**Quinolizidine-Derived Lucanthone and Amitriptyline Analogues Endowed with Potent Antileishmanial Activity**

**Michele Tonelli 1,\* , Anna Sparatore 2,\* , Nicoletta Basilico 3,\* , Loredana Cavicchini <sup>3</sup> , Silvia Parapini <sup>4</sup> , Bruno Tasso <sup>1</sup> , Erik Laurini <sup>5</sup> , Sabrina Pricl 5,6 , Vito Boido <sup>1</sup> and Fabio Sparatore <sup>1</sup>**


Received: 4 October 2020; Accepted: 21 October 2020; Published: 25 October 2020

**Abstract:** Leishmaniases are neglected diseases that are endemic in many tropical and sub-tropical Countries. Therapy is based on different classes of drugs which are burdened by severe side effects, occurrence of resistance and high costs, thereby creating the need for more efficacious, safer and inexpensive drugs. Herein, sixteen 9-thioxanthenone derivatives (lucanthone analogues) and four compounds embodying the diarylethene substructure of amitriptyline (amitriptyline analogues) were tested in vitro for activity against *Leishmania tropica* and *L. infantum* promastigotes. All compounds were characterized by the presence of a bulky quinolizidinylalkyl moiety. All compounds displayed activity against both species of *Leishmania* with IC<sup>50</sup> values in the low micromolar range, resulting in several fold more potency than miltefosine, comparable to that of lucanthone, and endowed with substantially lower cytotoxicity to Vero-76 cells, for the best of them. Thus, 4-amino-1-(quinolizidinylethyl)aminothioxanthen-9-one (**14**) and 9-(quinolizidinylmethylidene)fluorene (**17**), with selectivity index (SI) in the range 16–24, represent promising leads for the development of improved antileishmanial agents. These two compounds also exhibited comparable activity against intramacrophagic amastigotes of *L. infantum*. Docking studies have suggested that the inhibition of trypanothione reductase (TryR) may be at the basis (eventually besides other mechanisms) of the observed antileishmanial activity. Therefore, these investigated derivatives may deserve further structural improvements and more in-depth biological studies of their mechanisms of action in order to develop more efficient antiparasitic agents.

**Keywords:** *Leishmania tropica* and *infantum*; antileishmanial agents; lucanthone analogues; amitriptyline analogues; quinolizidine-derived compounds; molecular modelling studies

#### **1. Introduction**

Leishmaniases are neglected diseases that are endemic in many tropical and sub-tropical countries, leading annually to an estimated 700,000–1,000,000 new cases and 20,000–30,000 deaths [1]. Leishmaniases are caused by more than 20 species of protozoan parasites belonging to the genus *Leishmania*, which together with the genus *Trypanosoma,* belongs to the order *Trypanosomatidae*. Recent advances in the taxonomy, genetics, molecular and cellular biology and biochemistry of these organisms are well illustrated in two reviews [2,3]. The pathogen parasites (promastigotes) are transmitted to human and other mammalian hosts by the bites of infected female phlebotomine sandflies. In the mammalian host, the parasites differentiate into amastigote forms and affect skin, mucosa or internal tissues and organs to produce cutaneous (CL), muco-cutaneous (MC) and visceral (VL) leishmaniasis. The last form is fatal in absence of treatment. Current therapy is based on pentavalent antimonials; pentamidine; amphotericin B and its liposomal formulations used by parenteral route; and miltefosine, as the only oral agent. These drugs are burdened by heavy side effects, the occurrence of resistance and high costs, so that the need for novel, more efficacious, safe and inexpensive drugs is very stringent. Indeed, to meet this need, a number of studies are presently ongoing, exploring a very wide chemical space and also the possibility of the repositioning of known drugs. The structures of many interesting examples of investigational antileishmanial agents, able to hit different cellular targets, are illustrated in several reviews [4–9]. Interestingly, highly potent antileishmanial chemical classes, as the nitroimidazoles, the benzooxaboroles and the aminopyrazoles/pyrazolopyrimidines, have been identified, mainly within the Drugs for Neglected Diseases initiative (DNDi), Geneva, Switzerland. Orally active compounds from these series (such as DNDi-0690, DNDi-6148 and GSK-3186899) [10–12] are presently in clinical development (Figure 1).

**Figure 1.** Antileishmanial drugs in clinical development.

Additionally, many tricyclic antidepressant and antipsychotic drugs (and structurally related compounds) have been shown to display various degrees of activity against different species of *Leishmania*, and/or to inhibit enzymes (such as trypanothione reductase) playing essential roles in parasite development and virulence [13–16]. Among these, clomipramine and cyclobenzaprine have been recently repurposed for the treatment of visceral leishmaniasis [15,16]. All these compounds are characterized by different kinds of linear tricyclic systems, such as phenothiazine, iminodibenzyl, thioxanthene, dibenzocycloheptane and sulphur isosteric analogues, to which an aliphatic basic side chain is attached to the central ring (Figure 2A).

**Figure 2.** (**A**) Examples of basic derivatives of tricyclic systems displaying antileishmanial activity and inhibition of trypanothione reductase. (**B**) Basic derivatives of thioxanthene-9-one (lucanthone analogues) displaying antileishmanial activity. (**C**) Lucanthone analogues displaying dual inhibition of P-glycoprotein and cancer cell growth.

Thioxanthen-9-one is another tricyclic system whose derivatives exhibited antileishmanial activity, even featuring the basic side chain linked to a lateral ring (Figure 2B). Indeed, lucanthone, a drug largely used in the past for the treatment of schistosomiasis and presently (together with various analogues) under investigation as an antitumor agent, was shown to display activity against *L. major amazonensis* in tissue cultures a long time ago (1975) [17]. Although it was less powerful, this activity was also confirmed in vivo [18]. A few years later, this compound was also found to be active against intramacrophagic *L. tropica* amastigotes with IC<sup>50</sup> = 0.93 µg/mL (2.47 µM) [19]. More recently (2011), through the screening of more than 4000 compounds, using a novel ex vivo splenic explant model system, an 1-(piperidinoethylamino)-analog of hycanthone (the active metabolite of lucanthone) was identified as a lead compound against *L. donovani*, with IC<sup>50</sup> = 9.1 µM [20]. Hycanthone and its prodrug lucanthone are burdened with hepatic toxicity and mutagenicity [21–23], which were related to the presence or formation of a 4-hydroxymethyl group capable of alkylating the deoxyguanosine residue of DNA passing through the formation of a strongly electrophilic carbocation. Modification of the basic side chain and/or the introduction of substituents in position 6 of the thioxanthenone ring were shown to alter the mutagenicity, while retaining appreciable anti-schistosomal activity [24,25]. Even more, the replacement of the 4-methyl and 4-hydroxymethyl groups with other functional moieties limited the toxicological issues, as observed for the 4-propoxy analog of hycanthone (TXAI), investigated as

an antitumor agent (Figure 2C) [26]. These data suggest that genetic and anti-parasitic activities of thioxanthenone derivatives may be dissociated from each other.

#### **2. Results and Discussion**

Thus, when pursuing our investigation on antiplasmodial [27–29] and antileishmanial [30,31] agents we deemed worthwhile the study of the antileishmanial activity of a set of lucanthone analogues with a modified substitution pattern, and of a few compounds embodying the diarylethene substructure of, e.g., chlorprothixene, amitriptyline and cyclobenzaprine. The studied compounds were characterized by the presence of the bulky quinolizidinylalkyl moieties, which were shown to improve the antiplasmodial and/or antileishmanial activity in the corresponding chloroquine and clofazimine analogs [30] and also in the set of 1-basic substituted 2-phenyl/benzyl benzimidazoles [31]. In position 4 of the thioxanthen-9-one nucleus, besides the methyl, a nitro or amino group was introduced, which were shown to improve the inhibitory activity in another field (against the lymphocytic leukemia P388 [32]). The considered compounds were obtained from our in-house library, having been synthesized and studied in the past as antimicrobial and anti-leukemia P388 agents [33,34]; as modulators of uptake and release of neurotransmitters [35–37]; and more recently, as dual inhibitors of cholinesterases and Aβ aggregation [38,39].

The structures of the presently investigated compounds are depicted in Figures 3 and 4.

**Figure 3.** Investigated thioxanthen-9-one derivatives (lucanthone analogues) bearing quinolizidine-alkyl side chains.

**Figure 4.** Investigated lupinylidene (quinolizidinyl-methylidene) derivatives of planar and corrugated tricyclic systems.

#### *2.1. Chemistry*

With the exception of **4** and **20**, all the compounds of Figures 3 and 4 were previously described according to the following references: **1**–**3**, **5**–**10** and **13**–**15** [34]; **11**, **12** and **16** [38]; **17** and **18** [37]; **19** [33]. The novel compounds **4** and **20** were prepared by treating compounds **1** and **18** with methyl iodide (Scheme 1).

**Scheme 1.** Reagents and conditions: (a) CH<sup>3</sup> I (excess), 24 h, r.t.

#### *2.2. Biological Studies and SAR*

#### 2.2.1. Antileishmanial Activity Against *L. tropica* and *L. infantum* Promastigotes

Compounds in Figures 3 and 4 were tested in vitro against *Leishmania tropica* and *L. infantum* promastigotes using the MTT assay. Results are expressed as IC<sup>50</sup> ± SD (µM) and reported in Table 1, together with the ratios between the IC<sup>50</sup> of the reference drug (miltefosine) and that of each tested compound. All tested compounds displayed activity against both species of *Leishmania* and many of them exhibited IC<sup>50</sup> less than 10 µM (65% and 50% versus *L. tropica* and *L. infantum*, respectively). Commonly, activity was higher for *L. tropica* than *L. infantum*. In comparison to miltefosine, all compounds were several–fold more potent, up to 17-fold against *L. tropica* and up to 9-fold against *L. infantum*. These results indicate that the introduction of a quinolizidinyl alkyl moiety on

all the considered tricyclic systems is consistent with the expression of valuable antileishmanial activity, provided that, in the case of thioxanthenone, suitable substituents are present on position 4 and 7.


**Table 1.** In vitro data on antileishmanial activity against *Leishmania tropica* and *L. infantum* promastigotes of lucanthone and compounds **1**–**20**.

<sup>a</sup> The results are expressed as IC<sup>50</sup> ± SD of at least three different experiments performed in duplicate. <sup>b</sup> Ratios between the IC<sup>50</sup> of miltefosine and that of each compound against *L. tropica* or *L. infantum*.

Among the thioxanthenone derivatives, compounds **9** and **13**–**15** exhibited potency (IC<sup>50</sup> in the range 2.87–9.39 µM) comparable to that of lucanthone and 1-(piperidinoethyl)-4 hydroxymethylthioxanthen-9-one (the novel lead compound of Osorio et al. [20]), but differently from two of the others, they should not be associated with mutagenic activity because they lack groups potentially leading to alkylating species.

The elongation of the polymethylene linker exerted different effects, depending on the nature of the substituents, in position 4. Among the 4-methyl derivatives, the activity decreased with the increasing number of methylene groups in the side chain (IC50: from 3.36 to 8.89 µM and from 3.87 to 17.19 µM for the two *Leishmania* species), while among the 4-amino derivatives the activity remained practically unchanged (IC<sup>50</sup> in the range 3–5 µM). In the set of 4-nitro derivatives, the activity varied for both species in an unaccountable way, with the worst IC<sup>50</sup> values being for *n* = 1 (19.24 for *L. tropica* and 38.72 µM for *L. infantum*) and the best one being for *n* = 2 (IC50= 4.64 and 9.39 µM for *L. tropica* and *L. infantum*, respectively, Table 1). In other words, for a given number of CH<sup>2</sup> in the side chain, the substitution of the 4-methyl with the 4-amino group, besides eliminating the risk of mutagenicity, always improved the activity of the relevant derivatives, while the presence of a 4-nitro moiety produced either a slight increase (*n* = 2) or a decrease (*n* = 1 and 3) of activity. In particular, the decrement of activity was very pronounced for *n* = 1, with IC50= 19.24 and 38.72 µM against *L. tropica* and *L. infantum*, respectively. Additionally, the introduction of a methoxy group in position 7 had variable effects on the activity, which was enhanced in the case of 4-nitro derivatives and decreased for the 4-methyl derivatives. The decreasing effects was striking when *n* = 2 (compound **6**), with IC<sup>50</sup> = 28.35 µM and >46 µM for *L. tropica* and *L. infantum*, respectively. Thus compound **6** displayed the lowest activity among all

tested compounds, although it was by far the least toxic (CC<sup>50</sup> = 137 µM) on rat skeletal myoblast cells (L6), with a still valuable selectivity index (SI = 4.8 and ≤3) for the two *Leishmania* species (Reto Brun, personal communication to A.S.).

When comparing compounds **1** and **16**, it was observed that the antileishmanial activity was maintained with only modest loss of potency, indicating that even the exchange of the sulfur bridge with an oxygen atom did not modify significantly the thioxanthenone physico-chemical interactions with the *Leishmania* target(s), while it was able to abolish the antileukemic activity of lucanthone, as observed by Blanz and French [40]. On the contrary, the quaternarization of compounds **1** to give **4** produced a strong reduction of the antileishmanial activity, suggesting that the presence of a fixed charge, while hampering the usual target interactions, might shift the molecule towards a different cellular target. Supposing that the antileishmanial activity of the thioxanthenone derivatives could be related to one or some of the mechanisms previously described for the anti-schistosomal and anti-tumor activities of lucanthone (DNA intercalation, inhibition of nucleic acid biosynthesis, inhibition of topoisomerase II and apurinic endonuclease-1 [41], induction of autophagy and apoptosis [42], alteration of cholesterol biosynthesis and localization [43]), the quaternization of the quinolizidine nitrogen of **1** might produce the shift from some of these mechanisms to the inhibition of choline uptake, as it is known for common quaternary ammonium surfactants [44,45] and for the peculiar ammonium salts bearing bulky moieties on the charged head or on the lipophilic tail [46,47].

It is worth noting that compound **1** was shown in the past [33] to exhibit a large spectrum of antibacterial activity with MIC in the same micromolar range of IC<sup>50</sup> against *Leishmania* species; f.i. MIC: 2.5 µg/mL = 6.63 µM against *Staphylococcus aureus* and *Bacillus subtilis*, and 1.25 µg/mL = 3.32 µM against *Mycobacterium tuberculosis*. Lucanthone also displayed antibacterial activity but with MIC 10-fold higher (30–60 µM) [48,49].

The four *lupinylidene derivatives* **17**–**20** embody the 1,1-diarylethene substructure that characterizes amitriptyline, cyclobenzaprime, chlorprothixene and other antidepressant and antipsychotic drugs. Like most of these drugs, compounds **17**–**20** exhibited antileishmanial activity in the low micromolar range; in particular, the 9-lupinylidene fluorene **17** (with IC<sup>50</sup> = 3.5 and 4.35 µM against *L. tropica* and *L. infantum*, respectively) was 6/7-fold more potent than the lupinylidene dibenzocycloheptadiene **18** (IC<sup>50</sup> = 24.25 and 25.21 µM), possibly for being endowed with a more appropriate lipophilicity. The quaternization of these compounds produced a levelling effect on activity, decreasing the potency of the former and increasing that of the latter, so that the corresponding methyl iodides **19** and **20** were almost equipotent.

The antileishmanial activity of the abovementioned drugs has been shown to be related to the inhibition of trypanothione reductase (TryR) in the parasite [50–52], but also to modulation of the immune response in the host. It is reasonable to suppose that also the tertiary compounds **17** and **18** act through inhibition of TryR; however, in the quaternized compounds **19** and **20**, other mechanisms might modulate or replace the former, as discussed above for quaternized thioxanthenone **4**. Interestingly, the tertiary amitriptyline analog **18** (compound **17** was not tested) was shown in the past [35] to inhibit the choline uptake into rat brain synaptosomes at 1 µM concentration, at which amitriptyline was still completely ineffective. The inhibition of choline transport across the synaptosome membrane might be expected in some measure also at the level of the parasite cell membrane in an interplay with the intracellular TryR inhibition. With the quaternization of **18** to **20**, cell penetration and consequent TryR inhibition could be strongly reduced, but the basal activity on choline transport could be improved.

Lupinylidene fluorene **17** was not investigated as an inhibitor of choline uptake, but it was shown to display a large spectrum of anti-microbial activity with outstanding potency against *Mycobacterium tuberculosis H37Ra* (MIC = 0.49 µM) very close to that of isoniazide (MIC: 0.14–0.28 µM) [33]. The quaternization of **17** to **19** strongly affected the antimycobacterial activity (MIC = 83.9 µM) quite probably by reducing the cell wall crossing capability.

Even if the definition of the mechanism of the antileishmanial activity of the tested compounds is beyond the scope of this exploratory work, one cannot overlook the previous observation that lucanthone [53], the quinolizidinylalkylamino thioxanthenones [38] and the lupinylidene dibenzocycloheptadiene [39] inhibit AChE, and particularly BChE, with IC<sup>50</sup> in the low micromolar and sub-micromolar range. The availability of choline for building up the phosphatidylcholine, the main component of *Leishmania* promastigote membranes [54,55], might be compromised by the inhibition of cholinesterases. Cholinesterases are known to be present even in non-motile unicellular organisms, where besides or instead of the hydrolytic function, they may play non-classical roles that are fundamental for cell survival [56–58]. The inhibition of cholinesterases has been recently claimed as another mechanism of action for some antileishmanial agents extracted from several plants [59–61], and the present results add further support to this hypothesis.

#### 2.2.2. Cytotoxicity

To identify new potential candidates for the development of safe and effective antileishmanial drugs, the cytotoxicities of representative (most effective or structurally peculiar) compounds (**1**, **2**, **14**, **17**, **19** and **20**) were evaluated in a Vero-76 cell line. The cytotoxicity of lucanthone was also tested for comparison. The results in Table 2 show that all compounds exhibited CC<sup>50</sup> higher than their corresponding IC<sup>50</sup> values. The relevant selectivity index (SI)—a parameter that quantifies the preferential antileishmanial activity of a compound in relation to mammalian cell toxicity (CC50/IC50)—is tabulated in Table 2. As can be seen from this Table, the 4-methyl-thioxanthen-9-ones (**1**, **2**, and lucanthone) exhibited, as expected, the highest cytotoxicity among the tested compounds. It is worth noting that compounds **1** and **2** were somewhat less toxic than lucanthone, suggesting a possible toxicity-lowering effect of the cumbersome quinolizidinylalkyl side chain in comparison to an open chain substituent. The replacement of the 4-methyl substituent with an amino group in compound **14,** while preserving good antileishmanial activity, led to a further substantial decrease of cytotoxicity, with a resultantly safer profile (SI = 16.2 and 16.9 for the two *Leishmania* species).


**Table 2.** In vitro cytotoxicity data against Vero-76 cells and selectivity index (SI) values for selected antileishmanial compounds of Table 1.

<sup>a</sup> Compound concentration (µM) required to reduce the viability of mock-infected VERO-76 (monkey normal kidney) monolayers by 50%. The results are expressed as CC<sup>50</sup> ± SD of three different experiments performed in duplicate. <sup>b</sup> See Table 1. <sup>c</sup> The selectivity index (SI) is expressed as the ratio between the CC<sup>50</sup> value of each compound against Vero-76 cell line and the IC<sup>50</sup> of each compound against *L. tropica* or *L. infantum* promastigotes.

On the other hand, the lupinylidene tricyclic derivatives **17**, **19** and **20** (the amitriptyline analogues, broadly speaking) displayed the lowest cytotoxic effects, with CC<sup>50</sup> values in the range 80–90 µM, in the presence of either tertiary or quaternized nitrogen. However, taking into account the effects of quaternization on the activity, only the tertiary 9-lupinylidene fluorene **17** displayed a quite valuable SI value for both *Leishmania* species (23.9 and 16.2, respectively), resulting the most promising antileishmanial candidate among the whole set of studied compounds.

The cytotoxic concentration (CC50) against THP-1 differentiated into macrophages (Table 3), used for testing the activity against the amastigote stage, is also reported. Interestingly, compounds **1**, **14**, **17** and lucanthone showed a SI trend against THP-1 cells (Table 3) comparable to that against Vero-76 cells.


**Table 3.** In vitro antileishmanial activity against intramacrophagic amastigotes of *L. infantum* and cytotoxicity against PMA-differentiated THP-1 (human acute monocytic leukemia cell line).

<sup>a</sup> Data are expressed as mean ± SD of three experiments in triplicate. <sup>b</sup> Compound concentration (µM) required to reduce the viability of PMA-differentiated THP-1 by 50%. <sup>c</sup> The selectivity index (SI) is expressed as the ratio between the CC<sup>50</sup> value of each compound against THP-1 cell line and the IC<sup>50</sup> against *L. infantum* amastigotes.

It is additionally observed that, based on the results of their previous testing for antileukemic activity [34], only a low to moderate level of in vivo toxicity should be expected. Indeed, no mortality was observed even when compounds **1**, **5**, **6**, **9** and particularly the amino derivatives **13**–**15** were injected i.p. at doses up to 200–350 mg/kg, once a day for five consecutive days, in a group of six mice previously inoculated with leukemia P388 cells. Particularly, for compound **14** no mortality was observed at a dose of 260 mg/kg (638 µmol/kg), and therefore this compound represents an interesting lead for improved lucanthone analogues.

#### 2.2.3. Antileishmanial Activity against L. Infantum Amastigotes

Finally, for a better idea of their real value as antileishmanial agents, compounds **14** and **17**, displaying the most promising activity against promastigote stage and the highest SI values, together with lucanthone and the corresponding quinolizidine analog **1**, were tested against the intramacrophagic amastigote stage of *L. infantum*. The results indicate that the activity observed against *L. infantum* promastigotes is conserved (lucanthone and compound **1**) or even improved (**14** and **17**) against the corresponding amastigote stage (Table 3).

#### *2.3. Molecular Modelling Studies*

Amitriptyline and other related compounds have been identified as plausible inhibitors of the trypanothione reductase (TryR), an essential enzyme belonging to the antioxidant machinery of parasitic *Leishmania* [62,63]. TryR is a homodimer and it is active only in this aggregated form. Given some structural properties in common, we reasoned that the most potent compounds of this series, **14** and **17**, can exert their antileishmanial properties (mainly or besides other mechanisms) by efficiently binding TryR and consequently blocking its activity. In order to test our hypothesis, molecular dynamics (MD) simulations were performed on the corresponding complexes with TryR to shed light on the binding mechanisms of these compounds against their putative parasitic target (Figure 5). To validate our procedure and for comparison purposes, we applied the same computational procedure also to TryR in complex with lucanthone (Figure 5) and amitriptyline (Figure S1).

A putative binding site for these compounds was initially recognized on TryR following a consolidated protocol [64–66]. Through the MM/PBSA (molecular mechanics/Poisson–Boltzmann surface area) approach [67], we calculated each inhibitor/enzyme free energy of binding (∆Gbind) and its enthalpic and entropic components (∆Hbind and -T∆Sbind, respectively). The obtained values are in good agreement with their antileishmanial activity (Figure 5D–F and Table S1) yielding the following TryR affinity ranking: **14** (∆Gbind = −8.73 kcal/mol) < **17** (∆Gbind = −8.54 kcal/mol) <= lucanthone (∆Gbind = −8.45 kcal/mol) << amitriptyline (∆Gbind = −7.46 kcal/mol). Interestingly, all compounds share a common thermodynamics pattern; actually, their binding is robustly enthalpy driven characterized by favorable electrostatic and van der Waals interactions. On the other hand, the entropic components penalize the binding, as often detected in cases of small molecule/protein complexes. The precise binding mechanism and the specific ligand/protein interactions were elucidated

through the per-residue binding free energy deconvolution (PRBFED) of the enthalpic terms (∆Hres). The PRBFED analysis allowed us to identify the main aminoacid residues of TryR involved in the putative binding pocket (Figure 5D–F, Figure S1 and Table S2).

′ ′ − **Figure 5.** (top panel) Details of compounds **14** (**A**), **17** (**B**) and lucanthone (**C**) in the binding pocket of TryR. Compounds are shown as atom-colored sticks-and-balls (C, grey, N, blue, O, red). The side chains of the mainly interacting TryR residues are depicted as colored sticks and labeled as following: E466′ and T470′ , firebrick; E18, W21, I339, N340 and A343, gold; L17, I106 and Y110, goldenrod. The hydrophobic pockets are also highlighted by their transparent van der Waals surface. Hydrogen atoms, water molecules, ions and counterions are omitted for clarity. (bottom panel) Calculated free energy of binding (∆Gbind, forest green), and enthalpic (∆Hbind, lime green) and entropic (−T∆Sbind, chartreuse) components (upper row) and PRBFED of the main involved amino acids (bottom row) of TryR in complex with **14** (**D**), **17** (**E**) and lucanthone (**F**).

− − − − The most peculiar interaction is definitively performed by the charged nitrogen atom present in all four of the compounds. Indeed, this protonated tertiary amine group is involved in a virtuous, interactive triangle with the side chain of the second monomer of TryR residues E466′ and T470′ through stable hydrogen bonds and a salt bridge. The length and the rigidity of the spacer between the nitrogen atom and the tricyclic moiety of the inhibitor can affect the efficiency of these interactions. In effect, the ylidenepropyl-amino spacer of amitriptyline is not able to provide the optimal addressing of the charged group towards the side chain of E466′ and T470′ with respect to the bulkiest quinolizidine moiety of **14** and **17** or the most flexible amino-ethyl spacer of lucanthone; accordingly, its ∆Hres values resulted the less favorable of the series (Figure 5D–F, Figure S1, and Table S2). On the other hand, it is already established that the dibenzocycloheptene ring of the amitriptyline can be aptly accommodated in the so-called TryR hydrophobic wall formed by residues L17, W21 and Y110 [62,63]. Our MD

simulation confirmed this data (Figure S1) and our computational analysis allowed us to find other important TryR residues to improve the van der Waals interactions in this specific hydrophobic region (Figure S1). As listed in Table S2, the side chains of residues E18, I106, I339, N343 and A343 also had contributions to stabilizing the binding with the enzyme. It is worth to mention here that also the tricyclic scaffold of the other compounds can be encased in the same protein region with even better performance (Figure 5D–F, Table S2). In the case of compound **17**, this could be expected since the fluorene ring is very similar to the dibenzocycloheptene moiety, but for the thioxanthenone derivatives **14** and lucanthone this might not have been so obvious. Instead, our MD approach showed that both compounds can share a very similar TryR binding mode with the lupinylidene and amitriptyline derivatives. Finally, compound **14** was the best TryR binder and the amino substitution (-NH2) in position 4 of the thioxanthenone ring of 14 played an important role. Actually, this amino group can establish a further polar interaction with the side chain of E18 leading to a slight yet significative improvement of its binding capability, as shown in Figure 5D and Table S2.

#### **3. Materials and Methods**

#### *3.1. Chemistry*

#### 3.1.1. General Information

Chemicals, solvents and reagents used for the syntheses were purchased from Sigma-Aldrich or Alfa Aesar (Milan, Italy), and were used without any further purification. Melting points (uncorrected) were determined with a Büchi apparatus (Milan, Italy). <sup>1</sup>H NMR and <sup>13</sup>C NMR spectra were recorded with a Varian Gemini-200 spectrometer in CDCl3; the chemical shifts were expressed in ppm (δ), coupling constants (J) in Hertz (Hz). Elemental analyses were performed on a Flash 2000 CHNS (Thermo Scientific, Milan Italy) instrument in the Microanalysis Laboratory of the Department of Pharmacy, University of Genova. Q = quinolizidine ring; Ar= aromatic.

#### 3.1.2. General Procedure for the Synthesis of Quaternary Ammonium Iodides (**4** and **20**)

The compounds **1** [34] and **18** [37] (0.146 mmol) were reacted with iodomethane (0.5 mL, 8 mmol) at r.t. for 24 h with stirring. The reaction mixture was added with dry Et2O, and the collected compound was washed with dry Et2O affording the title quaternary ammonium salt.

(1*S*,9a*R*)-5-methyl-1-{[(4-methyl-9-oxo-9*H*-thioxanthen-1-yl)amino]methyl}-decahydroquinolizin-5 -ium iodide (**4**): orange crystals; yield: 92%. m.p. 147–150 ◦C (Et2O an.); <sup>1</sup>H NMR (200MHz, CDCl3): δ = 9.36 (s, NH–Ar, collapses with D2O); 8.40 (d, *J* = 9.2, 1 ArH), 7.43–7.20 (m, 4 ArH), 6.72 (d, *J* = 9.2, 1 ArH), 4.21–3.00 (m, 8H of Q and 3.51, s, CH3-N of Q), 2.95–2.83 (m, 1H, of Q), 2.40–1.40 (m, 9H of Q and 2.25, s, CH3-Ar superimposed); <sup>13</sup>C NMR (50 MHz, CDCl3): 182.4, 149.4, 137.1, 135.1, 134.7, 130.9, 128.6, 128.0, 125.1, 124.4, 119.7, 112.3, 107.3, 67.9, 65.0, 51.5, 49.8, 44.6, 31.7, 20.4, 19.7, 19.0, 18.7, 18.0. Anal. calcd for C25H31IN2OS: C 56.18, H 5.85, N 5.24, S 6.00, found: C 56.09, H 6.15, N 5.15, S 6.39.

(1*S*,9a*R*)-1-[(10,11-dihydro-5*H*-dibenzo[a,d][7]annulen-5-ylidene)methyl]-5-methyl-decahydroquinolizin -5-ium iodide (**20**): pale yellow crystals; yield: 98%. m.p. 283–286 ◦C (Et2O an.); <sup>1</sup>H NMR (200MHz, CDCl3): δ = 7.50–6.97 (m, 8 ArH), 6.58–6.40 (m, 1H, HC=), 3.58–3.20 (m, 2Hα near N of Q and 3.36, s, CH3-N of Q), 3.17–2.63 (m, 4H, 2CH2-Ar), 2.36–1.00 (m, 14H of Q); <sup>13</sup>C NMR (50 MHz, CDCl3): 139.8, 138.7, 138.6, 135.6, 135.4, 129.1, 128.9, 128.5, 128.1, 126.7, 126.6, 124.8, 124.6, 64.7, 64.4, 56.1, 38.0, 37.0, 32.9, 30.8, 29.7, 24.4, 23.7, 23.5, 20.5, 20.1. Anal. calcd for C26H32IN: C, 64.34; H, 6.64; N 2.89. Found: C, 64.27; H, 7.00; N 2.70.

#### *3.2. Biological Tests*

#### 3.2.1. Antileishmanial Activity

Promastigote stage of *L. infantum* strain MHOM/TN/80/IPT1 (kindly provided by Dr. M. Gramiccia, ISS, Roma) and *L. tropica* (MHOM/SY/2012/ISS3130) were cultured in RPMI 1640 medium (EuroClone)

supplemented with 10% heat-inactivated fetal calf serum (EuroClone, Milan Italy), 20 mM Hepes and 2 mM *L*-glutamine at 24 ◦C. To estimate the 50% inhibitory concentration (IC50), the MTT (3-[4.5-dimethylthiazol-2-yl]-2.5-diphenyltetrazolium bromide) method was used [68,69]. Compounds were dissolved in DMSO and then diluted with medium to achieve the required concentrations. Drugs were placed in 96 wells round-bottom microplates and seven serial dilutions made. Miltefosine was used as reference anti-*Leishmania* drug. Parasites were diluted in complete medium to 5 × 10<sup>6</sup> parasites/mL and 100 µL of the suspension was seeded into the plates, incubated at 24 ◦C for 72 h and then 20 µL of MTT solution (5 mg/mL) was added into each well for 3 h. The plates were then centrifuged, the supernatants were discarded and the resulting pellets were dissolved in 100 µL of lysing buffer consisting of 20% (*w*/*v*) of a solution of SDS (Sigma), 40% DMF (Merck, Milan Italy) in H2O. The absorbance was measured spectrophotometrically at a test wavelength of 550 nm and a reference wavelength of 650 nm. The results are all expressed as IC50, which is the dose of compound necessary to inhibit parasite growth by 50%; each IC<sup>50</sup> value is the mean of separate experiments performed in duplicate.

#### 3.2.2. In Vitro Intracellular Amastigote Susceptibility Assays

THP-1 cells (human acute monocytic leukemia) were maintained in RPMI supplemented with 10% FBS (EuroClone), 50 µM 2-mercaptoethanol, 20 mM Hepes and 2 mM glutamine, at 37 ◦C in 5% CO2. For *Leishmania* infections, THP-1 cells were plated at 5 × 10<sup>5</sup> cells/mL in 16-chamber Lab-Tek culture slides (Nunc, Milan, Italy) and treated with 0.1 µM phorbol myristate acetate (PMA, Sigma) for 48 h to achieve differentiation into macrophages. Cells were washed and infected with metacyclic *L. infantum* promastigotes at a macrophage/promastigote ratio of 1/10 for 24 h. Cell monolayers were then washed and incubated in the presence of test compounds for 72 h. Slides were fixed with methanol and stained with Giemsa. The percentages of infected macrophages among treated and non-treated cells were determined by light microscopy [70].

#### 3.2.3. Cell Cytotoxicity Assays

THP-1 cells were plated at 5 × 10<sup>5</sup> cells/mL in 96 wells flat bottom microplates and treated with 0.1 µM PMA for 48 h to achieve differentiation into macrophages. Cells were then treated with serial dilutions of test compounds and cell proliferation evaluated using the MTT assay described for promastigotes. The results are expressed as CC50, which is the dose of compound necessary to inhibit cell growth by 50%.

Vero-76 cells (ATCC CRL 1587 *Cercopithecus Aethiops*) were seeded at an initial density of 4 × 10<sup>5</sup> cells/mL in 24-well plates, in culture medium (Dulbecco's modified eagle's medium (D-MEM) with L-glutamine, supplemented with fetal bovine serum (FBS), 0.025 g/L kanamycin). Cell cultures were then incubated at 37 ◦C in a humidified, 5% CO<sup>2</sup> atmosphere in the absence or presence of serial dilutions of test compounds. Cell viability was determined after 48–96 h at 37 ◦C by the Crystal violet staining method. The results are expressed as CC50, which is the concentration of compound necessary to inhibit cell growth by 50%. Each CC<sup>50</sup> value is the mean and standard deviation of at least three separate experiments performed in duplicate.

#### *3.3. Computational Methods*

The 3D structure of the TryR from *Leishmania infantum* was obtained starting from the available Protein Data Bank file (pdb code: 2JK6 [71]) and optimized following a procedure previously described [64–66]. The optimized structures of the new tested compounds were docked into the putative binding pocket using Autodock 4.2.6/Autodock Tools1.4.61 [72]. The resulting complex was further energy minimized to convergence. The intermolecular complex was then solvated by a cubic box of TIP3P water molecules [73] and energy was minimized using a combination of molecular dynamics (MD) techniques [64–66]. Ten nanosecond molecular dynamics (MD) simulations at 298 K were then employed for system equilibration, and further, 50 ns MD simulations were run for data production. Following the MM/PBSA approach [67] each binding free energy value (∆Gbind) was

calculated as the sum of the electrostatic, van der Waals, polar solvation, nonpolar solvation, (∆Hbind) and entropic contributions (T∆Sbind). The PRBFED analysis was carried out using the molecular mechanics/generalized Boltzmann surface area (MM/GBSA) approach [74] and was based on the same snapshots used in the binding free energy calculation. All simulations were carried out using the pmemd and pmemd.CUDA modules of Amber 18 [75], running on our own CPU/GPU calculation cluster. Molecular graphics images were produced using the UCSF Chimera package (v.1.14) [76]. All other graphs were obtained using GraphPad Prism (v. 6.0, GraphPad, La Jolla, CA, USA).

#### **4. Conclusions**

Sixteen 9-thioxanthenone derivatives (lucanthone analogues) and four compounds embodying the diarylethene substructure of amitriptyline (amitriptyline analogues) were tested in vitro for activity against *Leishmania tropica* and *L. infantum* promastigotes, and in a few cases also against intramacrophagic amastigotes of *L. infantum*. All compounds were characterized by the presence of a bulky quinolizidinylalkyl moiety, while differing for the tricyclic system to which the basic chain was connected. All compounds displayed activity against both species of *Leishmania* and most of them exhibited IC<sup>50</sup> values lower than 10 µM, and were many-fold more potent than miltefosine. The six best compounds (**1**, **9**, **13**–**15** and **17**) displayed potency comparable to that of lucanthone (IC<sup>50</sup> = 2.5 and 3.5 µM for the two *Leishmania* species), but their cytotoxicity versus the Vero 76 cells was always lower, with significant improvement of SI from 4.8–3.5 (lucanthone) to 16.2–16.9 and 23.6–19.2 for compounds **14** and **17**, respectively. These compounds exhibited comparable activity (and selectivity against THP-1 cells) against intramacrophagic amastigotes of *L. infantum*, and thus represent promising, structurally distinct leads for the development of improved antileishmanial agents. Docking studies suggest that the antileishmanial activity of compounds **14** and **17** may be related to the inhibition of trypanothione reductase, as is the case for other tricyclic compounds. However, lucanthone and the tested compounds were previously shown to potently inhibit the AChE, and particularly the BChE; thus, it is now put forward that this inhibitory property may have a notable additional role in the antiparasitic mechanism, either by reducing the availability of choline to build up the main component of promastigote membrane, or by inhibiting other non-classical functions of cholinesterases in the unicellular organisms.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/1424-8247/13/11/339/s1. Table S1: In silico binding thermodynamics of compounds **14**, **17**, lucanthone and amitriptyline towards TryR. Table S2: Per-residue binding enthalpy decomposition (∆Hres) for compounds **14**, **17**, lucanthone and amitriptyline towards TryR, Figure S1: Amitriptyline in the binding pocket of TryR.

**Author Contributions:** Conceptualization, F.S.; methodology and validation, M.T., A.S. and N.B.; software, E.L., S.P. (Sabrina Pricl); S.P. (Silvia Parapini), L.C., B.T., E.L. and V.B. participated to methodology; investigation, F.S., M.T., A.S., N.B. and S.P. (Sabrina Pricl); resources, M.T., A.S., N.B. and S.P. (Sabrina Pricl); writing—original draft preparation, F.S. and M.T.; writing—review and editing, F.S., M.T., A.S., N.B. and S.P. (Sabrina Pricl); visualization, M.T.; supervision, M.T., A.S. and N.B.; project administration, F.S., M.T., A.S. and N.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** This work was financially supported by the University of Genoa. The Authors thank O. Gagliardo for performing elemental analyses.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

### **Antibacterial Drug-Release Polydimethylsiloxane Coating for 3D-Printing Dental Polymer: Surface Alterations and Antimicrobial E**ff**ects**

**Hang-Nga Mai 1,**† **, Dong Choon Hyun 2,**† **, Ju Hayng Park <sup>2</sup> , Do-Yeon Kim <sup>3</sup> , Sang Min Lee <sup>3</sup> and Du-Hyeong Lee 1,4,\***


Received: 4 September 2020; Accepted: 10 October 2020; Published: 12 October 2020

**Abstract:** Polymers are the most commonly used material for three-dimensional (3D) printing in dentistry; however, the high porosity and water absorptiveness of the material adversely influence biofilm formation on the surface of the 3D-printed dental prostheses. This study evaluated the effects of a newly developed chlorhexidine (CHX)-loaded polydimethylsiloxane (PDMS)-based coating material on the surface microstructure, surface wettability and antibacterial activity of 3D-printing dental polymer. First, mesoporous silica nanoparticles (MSN) were used to encapsulate CHX, and the combination was added to PDMS to synthesize the antibacterial agent-releasing coating substance. Then, a thin coating film was formed on the 3D-printing polymer specimens using oxygen plasma and thermal treatment. The results show that using the coating substance significantly reduced the surface irregularity and increased the hydrophobicity of the specimens. Remarkably, the culture media containing coated specimens had a significantly lower number of bacterial colony formation units than the noncoated specimens, thereby indicating the effective antibacterial activity of the coating.

**Keywords:** 3D-printing; dental polymer; antibacterial agent; coating; mesoporous silica nanoparticles; polydimethylsiloxane

#### **1. Introduction**

Three-dimensional (3D) printing technology is an additive manufacturing method in which a 3D object is formed by adding successive layers of material [1,2]. Dentistry has benefited from the rapid expansion of 3D-printing methods, especially in the field of prosthesis manufacturing [3–5]. The 3D-printing technology used in dentistry is classified into four main categories: extrusion printing, inkjet printing, laser melting/sintering, and stereolithography printing [3]. Those printing methods are based on the principle of layered manufacturing, which is more suitable than conventional casting and milling methods for producing individualized complex structures [6]. Moreover, using a machining process with computer-aided design/computer-aided manufacturing (CAD/CAM) reduces manual labor and material waste [1]. Recent studies have shown that 3D-printed dental prostheses had a clinically acceptable degree of precision [3,6,7].

Along with the increasing accuracy of 3D-printing technology, various materials, including polymers, ceramics, and metallic powders, have been developed [4]. Regarding dental materials used in 3D printers, polymers are the most commonly used in dentistry for interim and definitive prostheses due to their suitable mechanical strength, highly biocompatible properties, and ease of manipulation [3,4]. However, a major clinical complication associated with dental polymer prostheses is dental plaque surface deposition, which comprises numerous oral microorganisms due to the porosity and water absorptiveness of the polymer material [8]. Specifically, as the 3D-printed prosthesis is generated layer-by-layer, micropores are created when air is trapped between the layers during the printing process or when the individual layers are incompletely fused [9]. During clinical occlusal adjustment and clinical use, these micropores are exposed on the surface because of the abrasion of the restorations, potentially becoming a host for bacterial growth [6]. The risk of material contamination by microorganisms is a critical limitation for the longevity of 3D-printed polymer prostheses [3].

Several strategies have been reported for the fabrication of antimicrobial polymers for 3D-printed polymers [10]. Direct incorporation of an antibacterial agent into dental polymers has been used to reduce plaque accumulation [11]. Chlorhexidine (CHX) is an antiseptic agent widely used in dentistry for its broad-spectrum antibacterial effects and nontoxicity toward mammalian cells [12]. CHX is effective in managing infected oral mucositis; thus, it has commonly been used to prevent dental plaques and to control infections as a topical agent in daily mouth rinse and as an irrigation solution in endodontic treatment [13]. When CHX was directly mixed with the polymer and cured, CHX release at the therapeutic dose was maintained for 28 days [14–16]. However, direct mixing may negatively affect the mechanical and surface properties of polymers in terms of polymerization degree, surface porosity, and water absorption [11,17]. Moreover, commercial polymers used for the subtractive manufacturing method are provided in a completely polymerized state, making it difficult to directly add CHX inside the substance without jeopardizing material integrity [18].

An alternative technique for inhibiting bacterial adherence is to change the surface property of the object by creating a coating film [19–21]. Coating layer formation does not sacrifice the mechanical properties and integrity of the material. The major mechanisms underlying the antibacterial activity of the coating layers include antiadhesion/bacterial-repelling and contact-killing [22]. Antiadhesion coating reduces the adhesion force between bacteria and the substrate to enable the easy removal of bacteria before the biofilm layer is formed [21]. Alternatively, in the contact-killing approach, antibacterial compounds are attached to the surface of the material by flexible, hydrophobic polymeric chains, which can kill bacteria upon contact [21,22]. Remarkably, using a polymer coating significantly reduced plaque biofilm formation on polymeric restorations with the coating layer exhibiting acceptable mechanical and chemical durability [20]. Azuma et al. [19] reported that silica coating with silica nanoparticles of various sizes was effective in decreasing bacterial adherence to polymeric restorations. While these previous coating materials showed antiadhesion potential, they passively hindered the adherence of early colonizers by increasing polymer surface hydrophobicity [19,20]. This strategy may physically suppress dental plaque formation and maturation; however, it provides no active antimicrobial agent to inhibit the vitality and growth of pathogenic microorganisms.

With the development of new drug-carrier materials, active antibacterial agent-releasing coating is now a significant focus in biomedical research [22]. In its most advanced form, the coating mediates antibacterial activity by releasing loaded antibacterial compounds over time to kill the bacteria in the surrounding [21,22]. In drug delivery systems, materials with a porous structure are recommended for incorporating drug-loaded particles [23,24]. Among the porous polymeric materials, polydimethylsiloxane (PDMS) has been used in the medical field due to its flexibility, biocompatibility, transparency, low cost, and ease of fabrication [23,25]. The incompatibility between the coated substrate and PDMS could cause a dewetting that may induce a greater surface roughness; however, the subsequent treatments of oxygen plasma could enhance the wettability of precured PDMS, inducing its uniform coating [26,27]. Moreover, the drug release rate of PDMS can be controlled by surface modifications using plasma treatment, which effectively influence the water penetration

rate and functionalization of the polymer that contains drug [28]. The highly functionalized PDMS can be polymerized using thermal treatment or ultraviolet (UV) light activation for curing the coating layer, inducing a more chemically and physically stable coating layer on the substrate [27]. Thus, it would be an effective approach for tissue engineering or drug delivery systems in biological and biomedical applications [29]. In dentistry, PDMS coating has been used for tooth enamel and metallic dental implants as a hydrophobic layer to improve the antibacterial features [30]. However, no studies have examined the PDMS coating application for active antibacterial effects in dental 3D-printable polymers. Therefore, this study evaluated the effect of a newly developed CHX-loaded PDMS-based coating on the surface microstructure, surface wettability, and antibacterial activity of the 3D-printing dental polymer.

#### **2. Results**

#### *2.1. Surface Characteristics*

Figure 1 shows the microscopic images of specimens in different groups. Deep grooves and small defects were found on the surface of noncoated specimens, whereas no scratches were observed on the surface of coated specimens. Several different-sized silica particles were observed on the coating surface.

(**a**)

**Figure 1.** *Cont*.

**Figure 1.** Scanning electron microscope (SEM) images at 500 × (left) and 2000 × (right) magnification and the resulting histogram of pixel brightness levels (0 = black, 255 = white) of: (**a**) noncoated specimen; and (**b**) coated specimens.

(**b**)

Table 1 presents the SEM image analysis for surface roughness obtained from the surface plot histograms. The SEM roughness index (SRI) values were significantly lower in the coated group than in the noncoated group (*p* < 0.001). Regarding surface wettability, the mean contact angles were significantly higher in the coated group than in the noncoated group (*p* < 0.001) (Figure 2 and Table 1).



a, b Different superscript lowercase letters indicate a statistically significant difference within a column; SRI, SEM roughness index; CA, contact angle degree.

**Figure 2.** Contact angle of the specimens: (**a**) noncoated specimen; and (**b**) coated specimen.

#### *2.2. Antimicrobial Activity*

Because the specimen itself could affect bacterial growth, the incubation of *S. mutans* with noncoated specimen was a control in this experiment. Statistical analysis was performed using three technical replicates. Biological relevance of the results was confirmed using three independent experiments with similar results. Figure 3 shows the results of the *S. mutans* bacterial growth inhibition assay for the noncoated and coated groups. The numbers of bacterial colonies (×10<sup>4</sup> CFU/mL) of the noncoated and coated groups were 210.92 ± 8.02 and 70.76 ± 9.16, respectively. Independent *t*-test showed that the culture media containing the coated specimens had significantly lower CFU values than the culture media containing the noncoated specimens (*p* < 0.001).

**Figure 3.** Bacterial growth inhibition assay of noncoated and coated groups: (**a**) bacterial colonization on agar plates; and (**b**) colony-forming units. Statistical analysis was performed using three independent technical replicates, inducing similar biological results. Independent *t*-test showed that the culture media containing the coated specimens had significantly lower CFU values than the culture media containing the noncoated specimens (\* *p* < 0.001).

#### **3. Discussions**

This study evaluated the effect of a newly developed CHX-loaded PDMS-based coating substance on the surface properties and antibacterial ability of coated 3D-printing dental material. From the results of this study, applying the coating layer on the polymer specimens significantly reduced their surface irregularity while increasing the hydrophobicity and antibacterial activity.

Once a dental restoration is placed in the oral cavity, proteins from saliva cover the surface of the restoration as a film, followed by the attachment of free-floating bacteria to the film by microfilaments in the cell walls to form a biofilm on the restoration [31]. Restorations with greater surface roughness show higher biofilm formation because the irregular surface geometry offsets shear forces to the surface, thereby providing a favorable context for bacterial growth [32]. In the clinical context, surface smoothing of dental restoration by using polishing tools is necessary before restoration is placed in the mouth [33]. The microscopic images of this study showed sharp grooves and defects on the surface of specimens in the noncoated group due to the grinding motion during the polishing process. The result agrees with previous studies where the polishing tools caused some microdefects to the restoration surface as they removed material by abrasion [34]. In specimens given the coating substance, scratches and defects were unobserved, and the roughness value was significantly lower than that of the noncoated specimens. The results suggest that the coating decreases surface irregularity by filling the scratches resulting from polishing and the inherent micropores created by the 3D-printing process. Previous studies that evaluated the effect of different surface treatments on dental ceramic restoration also indicated a significant decrease in roughness when a thin layer of glazing material was coated on the restoration surface [35]. In addition, various micrometer-sized particles were observed on the surface of the coated specimens. Considering that the mesoporous silica nanoparticles (MSNs) dispersed in the solvent were nanosized, this result implies that some of the silica particles on the coated layer became aggregated and clustered. A silica coating layer needs to contain particles of

an optimum size to reduce microorganism adherence [19,36]. Thus, further technical optimization is needed to improve the homogeneous distribution of the silica nanoparticles and to minimize their uncontrolled aggregation in the coating layer.

Surface free energy greatly influences the initial step of biofilm formation [37]. Lower surface energy of the material has weakened bacterial adhesion; thus, the bacteria that adhere to a material with low surface energy are more easily removed by an external force [37,38]. The work of adhesion can be calculated by measuring the contact angle of the liquid to the solid surface [39]. For a low wetting surface, the surface energy of the solid is weaker than the surface tension of the liquid, allowing the liquid to easily retain its droplet shape. Therefore, a higher contact angle is related to lower surface energy and low interfacial tension of the solid surface. In this study, the coated group showed a significantly higher contact angle than the noncoated group, thereby indicating a lower surface energy of the coated specimens than that of the noncoated specimens. This is due to the hydrophobicity and low surface energy of PDMS, which is a beneficial feature of the material that contributes to the low bacterial adhesion [40].

*S. mutans* is an important etiologic agent for initiating dental caries [41]. The acid produced from this bacteria decays tooth structure and induces restorative treatment failure [41]. Accordingly, dental restorations require antibacterial qualities to ensure long-term success. For this purpose, the MSN was used to encapsulate CHX in this study, and the CHX@MSN was then combined with PDMS to unprecedentedly synthesize an antibacterial coating substance for polymeric restorative 3D-printing dental material. The bacterial growth inhibition assay results showed that the coated specimens had antimicrobial activity against *S. mutans.* The antimicrobial activity may be due to CHX release from the coating substance. Adding CHX to dental restorative material could increase its antibacterial activity [14,15]. Yan et al. [12] incorporated CHX@MSN into a glass ionomer cement powder and showed that CHX was continuously released, and the antibiofilm effect was maintained up to 30 days. Remarkably, dental resin composites with CHX@MSN showed controlled release of CHX over a prolonged time, providing strong inhibition against *S. mutans* adherence [11]. However, the antibacterial activity of previous coatings has been limited to the surface of the coated objects. In our study, an elastic porous material, PDMS, was used to store and release the CHX@MSN particles. The encapsulation and drug-loading efficiency of the CHX@MSN were recorded at rates of 25.22% and 63.04%, with a stable CHX releasing rate of approximately 1.56 µg/mL within the first 24 h in a pilot study [32]. Because the synthesized coating layer could release the loaded CHX over time [42], the coating exhibited antibacterial effects in the surrounding areas not directly in contact with the surface of the restoration. Moreover, the CHX@MSN and PDMS materials used in this study have been reported to be relatively noncytotoxic [32]. Therefore, this active antibacterial protective film formation is expected to be a novel method for actively inhibiting bacterial inhabitation around the coated surface of restorations.

Note that the human oral environment is more complicated than the in vitro experimental context because the oral cavity temperature, food intake, and the pH and composition of saliva vary between subjects and even within a subject [43]. In addition, the mechanical properties and material stability of the coating layer were not investigated in this study. However, the mechanical properties of PDMS strongly depend on the mixing ratio of a base to a curing agent in PDMS mixture [26,27,44]. In this study, the PDMS mixture comprises a base and a curing agent at a weight ratio of 5:1 with the relative elastic modulus number of 3.59 MPa [44]. However, the mechanical properties of the coating can be significantly improved by incorporating MSN into the PDMS as the elastic modulus of MSN is at least four orders of magnitude larger than that of PDMS [45]. Moreover, the hydrogen bonds between MSNs and covalent bonds between MSN and PDMS may induce a strong resistance to mechanical deformation [46]. The findings in those previous studies indicates that the mechanical strength of the PDMS coating layer can be tuned by controlling the input parameters such as the curing agent and MSN. In this study, the surface microstructure, surface wettability, and antibacterial activity were immediately evaluated after coating and within 24 h of drug releasing; thus, the material was expected to be stable during this stage. A further study that focuses on investigating the mechanical properties of the coating should be conducted in clinically relevant conditions to extend the understanding of this new material.

#### **4. Materials and Methods**

### *4.1. Synthesis of the* Coating Material *(CHX@MSN-Loaded PDMS)*

Figure 4 illustrates the fabrication process for the coating substance. Fifty milligrams of CHX (Sigma-Aldrich Co., St. Louis, MO, USA) were dissolved in 5 mL of absolute ethanol, and then dried MCM-41 mesoporous silica nanoparticles (MSN) (Sigma-Aldrich Co., St. Louis, MO, USA) with a pore volume of 0.98 cm<sup>3</sup> /g, pore size of approximately 2.5 nm, and Brunauer–Emmett–Teller (BET) surface area below 1000 m<sup>2</sup> /g were dispersed into the CHX solution. The mixture was sonicated for 10 min and incubated for 24 h at room temperature using a magnetic stirrer (Corning PC-420D, Fisher Scientific, Lowell, MA, USA) at a speed of 300 rpm. Next, to collect the CHX@MSN particles, the mixture was filtered (Labogene Scan Speed Mini, Lynge, Denmark) and then vacuum-dried (OV-11, JEIO Tech, Seoul, Korea).

**Figure 4.** Synthesis of the CHX@MSN-loaded PDMS coating substance. CHX, Chlorhexidine; MSN, Mesoporous silica nanoparticles; CHX@MSN, Chlorhexidine encapsulated in mesoporous silica nanoparticles; PDMS, Polydimethylsiloxane.

The CHX@MSN particles were then mixed with PDMS (Sylgard 184, Dow Corning, Midland, MI, USA) solution at 0.4 wt% relative to the total PMDS mass. The PDMS mixture comprised a base and a curing agent at a weight ratio of 5:1. All components were blended in a SpeedMixerTM (FlackTek Inc., Landrum, SC, USA) for 3 × 30 s to form a precured coating paste, which was stored in the dark at room temperature and vacuumed for 20 min to remove bubbles before use.

#### *4.2. Coating Procedure*

Sixty disk-shape specimens (*N* = 60) with a thickness of 1.0 mm and diameter of 13.0 mm were designed using CAD software (Geomagic Design X, 3D Systems, Inc., Rock Hill, SC, USA) and printed with the photopolymer (RAYDent C&B, Ray Co., Hwaseong, Korea) using a digital light processing 3D printer (RAM500, Ray Co., Hwaseong, Korea). All specimens were postcured with a curing unit (RPC500, Ray Co., Hwaseong, Korea) for 20 min and polished with a 1000-grit silicon carbide abrasive paper (Buehler GmbH, Dusseldorf, Germany) for 60 s. Table 2 provides the composition of the photopolymer.


**Table 2.** Composition of the 3D-printing photopolymer.

\* As provided by the manufacturer.

Surface functionalization and coating process were performed in the coated group specimens (*n* = 30). The polished specimens were cleaned with isopropyl alcohol and treated with oxygen plasma (CUTE, Femto Science Co., Seoul, Korea) for 5 min (Figure 5a). The specimen surfaces were functionalized by immersion into 5% (*v*/*v*) 3-aminopropyltriethoxysilane (APTES) (Sigma-Aldrich Co., St. Louis, MO, USA) solution at 85 ◦C for 10 min (Figure 5b). The specimens were then coated by dipping in the precured coating solution using a dip coating equipment (KD Scientific, Holliston, MA, USA) at a lowering speed of 6000 µm/s and lifting speed of 1000 µm/s (Figure 5c). Subsequently, the specimens underwent thermal treatment in an oven (OV-11, JEIO Tech, Seoul, Korea) at 80 ◦C for 2 h (Figure 5d). The noncoated group specimens were defined as control (*n* = 30).

**Figure 5.** Surface functionalization and coating treatment: (**a**) oxygen plasma treatment; (**b**) APTES treatment; (**c**) dip coating; and (**d**) Thermal curing. APTES, 3-aminpropyltriethoxysilane; PMMA, polymethyl methacrylate; CHX@MSN, Chlorhexidine encapsulated in mesoporous silica nanoparticles; PDMS, Polydimethylsiloxane.

#### *4.3. Evaluation of Surface Microstructure*

Surface microstructure was evaluated using a scanning electron microscope (SEM) (S-4500, Hitachi Co., Tokyo, Japan). Specimens were coated with spotted platinum using a sputter coater (E-1030, Hitachi Co., Tokyo, Japan), and scanned at 5 kV at 500× and 2000× magnification. Ten locations in each specimen were randomly selected for image acquisition. The images were converted to 8-bit grayscale (black = 0, white = 225) using ImageJ software (version 1.52k, National Institute of Health, Bethesda, MD, USA). The pixel brightness data of each image were plotted as 256-level histograms

to build a surface plot image, where the y-axis represents the 0–255 grayscale levels and the x-axis represents the pixel frequency. The histograms from all 10 images were combined, and the standard deviation of the pixel brightness was calculated as the SRI [47].

#### *4.4. Measurement of Surface Wettability*

Surface wettability was evaluated by measuring the contact angle (CA) in an air environment. A liquid droplet (5 µL) of distilled water was dispensed onto the specimen surface at a room temperature of 20 ◦C. The image of the droplet was immediately captured by a digital camera (Canon EOS 500D with Canon EF 100 mm f2.8 Macro USM Lens, Canon Inc., Tokyo, Japan) (Figure 6a), and the contact angles were determined from the corresponding pictures using the contact angle plugin of the ImageJ software. The CA degree was determined by measuring the tangent angle to the surface of the liquid droplet (Figure 6b). The mean value was calculated by averaging three individual measurements.

*θ* **Figure 6.** Measurement of surface wettability: (**a**) contact angle measurement system setting; and (**b**) contact angle (θ) measurement by the tangent angle of the liquid droplet to the surface.

#### *4.5. Assessment of Antimicrobial Activities*

To evaluate the antibacterial activity, a bacterial growth inhibition assay was performed. *Streptococcus mutans* ATCC 25175 (*S. mutans*) was inoculated into a 15-mL tube containing 10 mL of brain heart infusion (BHI) media (BHI, Mast Group, Bootle, UK) and incubated on a rotary shaker (150 rpm) at 37 ◦C. To determine the concentration of bacteria in culture media, the optical density at 600 nm (OD600) of the cell suspensions was recorded using a DS-11+ apparatus (DeNovix, Wilmington, DE, USA). After 12 h, OD<sup>600</sup> was checked to measure bacterial density in liquid media.

When the OD<sup>600</sup> of the *S. mutans* culture reached approximately 0.4, aliquots of 500 µL were added to each well in a 24-well plate and incubated with noncoated or coated specimens at 37 ◦C on a shaker at 15 rpm for 24 h. Cultures without any test specimens were used as the control condition. Following incubation, appropriate dilutions of the cultures were plated on BHI agar plates using an ethanol-flamed bacterial spreader, and the plate was incubated for 24 h at 37 ◦C. Then, the colony-forming units per milliliter (CFU/mL) were calculated using the number of colonies observed and the dilution factor (1:10<sup>4</sup> ) for each well.

#### *4.6. Statistical Analysis*

Statistical calculations were performed using SPSS software (SPSS version 25.0, IBM Inc., Chicago, IL, USA). The measured values were expressed as the mean ± standard deviation. The independent *t*-test was conducted to compare differences between the groups. The statistical significance level was set at 0.05.

#### **5. Conclusions**

This study evaluated the effect of a newly developed CHX@MSN-loaded PDMS-based coating substance on the surface properties and antibacterial ability of coated 3D-printing dental material. From our results, it can be concluded that applying the coating layer on the polymer specimens significantly reduced their surface irregularity, while increasing the hydrophobicity and antibacterial activity.

**Author Contributions:** Conceptualization, H.-N.M. and D.-H.L.; methodology, D.C.H., J.H.P., and D.-Y.K.; formal analysis, D.C.H. and S.M.L.; writing—original draft preparation, H.-N.M.; writing—review and editing, D.-H.L.; and supervision, D.-H.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (2020R1A2C4002518). The funders had no role in the study design, data collection and interpretation, or the decision to submit the work for publication.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

### **Cytotoxicity E**ff**ects of Water-Soluble Multi-Walled Carbon Nanotubes Decorated with Quaternized Hyperbranched Poly(ethyleneimine) Derivatives on Autotrophic and Heterotrophic Gram-Negative Bacteria**

**Nikolaos S. Heliopoulos 1,2 , Georgia Kythreoti 1,3, Kyriaki Marina Lyra <sup>1</sup> , Katerina N. Panagiotaki <sup>1</sup> , Aggeliki Papavasiliou <sup>1</sup> , Elias Sakellis <sup>1</sup> , Sergios Papageorgiou <sup>1</sup> , Antonios Kouloumpis <sup>4</sup> , Dimitrios Gournis <sup>4</sup> , Fotios K. Katsaros <sup>1</sup> , Kostas Stamatakis <sup>3</sup> and Zili Sideratou 1,\***


Received: 14 July 2020; Accepted: 1 October 2020; Published: 6 October 2020

**Abstract:** Oxidized multi-walled carbon nanotubes (oxCNTs) were functionalized by a simple non-covalent modification procedure using quaternized hyperbranched poly(ethyleneimine) derivatives (QPEIs), with various quaternization degrees. Structural characterization of these hybrids using a variety of techniques, revealed the successful and homogenous anchoring of QPEIs on the oxCNTs' surface. Moreover, these hybrids efficiently dispersed in aqueous media, forming dispersions with excellent aqueous stability for over 12 months. Their cytotoxicity effect was investigated on two types of gram(−) bacteria, an autotrophic (cyanobacterium *Synechococcus* sp. PCC 7942) and a heterotrophic (bacterium *Escherichia coli*). An enhanced, dose-dependent antibacterial and anti-cyanobacterial activity against both tested organisms was observed, increasing with the quaternization degree. Remarkably, in the photosynthetic bacteria it was shown that the hybrid materials affect their photosynthetic apparatus by selective inhibition of the Photosystem-I electron transport activity. Cytotoxicity studies on a human prostate carcinoma DU145 cell line and 3T3 mouse fibroblasts revealed that all hybrids exhibit high cytocompatibility in the concentration range, in which they also exhibit both high antibacterial and anti-cyanobacterial activity. Thus, QPEI-functionalized oxCNTs can be very attractive candidates as antibacterial and anti-cyanobacterial agents that can be used for potential applications in the disinfection industry, as well as for the control of harmful cyanobacterial blooms.

**Keywords:** carbon nanotubes; quaternary ammonium groups; hyperbranched dendritic polymers; antibacterial properties; anti-cyanobacterial properties

#### **1. Introduction**

Carbon nanotubes (CNTs) have attracted significant scientific and technological interest due to their unique structural characteristics and their excellent electronic, mechanical, and thermal properties [1,2]. Based on these properties, they have been used in a wide range of applications, including fillers in composite materials, sensors, drug delivery systems, antibacterial agents, and others [3,4]. However, their poor dispersibility in solvents, especially in water, has prevented their widespread industrial use, and reduced their great potential. Attempts to overcome this problem have focused on the functionalization of their surface, using a variety of covalent and non-covalent modification strategies [5]. On one hand, various organic molecules, such as dendritic and linear polymers, have been covalently conjugated onto the CNT's convex surfaces and tips by chemical reactions [6–8] in order to reduce aggregates and size polydispersity. However, covalent functionalization causes damage to the conjugated π-electrons, leading to degradation of their properties. On the other hand, non-covalent functionalization, based on π–π stacking and ionic interactions between various molecules and the CNTs graphitic surface [9–11] does not affect their electronic structure, and has been achieved using a multitude of surfactants [6,7] and polymers [8,10], resulting in modified CNTs, compatible with specific solvents or targeted applications. In this context, their functionalization with dendritic polymers such as dendrons, dendrimers, and hyperbranched polymers, is expected to be a very promising strategy, when aiming to achieve increased water solubility. This strategy has already been applied to single-walled carbon nanotubes (SWCNTs) that were functionalized using dendritic polymers through non-covalent interactions, achieving enhanced water solubility [8,12,13]. However, only a few studies have addressed the non-covalent functionalization of multi-walled carbon nanotubes (MWCNTs) with dendritic polymers for increased water solubility [14].

Dendritic polymers are highly branched macromolecules of nanosized dimensions, consisting of repeating units and surface end groups [15]. Their properties depend on both the structural characteristics of the branches in their interior, and the large number of surface end groups. These polymers can incorporate a variety of organic compounds as well as inorganic ions in their interior, while the surface end groups are primarily susceptible to functionalization or even multi-functionalization, to yield a variety of novel materials with diversified, tailor-made properties such as drug delivery systems, antibacterial agents, etc. [16–18]. Additionally, these terminal groups exhibit the so-called polyvalency effect [19], which enhances their binding with various substrates, due to their close proximity to the dendritic polymers' scaffold.

Recently, several studies have focused on the potential applications of carbon nanomaterials (CNMs), taking advantage of their antibacterial properties [20–22]. Specifically, functionalized single-walled carbon nanotubes (SWCNTs) and multi-walled carbon nanotubes (MWCNTs) were found to exhibit significant antibacterial activity towards both gram-positive and gram-negative bacteria [23,24]. CNTs' properties, such as carbon nanotube diameter [25], length [26], aggregation [27], concentration [28], surface functionalization [29–31], etc. have been reported to influence their antibacterial activity. To that respect, aqueous dispersibility can critically influence antibacterial efficiency, as highly dispersed CNTs enhance their interaction with cells, leading to increased antibacterial properties. Indeed, it was found [32] that individually dispersed CNTs were more toxic to bacteria than CNTs aggregates, due to increased contact with bacterial cells.

Nowadays, there is a growing concern about the possible effects of nanomaterials, as end of life products, on organisms and ecosystems [33,34]. Apart from studies investigating the mechanisms of interaction between CNTs and various biomacromolecules (DNA, RNA, etc.), to identify possible causes of undesired effects, [35] research efforts have also focused on the possible impact of MWCNTs on photosynthetic pathways. Remarkably, in photosynthetic organisms, unlike bacteria, a favorable effect of MWCNTs was observed, e.g., in the development of cereals, and the production of vegetative biomass [36]. Studies on algae revealed that MWCNTs did not influence photosynthesis, and any negative effects were due to turbidity and the resulting reduction of the available light [37]. However, oxidized MWCNTs were found to be toxic to the marine green alga *Dunaliella tertiolecta*, as at

concentrations between 1–10 mg/L they reduced algal growth, and affected the Photosystem (PS) II photochemical process and the cellular glutathione redox status [38]. Moreover, although cyanobacterial blooms have become a serious environmental problem, only recently, a novel nanomaterial based on MWCNTs, called Taunit, loaded with antibiotic chloramphenicol or herbicide diuron, was investigated as an effective anti-cyanobacterial agent [39]. It was found that the Taunit-diuron complex exhibited high biocide action against cyanobacterium *Synechocystis* sp. PCC 6803, higher than that of a Taunit-chloramphenicol complex.

In this study, aiming to develop water soluble MWCNTs with enhanced antibacterial/anticyanobacteria properties, oxidized multi-walled carbon nanotubes (oxCNTs) were non-covalently functionalized using a series of partially quaternized hyperbranched poly(ethyleneimine) derivatives, yielding novel water-soluble hybrid materials. Specifically, three positively charged derivatives of hyperbranched poly(ethyleneimine) (PEI) with a different degree of quaternization at the primary amino groups of PEI were prepared and interacted with the negatively charged oxidized CNTs through electrostatic interactions and van der Waals attraction forces. The obtained hybrid materials were physicochemically characterized by various techniques (FTIR, Raman, SEM, TEM, AFM, etc.). Their excellent aqueous stability was demonstrated using ζ-potential measurements, dynamic light scattering, and UV-vis spectroscopy. Additionally, their antibacterial and anti-cyanobacterial activity was investigated against two types of gram negative bacteria, an autotrophic (cyanobacterium *Synechococcus* sp. PCC 7942) and a heterotrophic (bacterium *Escherichia coli*), while their cytocompatibility was investigated on eukaryotic cell lines.

#### **2. Results and Discussion**

#### *2.1. Synthesis and Characterization of QPEI-Functionalized oxCNTs*

Positively charged stable aqueous suspensions of carbon nanotubes were prepared, applying quaternized hyperbranched poly(ethyleneimine) derivatives (QPEIs). A series of partially quaternized hyperbranched poly(ethyleneimine) derivatives with 30%, 50%, and 80% substitution degree of primary amino groups was prepared, following a method analogous to one previously described [40,41]. Initially, PEI was characterized by inverse-gate decoupling <sup>13</sup>C NMR. According to the literature [42] and comparing the integration of carbons of α-CH<sup>2</sup> relative to primary, secondary, and tertiary amine groups, the ratio of primary to secondary and to tertiary amines of PEI was found to be 1.00:1.18:1.01. The branching degree was found to be 0.68, and the average number of primary amine groups was determined to be 183. Based on the above, the introduction of α-hydroxyamine moieties, together with the trimethylammonium groups, to PEI was achieved by the reaction of the PEI primary amines with appropriate amounts of glycidyltrimethylammonium chloride, yielding three PEI derivatives, i.e., 30-QPEI, 50-QPEI, and 80-QPEI (Scheme 1). Their structures were confirmed by <sup>1</sup>H and <sup>13</sup>C NMR spectroscopy. Specifically, the new quintet appearing at 4.25 ppm was attributed to the proton of α-CH group relative to hydroxyl group, which was formed after an oxiran ring opening. Additionally, the α-CH<sup>2</sup> protons relative to the quaternary group appeared as a triplet at 3.45 ppm, and the protons of the quaternary methyl groups at 3.25 ppm, while a multiplet in the region 2.50–2.70 ppm was attributed to the PEI scaffold protons. Comparing the integration of peaks at 3.45 ppm and 2.70–2.50 ppm, the degree of quaternization at the primary amino group of PEI was calculated, and was found to be 30%, 50%, and 80% for 30-QPEI, 50-QPEI, and 80-QPEI, respectively.

Furthermore, <sup>13</sup>C NMR spectroscopy provided insights into the structure of QPEIs. The attachment of the α-hydroxyamine moiety at the PEI scaffold was confirmed by the peaks at 51.0 and 48.0 ppm, attributed to the α carbon of PEI, and relative to the newly formed amino group (α-CH2NH-Q), close to secondary (known as C1,2) and tertiary (known as C1,3) amines, respectively. Additionally, the α-methylene and methyl groups, attached at the quaternary center, were observed at 71.5 and 57.0 ppm, respectively, while a peak at 67.5 ppm was attributed to the α carbon relative to the newly formed hydroxyl group (CH–OH).

**Scheme 1.** Schematic representation of poly(ethyleneimine) (PEI) and the reaction scheme of quaternization.

Subsequently, these dendritic derivatives were interacted with oxidized CNTs in aqueous media (Scheme 2). The final hybrid nanomaterials, oxCNTs@30-QPEI, oxCNTs@50-QPEI, and oxCNTs@80-QPEI, were obtained after ultracentrifugation to remove the excess QPEIs, and physicochemically characterized using a variety of techniques, such as FTIR, RAMAN, TGA, SEM, TEM, AFM, etc.

**Scheme 2.** Schematic representation of oxidized multi-walled carbon nanotubes (oxCNTs) decorated with quaternized hyperbranched poly(ethyleneimine) derivative (QPEI).

− − − To investigate the successful attachment of QPEIs on the oxCNTs, initially, FTIR spectroscopy was employed. In Figures 1A and S1, the FTIR spectra of QPEIs, oxCNTs, and QPEI-functionalized oxCNTs are shown. The spectrum of oxCNTs shows a C=C stretching band at 1650 cm−<sup>1</sup> , attributed to the CNTs graphite structure. Additionally, the presence of oxygen containing groups (carboxylates, carbonyl, hydroxyl, and epoxy groups) on the oxCNTs was confirmed by the appearance of a C=O stretching band at 1740 cm−<sup>1</sup> , a broad OH stretching band centered at 3370 cm−<sup>1</sup> , and a strong C–OH

− −

−

stretching band at 1100 cm−<sup>1</sup> , as well as two peaks at 1565 and 1380 cm−<sup>1</sup> , which are associated with the carboxylate anion stretch mode (asymmetrical and symmetrical vibrations of COO- , respectively). Furthermore, the peaks at 2750 and 1255 cm−<sup>1</sup> , attributed to the stretching vibrations of OC–H and C–O–C, respectively, revealed the presence of aldehydes and epoxy groups on oxCNTs [43]. On the other hand, as expected, the FTIR spectra of all QPEIs exhibited the sample characteristic bands. Those included the characteristic vibrations of PEI, i.e., at 3360 and 3270 cm−<sup>1</sup> , attributed to the stretching vibration of primary and secondary amino groups, at 2940 and 2835 cm−<sup>1</sup> , assigned to the asymmetrical and symmetrical vibrations of CH2, at 1600 and 1560 cm−<sup>1</sup> , attributed to the NH deformation mode of the primary and secondary amino groups, respectively, at 1455 and 760 cm−<sup>1</sup> , corresponding to the bending and rocking mode of CH2, respectively, and at 1115 and 1050 cm−<sup>1</sup> , assigned to the asymmetrical and symmetrical vibrations of C–N [44]. Additionally, the most important stretching and deformation vibrations of CH<sup>3</sup> in the quaternary ammonium group (CH3)3N<sup>+</sup> at 3030 cm−<sup>1</sup> and 1485 cm−<sup>1</sup> , the asymmetrical stretching vibration of the whole (CH3)3N<sup>+</sup> group at 970 cm−<sup>1</sup> , and the stretching vibrations of C–OH groups at 1100 cm−<sup>1</sup> were observed [45]. The FTIR spectra of the QPEI-functionalized oxCNTs (Figures 1A and S1) revealed the existence of both oxCNTs and QPEIs, confirming their successful interaction. − − − − − − − − − −

The successful functionalization of oxCNTs was also confirmed by Raman measurements. The spectra of oxCNTs and QPEI-functionalized oxCNTs presented in Figure 1B, display the two main typical graphite bands at 1585 cm–1 (G band) and 1345 cm–1 (D band), attributed to the in-plane vibration of the sp<sup>2</sup> -bonded carbon atoms in graphite layers, and to the defects presented in carbon nanotubes due to the conversion of carbon atoms from an sp<sup>2</sup> to an sp<sup>3</sup> hybridization state, respectively. Additionally, a band at ~2700 cm–1 (G' band) is shown in Figure 1B, attributed to the D band overtone. As observed in Figure 1B, the Raman shifts of QPEI-functionalized oxCNTs compared to those of oxCNTs did not change, revealing that the graphitic structure of oxCNTs does not significantly alter after functionalization. Only the value of the intensity ratio of D- to G- bands (ID/IG), a measure of the defects present in a graphene structure during functionalization, slightly increased from 1.03 to 1.16, revealing successful polymer wrapping all over the graphite layer of CNTs [46]. In analogous studies involving covalent functionalization of multi-walled carbon nanotubes via third-generation dendritic poly(amidoamine) or amphiphilic poly(propyleneimine) dendrimers, a larger increase of the intensity ratio (ID/IG), was reported indicating that functionalization caused a larger defect of the graphitic network [47,48]. In contrast, in the present study oxCNTs were non-covalently functionalized with QPEI derivatives, retaining their surface almost intact.

**Figure 1.** (**A**) FTIR spectra of oxCNTs, oxCNTs@80-QPEI, and 80-QPEI. (**B**) Raman spectra of oxCNTs (black), oxCNTs@30-QPEI (red), oxCNTs@50-QPEI (blue), and oxCNTs@80-QPEI (magenta).

The results of thermogravimetric analysis (TGA) provided further information on the QPEI content on the surfaces of oxCNTs (Figure 2). In the TGA curve of oxCNTs, two distinct decomposition regions were observed. Specifically, a weight loss corresponding to ~4% of the initial weight was recorded in the temperature range of 180–400 ◦C, due to the removal of oxygen-containing functional groups present on the graphitic framework. A second significant loss was observed at higher temperatures (>500 ◦C), and was attributed to the thermal degradation of the graphitic framework. In contrast, the QPEI-functionalized oxCNTs exhibited a significant weight loss (10–20%) up to 250 ◦C, due to both the removal of oxCNTs' oxygen-containing groups, and the partial PEI degradation. The weight loss for QPEI-functionalized oxCNTs in the temperature range 250–400 ◦C was significantly higher in comparison to oxCNTs (up to 60%), since together with the decomposition of graphitic lattice, QPEI molecules were removed from the graphitic framework. Therefore, TGA measurements provided further qualitative experimental evidence that functionalization had occurred, however, due to the oxCNT contribution to the thermal phenomena, the polymer content could not be quantified. Again, above ~500 ◦C, the sharp weight loss indicated the total thermal destruction of the graphitic network.

**Figure 2.** Thermogravimetric curves of oxCNTs and QPEI-functionalized oxCNTs.

Elemental analysis measurements were additionally performed to confirm and quantify the polymer content in the QPEI-functionalized oxCNTs. Given that the nitrogen signal in the final hybrid originated mainly from QPEI, the difference compared to the starting oxCNTs represented the amount of polymer attached to the CNTs. Therefore, the QPEI content in hybrids was calculated from the following formula:

$$\text{QPEI (\% w/w)} = (N\_{\text{s}} - N\_{\text{CNTs}}) / (N\_{\text{QPEI}} - N\_{\text{CNTs}}) \times 100$$

− − where *N*s, *N*QPEI, and *N*CNTs, were the nitrogen elemental mass fraction in QPEI-functionalized oxCNTs, QPEI, and oxCNTs, respectively [49]. The results are summarized in Table S1. According to the elemental analysis results, the actual value of QPEI weight fraction in oxCNTs@30-QPEI, oxCNTs@50-QPEI, and oxCNTs@80-QPEI was found to be 16.05%, 19.92%, and 23.23%, respectively.

The morphology of the QPEI-functionalized oxCNTs was studied by scanning electron (SEM), transmission electron (TEM), and atomic force (AFM) microscopies. Representative SEM micrographs are shown in Figure 3. It is clear that after functionalization with QPEIs, the morphology of the oxCNTs did not change significantly. Additionally, oxCNTs@QPEIs are shown to be well-dispersed and no aggregation of nanotubes was observed, as in the case of oxCNTs (Figure S2). Especially, functionalization of oxCNTs with 80-QPEI rendered them fully isolated, as shown in Figure 3 (images C and D).

**Figure 3.** SEM images of oxCNTs@30-QPEI (**A**), oxCNTs@50-QPEI (**B**), and oxCNTs@80-QPEI (part **C,D**). The scale bar is 100 nm.

The morphology of QPEI-functionalized oxCNTs, as well as the presence of QPEIs on their surface, was studied by combining TEM bright-field imaging, EFTEM elemental mapping, and EELS spectroscopy. In Figure 4A functionalized carbon nanotubes are observed isolated, without any aggregation, as in the SEM images. These observations suggest that the QPEIs covered the surface of the nanotubes, improving their aqueous dispersibility and debundling. In the HRTEM images the structured graphite walls of oxCNTs can be observed, covered with an amorphous layer of QPEI polymer (Figure 4D,E). For this reason, electron energy loss spectroscopy (EELS) was employed to investigate the spatial distribution of nitrogen, observed in the bright field images of oxCNTs. An energy-filtered TEM (EFTEM) image (utilizing the three-window method), using the nitrogen K-edge at 401 eV electron energy loss, can be seen in Figure 4C, while Figure 4B is the bright field image of the same area. It is obvious that since the intensity of the maps corresponds to the concentration of N (red) that exclusively originated from QPEI, it can be concluded that oxCNTs were uniformly covered by QPEI. Additionally, in a typical background subtracted EELS spectrum the nitrogen K-edges recorded for oxCNTs@80-QPEI (Figure 4) are evidence for the presence of nitrogen atoms and the successful attachment of QPEI on the surface of oxCNTs.

**Figure 4.** TEM bright field image of oxCNTs@50-QPEI (**A**). Bright field image (**B**) and the corresponding EFTEM compositional nitrogen N map (red, **C**) of oxCNTs@50-QPEI, HRTEM images: images of oxCNTs@80-QPEI (**D**–**F**) and a typical background subtracted EELS spectrum of nitrogen K- edges, recorded for oxCNTs@80-QPEI (**G**).

AFM images of oxCNTs@50-QPEI and oxCNTs@80-QPEI, deposited on Si-wafer (Figure 5) show the morphological features of oxCNTs at the nanoscale after the interaction with the polymers. The AFM images (height and 3D) of nanocomposites reveal the successful attachment (wrapping) of polymer on the oxCNTs sidewalls. As derived from topographical section analysis, an overlay of 10–25 nm is observed from each side of nanotube in the case of oxCNTs@50-QPEI, while the size of the polymeric coating is much higher, and easily distinguishable in the case of oxCNTs@80-QPEI, corresponding to an average of 25–40 nm (Figure 5). Moreover, the average diameter of oxCNTs@80-QPEI, as derived from height analysis, is about 40–50 nm, a value much higher than that of oxCNTs in the absence of 80-QPEI, which ranges between 15 and 25 nm (Figure S2).

**Figure 5.** AFM images (height, profile section analysis, and 3D) of oxCNTs@50-QPEI (**A**) and oxCNTs@80-QPEI (**B**), showing the coverage of QPEI derivatives in the sidewalls of oxCNTs.

#### *2.2. Colloidal Stability of the CNTs Dispersions*

CNTs have an extremely strong tendency to aggregate in water due to their high surface energy, making them difficult to disperse in aqueous media resulting in the formation of large bundles [50]. Although the dispersibility of CNTs in aqueous media has been shown to increase following (i) various oxidation processes [51], and (ii) using high concentrations of various surfactants [6,7] or polymers [8,9], the resulting dispersions are only stable for short time. In this study, the negatively charged oxidized CNTs, modified with positively charged QPEIs through electrostatic interactions and van der Waals attraction forces, resulted in functionalized oxCNTs with high positive charge contents, and able to form stable aqueous dispersions. All QPEIs derivatives enhance the aqueous dispersibility of oxCNTs, as revealed by visual observation over time (Figure 6, upper part). It is obvious that stable dispersions of oxCNTs were obtained after functionalization with QPEIs for at least twelve months (Figure 6, upper part), while oxCNTs had precipitated within one month after the sonication process. This was achieved thanks to the presence of quaternary ammonium groups on the surface of the oxCNTs that

provide a high compatibility with aqueous media due to their strong hydrophilicity, while preventing the CNTs' aggregation due to electrostatic repulsion. In an analogous study, involving SWCNTs, Grunlan, J.C. et al. observed that after non-covalent functionalization with PEI, SWCNTs exhibited poor aqueous stability, attributed to PEIs hyperbranched structure that sterically reduced electrostatic interactions [52]. The same behavior was observed in this study for PEI functionalized oxCNTs. In contrast, quaternized PEI derivatives behave differently, probably due to their higher positively charged moieties content.

**Figure 6.** Dispersion state of (1) oxCNTs, (2) oxCNTs@30-QPEI, (3) oxCNTs@50-QPEI, and (4) oxCNTs@80-QPEI in water (5 mg/mL), (**A**) immediately after sonication, and after quiescent settling for (**B**) 2 weeks, (**C**) 1 month, (**D**) 3 months, (**E**) 6 months, and (**F**) 12 months (upper part). (**G**) Sedimentation behavior of oxCNTs and QPEI-functionalized oxCNTs at different aging times (lower part).

It is known that bundled CNTs, unlike individual ones, are not active in the UV–vis region [53] allowing the investigation of their dispersibility using UV–vis absorption spectroscopy. Figure S3 shows the UV–vis spectra of the oxCNTs@30-QPEI, oxCNTs@50-QPEI, and oxCNTs@80-QPEI aqueous dispersions. The dispersions have a characteristic absorption peak at 263 nm, affected by the p-plasmon absorption of carbon nanomaterials. The higher absorption, caused by the p-plasmon from the oxCNTs@80-QPEI, demonstrates more efficient debundling of oxCNTs by 80-QPEI, compared to other carbon nanomaterials [54]. Furthermore, the evaluation of colloidal stability of the CNTs was attained by UV–vis spectroscopy, again after investigation of the characteristic absorption of CNTs at 263 nm. Figure 6G presents the optical density (O.D.) of the oxCNTs and QPEI-functionalized oxCNTs dispersed

in water within the storage periods. It was obvious that the dispersions of all QPEI-functionalized oxCNTs were stable for at least twelve months, since their optical densities were reduced only by 10–20% compared to the initial O.D. On the other hand, the O.D. of the oxCNTs dispersion was reduced by 90% after 12 months storage. The O.D. reduction of oxCNTs dispersion was attributed to the gradual formation of CNTs agglomerates, some of which subsequently aggregated and finally precipitated. These findings are in line with the visual inspection of the CNT dispersions presented in Figure 6A–F. Moreover, the dispersion of oxCNTs@80-QPEI is the most stable since only a 10% reduction of O.D. was observed after 12 months storage. To the best of our knowledge the stability achieved is one of the highest reported in the literature, and its importance lies in the fact that the aqueous stability of such hybrid nanomaterials is of paramount importance in several industrial applications. *ζ*

#### *2.3. Characterization of the QPEI-Functionalized oxCNTs Dispersions*

Dynamic light scattering (DLS) and ζ-potential measurements can provide information on nanomaterials regarding their size distribution, and also their surface charge. Although, DLS measurement is appropriate for determination of the spherical particle diameter, it can also be used to determine the hydrodynamic diameter of nanotubes, based on the assumption that an equivalent hydrodynamic diameter (D*h*) of a sphere has the same diffusion properties as the CNT. [55] Even though one cannot determine absolute values, a relative size comparison can be obtained for similarly shaped materials [56]. Thus, comparing the aggregate size of oxCNTs to those of QPEI-functionalized oxCNTs, it is obvious that debundling of the oxCNTs took place after their interaction with the QPEIs, since the value of hydrodynamic diameter of the oxCNTs decreased from 1300 nm to 150 nm when 80-QPEI was used (Figure 7). *ζ* − *ζ ζ*

*ζ* **Figure 7.** Mean hydrodynamic diameter (solid line) and ζ-potential values (dot line) of oxCNTs and QPEI-functionalized oxCNTs dispersions (0.05 mg/mL, pH = 7.0).

Zeta potential values of the oxCNTs dispersions at pH = 7.0 are given in Figure 7. As expected, the aqueous dispersion of oxCNTs has a ζ-potential value around −34 mV, due to their negative surface charges. After modification with QPEI, the ζ-potential values of the QPEI-functionalized oxCNTs dispersions increased to positive, reaching the value of +60 mV for oxCNTs@80-QPEI, offering further evidence that the positively charged QPEIs were successfully attached onto the oxCNTs surface. It should be noted that all ζ-potential values were higher than +30 mV, indicating stable aqueous colloidal suspensions, in which the CNTs' aggregation was prevented due to electrostatic repulsion, in line with the results of UV–vis and DLS measurements (see above) [57].

#### *2.4. Evaluation of Antibacterial and Anti-Cyanobacterial Activity*

The cytotoxicity of QPEI-functionalized oxCNTs was assessed against two types of Gram-negative bacteria, i.e., the heterotrophic bacterial strain *Escherichia coli* XL1-blue and the autotrophic cyanobacterium *Synechococcus* sp. PCC 7942.

#### 2.4.1. Cytotoxicity Effects of oxCNTs@QPEIs on *Escherichia coli* XL1-Blue Bacteria

*Escherichia coli* growth was investigated by monitoring the fluorescence intensity of bacterial cells suspensions at 37 ◦C that express red fluorescent protein (RFP). The excellent dispersibility of oxCNTs@QPEIs renders the commonly used turbidity measurement inapplicable, as functionalized CNTs, especially at high concentrations, contribute to the final measurement. Thus, by employing the inherent fluorescence of RFP, the antibacterial activity can be precisely assessed, even in the presence of nanoparticle dispersions, as in the case of oxCNTs@QPEIs. Comparing the fluorescence intensity of each bacterial suspension containing oxCNTs@QPEIs at a certain time, with the initial fluorescence intensity corresponding to the initial *Escherichia coli* population (at OD<sup>600</sup> = 0.4), bacterial growth could be directly determined. Figure 8 depicts the *Escherichia coli* growth after 6 h incubation in the presence of oxCNTs and oxCNTs@QPEIs at different concentrations, ranging from 5 to 400 µg/mL, as a function of the fluorescence intensity of RFP at 590 nm (excitation: 545 nm), and normalized with the initial fluorescence intensity of control (100% fluorescence intensity). Contrary to the increase of the fluorescence intensity upon untreated bacteria growth (Figure S4), a decrease in the intensity was observed, for all samples, revealing bacterial growth inhibition. However, as shown in Figure 8, the oxCNTs exhibited low antibacterial activity, which is in accordance with the literature [58]. On the other hand, all QPEI-functionalized oxCNTs inhibited bacterial growth in a dose-dependent manner, displaying significantly higher antibacterial activity than the oxCNTs, and which increased upon substitution of PEI from 30% to 80%. In a further attempt to quantify the antibacterial activity of oxCNTs@QPEIs, the 50% inhibitory concentrations (IC-50) were calculated (Table 1). It was found that the lowest IC-50 was observed in the case of oxCNTs@80-QPEI (28.4 µg/mL), which showed that oxCNTs@80-QPEI exhibited the highest antibacterial activity amongst the other hybrid materials, due to both the increased aqueous dispersibility and the higher positive quaternary ammonium group content.

**Figure 8.** Cytotoxicity of oxCNTs and QPEI-functionalized oxCNTs against gram-negative *Escherichia coli* XL1-blue bacteria. Fluorescence intensity of RFP at 590 nm (excitation: 545 nm) after bacteria incubation for 6 h with oxCNTs@QPEIs at concentrations ranging from 5 to 400 µg/mL, normalized with the initial fluorescence intensity, corresponding to initial *E.coli* population (at OD<sup>600</sup> = 0.4). Error bars represent mean ± SD for at least three independent experiments.

**μ**


**Table 1.** IC-50 values of QPEI-functionalized oxCNTs on *Escherichia coli* XL1-blue bacteria.

The morphology of *Escherichia coli* after 6-h incubation at 37 ◦C with oxCNTs@QPEIs was investigated by scanning electron microscopy (SEM). In Figure 9, SEM images of control (untreated cells) and cells treated with oxCNTs@QPEIs at 50% inhibitory concentrations are presented, showing significant changes in cell morphology. Specifically, the treated cells lost their cellular integrity, and are shown more clustered, while their cell walls seem rougher and damaged in all cases (Figure 9B–D) compared to the untreated cells (Figure 9A), which appear to be intact, with a smooth surface. Moreover, Figure 9D shows the most severe effect of oxCNTs@80-QPEI on the bacterial cell wall and membrane, in which the cell walls seem to be ruptured and bacterial cell lysis is clearly observed probably due to membrane damage.

**Figure 9.** SEM images of *Escherichia coli* bacteria: untreated cells (**A**) and cells after 6-h incubation time at 37 ◦C with oxCNTs@30-QPEI (**B**), oxCNTs@50-QPEI (**C**), and oxCNTs@80-QPEI (**D**) at 50% inhibitory concentrations. The scale bar is 1 µm.

It is known that highly dispersed carbon nanotubes, mainly single wall carbon nanotubes, are able to interact strongly with bacteria through van der Waals forces, forming bacteria-CNTs aggregations [20,25]. This fact results in bacterial death due to either inhibition of transmembrane electron transfer, or to penetration leading to rupture or deformation of cell walls and membranes, which alter the bacterial metabolic processes [59]. Moreover, SWCNTs and MWCNTs containing various types of surface groups were investigated [29] regarding their antibacterial activity towards gram-negative and gram-positive bacteria. It was found that SWCNTs functionalized with hydroxyl and carboxyl surface groups exhibited improved antimicrobial activity against both gram-positive and gram-negative bacteria. However, MWCNTs containing the same functional surface groups did not exhibit any significant antibacterial effect [29]. On the contrary, covalently functionalized MWCNTs with positive moieties such as amines, arginines, and lysines, [60,61] or MWCNTs combined with surfactant molecules, such as dioctyl sodium sulfosuccinate [32], hexadecyltrimethylammonium bromide, triton X-100, and sodium dodecyl sulfate [7], exhibited enhanced antibacterial properties, due to enhanced interactions with bacterial membranes and the improved aqueous dispersibility and stability. In this study, similar antibacterial behavior of QPEI-functionalized oxCNTs was observed due to the high positive quaternary ammonium group content. These positive groups, as in the case of surfactant molecules or other positive functional groups, not only improved the debundling of

MWCNTs, which favors the strong interaction between the bacteria and MWCNTs, but also enhanced the penetration of MWCNTs though cell membranes, resulting in cell lysis and death.

Moreover, it is known that the quaternary ammonium moieties efficiently interact with the negatively charged groups of bacterial walls or cytoplasmic membranes, mainly through electrostatic as well as with secondary hydrophobic interactions, leading to dysfunction in cellular processes and probably to cell death [62]. Highly functionalized polymers with quaternary ammonium groups have been found to be more effective antibacterial agents than their low molecular weight analogues, as their higher charge densities lead to stronger interactions with the negatively charged bacteria walls [63]. For example, poly(propylene imine) dendrimers bearing 16 quaternary ammonium groups per molecule were found to exhibit two orders of magnitude greater antimicrobial activity than their mono-functional counterparts [64]. In agreement with this, the 80-QPEI derivative containing the highest content of quaternary ammonium groups provided the highest polycationic character to oxCNTs, in regards to the other QPEI derivatives. This effect probably induces the strongest interaction with bacteria, and the highest permeability of the cell membrane, and thus oxCNTs@80-QPEI exhibited the best antibacterial activity against the *Escherichia coli* bacteria compared to the other hybrid materials.

#### 2.4.2. Cytotoxicity Effects of oxCNTs@QPEIs on *Synechococcus* sp. PCC 7942 Cyanobacteria

The antibacterial activity of oxCNTs@QPEIs was further assessed against the cyanobacterium *Synechococcus* sp. PCC 7942, a very widespread bacterial strain in the aquatic environment. In general, cyanobacteria (gram negative bacteria) are prokaryotic organisms that perform oxygenic photosynthesis similar to higher plants. They are the oldest and one of the largest and most important groups of bacteria on earth. Cyanobacteria (except prochlorophytes) contain only Chl α, the molecule which makes photosynthesis possible, by passing its energized electrons on to molecules during sugar synthesis [65]. However, several cyanobacterial strains are known to produce a wide range of toxic secondary metabolites (hepatotoxins, neurotoxins, cytotoxins, dermatotoxins, and irritant toxins), which could be harmful to animals and potentially dangerous to humans [66].

In this study, cyanobacteria *Synechococcus* sp. PCC 7942 cell proliferation was monitored by measurement of the Chl α concentration every 24 h, for seven days. Figure 10 shows the cell proliferation of the unicellular cyanobacterium *Synechococcus* sp. PCC 7942 cells in the presence of increasing concentrations of oxCNTs@30-QPEI, oxCNTs@50-QPEI, and oxCNTs@80-QPEI. For comparison reasons, analogous experiments were performed using oxCNTs, and their effect on the cyanobacteria cell proliferation is shown in Figure S5. It is clear that oxCNTs did not inhibit the cyanobacterial cell proliferation in contrast to all oxCNTs@QPEIs. This can be attributed again to the strong positive character of oxCNTs@QPEIs that intensified the interaction with the cyanobacteria membrane, and resulted in higher cell penetration compare to oxCNTs. In order to quantify these results, the effective concentration for 50% inhibition (IC-50), which shows the ability of cells to proliferate under the toxic effect of the oxCNTs@QPEIs, was calculated from the cyanobacterial cell proliferation curves in the presence of each hybrid material (Figure S6 and Table 2). Based on the interpretation of the experimental data using a non-linear regression of the four-parameter logistic function (Figure S6), it is obvious that cell proliferation is concentration dependent, while oxCNTs@80-QPEI exhibited the most promising anti-cyanobacterial activity, compared to the other two hybrid materials. This implies that upon increasing the degree of quaternization, the proliferation rate of cyanobacteria cells decreases. Therefore the anti-cyanobacterial properties, as in case of *Escherichia coli*, depend, not only on the concentration, but also on the degree of quaternization.

*α*

*α*

**Figure 10.** Effect of oxCNTs@QPEIs on cell proliferation of cyanobacteria *Synechococcus* sp. PCC 7942. Growth curves of cyanobacteria in the presence of different concentrations of: oxCNTs@30-QPEI (**A**), oxCNTs@50-QPEI (**B**), and oxCNTs@80-QPEI (**C**). Error bars represent mean ± SD for at least three independent experiments.

**Table 2.** IC-50 values of oxCNTs@QPEIs on cyanobacterium *Synechococcus* sp. PCC 7942.


Triggered by these results, it was interesting to evaluate the effect of oxCNTs@QPEIs on the photosynthetic apparatus of cyanobacteria. Therefore, the activity of Photosystem (PS) I and II was assessed in the presence of oxCNTs@80-QPEI, the material exhibiting the best antibacterial performance, to investigate the consequences of oxCNTs@QPEIs on the integrity of the photosynthetic apparatus in terms of photoinduced electron transport.

Specifically, selective detection of the PSII [67] and PSI electron transporting activities [68] was performed on *Synechococcus* sp. PCC 7942 bacteria treated with lysozyme (permeaplasts) at room temperature [69]. Using *Synechococcus* permeaplasts, oxymetrically photoinduced electron transport activities, across both PSII (electron donor: water; post-PSII electron acceptor: p-benzoquinone) and PSI (post- PSII inhibitor: DCMU; post-PSII electron donor: diaminodurene and ascorbate; post-PSI electron acceptor: methyl viologen) were measured. It was found that upon increase in concentration of oxCNTs@80-QPEI, the rate of oxygen evolution decreases (Table S2), indicating that PSII and PSI electron transport activities depend on the QPEI-functionalized oxCNT concentration. The inhibition of PSI and PSII by oxCNTs and oxCNT@80-QPEI is shown in Figure 11. Similarly to results previously reported in the literature, [70], at high concentrations (250 µg/mL) the oxCNTs used in this study inhibited the PSII and PSI by 37.6% and 95.7%, respectively, while at lower concentrations (20 µg/mL) the PSII is practically unaffected and the PSI activity is reduced by almost 44.8%. However, the impact of oxCNT@80-QPEI was much higher. The novel hybrid with an 80% quaternization degree inhibited the PSII by 16.7% at concentration 20 µg/mL, and by around 53.9% at 250 µg/mL. In the case of the PSI, the effect was even more significant, exhibiting almost complete inhibition (more than 97%), even at low concentrations (20 µg/mL).

**μ**

≤

**Figure 11.** Effect of oxCNTs and oxCNTs@80-QPEI on the photosynthetic electron transport activities of PSII (**A**) and PSI (**B**) in *Synechococcus* sp. PCC 7942 permeaplasts.

The remarkable decrease in the IC-50 values of oxCNTs@QPEIs on cyanobacterial cell proliferation (Table 2) indicates that photosynthetic electron transport of both PSII and PSI is functionally impaired in cyanobacterial cells. Although, oxCNTs inhibit the photosynthetic redox reactions, and in the case of PSI almost fully prevent its activity, the effect of QPEI is significant. This may be attributed to the polycationic character of QPEI, as such compounds are known to completely inhibit PSI reactions, while leaving PSII relatively unaffected [71]. However, under the test conditions oxCNTs did not inhibit cyanobacterial cell proliferation (Figure S5), implying that the oxCNTs could not penetrate the cyanobacteria membranes. On the contrary, as mentioned above, the oxCNTs@QPEIs, exhibited high toxicity against cyanobacteria as a result of their efficient cell penetration (Figure 10).

Furthermore, to elucidate the potential alternative patterns for electron flow to and from PSI, the P700<sup>+</sup> transients in the presence of oxCNTs@80-QPEI were investigated. Table S3 shows the amounts of functional PSI complexes, estimated as the photooxidizable form of the PSI (P700+) reaction center, measured as ∆A820/A820 [72], where an 80% inhibition by oxCNTs@80-QPEI at high concentrations (250 µg/mL) can be observed. In a previous study, MWCNTs were successfully applied for direct transfer of electrons in isolated spinach thylakoids and cyanobacteria Nostoc sp. [73], pointing out that the results obtained in this study may also be associated with an interruption of the electron transport, due to the presence of oxCNTs. Concomitantly, the lower steady state photooxidation of P700 by far red light, might also be considered as an indication that oxCNTs@QPEIs quench the P700+. Based on the above, the observed enhanced anti-cyanobacterial effect of QPEI-functionalized oxCNTs on cyanobacterium *Synechococcus* sp. PCC 7942 is ascribed to the selective inhibition of PSI.

#### *2.5. Cell Viability Assay*

To investigate the cytotoxicity of QPEI-functionalized oxCNTs, the human prostate carcinoma DU145 cell line and the 3T3 mouse fibroblasts were employed. For this purpose, these cells were incubated for 24 h with oxCNTs and oxCNTs@QPEIs at concentrations below and above their IC-50 values and cell viability was assessed, employing the standard MTT assay. The percent cell viability caused by all derivatives is presented in Figure 12. It is obvious that all oxCNTs@QPEIs were not toxic at their IC-50 values, and significantly less lethal than the parent oxCNTs, while at higher concentrations (100–200 µg/mL) only oxCNTs@30-QPEI exhibited slight toxicity (~70% cell survival). It should be noted that at 200 µg/mL, a concentration much higher than their IC-50 values, oxCNTs@50-QPEI and oxCNTs@80-QPEI did not display any cytotoxicity. These results suggest that all oxCNTs@QPEIs simultaneously exhibited both low cytotoxicity and enhanced antibacterial/anti-cyanobacterial properties.

Δ

**Figure 12.** Cytotoxicity of oxCNTs and QPEI-functionalized oxCNTs on DU145 and 3T3 cells following incubation at various concentrations for 24 h, as determined by MTT assays. Data are expressed as mean ± SD of eight independent values obtained from at least three independent experiments.

#### **3. Materials and Methods**

#### *3.1. Chemicals and Reagents*

Hyperbranched poly(ethyleneimine) (PEI) with molecular weight 25 KDa (Lupasol® WF, water-free, 99%) and oxidized multi-walled carbon nanotubes were kindly donated by BASF (Ludwigshafen, Germany) and Glonatech S.A. (Athens Greece), respectively. Glycidyltrimethylammonium chloride, dialysis tubes (molecular weight cut-off: 1200) and triethylamine were obtained from Sigma-Aldrich (St. Louis, MA, USA). D-MEM low glucose with phenol red, l-glutamine, phosphate buffer saline (PBS), fetal bovine serum (FBS), penicillin/streptomycin, and trypsin/EDTA were purchased from BIOCHROM (Berlin, Germany). Thiazolyl blue tetrazolium bromide (MTT) and isopropanol were purchased from Merck KGaA (Calbiochem®, Darmstadt, Germany).

#### *3.2. Synthesis of Quaternized Hyperbranched Poly(ethyleneimine) Derivatives*

Quaternized derivatives of hyperbranched poly(ethyleneimine), with different substitution degrees of primary amino groups were prepared by a method previously described [40,41]. In brief, to an aqueous solution (20 mL) of PEI (5 mM), an aqueous mixture (10 mL) containing, 8, 12, or 18 mmol glycidyltrimethylammonium chloride and 16, 24, or 36 mmol triethylamine, respectively, was added. The reaction was completed after two days at room temperature and the final quaternized derivatives with 30% (30-QPEI), 50% (50-QPEI), and 80% (80-QPEI) degree substitution of primary amino groups were received after dialysis against deionized water and lyophilization. The introduction of the quaternary moieties at the external surface of the parent PEI was confirmed by <sup>1</sup>H and <sup>13</sup>C NMR spectroscopy. Additionally, the degree of quaternization at the primary amino groups of PEI was calculated by the integration of peaks at 3.45 ppm and 2.70–2.50 ppm in the <sup>1</sup>H NMR spectra.

<sup>1</sup>H NMR: (500 MHz, D2O) δ (ppm) = 4.25 (broad s, CH–OH), 3.45 (m, CH2N+(CH3)3), 3.25 (s, CH3), 2.70–2.50 (m, CH<sup>2</sup> of PEI scaffold).

<sup>13</sup>C NMR (125.1 MHz, D2O): δ (ppm) = 71.5 (CH2N+CH3)3), 67.5 (CH–OH), 57.0 (CH3), 55–51.0 (CH<sup>2</sup> of PEI scaffold), 51.0 and 48.0 (CH2NH-Q primary and secondary, respectively), 42.0 and 40.0 (CH2NH<sup>2</sup> close to secondary (C1,2) and tertiary (C1,3) amine, respectively).

#### *3.3. Preparation of QPEI-Functionalized oxCNTs*

The functionalization of oxCNTs was achieved by a method previously described [41]. Specifically, 50 mg of oxCNT powder was dispersed in 50 mL of an aqueous solution, containing an excess quantity (150 mg) of each quaternized derivative, and the resulting dispersions were ultrasonicated for 15 min (Hielscher UP200S high intensity ultrasonic processor equipped with a standard sonotrode (3 mm tip-diameter) at 50% amplitude and 0.5 cycles) and stirred for a further 12 h, at room temperature. The final hybrid materials, oxCNTs@30-QPEI, oxCNTs@50-QPEI, and oxCNTs@80-QPEI, were received after ultracentrifugation at 45,000 rpm, followed by thorough washing with water to remove the unreacted QPEI derivatives and lyophilization.

#### *3.4. Characterization of QPEI-Functionalized oxCNTs*

FTIR spectra were recorded using a Nicolet 6700 spectrometer (Thermo Scientific, Waltham, MA, USA) equipped with an attenuated total reflectance accessory with a diamond crystal (Smart Orbit, Thermo Electron Corporation, Madison, WI, USA). Raman spectra were obtained using a micro-Raman system RM 1000 Renishaw (laser excitation line at 532 nm, Nd-YAG) in the range of 400–2000 cm−<sup>1</sup> . AFM images were obtained in tapping mode, with a 3D Multimode Nanoscope, using Tap-300G silicon cantilevels with a <10 nm tip radius and a ≈20–75 N/m force constant. Samples were deposited onto silicon wafers (P/Bor, single side polished) by drop casting from ethanol solutions. Scanning electron microscopy (SEM) images were recorded using a Jeol JSM 7401F field emission scanning electron microscope equipped with a gentle beam mode. Transmission electron micrographs were taken using a Philips C20 TEM instrument equipped with a Gatan GIF 200 energy filter for electron energy loss elemental mapping. For the sample preparation, a drop of oxCNTs@QPEIs aqueous solution (0.1 mg/mL) was casted on a PELCO® Formvar grid and was left to evaporate. Thermogravimetric analyses (TGA) were carried out on a Setaram SETSYS Evolution 17 system at a 5 ◦C/min heating rate under oxygen atmosphere. Elemental analysis (EA) was measured by a Perkin Elmer 240 CHN elemental analyzer.

#### *3.5. Preparation and Characterization of QPEI-Functionalized oxCNTs Aqueous Dispersions*

The suspensions of polymer-functionalized oxCNTs were prepared by adding 10 mg of oxCNTs@30-QPEI, oxCNTs@50-QPEI, or oxCNTs@80-QPEI into 2 mL pure water. The suspensions were then ultra-sonicated for 5 min using a Hielscher UP200S high intensity ultrasonic processor at 40% amplitude and 0.5 cycles. Each sample was centrifuged at 1500 rpm for 15 min, and then the supernatant was diluted with pure water before measurement.

ζ-potential measurements were performed using a ZetaPlus -Brookhaven Instruments Corp. In a typical experiment, an aqueous 0.05 mg/mL dispersion of QPEI-functionalized oxCNTs was used, ten ζ-potential measurements were collected, and the results were averaged. Dynamic light scattering studies were carried out on an AXIOS-150/EX (Triton Hellas) system equipped with a 30 mW laser source, and an avalanche photodiode detector at 90◦ angle. In a typical experiment, an aqueous 0.05 mg/mL dispersion of QPEI-functionalized oxCNTs was used, at least ten measurements were collected, and the data were analyzed using the CONTIN algorithm to obtain the hydrodynamic radii distribution.

UV-vis spectra of the aqueous dispersions (1 mg/mL) were obtained by a Cary 100 Conc UV-visible spectrophotometer (Varian Inc., Mulgrave, Australia) in the range of 200–600 nm. Additionally, the colloid stability of the functionalized carbon nanotubes was evaluated at static conditions for 1, 6, and 12 months. Specifically, the dispersions obtained as described previously, were placed in vertically standing tubes and stored at room temperature. At each time point, 100 µL of the stock solutions from the very upper part was taken diluted in 1 mL water and the optical density (O.D.) of these dispersions was measured using UV-vis spectroscopy.

#### *3.6. Escherichia coli Growth Inhibition Assay*

The antibacterial activity of QPEI-functionalized oxCNTs was obtained by a bacteria growth inhibition assay. *Escherichia coli* XL1-blue bacteria expressing red fluorescent protein (RFP), from a plasmid-encoded gene, were grown in Luria–Bertani (LB) broth at 37 ◦C overnight, in a Stuart SI500 orbital shaker at approximately 200 rpm shaking speed in aerobic conditions. The culture was subsequently diluted to an optical density (O.D.) of 0.4 at 600 nm. The QPEI-functionalized oxCNTs were homogeneously dispersed in distilled water by sonication and added to the bacterial culture at concentrations ranging from 5 to 400 µg/mL. The assay was performed in a 96-well plate format in a 200 µL final volume. Fluorescence intensity over growth of untreated bacteria revealed that the optimum incubation time was 6 h, as the intensity reached a plateau (Figure S4). Thus, plates were incubated at 37 ◦C, shaking at 100 rpm in aerobic conditions for 6 h, and bacterial growth was monitored using the fluorescence intensity of red fluorescent protein, which was recorded at an emission wavelength of 590 nm by an Infinite M200 plate reader (Tecan group Ltd., Männedorf, Switzerland) at an excitation wavelength of 545 nm. In order to eliminate the effect of CNTs in the measured intensities, the initial values (at 0 h), although minor compared to those obtained after 6 h, were subtracted from the final measurements. For each treatment eight replicates were used and three independent experiments were performed. Untreated bacteria were used as control, representing 100% fluorescence intensity in Figure 8.

#### *3.7. SEM Analysis of the Cellular Morphology*

The morphology of the *Escherichia coli* bacteria after treatment with QPEI-functionalized oxCNTs was characterized by scanning electron microscopy (Jeol JSM 7401F Field Emission SEM). Specifically, cells were incubated with QPEI-functionalized oxCNTs at their 50% inhibitory concentration (IC-50), fixed with 3% glutaraldehyde in sodium cacodylate buffer (100 mM, pH = 7.1) for 6 h, transferred to a poly(l-lysine) coated glass cover slip, dehydrated using ethanol gradient (twice of 50%, 70%, 95%, and 100% ethanol for 10 min each), drying, and coated with gold in a sputter coater [74].

#### *3.8. Synechococcus sp. PCC7942 Cyanobacteria Growth Inhibition Assay*

*In vitro* anti-cyanobacterial activity of QPEI-functionalized oxCNTs was screened against *Synechococcus* sp. PCC 7942 bacteria. The unicellular cyanobacterium *Synechococcus* sp. PCC7942 was purchased from the Collection Nationale de Cultures de Microorganismes (CNCM), Institut Pasteur, Paris, France. The cyanobacterial cells were cultured in BG11 that additionally contained 20 mM HEPES-NaOH (pH = 7.5). The cultures were illuminated with white light from fluorescent lamps, providing a photosynthetic active radiation (PAR) of 100 µmol photons m-2 s -1, and were aerated with air containing 5% (*v*/*v*) CO<sup>2</sup> in an orbital incubator (Galenkamp INR-400) at 31 ◦C [75]. QPEI-functionalized oxCNTs dispersed in distilled water using ultrasonication were added to the bacterial culture at concentrations ranged from 10 to 100 µg/mL. Cyanobacterial cell proliferation was monitored in terms of concentration of Chl α, determined in *N*,*N*-dimethylformamide (DMF) extracts [76]. To extract Chl α, the cell suspensions were centrifuged, DMF was added to the residue, and the resulting clear supernatant DMF extract was obtained after a second centrifugation.

Toxicity tests were performed in three replicate experiments using at least five geometrically scaled dilutions for each compound concentration. The cyanobacteria culture was inoculated in each test solution in the exponential growth phase at concentrations of approximately 1 µg Chl α/mL. The toxicity of the oxCNTs@QPEIs was evaluated as the effective concentrations (µg/mL) of the test substance inhibiting cell proliferation by 50% (IC-50) relative to the control cultures; in this test, the IC-50 values were calculated by the area under the growth curves (biomass) for each concentration of the hybrid materials, using non-linear regression of a 4-parameters logistic function. The related data are presented in Figure S6.

#### *3.9. Measurements of Photosystem I and II Electron Transport Activities*

Photo-induced electron transport rates were determined in *Synechococcus* permeaplasts [68] at room temperature oxymetrically (for each Photosystem) with a Clark-type oxygen electrode (DW1; Oxygraph, Hansatech, King's Lynn, UK). *Synechococcus* sp. PCC 7942 bacteria were treated with lysozyme to obtain permeaplasts before the measurement of the photosynthetic electron transport activities [69]. The instrument was fitted with a slide projector to provide actinic illumination of samples. PSI activity was determined by measuring the rate of oxygen uptake, in the presence of the post-PSII electron transfer inhibitor 3-(3,4-dichlorophenyl)-1,10-dimethylurea (DCMU), using Na ascorbate/diaminodurene as an electron donor to PSI and methyl viologen as a post-PSI electron acceptor and mediator of oxygen uptake [77]. The reaction mixture (1 mL, in buffered BG11) contained permeaplasts (5 µg Chl α), diaminodurene (1 mM), Na-ascorbate (2 mM), methyl viologen (0.15 mM), and DCMU (0.01 mM). PSII activity was determined by measuring the rate of oxygen evolution, with water as electron donor and p-benzoquinone as post-PSII electron acceptor. The reaction mixture (1 mL in buffered BG11) contained permeaplasts (5 µg Chl α/mL) and p-benzoquinone (1 mM).

The redox state of P700 was determined *in vivo* using a PAM-101-modulated fluorometer (Heinz Walz GmbH, Effeltrich, Germany), equipped with an ED-800T emitter-detector, and PAM-102 units, following the procedure of Schreiber at al. [64]. The redox state of P700 was evaluated as the absorbance change around 820 nm (∆A820/A820).

#### *3.10. Cell Cytotoxicity*

In this study, human prostate carcinoma cell line DU145 and 3T3 mouse fibroblasts, obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA), were used. Cells were grown in low glucose supplemented D-MEM, containing 10% FBS, penicillin/streptomycin solution (100 U/mL + 100 µg/mL), and 2 mM L-Glutamine. Cells were incubated at 37 ◦C in a humidified atmosphere, containing 5% CO<sup>2</sup> and sub-cultured, twice a week after detaching with a solution containing 0.05% (*w*/*v*) trypsin and 0.02% (*w*/*v*) EDTA. The cytotoxicity of oxCNTs and the QPEI functionalized oxCNTs was assessed employing MTT assay. DU145 cancer cells and 3T3 mouse fibroblasts were inoculated (10<sup>4</sup> cells/well) into 96-well plates and incubated in complete media for 24 h. Cells were then treated with various concentrations of oxCNTs@QPEIs for 24 h. The mitochondrial redox function (translated as cell viability) of all cell groups was measured by the MTT assay. In brief, cell media was replaced with complete media containing MTT (10 µg/mL) and incubated at 37 ◦C in a 5% CO<sup>2</sup> humidified atmosphere for 3 h. Then, the supernatant containing MTT was discarded and the produced formazan crystals were dissolved in isopropyl alcohol (100 µL per well) under shaking for 10 min at 100 rpm in a Stuart SI500 orbital shaker. Finally, the endpoint absorbance measurements at 540 nm were carried out, employing an Infinite M200 plate reader (Tecan group Ltd., Männedorf, Switzerland). Eight replicates were performed for each concentration, and the experiment was repeated in triplicate. The relative cell viability was calculated as cell survival percentage compared to cells that were treated only with complete media (control). Blank values measured in wells with isopropyl alcohol and no cells, were in all cases subtracted.

#### **4. Conclusions**

In this study, negatively charged oxidized multi-walled carbon nanotubes (oxCNTs) were modified with positively charged quaternized hyperbranched poly(ethyleneimine) derivatives (QPEIs), through non-covalent functionalization. Specifically, three derivatives of hyperbranched poly(ethyleneimine), with a 30, 50, and 80% substitution degree of primary amino groups, were prepared and, subsequently, physically interacted with oxCNTs, yielding three novel QPEI functionalized oxCNTs, with QPEI loading ranged between 16–23%, approximately. Structural characterization of these hybrid materials using a variety of techniques, such as FTIR, RAMAN, SEM, TEM, AFM, etc., revealed the successful and homogenous anchoring of QPEIs on the oxCNTs surface. Furthermore, the microscopic techniques

revealed the effective wrapping of the QPEI over the ox-CNTs. Contrary to previous studies on non-covalent functionalization of CNTs with PEI, the obtained hybrids efficiently dispersed in aqueous media, forming dispersions with excellent aqueous stability for over 12 months. To evaluate the antibacterial and anti-cyanobacterial properties of these hybrids, two types of gram(−) bacteria, an autotrophic (cyanobacterium *Synechococcus* sp. PCC 7942) and a heterotrophic (bacterium *Escherichia coli*), were used. It was found that all materials exhibited an enhanced, dose-dependent antibacterial and anti-cyanobacterial activity against both test organisms. The obtained IC-50 values were much lower compared to oxidized MWCNTs, revealing that the non-covalent attachment of QPEIs strongly induces the antibacterial/anti-cyanobacterial properties of the hybrid materials. These improved properties were attributed to the polycationic character of the oxCNTs@QPEIs, which enables the effective interaction of the hybrids with the bacteria membranes, facilitating their internalization into the cells. Moreover, the excellent aqueous dispersibility and stability of the hybrids, upon increasing the quaternization degree, further enhanced their activity. Indeed, the QPEI derivative containing the highest content of quaternary ammonium groups (80-QPEI) exhibited the highest performance, compared to the other QPEI derivatives. In the case of the photosynthetic bacteria, it was shown that the hybrid materials affect their photosynthetic apparatus by selective inhibition of the Photosystem (PS) I electron transport activity, while also reducing the photosynthetic electron transport in PSII. To the best of our knowledge, the QPEI functionalized hybrids are the first materials exhibiting strong anti-cyanobacterial properties, without the use of any antibiotic/herbicide. Furthermore, cytotoxicity studies on human prostate carcinoma DU145 cell line and the 3T3 mouse fibroblasts were performed, revealing that all hybrids exhibit high cytocompatibility in the concentration range in which they also exhibit high antibacterial and anti-cyanobacterial properties. These results suggest that QPEI-functionalized oxCNTs can be very attractive candidates as antibacterial and anti-cyanobacterial agents that can be used for potential applications in the disinfection industry, as well as for control of harmful cyanobacterial blooms.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/1424-8247/13/10/293/s1, Figure S1: FTIR spectra of oxCNTs, 30-QPEI, oxCNTs@30-QPEI, 50-QPEI and oxCNTs@50-QPEI, Figure S2: SEM images (upper part), AFM image and profile section (lower part) of oxCNTs, Figure S3: UV–vis absorption spectra of oxCNTs (a), oxCNTs@30-QPEI (b), oxCNTs@50-QPEI (c) and oxCNTs@80-QPEI (d) in aqueous solution (1 mg/mL), Figure S4: Fluorescence intensity change of RFP at 590 nm (excitation: 545 nm) upon *Escherichia coli* XL1-blue bacteria growth. Figure S5: Effect of oxCNTs on cell proliferation of cyanobacteria *Synechococcus* sp. PCC 7942 in the presence of different concentrations. Error bars represent mean ± SD for at least three independent experiments. Figure S6: Plot of the area under the growth curves of *Synechococcus* sp. PCC 7942 cells for each concentration of oxCNTs@PEIs versus the corresponding concentration as well as the relevant IC-50 calculations. Table S1: Elemental analysis results of ox-CNTs, QPEI and QPEI-functionalized ox-CNTs. Table S2: Photosystem II and I electron transport activities measured on *Synechococcus* sp. PCC 7942 permeaplasts in the presence of oxCNTs@80-QPEI. Table S3: Effects of oxCNTs@80-QPEI on the steady state oxidation of P700 (∆A820/A820) by FR light in *Synechococcus* sp. PCC 7942 cells.

**Author Contributions:** Conceptualization, Z.S.; Data curation, N.S.H., G.K., K.M.L., K.N.P., A.P., E.S. and A.K.; Formal analysis, N.S.H., G.K., K.M.L., K.N.P., A.P., E.S., S.P. and A.K.; Investigation, D.G. and F.K.K.; Methodology, N.S.H., G.K., K.M.L., K.N.P., A.P., E.S., S.P., A.K., F.K.K. and K.S.; Project administration, D.G. and Z.S.; Resources, K.S.; Supervision, F.K.K. and Z.S.; Validation, K.N.P.; Visualization, S.P.; Writing—original draft, F.K.K., K.S. and Z.S.; Writing—review & editing, S.P., D.G., F.K.K. and Z.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was partially financed by the Greek General Secretariat for Research and Technology, under the frame of EuroNanoMed III, ANNAFIB project (MIS 5053890) and by the internal project entitled: "Synthesis and characterization of nanostructured materials for environmental applications" (EE11968). K.M.L. also thanks the Greek State Scholarships Foundation (MIS 5000432, contract number: 2018-050-0502-13820).

**Acknowledgments:** This research was supported by the Greek General Secretariat for Research and Technology, under the frame of EuroNanoMed III, ANNAFIB project (MIS 5053890) and by the internal project entitled: "Synthesis and characterization of nanostructured materials for environmental applications" (EE11968). K.M.L. acknowledges financial support from the Greek State Scholarships Foundation, program "Enhancement of human scientific resources through implementation of PhD research" with resources of the European program "Development of human resources, Education and lifelong learning", 2014–2020, co-funded by the European Social Fund and Greek State (MIS 5000432, contract number: 2018-050-0502-13820).

#### **Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

### **Potential of Cell-Free Supernatant from** *Lactobacillus plantarum* **NIBR97, Including Novel Bacteriocins, as a Natural Alternative to Chemical Disinfectants**

**Sam Woong Kim <sup>1</sup> , Song I. Kang <sup>1</sup> , Da Hye Shin <sup>1</sup> , Se Yun Oh <sup>1</sup> , Chae Won Lee <sup>2</sup> , Yoonyong Yang <sup>2</sup> , Youn Kyoung Son <sup>2</sup> , Hee-Sun Yang <sup>2</sup> , Byoung-Hee Lee <sup>2</sup> , Hee-Jung An <sup>3</sup> , In Sil Jeong 4,\* and Woo Young Bang 2,\***


Received: 6 August 2020; Accepted: 21 September 2020; Published: 23 September 2020

**Abstract:** The recent pandemic of coronavirus disease 2019 (COVID-19) has increased demand for chemical disinfectants, which can be potentially hazardous to users. Here, we suggest that the cell-free supernatant from *Lactobacillus plantarum* NIBR97, including novel bacteriocins, has potential as a natural alternative to chemical disinfectants. It exhibits significant antibacterial activities against a broad range of pathogens, and was observed by scanning electron microscopy (SEM) to cause cellular lysis through pore formation in bacterial membranes, implying that its antibacterial activity may be mediated by peptides or proteins and supported by proteinase K treatment. It also showed significant antiviral activities against HIV-based lentivirus and influenza A/H3N2, causing lentiviral lysis through envelope collapse. Furthermore, whole-genome sequencing revealed that NIBR97 has diverse antimicrobial peptides, and among them are five novel bacteriocins, designated as plantaricin 1 to 5. Plantaricin 3 and 5 in particular showed both antibacterial and antiviral activities. SEM revealed that plantaricin 3 causes direct damage to both bacterial membranes and viral envelopes, while plantaricin 5 damaged only bacterial membranes, implying different antiviral mechanisms. Our data suggest that the cell-free supernatant from *L. plantarum* NIBR97, including novel bacteriocins, is potentially useful as a natural alternative to chemical disinfectants.

**Keywords:** AMP; antimicrobial activity; antiviral activity; bacteriocin; COVID-19; disinfectant; *Lactobacillus plantarum*; plantaricin

#### **1. Introduction**

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), responsible for the global pandemic of coronavirus disease 2019 (COVID-19), is the foremost concern among recent global health issues [1]. For prevention of this infection, disinfectants have been widely used—mainly because SARS-CoV-2, like other coronaviruses and enveloped viruses, is surrounded by a fragile outer lipid envelope, which makes it more susceptible to disinfectants than non-enveloped viruses such as

rotavirus, norovirus, and poliovirus [2]. Accordingly, the pandemic of COVID-19 has led to a large surge in demand for disinfectants, especially chemical disinfectants such as alcohol- or chlorine-based formulas for the disinfection of hands or environmental surfaces [3–5]. Although chemical disinfectants are considered very effective, they could be hazardous to users if they are not properly handled; for example, alcohol-based disinfectants are flammable and can be harmful to humans if they enter the body [3]. For this reason, there is increasing interest in disinfectants based on natural products.

Lactic acid bacteria, traditionally used in fermented foods, have been considered as interesting resources to contribute to developing a safe alternative to biocides, which are potentially hazardous to humans, because they produce diverse antimicrobial substances and are seldom hazardous to humans [6,7]; most are approved by the U.S. Food and Drug Administration as GRAS (Generally Recognized as Safe). As typical antimicrobial substances, they secrete lactic acid with bacteriocins and antimicrobial peptides (AMPs), which are produced by most microbes [6,7]. In particular, bacteriocins, such as nisin, sakacin, plantaricin, and leucocin from lactic acid bacteria have been reported to have antibacterial activity against foodborne bacteria, such as *Escherichia coli*, *Salmonella enterica*, and *Listeria monocytogenes*, and thus many studies have highlighted their application as natural alternatives to artificial preservatives and antibiotics [6,8–10]. In addition, several bacteriocins have shown antiviral activities against pathogenic viruses such as poliovirus, herpes simplex virus, and influenza viruses [10–12]. Accordingly, the cell-free supernatant, including the bacteriocins and lactic acid, has potential as a natural alternative to chemical disinfectants, although there have been no attempts to apply it as a disinfectant as of yet. To the best of our knowledge, this report is the first that addresses these issues.

In this study, we first suggest that the cell-free supernatant from *Lactobacillus plantarum* NIBR97, a lactic acid bacterium isolated from kimchi, a Korean fermented food, could potentially be useful for disinfection against both pathogenic bacteria and viruses, mediated by bacteriocins as well as lactic acid. Through the genomic analysis of the NIBR97 strain, we discovered novel bacteriocins functioning as antibacterial and antiviral peptides. Our study will provide important information that will guide new strategies to replace chemical disinfectants with natural substances.

#### **2. Results**

#### *2.1. Antibacterial Activity of Cell-Free Supernatant from L. plantarum NIBR97*

*Lactobacillus plantarum* NIBR97 were screened from kimchi as a strain with superior antibacterial activity, andits cell-free supernatantwas further used for the examination of antibacterial activity, as shownin Figure 1. The minimum inhibitory concentrations (MIC50 and MIC90) were determined as 30.04 and 67.43 µg total proteins/mL against *Salmonella enterica* Serovar Enteritidis (*S*. Enteritidis), respectively, which indicates significantly higher antibacterial activity than the three *Lactobacillus plantarum* strains, KCTC33131, KCTC21004, and KCTC13093, with higher MIC50 and MIC90 values than NIBR97 (Figures 1A and S1A). The cell-free supernatant also showed MIC50s and MIC90s against *Salmonella* Gallinarum, *Edwardsiella tarda*, *Pasteurella multocida*, and *Streptococcus iniae* (Figures 1B and S1B), implying antibacterial activity against broad pathogenic bacteria. In addition, when *Escherichia coli* and *Staphylococcus aureus* were treated with the cell-free supernatant for 5 min, they showed a reduction of at least 99.9% (≥3log10) of the total count in the original inoculum (Figure 1C), indicating bactericidal activity and potential as a disinfectant.

μg 70.8 μg 70.8 μg **Figure 1.** Antibacterial activity of cell-free supernatant from *Lactobacillus plantarum* NIBR97. Antibacterial activities of the cell-free supernatant from the *L. plantarum* strains NIBR97, KCTC33131, KCTC21004, and KCTC13093 were examined against *Salmonella* Enteritidis, whose MIC50s and MIC90s were determined (**A**). The MIC50s and MIC90s of the cell-free supernatant were determined against *Salmonella* Gallinarum (SG), *Edwardsiella tarda* (ET)*, Pasteurella multocida* (PM), and *Streptococcus iniae* (SI), as well as *S.* Enteritidis (SE) (**B**). For bactericidal activity, *Escherichia coli* and *Staphylococcus aureus* were treated with the cell-free supernatant (126.6 µg total proteins/mL) for 5 min, and then were counted to determine the titer (Log<sup>10</sup> (colony-forming unit (CFU)/mL) and reduction rate (%) (**C**). For scanning electron microscopy, *S.* Enteritidis was treated without (control) or with the cell-free supernatant (70.8 µg total proteins/mL, MIC against *S.* Enteritidis) for 1 h and 8 h (**D**). The red arrows indicate the pores forming in the *Salmonella* membrane. (**E**) To investigate the effect of protease on the antibacterial activity of the cell-free supernatant, we added the proteinase K (100 µg/mL) to the cell-free supernatant at 70.8 µg total proteins/mL and the treated sample was used to examine its antibacterial activity against *S.* Enteritidis. In (**E**), the plus (+) mark indicates the treatment of cell-free supernatant or proteinase K, whereas the minus (−) mark does no treatment, and the proteinase K was inactivated at 80 ◦C for 10 min (+) or not (−). The different letters (A, B, C, a, b and c) in the graphs ((**A**), (**B**), (**C**) and (**E**)) represent significant differences (*p* < 0.05) and in (**A**) and (**B**), the capital (A, B and C) and small letters (a, b c) indicate the significant differences in MIC50 and MIC90 data, respectively.

In order to prove the antibacterial activity against pathogenic bacteria with the cell-free supernatant from *L. plantarum* NIBR97, we observed the *S*. Enteritidis treated with the cell-free supernatant using scanning electron microscopy (SEM). As shown in Figure 1D, the SEM images revealed that the cell-free supernatant effectively caused the Salmonella death via pore formation by cellular penetrating peptides, as is the case for typical AMPs [13]. Furthermore, when the cell-free supernatant was treated with proteinase K, its antibacterial activity against *S.* Enteritidis decreased by about 50% compared with the control without the proteinase K treatment (Figure 1E). Therefore, we suggest that proteins or peptides play major roles for the antibacterial activities of cell-free supernatant from *L. plantarum* NIBR97.

#### *2.2. Antiviral Activity of Cell-Free Supernatant from L. plantarum NIBR97*

To assess its antiviral activity, the cell-free supernatant from *L. plantarum* NIBR97 was exposed to green fluorescent protein (GFP)-labeled lentiviruses, based on human immunodeficiency virus (HIV), which causes acquired immunodeficiency syndrome (AIDS), for 5 min and 24 h, as shown in Figure 2. When the GFP-labeled lentiviruses, treated with the cell-free supernatant, infected the HEK-293T cells (human host cells), they were observed by fluorescence microscopy to decrease dose- and time-dependently within the host cells (Figure 2A, the GFP images) without any cytotoxic effect on the human host cells (Figure 2A, the Bright images). SEM also confirmed its antiviral activity by showing that the cell-free supernatant effectively causes lentiviral lysis through the collapse of envelopes after 5 min (Figure 2B). In addition, when the human influenza A virus subtype H3N2 (A/H3N2) was treated with the cell-free supernatant, it showed a reduction of at least 99.5% of the total count of its original inoculums, which increased until 99.999% with treatment time (Table 1). These results indicate that the cell-free supernatant from *L. plantarum* NIBR97 has superior antiviral activity, as well as potential as a disinfectant.

 μg μg **Figure 2.** Antiviral activity of cell-free supernatant from *L. plantarum* NIBR97. (**A**) For fluorescence microscopy, we treated GFP-labeled lentiviruses with the cell-free supernatant for 5 min and 24 h, and then were infected in HEK-293T human host cells. The 1× and 5× correspond to the concentrations treated to the lentiviruses, 79.15 and 395.75 µg total proteins/mL, respectively. The bright-field images (Bright) indicate the HEK-293T cells, and the green signals in the fluorescent images (GFP) represent the GFP-labeled lentiviruses. (**B**) For scanning electron microscopy, the GFP-labeled lentiviruses were treated without (**a**) or with the cell-free supernatant (395.75 µg total proteins/mL) (**b**) for 5 min.


**Table 1.** Disinfection activity of the cell-free supernatant from *L. plantarum* NIBR97 against A/H3N2.

<sup>1</sup> The A/H3N2 viruses were treated with water, a negative control, or the cell-free supernatant (NIBR97) for 10 min, 30 min, and 18 h; <sup>2</sup> and <sup>3</sup> indicate the viral titer (log10CCID50) and reduction (%), respectively.

#### *2.3. Discovery of Novel Bacteriocins by the Genomic Analysis of L. plantarum NIBR97*

Analysis of the whole-genome sequence for the *L. plantarum* NIBR97 was carried out by the PacBio RS II (Pacific Biosciences, Menlo Park, CA, USA) sequencing platform to identify the AMPs from the NIBR97. The NIBR97 genome identified from de novo assembly was composed of a single circular bacterial chromosome and four plasmids, containing 2927 predicted open reading frames (ORFs), 68 tRNAs, and 16 rRNAs (Table 2 and Figure S2). Among the ORFs, 10 were identified to encode homologous proteins with known AMPs via an NCBI (National Center for Biotechnology Information) homology BLAST (Basic Local Alignment Search Tool) (Table S1). Furthermore, their expression in *L. plantarum* NIBR97 was confirmed by the transcriptomic data (Table S2). In detail, the five ORFs—orf02155, orf02163, orf02164, orf02421, and orf00645—were found to have 100% identities with plantaricin N, F, and E, as well as bacteriophage holing and lysozyme, known previously as AMPs from the *Lactobacillus* genus (Table S1). Five ORFs—orf00467, orf01336, orf01363, orf01599, and orf01790—which were previously uncharacterized until now, were discovered in this study to consist of amino acid sequences with high positives (>60%) with AMPs undiscovered in *L. plantarum* strains: grammistin Pp3, indolicidin, bactofencin A, hymenochirin-5B, and latarcin-2a, (Table S1). Thus, we herein designated the AMPs as novel bacteriocins called plantaricin (Pln) 1, 2, 3, 4, and 5 (Table S1). Interestingly, their structural models revealed that the three plantaricins—Pln 1, 4 and 5—form helix structures, and the two plantaricins—Pln 2 and 3—form random coil structures (Figure 3), similar to typical AMPs [14,15], implying that they may have antibacterial activities.

#### *2.4. Antibacterial and Antiviral Activities of Plantaricins from L. plantarum NIBR97*

To confirm whether the five Plns function as AMPs, we assessed their synthetic peptides for antibacterial activity against *Salmonella* Typhimurium (Figure S3). Among them, Pln 5 exhibited the highest antimicrobial activity, showing the lowest MIC50 compared with others, whereas Pln 4 showed the lowest antimicrobial activity (Figure S3). In addition, the Pln 3 and 5 were identified to inhibit the growth of *Salmonella* Enteritidis (Figure 4A) and were observed by SEM to effectively cause cellular lysis by damaging the membrane of *S*. Enteritidis via pore formation (Figure 4B), as did the cell-free supernatant from *L. plantarum* NIBR97 (Figure 1D).


**Table 2.** Summary of the de novo genome assembly of *L. plantarum* NIBR97.

<sup>1</sup> and <sup>2</sup> indicate the length (bp, base pair) and GC (guanine-cytosine) contents (%) of contig in the form, respectively; 3 , 4 , and <sup>5</sup> represent the number of predicted open reading frames (ORFs), rRNA, and tRNA, respectively.

—

— —

— —

—

— —

**Figure 3.** Structural models of plantaricins. Pln 1, 2, 3, 4, and 5 comprise the amino acid sequences **Figure 3.** Structural models of plantaricins. Pln 1, 2, 3, 4, and 5 comprise the amino acid sequences VLGSLIGSVGIGVLSSLAARYK, IYPEKQPEEPVRR, KKSRRCQVYNNGMPTGMYTSC, PIVREPFKAMAVGIILAVMSGLLVT, and KAKKRFLRNRLSQQARKARTK, respectively. Pln 1, 4, and 5 form helix structures, and Pln 2 and 3 form random coil structures. The structures of Pln 1, 2, 3, 4, and 5 were predicted by the automated I-TASSER server (https://zhanglab.ccmb.med.umich.edu/I-TASSER/).

**Figure 4.** Antibacterial activity of plantaricin 3 and 5 against *S*. Enteritidis. Pln 3 and 5 were synthesized according to the amino acid sequences in Figure 3, and further examined for their antibacterial activity against *S*. Enteritidis (**A**). The *y*-axis and different letters (A, B, C, a and b) in the graphs represent the relative bacterial growth (%) and significant differences (*p* < 0.05), respectively. In (**A**), the capital (A, B and C) and small letters (a and b) indicate the significant differences between different concentrations (0-5 mg/mL) and the ones between Pln 3 and 5, respectively. (**B**) For scanning electron microscopy, *S.* Enteritidis was treated without or with synthetic Pln 3 or 5 (5 mg/mL) for 1 h and 8 h. The red arrows indicate the pores forming in the *Salmonella* membrane.

 ≈ The synthetic Pln 3 and 5 were further examined for antiviral activity against GFP-labeled lentiviruses. The synthetic peptides exhibited a cytotoxicity on the human host cells when the lentiviruses were treated with 5 µg/µL of synthetic peptides, but not with ≈2.5 µg/µL of synthetic peptides (Figure 5, the Bright images). The fluorescence microscopy revealed that the lentiviruses decreased considerably within the host cells when they were treated with the Pln 3 or 5 for 24 h, but not for 5 min (Figure 5, the GFP images). This suggests that Pln 3 and 5 can considerably suppress viral infection in host cells. Interestingly, SEM revealed that Pln 3 effectively caused lentiviral lysis through the collapse of the envelopes (Figure 6), as the cell-free supernatant did (Figure 2B), whereas Pln 5 did not (Figure 6). This implies that Pln 3 and 5 may exert their antiviral role through different mechanisms.

≈

 **Figure 5.** Fluorescence micrographs of HEK-293T cells infected with GFP-labeled lentiviruses treated with synthetic Pln 3 and 5. The lentiviruses were treated without (control) or with the synthetic peptides (1 to 5 µg/µL) Pln 3 (**A**) and 5 (**B**) for 5 min or 24 h, and then the HEK-293T human host cells were infected. The bright-field images (Bright) indicate the HEK-293T cells, and the green signals in the fluorescent images (GFP) represent the GFP-labeled lentiviruses.

 **Figure 6.** Scanning electron micrographs of the GFP-labeled lentiviruses treated with synthetic Pln 3 and 5. The lentiviruses were treated without or with the synthetic peptides at 5 µg/µL for 24 h.

#### **3. Discussion**

– *Lactobacillus plantarum* is one of the most widespread lactic acid bacteria species and is largely used for the production of fermented products of animal and plant origin [16]. Moreover, some strains are known to produce several natural antibacterial substances, such as bacteriocins, organic acids (mainly lactic and acetic acid), and hydrogen peroxide [17,18], and thus many studies have highlighted their application as preservatives and antibiotics [6,8–10]. Here, we investigated their potential as a natural alternative to chemical disinfectants.

In this study, the NIBR97 strain was screened from kimchi, a Korean fermented food, and its cell-free supernatant was identified to have higher antibacterial activity against *Salmonella* bacteria than other *L. plantarum* strains (Figure 1A), as well as possessing antibacterial activities against a broad range of pathogenic bacteria (Figure 1B). It exhibited significant disinfection activities against the human pathogens influenza A virus H3N2, *Escherichia coli*, and *Staphylococcus aureus*, reducing them by at least 99.9% of the total count of their original inoculums within 30 min (Figure 1C and Table 1). These results indicate that the cell-free supernatant from *L. plantarum* NIBR97 has potential as a natural disinfectant, and thus further investigations were performed to identify the antimicrobial substances, such as AMPs, in the NIBR97 strain.

AMPs are small peptides composed of 10 to 40 amino acids, which cause microbial membrane modification via either pore formation by cell-penetrating property through a barrel stave or a toroidal pore mechanism, or through a non-pore carpet-like mechanism [13,19]. Our scanning electron micrographs of *S*. Enteritidis showed clearly that the cell-free supernatant from the NIBR97 formed a pore on the *Salmonella* surface (Figure 1D), as do typical AMPs [13]. Proteinase K treatment of the cell-free supernatant led to a considerable decrease in its antibacterial activity against both *S*. Enteritidis and *S*. Gallinarum (Figure 1E). Thus, these results confirm that the antibacterial activities of the cell-free supernatant from the NIBR97 are mediated mainly by its proteins or peptides, functioning as AMPs. The scanning electron micrographs of GFP-labeled lentivirus showed that the cell-free supernatant causes lentiviral lysis through envelope collapse (Figure 2A), but it was unclear whether the AMPs were involved in the envelope collapse of the virus.

Finally, to identify AMPs from the NIBR97 strains, we performed whole-genome sequencing, which revealed that the 10 ORFs encoded AMPs, including known forms (plantaricin E, F, N; bacteriophage holin; lysozyme) and novel forms (Pln 1 to 5 (Table S1)). In the case of the known AMPs, plantaricin E, F, and N are bacteriocins produced in *Lactobacillus plantarum* C11 [20]; holin, produced by bacteriophages, triggers and controls the degradation of the cell wall of the host bacteria [21]; and lysozyme functions as 1,4-beta-N-acetylmuramidase, an antimicrobial enzyme, and has been found mainly in *Lactobacillus rhamnosus* strains [22]. Interestingly, five ORFs were discovered as novel bacteriocins in this study (Figure 3) and were designated as Pln 1, 2, 3, 4, and 5. They were further confirmed as being expressed in the NIBR97 strain through transcriptomic sequencing (Table S2), and even their synthetic peptides exhibited antibacterial activity against *Salmonella* Typhimurium (Figure S3). The synthetic Pln 3 and 5 also inhibited the growth of *S.* Enteritidis and effectively caused cellular lysis through damage to the *Salmonella* membrane via pore formation (Figure 4), suggesting that they function as AMPs. However, the synthetic Plns showed overall lower antibacterial activities than antibiotics such as octenidine when their MICs were compared with each other (Figure S3) [23]. This is presumably because the Plns were not synthesized on the basis of complete amino acid sequences for the optimal antibacterial activity but were done on the basis of minimal sequences for the activity; thus, it is further necessary to identify the mature peptide sequence responsible for the optimal antibacterial activity, following the signal peptide cleavage. Moreover, Pln 3 and 5 were identified to suppress lentiviral infection in human host cells (Figure 5). Collectively, these results suggest that the cell-free supernatant from *L. plantarum* NIBR97 may include AMPs, such as Pln 3 and 5, exhibiting antibacterial and antiviral activities. However, Pln 3 and 5 were observed by SEM to act differentially in the suppression of viral infection; Pln 3 had a significant effect on the viral shape through the collapse of the viral envelope, which suggests that it may cause direct damage to the envelope. In contrast, Pln 5 had little effect on it (Figure 6), which implies that it may interfere with the interaction between viruses and host cells [24,25].

Noticeably, Pln 3 and 5 suppressed viral infection when used against lentivirus for 24 h, but not for 5 min (Figure 5), which indicates that long exposure is required for their antiviral role. Although the Plns exhibited low antibacterial activities as mentioned above, during long expose (i.e., 24 h), they may also contribute significantly to the antibacterial activities of cell-free supernatant, together with other AMPs discovered by genomic analysis of NIBR97, which is strongly supported by the proteinase K treatment leading to a considerable decrease (>50%) in antibacterial activity of the cell-free supernatant (Figure 1E). Furthermore, this is confirmed by Figure S4—the cell-free supernatant from the *E. coli* Top10 strain (Invitrogen, Carlsbad, CA, USA), harboring each *Pln* gene cloned, showed significant antibacterial activities against both Gram-negative and Gram-positive bacteria, whereas very little antibacterial activity was detected in the negative control, that is, treatment with the cell-free supernatant from the strain without the *Pln* genes (Figure S4). Meanwhile, the disinfection activity of the cell-free supernatant during short exposures (i.e., within 30 min), as shown in Figure 1C and Table 1, was presumably because the lactic acid may have functioned as a disinfectant during the short exposure. This is supported by the data, showing that the cell-free supernatant contained considerable lactic acids (≈2%) when the NIBR97 strain was cultured in the de Man, Rogosa and Sharpe (MRS) medium, consisting of 5% solutes and 95% water, for 24 h (Figure S5), and by a previous report stating that they induce sudden severe acid stress, leading to a shock of oxidative stress and resulting in the destabilization of

the bacterial membrane [26]. Therefore, the cell-free supernatant may exert its role as a disinfectant, mainly through lactic acid during short exposure (i.e., within 30 min), while it does so through an integrated effect between the lactic acid and the various AMPs during long exposure (i.e., 24 h).

#### **4. Materials and Methods**

#### *4.1. Materials*

As susceptible bacteria to AMPs, *S.* Enteritidis, *S*. Gallinarum, *Edwardsiella tarda, Pasteurella multocida,* and *Streptococcus iniae* were obtained from Dr. Jin Hur (Chonbuk National University, Iksan, Korea) and Dr. Tae Sung Jung (Gyeongsang National University, Jinju, Korea). The *Lactobacillus plantarum* strains KCTC33131, KCTC21004, and KCTC13093, as well as the susceptible bacteria *Escherichia coli* ATCC 10536 and *Staphylococcus aureus* ATCC 6538, were purchased from KCTC (Korean Collection for Type of Cultures, Daejeon, Korea). The human influenza A/H3N2 was provided by the Korea Centers for Disease Control and Prevention (KCDC, Chungcheongbuk-do, Korea). The plasmids for lentiviral packaging (two packaging vectors, pRSV-Rev and pCgpV, and an envelope vector, pCMV-VSV-G) and for a positive control of transduction (pSIH1-H1-siLUC-copGFP) were purchased from Cellbiolab (San Diego, CA, USA) and System Biosciences (Palo Alto, CA, USA), respectively. The five synthetic peptides—plantaricin 1 to 5—were purchased from Cosmogenetech Inc. (Seoul, Korea).

#### *4.2. Analysis of the Minimal Inhibitory Concentration (MIC50 and MIC90)*

*L. plantarum* NIBR97 was incubated at 37 ◦C for 24 h in an MRS liquid medium (10 g/L peptone, 8 g/L meat extract, 4 g/L yeast extract, 20 g/L d(+)-glucose, 2 g/L dipotassium hydrogen phosphate, 5 g/L sodium acetate trihydrate, 2 g/L triammonium citrate, 0.2 g/L magnesium sulfate heptahydrate, and 0.05 g/L manganous sulfate tetrahydrate). The cultural broth was centrifuged for 20 min at 2000× *g*, and the centrifugal supernatant was collected and then sterilized by a 0.22 µm filtration. The sterilized fluid was either applied directly for the examination of antimicrobial activity or fractionated and stored at −80 ◦C until use. The assessment of antimicrobial activity on a microtiter plate was performed by some modification of the dilution assay of Wiegand et al. [27]. The MIC50 and MIC90 were expressed as total proteins equivalent (µg) per volume (mL) of the sample, and the effect of proteinase K treatment was examined by a previously described procedure [28].

#### *4.3. Measurement of Bactericidal Activity*

The susceptible bacterial strains *Escherichia coli* ATCC 10536 and *Staphylococcus aureus* ATCC 6538 were adjusted into 1.5 to 5.0 × 10<sup>8</sup> CFU/mL after pre-culture, and 10% sucrose was used as an interfering agent, 0.25 M KH2PO<sup>4</sup> (pH 7.2) was used as a neutralizing agent, and 20 mg/mL proteinase K was used to degrade the AMPs. For the bactericidal activity assay, we mixed 100 µL of prepared susceptible bacterial solution, 100 µL 10% sucrose, and 800 µL cell-free supernatant (126.6 µg total proteins/mL) from *L. plantarum* NIBR97 and reacted the mixture at 20 ◦C for 5 min. An aliquot (100 µL) of the reaction solution was mixed with 800 µL 0.25 M KH2PO<sup>4</sup> (pH 7.2), 5 µL proteinase K, and 100 µL distilled water, and then reacted at 20 ◦C for 5 min. The surviving cells were counted by serial dilution of the treated solution and incubation on an Luria-Bertani (LB) plate.

#### *4.4. Scanning Electron Microscopy (SEM)*

The *S.* Enteritidis was treated with the cell-free supernatant (70.8 µg total proteins/mL, MIC against *S.* Enteritidis) from *L. plantarum* culture or the synthetic peptides, Pln 3 (1 µg/µL) or Pln 5 (1 µg/µL), for 0, 1, and 8 h, and the lentivirus was assessed with the cell-free supernatant (15.8 µg total proteins/mL) for 5 min and with the synthetic peptides Pln 3 (5 µg/µL) or Pln 5 (5 µg/µL) for 24 h. The treated bacteria and viruses were observed by a scanning electron microscope according to previously described procedures [28].

#### *4.5. Antiviral Analysis Against Influenza A*/*H3N2*

For the antiviral test, we co-incubated 0.1 mL of the A/H3N2 soup (2–4 × 10<sup>5</sup> viruses/µL) with 0.9 mL of the cell-free supernatant (142.5 µg total proteins/mL) for 10 min, 30 min, and 18 h at 25 ◦C. After the co-incubation, the cell-free supernatant-A/H3N2 mixture was 10-fold serially diluted to infect Madin–Darby canine kidney (MDCK) cells (3 × 10<sup>4</sup> cells per well) and, thereafter, the cell culture infectious dose (CCID50) and the viral reduction were determined by cytopathic effect (CPE) and plaque assays, as previously described [29].

#### *4.6. Antiviral Analysis Against GFP-Labeled Lentivirus*

For the production of GFP-labeled lentivirus, we transfected 5 µg of pRSV-Rev, 5 µg of pCMV-VSV-G, 5 µg of pCgpV, and 15 µg of pSIH1-H1-siLUC-copGFP plasmids into HEK-293T cells (6 × 10<sup>6</sup> cells per well) using lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA). The lentiviral supernatants were harvested 72 h after transfection, filtered through Millex-GP 0.45 µm filters (Millipore, Schwalbach, Germany), and concentrated using Retro-Concentin Retroviral Concentration Reagent (System Biosciences, Palo Alto, CA, USA). The titer of lentiviruses was measured with a QuickTiter Lentivirus Titer Kit (Cellbiolabs, San Diego, CA, USA) and stored at −80 ◦C.

For the anti-viral test, we co-incubated 2 µL of lentivirus soup (2.8 × 10<sup>6</sup> lentiviruses/µL) with 2 µL of test sample for 5 min and 24 h at 25 ◦C. After the co-incubation, 2 µL from the total 4 µL of the test sample–lentivirus mixture was infected in HEK-293T cells (1 × 10<sup>4</sup> cells per well). Expression of the copGFP protein was observed at day 3 after infection with a Zeiss 510 fluorescence microscope (Carl Zeiss Co., Oberkochen, Germany).

#### *4.7. Analysis of the Genome*

Genomic analysis of *L. plantarum* NIBR97 was performed by previously described procedures. In detail, genomic DNA from the NIBR97 was extracted and sequenced by previously described procedures [28]. De novo assembly and putative gene coding sequences (CDSs) from the assembled contigs was performed by the hierarchical genome assembly process (HGAP, Version 3) workflow [30] and the bacterial genome was checked by MUMmer 3.5 [31], identifying that the genome comprises a single circular DNA chromosome of 3,022,780 bp with four plasmids by trimming one of the self-similar ends for manual genome closure (Table 2). Putative gene coding sequences (CDSs) from the assembled contigs were identified by Glimmer v3.02 [32], and the obtained ORFs were examined by Blastall alignment (http://www.ncbi.nlm.nih.gov/books/NBK1762). Gene ontology annotations of the ORFs were assigned by Blast2GO software [33]. In addition, ribosomal RNAs and transfer RNAs were separated by RNAmmer 1.2 and tRNAscan-SE 1.4 [34,35]. Finally, the whole-genome sequence data were deposited as Sequence Read Archive (SRA) data in GenBank (SRA no., SRR12344691; BioProject no., PRJNA647132).

#### *4.8. Statistical Analysis*

Themean valueswere separated by the probability difference option according to significant differences. The results are exhibited as least square means with standard deviations. Duncan's multiple range tests (MRT) were applied for verification of significant differences (*p* < 0.05) between sample types. All the analyses were performed by the SAS statistical software package (version 9.1, SAS Inst., Inc., Cary, NC, USA), for which differences were considered significant at *p* < 0.05.

#### **5. Conclusions**

Together, our data showed that the cell-free supernatant from *L. plantarum* NIBR97, producing novel bacteriocins, has superior antibacterial and antiviral activities during both short and long exposures, which suggests that it is potentially useful as a natural material to completely or partially replace chemical disinfectants.

*Pharmaceuticals* **2020**, *13*, 0266

**Supplementary Materials:** The following are available online at http://www.mdpi.com/1424-8247/13/10/0266/s1, Figure S1. Antibacterial activity of cell-free supernatant from *L. plantarum* NIBR97. Figure S2. Overall features of the *L. plantarum* NIBR97 genome (contig 1) and plasmids (contig 2 to 5). Figure S3. Antibacterial activity of synthetic plantaricins identified from the *L. plantarum* NIBR97 genome. Figure S4 Antibacterial activity of the cell-free supernatant from *E. coli.* Top10 strain, harboring each *Pln* gene. Figure S5. The content of lactic acid in the cell-free supernatant from *L. plantarum* NIBR97. Table S1. Identification of ORFs predicted as antimicrobial peptides (AMPs) from the genome assembly data of *L. plantarum* NIBR97. Table S2. Transcriptomic analysis results of AMPs from *L. plantarum* NIBR97.

**Author Contributions:** W.Y.B., I.S.J., and S.W.K. conceived and designed the experiments; S.I.K., D.H.S., S.Y.O., Y.Y., and S.W.K. performed the experiments; C.W.L., Y.K.S., H.-S.Y., and B.-H.L. analyzed the data; S.W.K., B.-H.L., H.-J.A., I.S.J., and W.Y.B contributed reagents/materials/analysis tools; I.S.J. and W.Y.B. wrote the paper. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded mainly by the National Institute of Biological Resources (NIBR), the Ministry of Environment (MOE) of the Republic of Korea, grant number NIBR202019103. I.S.J. was supported by the National Research Foundation of Korea (NRF) grant funded by the Korean government (Ministry of Science and ICT) (no. 2020R1C1C1007796).

**Acknowledgments:** The *S.* Gallinarum, pathogenic *E. coli*, and *S. iniae*that are susceptible to AMPs were obtained from Jin Hur (Chonbuk National University, Iksan, Republic of Korea) and Tae Sung Jung (Gyeongsang National University, Jinju, Republic of Korea).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
