was 2.7 log CFU/mL with the phage cocktail, which was statistically different (ANOVA, p < 0.05) from the values obtained in the treatment with phages ELY-1 and phSE-5 (1.6 and 2.0 log CFU/mL, respectively). Bacterial density in the BC increased by 3.0 log CFU/mL (ANOVA, p < 0.05, Figure 3A) during 12 h of incubation.
In the case of phages phSE-5, ELY-1, and the phage cocktail controls (PC) no decrease of the phage survival (ANOVA, p > 0.05) was observed during the 12 h of the experiments ( Figure 3B). When the phages ELY-1 and phSE-5 were incubated in the presence of the mixture of bacteria, a significant increase (ANOVA, p < 0.05) of 1.7 and 0.5 PFU/mL was observed, respectively, after 6 h of incubation ( Figure 3B). No significant differences were observed between the phage concentration and the phage control during the experiment with the phage cocktail (ANOVA, p > 0.05). 9 CFU/mL, respectively). Bacterial density in the BC increased by 3.0 log CFU/mL (ANOVA, p < 0.05, Figure 3A) during 12 h of incubation.
In the case of phages phSE-5, ELY-1, and the phage cocktail controls (PC) no decrease of the phage survival (ANOVA, p > 0.05) was observed during the 12 h of the experiments ( Figure 3B). When the phages ELY-1 and phSE-5 were incubated in the presence of the mixture of bacteria, a significant increase (ANOVA, p < 0.05) of 1.7 and 0.5 PFU/mL was observed, respectively, after 6 h of incubation ( Figure 3B). No significant differences were observed between the phage concentration and the phage control during the experiment with the phage cocktail (ANOVA, p > 0.05).

Determination of the Emergence Rate of Bacterial Mutants
E. coli, S. Typhimurium, and the mixture with the two bacteria showed different rates of resistant mutants when subjected to phages ELY-1, phSE-5, and the phage cocktail ELY-1/phSE-5 ( Table 2).

Determination of the Emergence Rate of Bacterial Mutants
E. coli, S. Typhimurium, and the mixture with the two bacteria showed different rates of resistant mutants when subjected to phages ELY-1, phSE-5, and the phage cocktail ELY-1/phSE-5 ( Table 2).

Fitness of Phage Resistant Mutants
In all experiments, in the presence of phages ELY-1, phSE-5, and the cocktail, differences were observed ( Figure 4A-C, ANOVA, p < 0.05) between the growth profiles of sensitive bacteria and resistant bacteria. Higher concentrations of resistant bacteria to the phages or to the phage cocktail than sensitive bacteria at 6 and 12 h were observed. In the absence of phages, no differences ( Figure 4A-C, ANOVA, p > 0.05) were found between the concentration of resistant bacteria, using the single phage suspensions or the cocktail, and the concentration of sensitive bacteria.   The results obtained by the OD measurements were in concordance with those obtained by the colony counts for all cases ( Figure 5A-C). The rate of bacterial growth of the two pure cultures and of the mixed culture of the two bacteria in the presence of single phage suspensions and of the cocktail was significantly different (ANOVA, p < 0.05) from that obtained with the sensitive and resistant bacteria in the absence of the phages and resistant bacteria with phages.

Discussion
Even though the major advantage of phage treatment is the phage specificity, since the nontarget bacterial populations should remain undisturbed, phages should be capable to lyse the majority of strains of a given bacterial species [64]. Moreover, phages with a broader host range, that affect pathogenic bacteria of other species or even of other genera, can also be useful when the involved bacterium is not yet identified. Consequently, before using phages to treat infections caused by Enterobacteriaceae, like E. coli and S. Typhimurium, which are by far most commonly isolated microorganism in the clinical laboratory, the in vitro dynamics of phage-host replication characterization is important to decide if phage cocktails including strains specific to these bacteria can be effective to control infections caused by E. coli or S. Typhimurium.

Discussion
Even though the major advantage of phage treatment is the phage specificity, since the non-target bacterial populations should remain undisturbed, phages should be capable to lyse the majority of strains of a given bacterial species [64]. Moreover, phages with a broader host range, that affect pathogenic bacteria of other species or even of other genera, can also be useful when the involved bacterium is not yet identified. Consequently, before using phages to treat infections caused by Enterobacteriaceae, like E. coli and S. Typhimurium, which are by far most commonly isolated microorganism in the clinical laboratory, the in vitro dynamics of phage-host replication characterization is important to decide if phage cocktails including strains specific to these bacteria can be effective to control infections caused by E. coli or S. Typhimurium.
The results of this study indicate that both phages, phSE-5 (produced on Salmonella) and ELY-1 (produced on E. coli), can control bacterial strains of E. coli and Salmonella, two closely related genera of the Enterobacteriaceae family. However, none of the two phages infected other bacterial strains of the Enterobacteriaceae genera, nor other strains belonging to other bacterial families. Phages ELY-1 and phSE-5 formed completely cleared zones (spot test results) on 9 and 25 of the 50 tested strains, respectively. However, the efficiency of plating (EOP) results indicate that phages ELY-1 and phSE-5, besides their hosts, only formed lysis plaques on 4 strains of the tested bacterial strains. Phage ELY-1, besides its host, also infected 3 E. coli strains (E. coli ATCC 25922, E. coli Scc 34, and E. coli Scc69) and 1 Salmonella strain (S. Typhimurium ATCC 13311). Phage phSE-5, besides its host, infect also four non-Salmonella strains (bioluminescent E. coli, E. coli AE11, E. coli Scc 34, and E. coli Scc 69), which belong to different bacteria genera, but belong to the family of the Enterobacteriaceae. Mirzaei and Nilsson (2015) obtained analogous results and stated that the selection of phages cannot be performed by spot test and should be replaced by the EOP assays [65]. Negative EOP and positive spot test outcomes can occur when an overload of phages simultaneously infects a bacterium, leading to lysis due to the presence of high concentrations of lysins ("lysis from without") [66], avoiding the replication of the phages by the bacteria [67][68][69][70]. Phages ELY-1 and phSE-5 respect the specificity criterion, so important in phage therapy, however, both phages have quite narrow host ranges. In the future, new phages need to be isolated and tested together with phages ELY-1 and phSE-5 to produce a cocktail with a broader spectrum of activity towards E. coli and S. Typhimurium Although ELY-1 and phSE-5 showed low efficiency of plating against S. Typhimurium and E. coli, respectively, the obtained rate of inactivation was high. The maximum inactivation of phage ELY-1 was 4.5 log CFU/mL for E. coli and 2.2 log CFU/ml for S. Typhimurium. When phage phSE-5 was used to inactivate S. Typhimurium, after 4 h of incubation, the rate of inactivation was higher (2.2 log CFU/mL) than that observed against E. coli. However, the maximum inactivation of phage phSE-5 was higher against E. coli than against Salmonella (~3.8 log CFU/mL reduction for E. coli and 2.6 log CFU/mL for S. Typhimurium), but its effect on E. coli starts later (only after 8 h of treatment) than that of phage ELY-1 (inactivation starts after 2 h of treatment).
The use of several phages in the form of cocktails increases their potential against pathogenic bacteria. Moreover, if more phages were included in the cocktail, the greater its potential would be for presumptive usage, i.e., prior to identification of the pathogens. However, as the number of phages in the cocktail increases, the possible impact on non-target bacteria is higher. Even so, like in most cases, this impact will be minimal when compared to what is expected from typical commercial antibiotics [49]. We also have to consider that too many phages per formulation can result in higher development and manufacturing costs and using multiple phages of the same genus with different host ranges in the cocktail may result in phage recombination and generate new host specificities [71].
The phage cocktail used in this study efficiently inactivated the two tested bacteria, either in mixtures or individually. The phage cocktail was more efficient to control S. Typhimurium and the mixture of bacteria than the single phage suspensions. Moreover, in both cases, the inactivation was observed sooner when the cocktail was used. When the phage cocktail was used to inactivate S. Typhimurium and the mixture of bacteria, after 2 h of treatment the rate of inactivation was higher (2 and 1.2 log CFU/mL, respectively) than the single phage suspensions (bacterial concentration similar to the bacterial control). These results are in accordance with other studies [50,[72][73][74] that achieved a faster and higher inactivation by using phage cocktails over single phage suspensions. However, the phage cocktail was no more effective to inactivate the E. coli than the single suspension of phage ELY-1, but the efficiency of E. coli inactivation by the cocktail was similar to that obtained with the single phage suspension. As the purpose of using the cocktail resides in its application to inactivate both bacteria, that is, to be used prior to the identification of the pathogen involved in an infection, this is not a negative aspect.
Phage therapy with phage cocktails, as well as with single-phage suspensions, did not prevent the bacterial regrowth after treatment. However, as stated before, for S. Typhimurium and for the mixture of bacteria, the use of the cocktail retarded the regrowth of bacteria, which is an important achievement. Moreover, the use of the phage cocktail limited the emergence of phage-resistant mutants in the mixture of the two bacteria. The development of resistant mutants in the mixture of bacteria against the phage cocktail ELY-1/phSE-5 (2.75 × 10 −4 ) was lower than those obtained with phages ELY-1 (4.66 × 10 −4 ) and phSE-5 (5.09 × 10 −4 ). However, when the phage phSE-5 was used alone against its host as well as against E. coli, the emergence of resistance was higher than when the phage ELY-1 was used alone, reaching values of 10 −3 CFU/mL. These results suggest that the presence of the phage phSE-5 in the phage cocktail increases the rate of resistant mutants against its host and E. coli. The frequency of emergence of phage resistant mutants using the phage ELY-1 against E. coli and against S. Typhimurium was significantly lower, 10 −5 CFU/mL, a value similar to those observed before in other studies [32,75].
Some authors have suggested that phage exposure could cost bacteria their fitness, which can lead to their removal from the environment at a faster rate than their wild-type parents [76,77]. In this study, the experiment results of the fitness of phage resistant mutants showed that the concentration of sensitive bacteria and resistant mutants, when grown in the absence of phages ELY-1, phSE-5, and the phage cocktail ELY-1/phSE-5, are not different. However, these experiments were done in nutrient rich medium (culture medium) and without the presence of competition, from which, according to some authors, the cost of resistance can vary across environmental factors and degree of competition for resources [78,79]. Further studies are necessary to evaluates the cost of resistance in vivo.
In the near future, it will be essential to understand the efficiency of these single phage suspensions and cocktails in vivo, using, for instance, G. mellonella larvae as an animal model. According to several authors, in vitro experiments are not satisfactory to distinguish phage-bacteria interactions in vivo [38,44,80,81]. Numerous factors can affect the survival and lytic properties of the phages in vivo, e.g., bacteriophages/target bacteria ratio, the way and moment of treatment, physical and chemical properties, the neutralization of phages, and accessibility to target bacteria [82].

Conclusions
The use of a phage cocktail against the Enterobacteriaceae increases their potential to be used prior to the identification of the pathogenic bacterium, retarding even the regrowth of bacteria. However, the development of phage-resistant mutants is not prevented. Nevertheless, further studies are needed to understand the true potential of the use of these two phages and of the cocktail phages to control E. coli and S. Typhimurium, namely in vivo studies using animal models.
Author Contributions: P.C. performed the experiments. C.P. and A.T.P.C.G. wrote the paper. A.A. supervised the work, revised the paper, and contributed with reagents and analysis tools.